University of Southern California Dissertations and Theses

Molecular mechanisms of young-onset type 2 diabetes: integration of diet and multi-omics biomarkers

(USC Thesis Other)

Molecular mechanisms of young-onset type 2 diabetes: integration of diet and multi-omics biomarkers

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Molecular mechanisms of young-onset type 2 diabetes: integration of diet and multi-omics
biomarkers
by
Elizabeth Costello
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(EPIDEMIOLOGY)
December 2023

ii
Acknowledgments
What an adventure this has been. It’s no exaggeration to say that this would have been
impossible to do alone, and there are so many colleagues, mentors, and friends throughout the
last four and a half years whose support made this possible for me. There are far too many
people for me to thank everyone as they deserve, but I am especially grateful to the following:
First, to my dissertation chair and mentor, Dr. Leda Chatzi: Thank you for guiding me
through this journey, and for always expecting more of me that I would for myself. To just say
thank you doesn’t come close, and I never would have made it through this without you. Thank
you for choosing me as your first PhD student at USC, thank you for showing me how to
navigate academia, and thank you for this incredible opportunity to work on anything and
everything that interests me.
To my dissertation committee, Dr. David Conti, Dr. Tanya Alderete, Dr. Zhanghua Chen,
and Dr. Michael Goran: thank you so much for lending me your expertise in advanced statistical
methods, study design, metabolic disease, and uncountable other things. I am forever grateful
for your patience and flexibility though everything. Your support and feedback throughout the
dissertation process especially have been invaluable.
To Dr. Rob McConnell, Dr. Jim Gauderman, and Nicole Avila for their leadership and
support during my time on the T32. Forever thankful for the opportunity to work with and learn
from so many accomplished scientists and for the opportunity to present my work around the
world.
To my first lab group: Dr. Brittney Baumert, for your mentorship in both science and life,
your friendship, and for letting me share your dog. Dr. Jesse Goodrich for being the go-to
person for analysis advice and edits (and for lending me your R mastery!). And Sarah Rock, for
so many things, including helping me through my first ever manuscript submission, and for
always having the answer to everything.

iii
To my fellow PhD students, colleagues, and friends: Yijie Li, Emily Beglarian, Shudi Pan,
Hailey Hampson, Carmen Chen, and Will Patterson. Thank you all for your feedback,
cheerleading, and, of course, willingness to commiserate (when necessary).
Finally, to my family and friends from “regular” life, thank you thank you thank you for
supporting me throughout this journey in all ways big and small.

iv
Table of Contents
Acknowledgments ........................................................................................................................ ii
List of Tables................................................................................................................................. v
List of Figures .............................................................................................................................. vi
Abstract........................................................................................................................................ vii
Chapter 1: Introduction................................................................................................................ 1
1.1. Epidemiology of Young-Onset Type 2 Diabetes................................................................. 1
1.2. Pathogenesis of Type 2 Diabetes ....................................................................................... 2
1.3. Risk Assessment Using Omics ........................................................................................... 3
1.4. Precision Prevention Approaches for Type 2 Diabetes ...................................................... 5
1.5. Overview .............................................................................................................................. 6
1.6. References........................................................................................................................... 8
Chapter 2: Diet quality is associated with glucose regulation.............................................. 14
2.1. Abstract.............................................................................................................................. 14
2.2. Introduction ........................................................................................................................ 14
2.3. Methods ............................................................................................................................. 16
2.4. Results ............................................................................................................................... 21
2.5. Discussion.......................................................................................................................... 29
2.6. Conclusion ......................................................................................................................... 33
2.7. References......................................................................................................................... 33
2.8. Supplemental Material....................................................................................................... 39
Chapter 3: Proteomic and metabolomic signatures of diet quality...................................... 44
3.1. Abstract.............................................................................................................................. 44
3.2. Introduction ........................................................................................................................ 44
3.3. Methods ............................................................................................................................. 47
3.4. Results ............................................................................................................................... 50
3.5. Discussion.......................................................................................................................... 55
3.6. References......................................................................................................................... 60
3.7. Supplemental Material....................................................................................................... 66
Chapter 4: Metabolites and proteins mediate the relationship between diet and insulin
sensitivity..................................................................................................................................... 85
4.1. Abstract.............................................................................................................................. 85
4.2. Introduction ........................................................................................................................ 87
4.3. Methods ............................................................................................................................. 88
4.4. Results ............................................................................................................................... 95
4.5. Discussion........................................................................................................................ 100
4.6. Conclusion ....................................................................................................................... 102
4.8. Supplemental Material..................................................................................................... 107
Chapter 5: Summary................................................................................................................. 111
5.1. Implications and Next Steps............................................................................................ 113
5.2. Future Directions: Precision Health Approaches ............................................................ 115
5.3. Conclusion ....................................................................................................................... 116
5.4. References....................................................................................................................... 117

v
List of Tables
Table 2.1. Descriptive statistics for participant demographics at baseline and follow up………...22
Table 2.2. Descriptive statistics for diet at baseline, follow up, and change between visits…..…23
Table 2.3. Descriptive statistics for glucose outcomes at baseline, follow up, and change
between visits……………………………………………………………………………………………23
Table 2.4. Descriptive statistics for body composition at baseline, follow up, and change
between visits……………………………………………………………………………..…………..…24
Table 2.5. Estimated effect size and 95% CI for the effect of 1 standard deviation increase
in diet score on body composition………………………………………………………………….…..28
Table S2.1. Results (effects and 95% CIs) for sensitivity analyses at the baseline visit………….40
Table S2.2. Results (effects and 95% CIs) for sensitivity analyses at the follow up visit……….…41
Table S2.3. Results (effects and 95% CIs) for sensitivity analyses for the effects of change
in diet score between visits……………………………………………………………………………..42
Table S2.4. Estimated effect size and 95% CI for the relationship between body
composition and risk for prediabetes/type 2 diabetes……………………………………………..…43
Table 3.1. Descriptive statistics for MetaAIR participant demographics and diet indices………...51
Table S3.1. Proteins significantly associated with at least one diet index………………………….66
Table S3.2. Annotated metabolites significantly associated with at least one diet index……….69
Table S3.3. Proteomic and metabolomic pathways and their contributing features………………70
Table 4.1. Omics features associated with HEI and included in mediation analyses……………..93
Table 4.2. Descriptive statistics for participant demographics, exposure, and outcomes………95
Table 4.3. Associations between baseline HEI and insulin- and glucose-related outcomes……..96
Table 4.4. Comparisons between the base and mediator models………………………………….99
Table S4.1. Results from an omics-wide association study between miRNA and the
Healthy Eating Index 2015……………………………………………………………………….…...107
Table S4.2. Results from causal mediation analyses for HIMA-selected features and
their joint mediation effects…………………………………………………………………………....110

vi
List of Figures
Figure 2.1. Coefficient plots for the effects of diet quality on prediabetes………………………….25
Figure 2.2. Coefficient plots for the effects of diet quality on glucose measurements…………….26
Figure S2.1. Flowchart for study recruitment…………………………………………………………39
Figure 3.1. Chord diagram showing the significant (p < 0.05) associations between
each diet index and each of the proteins and annotated metabolites……………………………….53
Figure 3.2. Significantly enriched proteomics and metabolomics pathways……………………....55
Figure 4.1. Features selected from the early integration and late integration approaches……….98
Figure 4.2. Effect estimates and 95% confidence intervals from causal mediation
analysis for HIMA-selected features and their joint mediation effects…………………………..….99

vii
Abstract
Type 2 diabetes (T2D), traditionally a disease of adults in late middle age or older, is
becoming more prevalent in younger age groups. Young people who develop T2D tend to have
more severe disease that progresses more quickly than older age groups, and T2D is also
associated with more complication at younger ages. High quality diet can reduce the risk for
T2D, prediabetes, and other early signs of insulin sensitivity and glucose dysregulation in all age
groups. However, the biological mechanisms responsible for the protective effects of healthy
diet are not well understood. High-throughput technologies provide an opportunity to investigate
these mechanisms and include omics layer like genomics, transcriptomics, miRNA, proteomics,
and metabolomics. Omics methods have the potential to help track disease progression or
response to interventions and can be used to elucidate mechanisms underlying the relationship
between diet and disease. This dissertation aims to investigate the relationship between diet
quality and risk for prediabetes and insulin resistance, two metabolic states that increase the
risk for T2D, in a cohort of primarily Hispanic young adults in Southern California. I use both
traditional diet assessment methods and omics analyses to evaluate the protective effects of
healthy dietary patterns over four years of follow up, and perform an integrated analysis with
metabolomics, proteomics, and miRNA to investigate potential mechanisms responsible for the
effects of diet on insulin sensitivity.
Here I present results showing that 1) improvements in diet quality are associated with
reduced risk for prediabetes; 2) proteomics and metabolomics can be used to identify molecular
signatures of high-quality diet; and 3) proteins and metabolites may mediate more than half the
total effect of healthy diet on later insulin sensitivity. Diet quality, measured using the Healthy
Eating Index-2015 and the Dietary Approaches to Stop Hypertension (DASH) diet, were highly
protective against prediabetes and were linked to proteins, metabolites, and biological pathways
related to nutrient metabolism, oxidative stress, and inflammation. One coagulation protein and
one polyunsaturated fatty acid were identified as mediators of the relationship between diet and

viii
insulin sensitivity; both molecules have been previously linked to T2D development in older
adults and are involved in inflammation, suggesting that this is a primary pathway by which
insulin sensitivity is influenced by dietary intake. These findings confirm that healthy dietary
patterns reduce the risk for T2D in young adults and identify molecular features and pathways
that may be responsible for this effect. These pathways may be additional targets for precision
medicine approaches to design targeted interventions, monitor disease progression, or predict
future T2D. Future research directions include additional integration analyses for precision
health, including consideration of environmental chemical exposures, particularly those also
associated with dietary intake.

1
Chapter 1: Introduction
1.1. Epidemiology of Young-Onset Type 2 Diabetes
The prevalence of type 2 diabetes (T2D) in adolescents and young adults is a growing
public health concern. Despite T2D incidence appearing to stabilize or even decrease in middleaged and older adults [1], increasing numbers of youth are at risk for or are developing T2D in
the United States [2]. This trend is particularly pronounced in minority populations and mirrors
increases in pediatric obesity [2-4].Health outcomes for young people with T2D may be more
severe than in those diagnosed later in life; youth with T2D experience higher rates of
complications, more comorbidities, and higher mortality risk than older adults .
Young people with T2D are more likely to have significant comorbidities such as nonalcoholic fatty liver disease (NAFLD), obstructive sleep apnea, polycystic ovary syndrome
(PCOS), and psychiatric disorders [5]. Youth are also at higher risk than older adults for microand macrovascular T2D complications, including nephropathy, retinopathy, neuropathy,
hypertension, and microalbuminuria [4]. As a result, major complications like amputations and
kidney failure also occur earlier in those with young-onset T2D [5]. This leads to a substantial
loss of life years: being diagnosed with T2D between age 15 and 24 is associated with a 15
year decrease in life expectancy [6].
These health effects are significant, and their increasing prevalence may lead to
increases in death and disability worldwide. Though part of the increase in young-onset T2D is
due to parallel increases in the risk factors associated with T2D in older adults, youth have
additional unique risks and treatment concerns. Children with T2D progress to beta cell failure
faster than adults with T2D, and treatment is more likely to fail for those with young-onset T2D
[5]. Additionally, insulin resistance increases during puberty, and adolescence and early
adulthood are thus important windows for T2D prevention [7]. In order to intervene effectively, it

2
is critical to assess high-risk youth early in their disease progression and understand the
mechanisms behind both adult T2D and young-onset T2D.
1.2. Pathogenesis of Type 2 Diabetes
Progression from glucose dysregulation to prediabetes to T2D is similar in both older
adults and youth. T2D is characterized by both pancreatic beta cell dysfunction, in which insulin
secretion is insufficient, and insulin resistance, where glucose uptake is impaired [8]. In
someone with normoglycemia, glucose is produced by the liver during fasting or between meals.
After eating, beta cells secrete insulin, which stimulates glucose uptake in liver and muscle
cells. This suppresses hepatic glucose production, and blood glucose eventually returns to
fasting levels [9]. Glucose homeostasis is tightly controlled, and insulin resistance may develop
even as fasting glucose or postprandial glucose remain within normal limits [10]. Insulin
resistance is a sign of departure from normoglycemia often observed years before a T2D
diagnosis, and may occur primarily in hepatic cells or muscle cells [11].
The initial metabolic dysfunction underlying T2D typically appears in one of two ways: as
impaired fasting glucose (IFG) or impaired glucose tolerance (IGT). Both are intermediates in
the progression from normoglycemia to T2D, though there is limited overlap between them and
each represents a different pathophysiology [10]. In IFG, the site of insulin resistance is
primarily the liver whereas in IGT insulin resistance is primarily in muscle cells [9]. In either
scenario, affected people have higher fasting plasma insulin and impaired early insulin response
to glucose, indicating problems with insulin secretion and beta cell function [10]. If there is no
intervention, those with IFG will eventually develop IGT, and vice versa, and ultimately progress
to T2D. Having both IFG and IGT is associated with very high risk for progression to T2D, and
indicates the presence of both hepatic and muscle cell insulin resistance [10].

3
Prediabetes, or intermediate hyperglycemia, is defined by IFG (fasting glucose between
100 and 125 mg/dL), IGT (2-hour glucose between 140 and 199 mg/dL), or hemoglobin A1c (an
indicator of long-term glucose levels) between 5.7 and 6.4% [12]. Young people with
prediabetes are at overall higher risk of T2D than are older adults with prediabetes and are
more likely to have impaired insulin secretion as a result of beta-cell dysfunction [13]. In people
of all ages, prediabetes greatly increases the risk for T2D and is an important target for
prevention [14, 15]. The prevalence of prediabetes has also been increasing in youth in the
United States; within the last three decades, prevalence has increased such that almost one in
four people between the ages of 12 and 34 have prediabetes [16, 17]. The majority of those with
prediabetes are expected to eventually progress to T2D, and about 25% will do so within 3 to 5
years [18]. Even without progression to T2D, prediabetes is itself associated with elevated risk
for diabetic complications, including cardiovascular and kidney disease [15, 18]. It is important
to identify preventable sources of risk for young-onset prediabetes and T2D, and modifiable
behaviors are frequently targeted in preventive measures [15, 19].
1.3. Risk Assessment Using Omics
Novel analytic methods have recently emerged that can offer more information about the
biological effects of dietary intake, disease, pharmaceuticals, and other exposures than have
previously been possible. “Omics” refers to the molecular intermediates responsible for
biological processes, and includes many layers of molecules, including genomics, epigenetics,
proteomics, metabolomics, and lipidomics, among others [29]. While omics features may be
biomarkers of disease, understanding the biological effects and pathway activity of these
molecules may reveal the mechanisms of disease prevention, progression, and treatment.
Multiple omics layers have been examined for biomarkers associated with T2D and
intermediate hyperglycemic states. Genomics studies have identified hundreds of genetic

4
variants that may affect risk for T2D [30], while transcriptomics work has helped link at-risk
genotypes with T2D phenotypes and subtypes [31] and miRNA assessments have identified
potential biomarkers and drug targets [32]. Proteomics and metabolomics, two omics layers
closely related to phenotype, have also produced potential biomarkers for T2D [33].
Dietary intake can also be assessed using omics measurements. Traditional diet
assessment methods ask participants to either to recall their past diet or record it in real time; a
process that is subject to bias [34, 35]. Omics offers a more objective assessment, either by
identifying biomarkers indicative of specific food or food group intake [36] or by describing
metabolic, proteomics, or other omics profiles related to dietary patterns [37, 38]. These
methods can describe alterations in metabolic pathways associated with diet and provide insight
into the biological mechanisms of developing disease.
Data from three omics layers are analyzed in this dissertation: microRNA (miRNA),
proteomics, and metabolomics:
miRNA are short, non-coding RNA strands that control expression of mRNA [39]. They
are important regulatory molecules that can be affected by stress, hormones, and external
exposures like drugs or chemicals [40]. miRNA are involved in drug resistance [41] and may
have a role in carcinogenesis induced by environmental chemicals and exposures [42].
Beneficial nutrients found in the diet have also been linked to miRNA expression, suggesting
possible mechanisms of anti-oxidative or anti-carcinogenic processes associated with healthy
diet [43].
Proteomics is the characterization of the structure and function of all proteins in a given
tissue or biological system [44]. Unlike genomics, which contains the genes that code for
proteins and can be expected to remain constant across the life course, proteins are affected by
external conditions and changes in proteomics are expected to reflect environmental exposures
and health, making proteins good candidates for disease biomarkers [45]. Protein expression

5
can be measured using several techniques, including gel-based identification, mass
spectroscopy, and chromatography, which includes additional techniques to identify proteins
based on size, charge, or affinity for a ligand [44, 46].
Metabolomics aims to characterize all metabolites and their interactions within a
biological system [47]. High-resolution mass spectroscopy (HRMS) combined with ultra-high
performance liquid chromatography can detect thousands of metabolites in human tissues [48,
49]. Analyses may be targeted or untargeted: targeted metabolomics can offer quantification of
known metabolites based on analytical standards, while untargeted metabolomics may describe
the ‘metabolic fingerprint’ of a biological system and provide information about the activity of
metabolic pathways [47, 49]. These methods were originally used to identify biomarkers
predictive of disease, and more recently to describe the effects of environmental exposures on
metabolic pathways [47, 50].
1.4. Precision Prevention Approaches for Type 2 Diabetes
The advent of omics technologies and integrated data analysis methods has rapidly
increased the viability of “precision medicine” approaches to clinical practice. Precision medicine
refers to a clinical approach that takes into account an individual’s unique biology, exposures,
and behaviors to tailor personalized interventions [51]. Omics allow researchers and clinicians
to characterize patients’ gene expression, metabolic state, environmental exposures, and
response to treatment in ways that were previously unimaginable. Analytical methods that
integrate omics signatures into traditional epidemiology or clinical risk assessment create
opportunities to explore the mechanisms responsible for disease development or identify
subgroups of patients at increased risk for disease.
For several years genetics, epigenetics, metabolomics, and other omics have been
targets of investigation for T2D diagnosis, prevention, treatment, and monitoring. For example,
patients with certain genetic subtypes respond better to specific T2D treatments [52] and omics

6
biomarkers have been found to be predictive of T2D [53]. There is additional potential for omics
features to inform risk profiles for T2D and related diseases, using novel integrative analytical
approaches like LUCID [54], HIMA [55], and JIVE [56]. Causal mediation analyses may identify
molecular mechanisms that link environmental exposures, like diet, and pharmaceutical
interventions to T2D progression and prevention [57, 58]. However, integration of multiple omics
layers in T2D research is still limited.
1.5. Overview
The research described in this dissertation was conducted in a cohort of primarily
Hispanic (>50%) young adults in Southern California. Between 2014 and 2018, participants
were recruited from the Southern California Children’s Health Study (CHS) [59] to participate in
the MetaAIR study, which examined the impact of air pollution on metabolic disease [60]. In
MetaAIR, 155 participants between the ages of 17 and 22 provided information about their
health, lifestyle, and diet, and completed a clinical visit which included an oral glucose tolerance
test (OGTT) and dual x-ray absorptiometry (DEXA) scan. In 2020, 140 of these participants
were invited to return for a follow up visit for the “MetaCHEM” study, which would examine the
effects of chemical exposures on risk for T2D and other metabolic diseases. MetaCHEM
participants again provided information on lifestyle and diet, and completed another clinical visit
with OGTT and DEXA scan. MetaCHEM recruitment continued until early 2022, when 89
participants had returned for the follow up visit. Specific details of exposure and outcome
assessments are provided in each chapter.
The purpose of this dissertation is to assess the protective effects of diet in young-onset
T2D and explore mechanisms that may underly this relationship through the following aims: 1)
describe the relationship between dietary changes in young adults and risk for prediabetes and
related outcomes; 2) identify metabolomic and proteomic signatures of diet quality; and 3)

7
determine if omics biomarkers mediate the relationship between diet quality and insulin
resistance.
In Chapter 2, I evaluated the relationship between diet quality and markers of glucose
homeostasis and body composition in the MetaAIR-MetaCHEM cohort, in a study published in
Nutrients [61] and a reprint book, Diet Quality and Risk of Cardiometabolic and Diabetes
(ISBN 978-3-0365-8774-5). I calculated four diet indices, the Healthy Eating Index-2015 (HEI),
Dietary Approaches to Stopping Hypertension (DASH), Mediterranean Diet Score (MDS), and
Diet Inflammatory Index (DII), at both baseline and follow up in order to examine the effects of
diet on IFG, IGT, prediabetes, and adiposity. Both cross-sectional and longitudinal analyses
were conducted to examine both the relationships between diet and T2D-related outcomes at
the same time points and the relationship between changes in diet and changes in outcomes
during the four-year follow up period. These results provided evidence that higher diet quality is
protective against prediabetes, and that improvements in diet quality greatly reduced the risk for
prediabetes and glucose dysregulation.
In Chapter 3, two diet indices (HEI and DASH) were further investigated for associations
with proteomics and metabolomics features and their corresponding biological pathways.
Several proteins and metabolites were associated with both HEI and DASH, and may represent
biomarkers of overall diet quality. Selected metabolites included fatty acids, bile acids, and
amino acid metabolites, as well as several pesticides. Functional pathway analyses indicated
that these proteins and metabolites have biological functions related to macronutrient
metabolism, oxidative stress, inflammation, and immune function.
In Chapter 4, I examined the effects of the baseline HEI score on insulin sensitivity four
years later, and performed high-dimensional and traditional causal mediation analyses to
determine if omics biomarkers might mediate these relationships. Candidate mediators included
the proteins and annotated metabolites found to be associated with HEI in Chapter 3, as well as
miRNA associated with HEI. Both early and late integration approaches to high dimensional

8
mediation were used to select significant mediators of the relationship between HEI and the
Matsuda Index. Mediators selected by both approaches were further investigated using causal
mediation to determine the mediation effects and proportion of the HEI-Matsuda Index mediated
by each biomarker together and separately. Two biomarkers, a polyunsaturated fatty acid and a
coagulation protein, were found to mediate more than half of the effect of HEI on insulin
sensitivity.
In the final chapter, I summarize the overall conclusions of these projects and provide an
overview of future directions and implications. This study confirmed that diet is a strong risk
factor in the development of prediabetes and T2D in this cohort of young adults, and that
metabolites and proteins and may explain the majority of the total effect of diet on diabetes
development. Future studies in larger populations spanning greater age ranges are needed to
confirm these findings and use omics to bring precision medicine approaches for T2D from
theory into practice.
1.6. References
1. Magliano DJ, Islam RM, Barr ELM, Gregg EW, Pavkov ME, Harding JL, Tabesh M, Koye
DN, Shaw JE. Trends in incidence of total or type 2 diabetes: systematic review. BMJ
2019;366:l5003. doi: 10.1136/bmj.l5003.
2. Divers J, Mayer-Davis EJ, Lawrence JM, Isom S, Dabelea D, Dolan L, Imperatore G,
Marcovina S, Pettitt DJ, Pihoker C, et al. Trends in Incidence of Type 1 and Type 2
Diabetes Among Youths - Selected Counties and Indian Reservations, United States,
2002-2015. MMWR Morb Mortal Wkly Rep 2020;69(6):161-5. doi:
10.15585/mmwr.mm6906a3.
3. Shah AS, Nadeau KJ. The changing face of paediatric diabetes. Diabetologia
2020;63(4):683-91. doi: 10.1007/s00125-019-05075-6.
4. Magliano DJ, Sacre JW, Harding JL, Gregg EW, Zimmet PZ, Shaw JE. Young-onset
type 2 diabetes mellitus - implications for morbidity and mortality. Nat Rev Endocrinol
2020;16(6):321-31. doi: 10.1038/s41574-020-0334-z.
5. Savic Hitt TA, Katz LEL. Pediatric Type 2 Diabetes: Not a Mini Version of Adult Type 2
Diabetes. Endocrinology and Metabolism Clinics of North America 2020;49(4):679-93.
doi: https://doi.org/10.1016/j.ecl.2020.08.003.

9
6. Rhodes ET, Prosser LA, Hoerger TJ, Lieu T, Ludwig DS, Laffel LM. Estimated morbidity
and mortality in adolescents and young adults diagnosed with Type 2 diabetes mellitus.
Diabetic Medicine 2012;29(4):453-63. doi: https://doi.org/10.1111/j.1464-
5491.2011.03542.x.
7. Wilmot E, Idris I. Early onset type 2 diabetes: risk factors, clinical impact and
management. Therapeutic Advances in Chronic Disease 2014;5(6):234-44. doi:
10.1177/2040622314548679.
8. DeFronzo RA. The Triumvirate: β-Cell, Muscle, Liver: A Collusion Responsible for
NIDDM. Diabetes 1988;37(6):667-87. doi: 10.2337/diab.37.6.667.
9. Ferrannini E, Gastaldelli A, Iozzo P. Pathophysiology of prediabetes. Med Clin North Am
2011;95(2):327-39, vii-viii. doi: 10.1016/j.mcna.2010.11.005.
10. Abdul-Ghani MA, Tripathy D, DeFronzo RA. Contributions of beta-cell dysfunction and
insulin resistance to the pathogenesis of impaired glucose tolerance and impaired
fasting glucose. Diabetes Care 2006;29(5):1130-9. doi: 10.2337/diacare.2951130.
11. Thomas DD, Corkey BE, Istfan NW, Apovian CM. Hyperinsulinemia: An Early Indicator
of Metabolic Dysfunction. J Endocr Soc 2019;3(9):1727-47. doi: 10.1210/js.2019-00065.
12. ADA. 2. Classification and Diagnosis of Diabetes: Standards of Medical Care in
Diabetes—2020. Diabetes Care 2019;43(Supplement_1):S14-S31. doi: 10.2337/dc20-
S002.
13. Bacha F, Lee S, Gungor N, Arslanian SA. From Pre-Diabetes to Type 2 Diabetes in
Obese Youth: Pathophysiological characteristics along the spectrum of glucose
dysregulation. Diabetes Care 2010;33(10):2225-31. doi: 10.2337/dc10-0004.
14. Morris DH, Khunti K, Achana F, Srinivasan B, Gray LJ, Davies MJ, Webb D. Progression
rates from HbA1c 6.0-6.4% and other prediabetes definitions to type 2 diabetes: a metaanalysis. Diabetologia 2013;56(7):1489-93. doi: 10.1007/s00125-013-2902-4.
15. Tabák AG, Herder C, Rathmann W, Brunner EJ, Kivimäki M. Prediabetes: a high-risk
state for diabetes development. Lancet 2012;379(9833):2279-90. doi: 10.1016/s0140-
6736(12)60283-9.
16. Liu J, Li Y, Zhang D, Yi SS, Liu J. Trends in Prediabetes Among Youths in the US From
1999 Through 2018. JAMA Pediatrics 2022. doi: 10.1001/jamapediatrics.2022.0077.
17. Andes LJ, Cheng YJ, Rolka DB, Gregg EW, Imperatore G. Prevalence of Prediabetes
Among Adolescents and Young Adults in the United States, 2005-2016. JAMA Pediatr
2020;174(2):e194498. doi: 10.1001/jamapediatrics.2019.4498.
18. Hostalek U. Global epidemiology of prediabetes - present and future perspectives. Clin
Diabetes Endocrinol 2019;5:5-. doi: 10.1186/s40842-019-0080-0.
19. ADA. 3. Prevention or Delay of Type 2 Diabetes: Standards of Medical Care in Diabetes2020. Diabetes Care 2020;43(Suppl 1):S32-s6. doi: 10.2337/dc20-S003.

10
20. Kolb H, Martin S. Environmental/lifestyle factors in the pathogenesis and prevention of
type 2 diabetes. BMC Medicine 2017;15(1):131. doi: 10.1186/s12916-017-0901-x.
21. Lascar N, Brown J, Pattison H, Barnett AH, Bailey CJ, Bellary S. Type 2 diabetes in
adolescents and young adults. Lancet Diabetes Endocrinol 2018;6(1):69-80. doi:
10.1016/s2213-8587(17)30186-9.
22. Viner R, White B, Christie D. Type 2 diabetes in adolescents: a severe phenotype posing
major clinical challenges and public health burden. The Lancet 2017;389(10085):2252-
60. doi: https://doi.org/10.1016/S0140-6736(17)31371-5.
23. Kodama K, Tojjar D, Yamada S, Toda K, Patel CJ, Butte AJ. Ethnic Differences in the
Relationship Between Insulin Sensitivity and Insulin Response: A systematic review and
meta-analysis. Diabetes Care 2013;36(6):1789-96. doi: 10.2337/dc12-1235.
24. Ali O. Genetics of type 2 diabetes. World J Diabetes 2013;4(4):114-23. doi:
10.4239/wjd.v4.i4.114.
25. Qian F, Liu G, Hu FB, Bhupathiraju SN, Sun Q. Association Between Plant-Based
Dietary Patterns and Risk of Type 2 Diabetes: A Systematic Review and Meta-analysis.
JAMA Internal Medicine 2019;179(10):1335-44. doi: 10.1001/jamainternmed.2019.2195.
26. Neuenschwander M, Ballon A, Weber KS, Norat T, Aune D, Schwingshackl L,
Schlesinger S. Role of diet in type 2 diabetes incidence: umbrella review of metaanalyses of prospective observational studies. BMJ 2019;366:l2368. doi:
10.1136/bmj.l2368.
27. de Koning L, Chiuve SE, Fung TT, Willett WC, Rimm EB, Hu FB. Diet-quality scores and
the risk of type 2 diabetes in men. Diabetes Care 2011;34(5):1150-6. doi: 10.2337/dc10-
2352.
28. Salas-Salvadó J, Bulló M, Estruch R, Ros E, Covas MI, Ibarrola-Jurado N, Corella D,
Arós F, Gómez-Gracia E, Ruiz-Gutiérrez V, et al. Prevention of diabetes with
Mediterranean diets: a subgroup analysis of a randomized trial. Ann Intern Med
2014;160(1):1-10. doi: 10.7326/m13-1725.
29. Franks PW, Pomares-Millan H. Next-generation epidemiology: the role of high-resolution
molecular phenotyping in diabetes research. Diabetologia 2020;63(12):2521-32. doi:
10.1007/s00125-020-05246-w.
30. DeForest N, Majithia AR. Genetics of Type 2 Diabetes: Implications from Large-Scale
Studies. Curr Diab Rep 2022;22(5):227-35. doi: 10.1007/s11892-022-01462-3.
31. Tonyan ZN, Nasykhova YA, Danilova MM, Barbitoff YA, Changalidi AI, Mikhailova AA,
Glotov AS. Overview of Transcriptomic Research on Type 2 Diabetes: Challenges and
Perspectives. Genes (Basel) 2022;13(7). doi: 10.3390/genes13071176.
32. Kim M, Zhang X. The Profiling and Role of miRNAs in Diabetes Mellitus. J Diabetes Clin
Res 2019;1(1):5-23. doi: 10.33696/diabetes.1.003.

11
33. Chen ZZ, Gerszten RE. Metabolomics and Proteomics in Type 2 Diabetes. Circ Res
2020;126(11):1613-27. doi: 10.1161/circresaha.120.315898.
34. Bingham SA. Limitations of the Various Methods for Collecting Dietary Intake Data.
Annals of Nutrition and Metabolism 1991;35(3):117-27. doi: 10.1159/000177635.
35. Shim J-S, Oh K, Kim HC. Dietary assessment methods in epidemiologic studies.
Epidemiol Health 2014;36:e2014009-e. doi: 10.4178/epih/e2014009.
36. Lampe JW, Huang Y, Neuhouser ML, Tinker LF, Song X, Schoeller DA, Kim S, Raftery
D, Di C, Zheng C, et al. Dietary biomarker evaluation in a controlled feeding study in
women from the Women's Health Initiative cohort. Am J Clin Nutr 2017;105(2):466-75.
doi: 10.3945/ajcn.116.144840.
37. Andraos S, Beck KL, Jones MB, Han T-L, Conlon CA, de Seymour JV. Characterizing
patterns of dietary exposure using metabolomic profiles of human biospecimens: a
systematic review. Nutrition Reviews 2022;80(4):699-708. doi: 10.1093/nutrit/nuab103.
38. Bourgonje AR, Bolte LA, Vranckx LLC, Spekhorst LM, Gacesa R, Hu S, van Dullemen
HM, Visschedijk MC, Festen EAM, Samsom JN, et al. Long-Term Dietary Patterns Are
Reflected in the Plasma Inflammatory Proteome of Patients with Inflammatory Bowel
Disease. Nutrients 2022;14(12). doi: 10.3390/nu14122522.
39. Cai Y, Yu X, Hu S, Yu J. A brief review on the mechanisms of miRNA regulation.
Genomics Proteomics Bioinformatics 2009;7(4):147-54. doi: 10.1016/s1672-
0229(08)60044-3.
40. Gulyaeva LF, Kushlinskiy NE. Regulatory mechanisms of microRNA expression. Journal
of Translational Medicine 2016;14(1):143. doi: 10.1186/s12967-016-0893-x.
41. Egeland NG, Lunde S, Jonsdottir K, Lende TH, Cronin-Fenton D, Gilje B, Janssen EA,
Søiland H. The Role of MicroRNAs as Predictors of Response to Tamoxifen Treatment
in Breast Cancer Patients. Int J Mol Sci 2015;16(10):24243-75. doi:
10.3390/ijms161024243.
42. Chen T. The role of MicroRNA in chemical carcinogenesis. J Environ Sci Health C
Environ Carcinog Ecotoxicol Rev 2010;28(2):89-124. doi:
10.1080/10590501.2010.481477.
43. Milenkovic D, Jude B, Morand C. miRNA as molecular target of polyphenols underlying
their biological effects. Free Radic Biol Med 2013;64:40-51. doi:
10.1016/j.freeradbiomed.2013.05.046.
44. Al-Amrani S, Al-Jabri Z, Al-Zaabi A, Alshekaili J, Al-Khabori M. Proteomics: Concepts
and applications in human medicine. World J Biol Chem 2021;12(5):57-69. doi:
10.4331/wjbc.v12.i5.57.
45. He B, Huang Z, Huang C, Nice EC. Clinical applications of plasma proteomics and
peptidomics: Towards precision medicine. Proteomics Clin Appl 2022;16(6):e2100097.
doi: 10.1002/prca.202100097.

12
46. Aslam B, Basit M, Nisar MA, Khurshid M, Rasool MH. Proteomics: Technologies and
Their Applications. Journal of Chromatographic Science 2017;55(2):182-96. doi:
10.1093/chromsci/bmw167.
47. Johnson CH, Ivanisevic J, Siuzdak G. Metabolomics: beyond biomarkers and towards
mechanisms. Nat Rev Mol Cell Biol 2016;17(7):451-9. doi: 10.1038/nrm.2016.25.
48. Muthubharathi BC, Gowripriya T, Balamurugan K. Metabolomics: small molecules that
matter more. Mol Omics 2021;17(2):210-29. doi: 10.1039/d0mo00176g.
49. Alves S, Paris A, Rathahao-Paris E. Chapter Four - Mass spectrometry-based
metabolomics for an in-depth questioning of human health. Edtion ed. In: Makowski GS,
ed. Advances in Clinical Chemistry: Elsevier, 2020:147-91.
50. Rinschen MM, Ivanisevic J, Giera M, Siuzdak G. Identification of bioactive metabolites
using activity metabolomics. Nat Rev Mol Cell Biol 2019;20(6):353-67. doi:
10.1038/s41580-019-0108-4.
51. Goetz LH, Schork NJ. Personalized medicine: motivation, challenges, and progress.
Fertil Steril 2018;109(6):952-63. doi: 10.1016/j.fertnstert.2018.05.006.
52. Gloyn AL, Drucker DJ. Precision medicine in the management of type 2 diabetes. Lancet
Diabetes Endocrinol 2018;6(11):891-900. doi: 10.1016/s2213-8587(18)30052-4.
53. Carrasco-Zanini J, Pietzner M, Wheeler E, Kerrison ND, Langenberg C, Wareham NJ.
Multi-omic prediction of incident type 2 diabetes. Diabetologia 2023. doi:
10.1007/s00125-023-06027-x.
54. Peng C, Wang J, Asante I, Louie S, Jin R, Chatzi L, Casey G, Thomas DC, Conti DV. A
latent unknown clustering integrating multi-omics data (LUCID) with phenotypic traits.
Bioinformatics 2020;36(3):842-50. doi: 10.1093/bioinformatics/btz667.
55. Zhang H, Zheng Y, Zhang Z, Gao T, Joyce B, Yoon G, Zhang W, Schwartz J, Just A,
Colicino E, et al. Estimating and testing high-dimensional mediation effects in epigenetic
studies. Bioinformatics 2016;32(20):3150-4. doi: 10.1093/bioinformatics/btw351.
56. Lock EF, Hoadley KA, Marron JS, Nobel AB. JOINT AND INDIVIDUAL VARIATION
EXPLAINED (JIVE) FOR INTEGRATED ANALYSIS OF MULTIPLE DATA TYPES. Ann
Appl Stat 2013;7(1):523-42. doi: 10.1214/12-aoas597.
57. Xu M, Zheng J, Hou T, Lin H, Wang T, Wang S, Lu J, Zhao Z, Li M, Xu Y, et al. SGLT2
Inhibition, Choline Metabolites, and Cardiometabolic Diseases: A Mediation Mendelian
Randomization Study. Diabetes Care 2022;45(11):2718-28. doi: 10.2337/dc22-0323.
58. Dreyfuss JM, Yuchi Y, Dong X, Efthymiou V, Pan H, Simonson DC, Vernon A, Halperin
F, Aryal P, Konkar A, et al. High-throughput mediation analysis of human proteome and
metabolome identifies mediators of post-bariatric surgical diabetes control. Nat Commun
2021;12(1):6951. doi: 10.1038/s41467-021-27289-2.
59. Gauderman WJ, Avol E, Gilliland F, Vora H, Thomas D, Berhane K, McConnell R,
Kuenzli N, Lurmann F, Rappaport E, et al. The Effect of Air Pollution on Lung

13
Development from 10 to 18 Years of Age. New England Journal of Medicine
2004;351(11):1057-67. doi: 10.1056/NEJMoa040610.
60. Kim JS, Chen Z, Alderete TL, Toledo-Corral C, Lurmann F, Berhane K, Gilliland FD.
Associations of air pollution, obesity and cardiometabolic health in young adults: The
Meta-AIR study. Environment International 2019;133:105180. doi:
https://doi.org/10.1016/j.envint.2019.105180.
61. Costello E, Goodrich J, Patterson WB, Rock S, Li Y, Baumert B, Gilliland F, Goran MI,
Chen Z, Alderete TL, et al. Diet Quality Is Associated with Glucose Regulation in a
Cohort of Young Adults. Nutrients 2022;14(18):3734.

14
Chapter 2: Diet quality is associated with glucose regulation
2.1. Abstract
Young-onset type 2 diabetes and prediabetes is a growing epidemic. Poor diet is a known risk
factor for T2D in older adults, but the contribution of diet to risk factors for T2D is not welldescribed in youth. Our objective was to examine the relationship of diet quality with
prediabetes, glucose regulation, and adiposity in young adults. A cohort of young adults (n =
155, age 17–22) was examined between 2014–2018, and 89 underwent a follow-up visit from
2020–2022. At each visit, participants completed diet and body composition assessments and
an oral glucose tolerance test. Adherence to four dietary patterns was assessed: Dietary
Approaches to Stop Hypertension (DASH), Healthy Eating Index (HEI), Mediterranean diet, and
Diet Inflammatory Index (DII). Regression analyses were used to determine adjusted
associations of diet with risk for prediabetes and adiposity. Each one-point increase in DASH or
HEI scores between visits reduced the risk for prediabetes at follow-up by 64% (OR, 95% CI:
0.36, 0.17–0.68) and 9% (OR, 95% CI: 0.91, 0.85–0.96), respectively. The DASH diet was
inversely associated with adiposity, while DII was positively associated with adiposity. In
summary, positive changes in HEI and DASH scores were associated with reduced risk for
prediabetes in young adults.
2.2. Introduction
The prevalence of prediabetes, a condition where blood glucose levels are elevated but
below diagnostic cut-offs for type 2 diabetes (T2D) [1], is increasing in adolescents and young
adults in the United States (U.S.) [2,3]. Prediabetes greatly increases the risk for T2D [4];
therefore, T2D incidence is also increasing in the U.S., following a similar trend [5]. This trend is
of considerable concern because T2D is often more aggressive in youth than in older adults and

15
is associated with higher rates of complications, more comorbidities, and higher mortality risk
[6,7]. Disparities also exist in T2D risk, with Hispanics and other racial or ethnic minorities at
higher risk compared to non-Hispanic Whites [5,7,8]. Lifestyle is a source of modifiable risk
factors frequently targeted in preventive measures [1,9], of which diet is especially important.
Depending on quality, diet may be either protective against or a risk factor for prediabetes and
T2D [10–12]. Healthy dietary patterns high in fruits, vegetables, and whole grains and low in
sodium, saturated fat, and added sugars are associated with reduced risk for prediabetes and
T2D [10,13–15]. In middle-aged and older adults, adherence to healthy eating patterns such as
the Mediterranean diet, Dietary Approaches to Stop Hypertension (DASH) diet, and federal
dietary recommendations reduces the risk for T2D [13,14,16]. The Dietary Inflammatory Index
(DII), an alternative dietary pattern that quantifies the inflammatory effects of dietary intake, is
linked with prediabetes and T2D, where more pro-inflammatory diets are associated with
increased risk [17,18]. However, most studies evaluating the relationship between diet and T2D
risk have been conducted in middle-aged or older adults or only incidentally included young
adults [11,19,20]. Less is understood about the impact of diet quality or dietary changes on T2D
risk in young adulthood.
Few prospective studies have examined the relationships between the DASH diet,
Mediterranean diet, or other dietary patterns and T2D in youth [21–23]. Findings in children and
adolescents suggest that increased adherence to the DASH diet may improve cardiovascular
and metabolic risk factors [21] and that weight control is critical in T2D prevention [22,24].
Limited studies exist on the development of T2D in young adults [25–28] though this life stage
may represent a critical window for behavior change and diabetes prevention, as young people
transition from their adolescent years into independent adulthood [29].
The purpose of this study was to examine the relationship between diet quality and risk
for T2D in a cohort of primarily Hispanic young adults. Participants were evaluated for glucose
dysregulation and diet quality at age 17–22 and again after approximately four years. Glucose

16
regulation was assessed using hemoglobin A1c (HbA1c) and two-hour oral glucose tolerance
tests (OGTTs). We hypothesize that higher diet quality will be protective against glucose
dysregulation and that improvement in diet quality between visits will reduce the risk for
prediabetes and type 2 diabetes.
2.3. Methods
2.3.1 Cohort
Between 2014 and 2018, a subset of 158 participants between 17 and 22 years old were
recruited from the Children’s Health Study (CHS) [30] for the Meta-AIR study [31]. Subjects
were selected if they had overweight or obesity in early adolescence, had not been diagnosed
with type 1 or type 2 diabetes, had no medical conditions, and were taking no medications that
affect glucose metabolism [31]. Between January 2020 and March 2022, 140 of these
participants were invited to participate in a follow-up visit. All but 7 participants underwent
follow-up testing during the COVID-19 pandemic. All study visits were completed at the
Diabetes and Obesity Research Institute at the University of Southern California. This study was
approved by the University of Southern California Institutional Review Board. Written informed
consent was obtained from participants at both baseline and follow-up visits or by participants
and their guardians for those under age 18 at baseline.
Of the 158 participants at baseline, 155 had diet data and data for at least one outcome.
Eighty-six of these participated in the follow-up (Figure S1). An additional three CHS
participants without baseline data completed the follow-up visit.
2.3.2. Glucose outcomes
A two-hour oral glucose tolerance test (OGTT) was performed at each visit, using a
glucose load of 1.75 g of glucose per kg of body mass (max 75 g). Blood was sampled before
the glucose challenge and at 30-, 60-, 90- and 120-minute post-challenge. Glucose

17
concentrations at each time point were measured in plasma. Fasting glucose was also
measured using a glucometer before the glucose challenge, and the OGTT was not completed
if participants had a fasting value greater than 126 mg/dL. Hemoglobin A1c (HbA1c) was
measured in fasting whole blood samples. Glucose area under the curve (AUC) was calculated
from the five glucose measurements using the trapezoidal method [32].
Prediabetes and type 2 diabetes were based on clinical cutoffs for HbA1c, fasting
plasma glucose, or 2-hour plasma glucose [33]. Participants having either HbA1c values of
6.5% or higher, fasting glucose of 126 mg/dL or higher, or 2-hour glucose of 200 mg/dL or
higher were considered to have type 2 diabetes, while those with HbA1c between 5.7% and
6.4%, fasting glucose between 100 mg/dL and 125 mg/dL, or 2-hour glucose between 140
mg/dL and 199 mg/dL were categorized as prediabetic.
2.3.3. Adiposity outcomes
Weight and height were measured at each visit, and BMI calculated as kg per meter
squared (kg/m2). Body composition was assessed using dual-energy X-ray absorptiometry
(DEXA) whole body scans. Baseline scans were performed on either a Hologic QDR 4500W or
Horizon W machine at baseline, while all follow-up scans were performed on the Horizon W.
Body composition measures included total body fat percentage, fat mass to height ratio, fat free
mass index (FFMI, kg/m2), android to gynoid ratio, trunk to leg ratio, trunk to limb ratio, and
visceral adipose tissue (VAT, in3). Only body fat percentage and fat mass to height ratio were
measured on the QDR 4500W machine.
2.3.4. Diet assessment
At each visit, participants were asked to complete two 24-hour dietary recalls on nonconsecutive days: one weekday and one weekend day. Baseline recalls were completed by
trained interviewers using the Nutritional Data System for Research (NDSR) software version

18
2014, developed by the Nutrition Coordinating Center (University of Minnesota, Minneapolis,
MN, USA) [34], while follow-up recalls used the Automated Self-Administered 24-hour (ASA24)
Dietary Assessment Tool, version (2018), developed by the National Cancer Institute, Bethesda,
MD, USA [35]. An average of the values from both days was calculated for each diet
component. At baseline, 16 participants (10.3%) completed only one recall, and 9 (10.2%)
completed only one recall at follow-up. If a participant did not complete both recalls, values from
the single recall were used.
Four diet indices were calculated from the recall data at both the baseline and follow-up
visits: the 2015 Healthy Eating Index (HEI), DASH score, Mediterranean Diet Score (MDS), and
DII. The HEI ranges from 0–100, is based on adherence to the United States Department of
Agriculture (USDA) 2015 Dietary Guidelines [36] and contains the following thirteen elements
standardized to calorie intake: total fruit, whole fruit, total vegetables, greens and beans, whole
grains, refined grains, dairy, total protein foods, seafood and plant proteins, mono- and
polyunsaturated fatty acids, saturated fats, sodium, and added sugars. The DASH scoring
method follows the calculation proposed by Mellen et al. [37], using nutrient goals for DASH diet
adherence. This DASH score ranges from 0 to 8 and includes the following elements
standardized to calorie intake: protein, fiber, magnesium, calcium, potassium, total fat, saturated
fat, cholesterol, and sodium. One point was assigned if the nutrient goal was met, and half of a
point was assigned if an intermediate nutrient goal was met. The MDS was calculated using the
method developed by Trichopoulou et al. [38], which ranges from 0 to 9 with ten components:
vegetables, legumes, fruits and nuts, dairy, cereals, meat, poultry, fish, alcohol, and ratio of
mono- to saturated fats. For each component, one point was assigned for exceeding the sexspecific median. The DII was adapted from Shivappa et al. [39], with negative values indicating
an anti-inflammatory diet and positive values indicating a pro-inflammatory diet. For each
element, a centered percentile was calculated by comparing the reported intake to a global
mean and standard deviation of intake. This centered percentile was multiplied by the element’s

19
overall inflammatory effect score, and the scores for all elements were summed to produce the
DII score. Twenty-eight of the forty-five elements from Shivappa et al. [39] were included:
alcohol, beta-carotene, caffeine, carbohydrates, cholesterol, calories, total fat, fiber, iron,
magnesium, folic acid, mono- and polyunsaturated fatty acids, omega-3 and omega-6 fatty
acids, protein, saturated fat, selenium, zinc, and vitamins A, B1 (thiamin), B2 (riboflavin), B3
(niacin), C, D, and E. The remaining elements were excluded because they are not specifically
captured by the recalls systems used in this study. Trans fats were banned by the United States
Food and Drug Administration in 2015, with a 2018 deadline for implementation [40], and were
excluded from the DII calculation in the follow-up visit.
2.3.5. Covariates
Questionnaires were administered to collect sociodemographic information, including
age, sex, race and ethnicity, physical activity, and parental education. Ethnicity was categorized
as non-Hispanic White, Hispanic, or Other. Parental education was categorized as “Did not
complete high school”, “Completed high school”, “Completed more than high school”, or “Don’t
know”. At baseline, physical activity was assessed as a binary variable, where participants
responded yes or no to the question “Do you exercise?”. At follow-up, physical activity was
assessed using the International Physical Activity Questionnaire Short Form [41], and metabolic
equivalent of task (MET) minutes were calculated according to the scoring guidelines.
Participants were considered to have “high” physical activity if they met either of the following
criteria: (1) reported vigorous physical activity (VPA) three or more days per week and 1500 or
more MET minutes per week or (2) seven days of any combination of VPA, moderate physical
activity (MPA), and walking for at least 3000 MET minutes. Participants were considered to
have “moderate” physical activity if they (1) reported at least 3 days of VPA where the activity
lasted at least 30 min or (2) five or more days of MPA or walking where the activity lasted at
least 30 min or (3) five or more days of some combination of VPA, MPA, and walking for at least

20
600 MET minutes. Participants were categorized as having “low” physical activity if they did not
meet any of these criteria.
2.3.6. Statistical analysis.
Descriptive statistics were calculated for all outcomes and exposures. Pearson’s
correlations were calculated between the four diet scores at each visit separately and between
time points. Independent two-sample t-tests, chi-square tests, or Fisher’s exact tests were used
to test for differences in participant demographics between the baseline cohort and follow-up
cohort. Paired t-tests or McNemar–Bowker tests were used to test for differences in exposures
and outcomes between visits. Due to the small numbers of participants with values meeting the
diagnostic criteria for type 2 diabetes, prediabetes and diabetes were combined into one
category (prediabetes/T2D) for analysis. Primary outcomes of interest were those related to
glucose regulation: prediabetes/T2D, fasting glucose, 2-hour glucose, glucose AUC, and
HbA1c. Body composition measurements were secondary outcomes: BMI, body fat percent,
FFMI, fat mass to height ratio, android to gynoid ratio, trunk to leg ratio, trunk to limb ratio, and
VAT.
Cross-sectional analyses were performed for both baseline and follow-up visits, using
multivariable linear regression for continuous outcomes and logistic regression for
prediabetes/T2D. For longitudinal analyses, change in diet indices from baseline to follow-up
was modeled against change in outcome using linear regression for continuous outcomes, or
against diabetes at follow-up using logistic regression. Longitudinal models also adjusted for
baseline diet score. Beta coefficients for exposures were scaled to one standard deviation (SD)
of the exposure to account for the differing scales.
All analyses included the following covariates: age, ethnicity, physical activity, energy
intake, and parental education. Because these factors were not accounted for in the scoring
system, analyses with HEI, DASH, and DII scores additionally controlled for sex, and analyses

21
with MDS additionally controlled for energy intake. BMI and body fat percent were presumed to
be on the causal pathway between diet and prediabetes and T2D and were not included as
covariates in the main analyses to avoid overadjustment [42].
2.3.7. Sensitivity analyses
For all diet indices and glucose outcomes, two additional analyses were performed. The
first did not include physical activity in as a covariate to determine if it had the potential to
confound the relationship between diet and glucose regulation and if it was necessary to control
for this variable in the main analysis. The second analysis controlled for body fat percent.
Though we expect that body fat (or BMI) is on the causal pathway between diet and T2D, we
included it as a covariate to examine the possibility that body fat mediates the relationship
between diet and T2D.
We also performed additional logistic regression analyses to examine the association
between each adiposity measure and risk for prediabetes/T2D at each visit and to examine the
associations between changes in these measures between visits and risk for prediabetes/T2D
at the follow-up visit. Models were adjusted for age, sex, ethnicity, parental education, energy
intake, and physical activity as in the main analyses.
2.4. Results
Average length of follow-up was 4.1 years (SD = 1.1 years). There were no differences
in participant demographics at each visit (Table 2.1). HEI, DASH, and DII scores significantly
decreased from baseline to follow-up (Table 2.2), and mean fasting glucose and glucose AUC
increased (Table 2.3). Mean BMI and body fat percentage also increased between visits (Table
2.4).

22
Table 2.1. Descriptive statistics for participant demographics at baseline and follow up.
Baseline
(n = 155)
Follow-Up
(n = 88)1
Baseline vs. Follow-Up
p-Value2
Age (years), Mean (SD) 19.7 (1.2) 24.1 (0.8) -
Sex, n (%)
Female
Male
71 (45.8)
84 (54.2)
46 (52.3)
42 (47.7)
0.40
Ethnicity, n (%)
Hispanic/Latino
Non-Hispanic White
Other
94 (60.6)
52 (33.5)
9 (5.8)
50 (56.8)
30 (34.1)
8 (9.1)
0.60
Parental Education, n (%)
Did not complete high school
Completed high school
More than high school
Don’t know
31 (20.0)
23 (14.8)
96 (61.9)
5 (3.2)
15 (17.0)
12 (13.6)
56 (63.6)
5 (5.7)
0.76
Exercise3
, n (%)
Yes
No
118 (76.1)
37 (23.9)
- -
Physical Activity Category, n (%)
High
Moderate
Low
Missing, n (%)
-
50 (56.8)
21 (23.9)
16 (18.2)
1 (1.1)
-
1
Includes three participants who did not complete the baseline visit.
2 p-values calculated using chi-Square or Fisher’s exact tests.
3 Response to the question “Do you exercise?”.
Abbreviations: SD, standard deviation.

23
Table 2.2. Descriptive statistics for diet at baseline, follow up, and change between visits.
Baseline
(n = 155)
Follow-Up
(n = 88)
Change between
Baseline and
Follow-Up (n =
85)1
Baseline vs.
Follow-Up
p-Value2
HEI, Mean (SD)
Range: 0–100 52.7 (13.0) 49.7 (12.5) −4.9 (13.2) <0.001
MDS, Mean (SD)
Range: 0–9
5.03 (1.23) 4.92 (1.53) −0.22 (1.79) 0.25
DASH, Mean (SD)
Range: 0–8
2.26 (1.51) 1.74 (1.31) −0.45 (1.53) 0.009
DII, Mean (SD) 0.81(1.56) 0.29 (2.05) −0.44 (1.98) 0.044
Energy (kcal), Mean
(SD) 2053 (630) 2223 (773) 158 (792) 0.070
1 Three additional CHS participants participated in the second visit without having completed the
first.
2 p-values calculated using paired t-tests.
Abbreviations: HEI, Healthy Eating Index—2015; MDS, Mediterranean Diet Score; DASH, Dietary
Approaches to Stop Hypertension; DII, Dietary Inflammatory Index.
Table 2.3. Descriptive statistics for glucose outcomes at baseline, follow up, and change
between visits.
Baseline
(n = 155)
Follow-Up
(n = 88)
Change between
Baseline and FollowUp (n = 85)1
Baseline vs.
Follow-Up
p-Value2
Fasting Glucose, Mean (SD)
Missing: n (%)
91. (14)
1 (0.6)
95 (16)
1 (1.1)
5 (15)
1 (1.2%) 0.003
2-Hour Glucose, Mean (SD)
Missing: n (%)
123 (37)
1 (0.6)
122 (35)
4 (4.5)
3 (32)
4 (4.7) 0.39
HbA1c, Mean (SD)
Missing: n (%)
5.25 (0.53)
1 (0.6)
5.26 (0.51) 0.042 (0.46) 0.35
Glucose AUC, Mean (SD)
Missing: n (%)
267 (59)
1 (0.6)
269 (44)
6 (6.8)
11 (40)
6 (7.1) 0.023
Diabetes, n (%)
No Diabetes
Prediabetes
Type 2 Diabetes
Missing
109 (70.3)
42 (27.1)
3 (1.9)
1 (0.6)
54 (61.4)
30 (34.1)
4 (4.5)
0.17
1 Three additional CHS participants participated in the second visit without having completed the first.
2 p-values calculated using paired t-tests for continuous variables and McNemar–Bowker test for
diabetes categories.
Abbreviations: SD, standard deviation; HbA1c, hemoglobin A1c; AUC, area under the curve.

24
Table 2.4. Descriptive statistics for body composition at baseline, follow up, and change
between visits.
Baseline
(n = 155)
FollowUp
(n = 88)
Change between
Baseline and
Follow-Up
(n = 85)1
Baseline vs.
Follow-Up
p-Value2,3
BMI Category, n (%)
Normal Weight
Overweight
Obese
24 (15.5)
73 (47.1)
58 (37.4)
12 (13.6)
34 (38.6)
42 (47.7)
0.47
BMI (kg/m2
), Mean (SD) 29.9 (5.1) 31.7 (7.0) 1.8 (4.3) <0.001
Body Fat %, Mean (SD)
Missing: n (%)
34.8 (8.6) 38.5 (8.3)
2 (2.3)
3.1 (4.7)
2 (2.4) <0.001
FFMI (kg/m2
), Mean (SD)
Missing: n (%)
18.5 (2.5) 17.7 (2.9)
2 (2.3)
−0.6 (1.5)
2 (2.4) 0.001
Fat Mass:Height Ratio, Mean (SD)
Missing: n (%)
10.8 (4.3)
98 (63.2)
12.2 (4.7)
2 (2.3)
1.6 (2.1)
47 (55.3) <0.001
Android:Gynoid Ratio, Mean (SD)
Missing: n (%)
(0.14)
98 (63.2)
1.01 (0.15)
2 (2.3)
0.015 (0.085)
47 (55.3) 0.30
Trunk:Leg Ratio, Mean (SD)
Missing: n (%)
0.95
(0.13)
98 (63.3)
0.97 (0.13)
2 (2.3)
0.016 (0.077)
47 (55.3) 0.20
Trunk:Limb Ratio, Mean (SD)
Missing: n (%)
1.05
(0.20)
98 (63.3)
1.10 (0.23)
2 (2.3)
0.051 (0.11)
47 (55.3) 0.005
VAT Volume (in3
), Mean (SD)
Missing: n (%)
592 (301)
98 (63.3)
633 (325)
2 (2.3)
88 (148)
47 (55.3) <0.001
1 Three additional CHS participants participated in the second visit without having completed the
first. 2 p-values calculated using t-tests for continuous variables and McNemar–Bowker test for
BMI category. 3 Fifty-seven participants completed the DEXA scan on a machine that provided
additional body composition indices. Abbreviations: BMI, body mass index; FFMI, fat free mass
index; VAT, visceral adipose tissue; SD, standard deviation.
2.4.1. Prediabetes/T2D

25
Positive change in HEI and DASH scores between the baseline and follow-up visits was
associated with decreased risk for prediabetes/T2D at follow-up (Figure 2.1). A one-point
increase in DASH score over the follow-up period was associated with a 64% (OR = 0.36, 95%
CI: 0.17, 0.68) reduction in risk for prediabetes/T2D at follow-up, while a one-point increase in
HEI between visits was associated with a 9% decrease in risk (OR = 0.91, 95% CI: 0.85, 0.96).
When scaled by standard deviation of diet index, improvements in DASH diet score reduced the
risk for prediabetes/T2D by a greater extent than the HEI (OR = 0.14, 95% CI: 0.03, 0.46; OR =
0.83, 95% CI: 0.72, 0.93, respectively). In the cross-sectional analysis of the follow-up visit,
higher HEI and DASH scores were also associated with reduced risk for prediabetes/T2D. At
baseline, only MDS was associated with reduced risk for prediabetes/T2D.
Figure 2.1. Coefficient plots for the effects of diet quality on prediabetes. “Baseline” and “follow–
up” values are the result of cross–sectional analyses of diet quality score and risk of
prediabetes/T2D at the same visit. The value for “change between visits” represents the risk of
prediabetes/T2D at the follow–up visit associated with change in diet score between the
baseline and the follow–up visit. Effects are standardized to one standard deviation of exposure.
Covariates: Baseline and follow–up models. HEI, DASH, and DII models adjusted for age, sex,
ethnicity, physical activity, and parental education. MDS models adjusted for energy intake, age,

26
ethnicity, physical activity, and parental education. Change between visits models. Baseline and
follow–up model covariates + baseline diet score. Abbreviations: DASH, Dietary Approaches to
Stop Hypertension; DII, Dietary Inflammatory index; HEI, Healthy Eating Index—2015; MDS,
Mediterranean Diet Score.
2.4.2. Fasting glucose and glucose tolerance
There were no statistically significant cross-sectional associations between fasting
glucose and any dietary index at either visit or between change in diet scores and change in
fasting glucose between visits (Figure 2.2).
Figure 2.2. Coefficient plots for the effects of diet quality on glucose measurements. “Baseline”
and “follow–up” values are the result of cross–sectional analyses of diet quality score and each
outcome. The value for “change between visits” represents the association between the change
in diet score between the baseline and the follow–up visit on the change in outcome between
visits. Effects are scaled to one standard deviation of exposure. Covariates: Baseline and follow–
up models: HEI, DASH, and DII models adjusted for age, sex, ethnicity, physical activity, and
parental education. MDS models adjusted for energy intake, age, ethnicity, physical activity, and

27
parental education. Change between visits models: Baseline and follow-up model covariates +
baseline diet score. Abbreviations: DASH, Dietary Approaches to Stop Hypertension; DII, Dietary
Inflammatory Index; HEI, Healthy Eating Index—2015; MDS, Mediterranean Diet Score.
Higher HEI scores and higher MDS were associated with lower 2-hour glucose values at
baseline in the cross-sectional analyses (HEI: ß = −7.01, 95% CI: −12.86, −1.16; MDS: ß =
−7.43, 95% CI: −13.25, −1.61) (Figure 2). Follow-up HEI and DASH scores were inversely
associated with 2-hour glucose at the same visit (HEI: ß = −8.64, 95% CI: −16.16, −1.12; DASH:
ß = −8.25, 95% CI: −15.71, −0.78) and with glucose AUC (HEI: ß = −11.34, 95% CI: −20.84,
−1.84; DASH: ß = −10.99, 95% CI: −20.44, −1.53).
2.4.3. Hemoglobin A1c
There were no statistically significant associations between HbA1c and any dietary
index. However, there were consistent inverse relationships between higher HEI and DASH
scores and HbA1c at both visits and between change in HEI or DASH and change in HbA1c
between visits although these did not reach the threshold for statistical significance (Figure 2.2).
2.4.4. Body composition
The DASH diet was consistently associated with several adiposity measures (Table 2.5).
At the follow-up visit, higher DASH scores were associated with lower BMI (ß = −1.64, 95% CI:
−3.17, −0.11), body fat percent (ß = −1.79, 95% CI: −3.01, −0.57), and fat mass to height ratio
(ß = −1.09, 95% CI: −3.27, −0.61) at the same visit, and increases in DASH between visits were
also inversely associated with change in BMI (ß = −1.64, 95% CI: −2.92, −0.36) and body fat
percent (ß = −1.62, 95% CI: −2.02, −0.17). Similar inverse associations were observed between
DASH and measures of central adiposity, including trunk to limb ratio and VAT.

28
The DII was positively associated with body fat percent in the cross-sectional baseline
analyses (Table 2.5). Though not statistically significant, the DII was also positively associated
with most adiposity measurements at both visits, and positive change in DII was associated with
positive changes in adiposity from baseline to follow-up.
Table 2.5. Estimated effect size and 95% CI for the effect of 1 standard deviation increase in
diet score on body composition.
Outcome
Effect Estimate, β (95% CI)
Baseline 1 Follow–Up 1 Change between Visits 2
Healthy Eating Index—2015 (HEI)
BMI (kg/m2
) −0.62 (−1.45, 0.21) −1.33 (−2.89, 0.24) −0.38 (−1.62, 0.85)
Body Fat (%) −0.85 (−1.86, 0.16) −1.09 (−2.37, 0.18) 0.40 (−0.92, 1.73)
FFMI (kg/m2
) −0.14 (−0.46, 0.17) −0.46 (−1.04, 0.12) −0.23 (−0.64, 0.18)
Fat Mass:Height Ratio −0.56 (−1.74, 0.62) −0.73 (−1.68, 0.22) −0.36 (−1.50, 0.78)
Android:Gynoid Ratio −0.045 (−0.087,
−0.0036) −0.043 (−0.071, −0.014) −0.014 (−0.061, 0.034)
Trunk:Leg Ratio −0.040 (−0.077,
−0.0028) −0.035 (−0.060, −0.0087) −0.0013 (−0.043, 0.041)
Trunk:Limb Ratio −0.052 (−0.11, 0.010) −0.052 (−0.099, −0.0048) −0.036 (−0.092, 0.020)
VAT (in3
) −65.78 (−161.45, 29.49) −60.54 (−132.21, 11.13) −48.05 (−123.33, 27.23)
Dietary Approaches to Stop Hypertension (DASH) Score
BMI (kg/m2
) 0.067 (−0.80, 0.94) −1.64 (−3.17, −0.11) −1.63 (−2.91, −0.35)
Body Fat (%) 0.12 (−0.94, 1.18) −1.79 (−3.01, −0.57) −1.61 (−3.01, −0.21)
FFMI (kg/m2
) −0.036 (−0.36, 0.29) −0.49 (−1.06, 0.088) −0.41 (−0.85, 0.024)
Fat Mass:Height Ratio 0.50 (−0.89, 1.88) −1.09 (−2.02, −0.17) −1.50 (−2.73, −0.27)
Android:Gynoid Ratio −0.015 (−0.066, 0.035) −0.043 (−0.071, −0.015) −0.047 (−0.098, 0.0045)
Trunk:Leg Ratio −0.023 (−0.068, 0.022) −0.039 (−0.064, −0.014) −0.037 (−0.084, 0.0097)
Trunk:Limb Ratio −0.018 (−0.093, 0.057) −0.052 (−0.099, −0.0057) −0.073 (−0.13, −0.011)
VAT (in3
) 42.25 (−70.97, 155.46) −76.57 (−146.46, −6.68) −100.39 (−183.62, −17.17)
Mediterranean Diet Score (MDS)
BMI (kg/m2
) −0.090 (−0.91, 0.73) −0.71 (−2.28, 0.86) 0.27 (−0.95, 1.49)
Body Fat (%) −0.45 (−1.69, 0.79) −0.48 (−2.35, 1.39) 1.24 (−0.062, 2.55)
FFMI (kg/m2
) 0.078 (−0.32, 0.47) 0.075 (−0.57, 0.72) −0.00040 (−0.42, 0.42)
Fat Mass:Height Ratio −0.37 (−1.49, 0.75) −0.28 (−1.38, 0.83) −0.081 (−1.11, 0.95)
Android:Gynoid Ratio 0.00054 (−0.042, 0.043) −0.0049 (−0.039, 0.030) 0.021 (−0.015, 0.057)
Trunk:Leg Ratio −0.030 (−0.065, 0.0037) −0.0042 (−0.035, 0.027) −0.0030 (−0.041, 0.035)
Trunk:Limb Ratio −0.044 (−0.10, 0.014) −0.0073 (−0.062, 0.047) −0.011 (−0.064, 0.042)
VAT (in3
) −21.86 (−109.41, 65.68) −17.16 (−92.10, 57.79) −25.82 (−98.45, 46.81)
Dietary Inflammatory Index (DII)
BMI (kg/m2
) 0.86 (0.044, 1.67) −0.67 (−2.32, 0.97) −0.21 (−1.24, 0.83)
Body Fat (%) 2.04 (1.09, 2.99) 1.13 (−0.19, 2.45) 0.44 (−0.66, 1.54)
FFMI (kg/m2
) −0.073 (−0.38, 0.23) −0.60 (−1.20, −0.0068) −0.16 (−0.50, 0.18)
Fat Mass:Height Ratio 0.88 (−0.23, 1.99) −0.17 (−1.17, 0.84) 0.52 (−0.33, 1.37)
Android:Gynoid Ratio 0.031 (−0.010, 0.072) 0.014 (−0.017, 0.045) 0.035 (0.0025, 0.068)
Trunk:Leg Ratio 0.027 (−0.010, 0.063) 0.021 (−0.0070, 0.048) 0.017 (−0.014, 0.048)
Trunk:Limb Ratio 0.028 (−0.033, 0.089) 0.023 (−0.027, 0.074) 0.029 (−0.014, 0.071)
VAT (in3
) 47.00 (−44.96, 138.95) −22.50 (−97.94, 52.94) 17.77 (−42.53, 78.08)

29
1 Model A: outcome ~ diet score + covariates. 2 Model B: outcome ~ diet score + covariates. Model A
covariates: HEI, DASH, and DII models adjusted for age, sex, ethnicity, physical activity, and parental
education. MDS models adjusted for energy intake, age, ethnicity, physical activity, and parental
education. Model B covariates: Model A covariates + baseline diet score. Effects were scaled to 1 standard
deviation of exposure.
Abbreviations: BMI, body mass index; FFMI, fat-free mass index; VAT, visceral adipose tissue.
2.4.5. Sensitivity Analyses
Results from the sensitivity analyses are reported in Supplemental Tables S2.1–S2.3.
Models that did not adjust for physical activity had slightly larger effect estimates for the
relationship between HEI and DASH and impaired glucose tolerance compared to models that
did adjust for physical activity. There was little effect on risk for prediabetes/T2D, and the main
findings were the same in the physical activity-adjusted and -unadjusted models. Adjustment for
body fat percent also had little effect on the relationships between HEI or DASH and
prediabetes/T2D, suggesting that it may not mediate the relationship between diet and
prediabetes/T2D. However, in most cases, controlling for body fat percent attenuated the effects
of each diet on all other glucose outcomes.
BMI, body fat percent, FFMI, fat mass to height ratio, and VAT were significantly
associated with increased risk for prediabetes/T2D at all time points (Table S2.4). At the followup visit only, android to gynoid ratio, trunk to leg ratio, and trunk to limb ratio were also positively
associated with prediabetes/T2D.
2.5. Discussion
We observed strong inverse associations both in cross-sectional and longitudinal
analyses between the HEI and DASH diet and risk of prediabetes/T2D. We also found negative
associations between the HEI and DASH diet and two-hour glucose, HbA1c, fasting glucose,
and glucose AUC at both visits and in the longitudinal analysis though these relationships were
not all statistically significant. The MDS was not consistently associated with prediabetes/T2D,
glucose measurements, or body composition. We also observed inverse relationships between

30
HEI, DASH, and MDS with measures of adiposity and body composition, suggesting that high
diet quality may be protective against obesity and adverse accumulation of adipose tissue. The
period between late adolescence to early adulthood is one of transition, where young people
begin to live independently and gain more control of their lifestyles. However, there are limited
assessments of change in diet quality during this transition [43], and these results emphasize
the importance of considering diet quality in T2D risk within this age group.
To our knowledge, no other study has evaluated the longitudinal relationship between
glucose dysregulation and HEI, DASH, MDS, and DII in young adults. Several meta-analyses
have summarized the relationship between diet quality and type 2 diabetes, prediabetes, or
other measures of glucose dysregulation in older adults. These analyses consistently report
strong protective effects of healthy dietary patterns, including the DASH and HEI [10,13,15].
However, previous reviews found effects of similar magnitude between the HEI, DASH, and
MDS [14], whereas we report a larger protective effect associated with increases in DASH diet
adherence across both visits compared to either the HEI or MDS. The DII has been
inconsistently associated with risk of T2D in older adults [17,18] though inflammation is involved
in the pathogenesis of type 2 diabetes [44]. Like Vahid (2017), we observed positive
associations between DII and impaired glucose intolerance and prediabetes.
Diet is also a risk factor for obesity, which is itself a significant driver of the T2D epidemic
in both adults and youth [6,45,46], and increases in body fat greatly increase the risk for future
diabetes [47]. Accumulation of visceral fat is also linked to T2D development and severity
[48,49]. Our study found similar effects, with multiple adiposity indices significantly associated
with increased risk of prediabetes/T2D. Our findings also suggest an inverse relationship
between high diet quality and central obesity, with HEI and DASH consistently associated with
android to gynoid fat ratio, trunk to limb fat ratios, and VAT. There also appeared to be positive
associations between DII and adiposity and visceral fat measures. These findings suggest that

31
high quality diets may reduce the risk of type 2 diabetes in part by reducing total body and
visceral fat.
This study has several strengths. Participants were recruited from the Southern
California Children’s Health Study [30], which allowed detailed measures of glucose
metabolism, diet, body composition, and lifestyle factors. OGTT and DEXA provide highly
detailed information about glucose metabolism and body composition, respectively, beyond that
of fasting plasma glucose, HbA1c, or BMI alone [50,51]. Two-hour glucose and glucose AUC,
for example, assess glucose tolerance, and impaired glucose tolerance is an early sign of
glucose dysregulation and type 2 diabetes risk not often captured in clinical settings [52].
Additionally, exposures and outcomes were assessed at both visits, which allowed us to
examine associations across time. Despite this, we note some limitations. Two systems were
used to collect 24-hour dietary recalls: the NDSR at baseline and the ASA24 at follow-up. We
are not aware of any evidence that this difference would introduce bias away from the null, and
any misclassification of diet is expected to be nondifferential and independent of
prediabetes/T2D status. It is also common for studies involving multiple cohorts to integrate
different diet assessment measures [53,54]. There is a possibility that residual confounding
contributed to our reported effects; family history of T2D, maternal obesity, and low birthweight
are also associated with young-onset T2D though they are less likely to be associated with diet.
However, the magnitude of the relationships we report are large, and any confounding by these
or other factors are unlikely to account for the entire effect. Additionally, our sample size for the
longitudinal analysis was 85, limiting the statistical power to detect significant relationships.
Limitations of one of the DEXA machines used at baseline also limited the available sample size
for some adiposity measurements (e.g., android to gynoid fat ratio, trunk to limb fat ratio).
However, power was sufficient to identify strong, statistically significant, protective effects of
high-quality diets on prediabetes risk.

32
The COVID-19 pandemic may also have affected our recruitment efforts for the follow-up
visit. Our recruitment began as the SARS-CoV-2 virus (COVID-19) was declared first a Public
Health Emergency and then a pandemic [55]. The resulting disruptions to daily life would have
affected our participants and likely impacted lifestyle factors such as physical activity, sleep, and
eating habits as well as stress, social supports, and physical health, all of which may affect noncommunicable disease risk [55–57]. It is possible that the observed decreases in diet quality
between the baseline and follow-up visits may be, in part, due to the pandemic. Even if some of
the change in diet were due to changes in lifestyle associated with the COVID-19 pandemic, our
findings emphasize the importance of maintaining a healthy diet to reduce the risk for T2D.
Our results indicate that improvements in adherence to the HEI and DASH dietary
patterns may reduce risk for T2D. Though both measure diet quality, the construction of each
index emphasizes different nutrients and food groups, and there are several ways in which an
individual may improve their score and overall diet quality. For example, the HEI rewards
greater adherence to the USDA Dietary Guidelines for Americans with higher scores on a 100-
point scale [36]. To improve a HEI score, one has several options: (1) increase intake of one or
several food groups (fruit, vegetables, seafood, etc.) to the levels recommend by the USDA; (2)
reduce intake of added sugars and salt as recommended by the USDA; or (3) reduce the
proportion of total grains that come from refined sources or increase the proportion of dietary
fats that are mono- or polyunsaturated [58]. Similarly, improvements in DASH diet score could
be achieved by reducing consumption of saturated fat, cholesterol, or sodium, or by increasing
fiber, magnesium, potassium, and calcium intake [37]. By encouraging changes to overall
dietary patterns rather than emphasizing specific foods or nutrients (i.e., kilocalories, sugarsweetened beverages), individuals may have more flexibility in their choice of dietary habits to
alter or methods of alteration, leading to more successful behavior change [59,60].

33
2.6. Conclusion
Late adolescence to early adulthood is a period of significant change and represents an
important window in which to establish lifelong habits [29]. To our knowledge, this study is one
of few to evaluate the impact of dietary changes on glucose regulation in people between the
ages of 18 and 30. We found that adherence to the DASH diet and USDA Dietary Guidelines is
associated with reduced risk for prediabetes and better glucose tolerance. Improvement in
DASH or HEI scores over the follow-up period was also associated with lower risk for
prediabetes or type 2 diabetes, with the strongest effects observed for the DASH diet. These
findings indicate that the DASH dietary pattern may be a promising target for diabetes
prevention efforts in young adults.
2.7. References
1. Tabák, A.G.; Herder, C.; Rathmann, W.; Brunner, E.J.; Kivimäki, M. Prediabetes: A highrisk state for diabetes development. Lancet 2012, 379, 2279–2290.
https://doi.org/10.1016/s0140-6736(12)60283-9.
2. Andes, L.J.; Cheng, Y.J.; Rolka, D.B.; Gregg, E.W.; Imperatore, G. Prevalence of
Prediabetes Among Adolescents and Young Adults in the United States, 2005–2016. JAMA
Pediatr. 2020, 174, e194498. https://doi.org/10.1001/jamapediatrics.2019.4498.
3. Liu, J.; Li, Y.; Zhang, D.; Yi, S.S.; Liu, J. Trends in Prediabetes Among Youths in the US
From 1999 Through 2018. JAMA Pediatrics 2022, 176, 608–611.
https://doi.org/10.1001/jamapediatrics.2022.0077.
4. Morris, D.H.; Khunti, K.; Achana, F.; Srinivasan, B.; Gray, L.J.; Davies, M.J.; Webb, D.
Progression rates from HbA1c 6.0–6.4% and other prediabetes definitions to type 2
diabetes: A meta-analysis. Diabetologia 2013, 56, 1489–1493.
https://doi.org/10.1007/s00125-013-2902-4.
5. Divers, J.; Mayer-Davis, E.J.; Lawrence, J.M.; Isom, S.; Dabelea, D.; Dolan, L.; Imperatore,
G.; Marcovina, S.; Pettitt, D.J.; Pihoker, C.; et al. Trends in Incidence of Type 1 and Type 2
Diabetes Among Youths-Selected Counties and Indian Reservations, United States, 2002–
2015. MMWR Morb. Mortal. Wkly. Rep. 2020, 69, 161–165.
https://doi.org/10.15585/mmwr.mm6906a3.
6. Shah, A.S.; Nadeau, K.J. The changing face of paediatric diabetes. Diabetologia 2020, 63,
683–691. https://doi.org/10.1007/s00125-019-05075-6.

34
7. Magliano, D.J.; Sacre, J.W.; Harding, J.L.; Gregg, E.W.; Zimmet, P.Z.; Shaw, J.E. Youngonset type 2 diabetes mellitus-implications for morbidity and mortality. Nat. Rev.
Endocrinol. 2020, 16, 321–331. https://doi.org/10.1038/s41574-020-0334-z.
8. Lascar, N.; Brown, J.; Pattison, H.; Barnett, A.H.; Bailey, C.J.; Bellary, S. Type 2 diabetes in
adolescents and young adults. Lancet Diabetes Endocrinol. 2018, 6, 69–80.
https://doi.org/10.1016/s2213-8587(17)30186-9.
9. ADA. 3. Prevention or Delay of Type 2 Diabetes: Standards of Medical Care in Diabetes2020. Diabetes Care 2020, 43 (Suppl. 1), S32–S36. https://doi.org/10.2337/dc20-S003.
10. Neuenschwander, M.; Ballon, A.; Weber, K.S.; Norat, T.; Aune, D.; Schwingshackl, L.;
Schlesinger, S. Role of diet in type 2 diabetes incidence: Umbrella review of meta-analyses
of prospective observational studies. BMJ 2019, 366, l2368.
https://doi.org/10.1136/bmj.l2368.
11. Schwingshackl, L.; Hoffmann, G.; Lampousi, A.M.; Knüppel, S.; Iqbal, K.; Schwedhelm, C.;
Bechthold, A.; Schlesinger, S.; Boeing, H. Food groups and risk of type 2 diabetes mellitus:
A systematic review and meta-analysis of prospective studies. Eur. J. Epidemiol. 2017, 32,
363–375. https://doi.org/10.1007/s10654-017-0246-y.
12. Toi, P.L.; Anothaisintawee, T.; Chaikledkaew, U.; Briones, J.R.; Reutrakul, S.; Thakkinstian,
A. Preventive Role of Diet Interventions and Dietary Factors in Type 2 Diabetes Mellitus:
An Umbrella Review. Nutrients 2020, 12, 2722. https://doi.org/10.3390/nu12092722.
13. de Koning, L.; Chiuve, S.E.; Fung, T.T.; Willett, W.C.; Rimm, E.B.; Hu, F.B. Diet-quality
scores and the risk of type 2 diabetes in men. Diabetes Care 2011, 34, 1150–1156.
https://doi.org/10.2337/dc10-2352.
14. Jannasch, F.; Kröger, J.; Schulze, M.B. Dietary Patterns and Type 2 Diabetes: A
Systematic Literature Review and Meta-Analysis of Prospective Studies. J. Nutr. 2017, 147,
1174–1182. https://doi.org/10.3945/jn.116.242552.
15. Qian, F.; Liu, G.; Hu, F.B.; Bhupathiraju, S.N.; Sun, Q. Association Between Plant-Based
Dietary Patterns and Risk of Type 2 Diabetes: A Systematic Review and Meta-analysis.
JAMA Intern. Med. 2019, 179, 1335–1344.
https://doi.org/10.1001/jamainternmed.2019.2195.
16. Salas-Salvadó, J.; Bulló, M.; Estruch, R.; Ros, E.; Covas, M.I.; Ibarrola-Jurado, N.; Corella,
D.; Arós, F.; Gómez-Gracia, E.; Ruiz-Gutiérrez, V.; et al. Prevention of diabetes with
Mediterranean diets: A subgroup analysis of a randomized trial. Ann. Intern. Med. 2014,
160, 1–10. https://doi.org/10.7326/m13-1725.
17. Vahid, F.; Shivappa, N.; Karamati, M.; Naeini, A.J.; Hebert, J.R.; Davoodi, S.H. Association
between Dietary Inflammatory Index (DII) and risk of prediabetes: A case-control study.
Appl, Physol. Nutr. Metab. 2017, 42, 399–404. https://doi.org/10.1139/apnm-2016-0395.
18. Laouali, N.; Mancini, F.R.; Hajji-Louati, M.; El Fatouhi, D.; Balkau, B.; Boutron-Ruault, M.-
C.; Bonnet, F.; Fagherazzi, G. Dietary inflammatory index and type 2 diabetes risk in a
prospective cohort of 70,991 women followed for 20 years: The mediating role of BMI.
Diabetologia 2019, 62, 2222–2232. https://doi.org/10.1007/s00125-019-04972-0.

35
19. Esposito, K.; Chiodini, P.; Maiorino, M.I.; Bellastella, G.; Panagiotakos, D.; Giugliano, D.
Which diet for prevention of type 2 diabetes? A meta-analysis of prospective studies.
Endocrine 2014, 47, 107–116. https://doi.org/10.1007/s12020-014-0264-4.
20. Alhazmi, A.; Stojanovski, E.; McEvoy, M.; Garg, M.L. The association between dietary
patterns and type 2 diabetes: A systematic review and meta-analysis of cohort studies. J.
Hum. Nutr. Diet. 2014, 27, 251–260. https://doi.org/10.1111/jhn.12139.
21. Barnes, T.L.; Crandell, J.L.; Bell, R.A.; Mayer-Davis, E.J.; Dabelea, D.; Liese, A.D.; for the
SEARCH for Diabetes in Youth Study Group. Change in DASH diet score and
cardiovascular risk factors in youth with type 1 and type 2 diabetes mellitus: The SEARCH
for Diabetes in Youth Study. Nutr. Diabetes 2013, 3, e91.
https://doi.org/10.1038/nutd.2013.32.
22. Gow, M.L.; Garnett, S.P.; Baur, L.A.; Lister, N.B. The Effectiveness of Different Diet
Strategies to Reduce Type 2 Diabetes Risk in Youth. Nutrients 2016, 8, 486.
https://doi.org/10.3390/nu8080486.
23. Kim, R.J.; Lopez, R.; Snair, M.; Tang, A. Mediterranean diet adherence and metabolic
syndrome in US adolescents. Int. J. Food Sci. Nutr. 2021, 72, 537–547.
https://doi.org/10.1080/09637486.2020.1840533.
24. Cañete, R.; Gil-Campos, M.; Aguilera, C.M.; Gil, A. Development of insulin resistance and
its relation to diet in the obese child. Eur. J. Nutr. 2007, 46, 181–187.
https://doi.org/10.1007/s00394-007-0648-9.
25. Hirahatake, K.M.; Jacobs, D.R., Jr.; Shikany, J.M.; Jiang, L.; Wong, N.D.; Odegaard, A.O.
Cumulative average dietary pattern scores in young adulthood and risk of incident type 2
diabetes: The CARDIA study. Diabetologia 2019, 62, 2233–2244.
https://doi.org/10.1007/s00125-019-04989-5.
26. Cha, E.; Pasquel, F.J.; Yan, F.; Jacobs, D.R., Jr.; Dunbar, S.B.; Umpierrez, G.; Choi, Y.;
Shikany, J.M.; Bancks, M.P.; Reis, J.P.; et al. Characteristics associated with early- vs.
later-onset adult diabetes: The CARDIA study. Diabetes Res. Clin. Pract. 2021, 182,
109144. https://doi.org/10.1016/j.diabres.2021.109144.
27. Bantle, A.E.; Chow, L.S.; Steffen, L.M.; Wang, Q.; Hughes, J.; Durant, N.H.; Ingram, K.H.;
Reis, J.P.; Schreiner, P.J. Association of Mediterranean diet and cardiorespiratory fitness
with the development of pre-diabetes and diabetes: The Coronary Artery Risk Development
in Young Adults (CARDIA) study. BMJ Open Diabetes Res. Care 2016, 4, e000229.
https://doi.org/10.1136/bmjdrc-2016-000229.
28. Choi, Y.; Larson, N.; Gallaher, D.D.; Odegaard, A.O.; Rana, J.S.; Shikany, J.M.; Steffen,
L.M.; Jacobs, D.R., Jr. A Shift Toward a Plant-Centered Diet From Young to Middle
Adulthood and Subsequent Risk of Type 2 Diabetes and Weight Gain: The Coronary Artery
Risk Development in Young Adults (CARDIA) Study. Diabetes Care 2020, 43, 2796–2803.
https://doi.org/10.2337/dc20-1005.
29. Nelson, M.C.; Story, M.; Larson, N.I.; Neumark-Sztainer, D.; Lytle, L.A. Emerging
Adulthood and College-aged Youth: An Overlooked Age for Weight-related Behavior
Change. Obesity 2008, 16, 2205–2211. https://doi.org/10.1038/oby.2008.365.

36
30. McConnell, R.; Shen, E.; Gilliland Frank, D.; Jerrett, M.; Wolch, J.; Chang, C.-C.; Lurmann,
F.; Berhane, K. A Longitudinal Cohort Study of Body Mass Index and Childhood Exposure
to Secondhand Tobacco Smoke and Air Pollution: The Southern California Children’s
Health Study. Environ. Health Perspect. 2015, 123, 360–366.
https://doi.org/10.1289/ehp.1307031.
31. Kim, J.S.; Chen, Z.; Alderete, T.L.; Toledo-Corral, C.; Lurmann, F.; Berhane, K.; Gilliland,
F.D. Associations of air pollution, obesity and cardiometabolic health in young adults: The
Meta-AIR study. Environ. Int. 2019, 133, 105180.
https://doi.org/10.1016/j.envint.2019.105180.
32. Floch, J.-P.L.; Escuyer, P.; Baudin, E.; Baudon, D.; Perlemuter, L. Blood Glucose Area
Under the Curve: Methodological Aspects. Diabetes Care 1990, 13, 172–175.
https://doi.org/10.2337/diacare.13.2.172.
33. ADA. 2. Classification and Diagnosis of Diabetes: Standards of Medical Care in Diabetes—
2021. Diabetes Care 2020, 44 (Suppl. 1), S15–S33. https://doi.org/10.2337/dc21-S002.
34. Hoffmann, K.; Boeing, H.; Dufour, A.; Volatier, J.L.; Telman, J.; Virtanen, M.; Becker, W.;
De Henauw, S.; for the EFCOSUM Group. Estimating the distribution of usual dietary intake
by short-term measurements. Eur. J. Clin. Nutr. 2002, 56, S53–S62.
https://doi.org/10.1038/sj.ejcn.1601429.
35. Park, Y.; Dodd, K.W.; Kipnis, V.; Thompson, F.E.; Potischman, N.; Schoeller, D.A.; Baer,
D.J.; Midthune, D.; Troiano, R.P.; Bowles, H.; et al. Comparison of self-reported dietary
intakes from the Automated Self-Administered 24-h recall, 4-d food records, and foodfrequency questionnaires against recovery biomarkers. Am. J. Clin. Nutr. 2018, 107, 80–93.
https://doi.org/10.1093/ajcn/nqx002.
36. Krebs-Smith, S.M.; Pannucci, T.E.; Subar, A.F.; Kirkpatrick, S.I.; Lerman, J.L.; Tooze, J.A.;
Wilson, M.M.; Reedy, J. Update of the Healthy Eating Index: HEI-2015. J. Acad. Nutr. Diet.
2018, 118, 1591–1602. https://doi.org/10.1016/j.jand.2018.05.021.
37. Mellen, P.B.; Gao, S.K.; Vitolins, M.Z.; Goff, D.C., Jr. Deteriorating Dietary Habits Among
Adults with Hypertension: DASH Dietary Accordance, NHANES 1988–1994 and 1999–
2004. Arch. Intern. Med. 2008, 168, 308–314.
https://doi.org/10.1001/archinternmed.2007.119.
38. Trichopoulou, A.; Costacou, T.; Bamia, C.; Trichopoulos, D. Adherence to a Mediterranean
diet and survival in a Greek population. N. Engl. J. Med. 2003, 348, 2599–2608.
https://doi.org/10.1056/NEJMoa025039.
39. Shivappa, N.; Steck, S.E.; Hurley, T.G.; Hussey, J.R.; Hébert, J.R. Designing and
developing a literature-derived, population-based dietary inflammatory index. Public Health
Nutr. 2014, 17, 1689–1696. https://doi.org/10.1017/S1368980013002115.
40. FDA. Final determination regarding partially hydrogenated oils. Fed. Regist. 2015, 80,
34650–34670.
41. Craig, C.L.; Marshall, A.L.; Sjöström, M.; Bauman, A.E.; Booth, M.L.; Ainsworth, B.E.; Pratt,
M.; Ekelund, U.; Yngve, A.; Sallis, J.F.; et al. International Physical Activity Questionnaire:
12-Country Reliability and Validity. Med. Sci. Sports Exerc. 2003, 35, 1381–1395.
https://doi.org/10.1249/01.Mss.0000078924.61453.Fb.

37
42. Schisterman, E.F.; Cole, S.R.; Platt, R.W. Overadjustment bias and unnecessary
adjustment in epidemiologic studies. Epidemiology 2009, 20, 488–495.
https://doi.org/10.1097/EDE.0b013e3181a819a1.
43. Winpenny, E.M.; Penney, T.L.; Corder, K.; White, M.; van Sluijs, E.M.F. Change in diet in
the period from adolescence to early adulthood: A systematic scoping review of longitudinal
studies. Int. J. Behav. Nutr. Phys. Act. 2017, 14, 60. https://doi.org/10.1186/s12966-017-
0518-7.
44. Esser, N.; Legrand-Poels, S.; Piette, J.; Scheen, A.J.; Paquot, N. Inflammation as a link
between obesity, metabolic syndrome and type 2 diabetes. Diabetes Res. Clin. Pract.
2014, 105, 141–150. https://doi.org/10.1016/j.diabres.2014.04.006.
45. Pulgaron, E.R.; Delamater, A.M. Obesity and type 2 diabetes in children: Epidemiology and
treatment. Curr. Diab. Rep. 2014, 14, 508. https://doi.org/10.1007/s11892-014-0508-y.
46. Maggio, C.A.; Pi-Sunyer, F.X. Obesity and type 2 diabetes. Endocrinol. Metab. Clin. N. Am.
2003, 32, 805–822, https://doi.org/10.1016/s0889-8529(03)00071-9.
47. Zhao, T.; Lin, Z.; Zhu, H.; Wang, C.; Jia, W. Impact of body fat percentage change on
future diabetes in subjects with normal glucose tolerance. IUBMB Life 2017, 69, 947–955.
https://doi.org/10.1002/iub.1693.
48. Chen, Y.; He, D.; Yang, T.; Zhou, H.; Xiang, S.; Shen, L.; Wen, J.; Chen, S.; Peng, S.; Gan,
Y. Relationship between body composition indicators and risk of type 2 diabetes mellitus in
Chinese adults. BMC Public Health 2020, 20, 452. https://doi.org/10.1186/s12889-020-
08552-5.
49. Gupta, P.; Lanca, C.; Gan, A.T.L.; Soh, P.; Thakur, S.; Tao, Y.; Kumari, N.; Man, R.E.K.;
Fenwick, E.K.; Lamoureux, E.L. The Association between Body Composition using Dual
energy X-ray Absorptiometry and Type-2 Diabetes: A Systematic Review and MetaAnalysis of Observational studies. Sci. Rep. 2019, 9, 12634.
https://doi.org/10.1038/s41598-019-49162-5.
50. Jesudason, D.R.; Dunstan, K.; Leong, D.; Wittert, G.A. Macrovascular Risk and Diagnostic
Criteria for Type 2 Diabetes: Implications for the use of FPG and HbA1c for cost-effective
screening. Diabetes Care 2003, 26, 485–490. https://doi.org/10.2337/diacare.26.2.485.
51. Messina, C.; Albano, D.; Gitto, S.; Tofanelli, L.; Bazzocchi, A.; Ulivieri, F.M.; Guglielmi, G.;
Sconfienza, L.M. Body composition with dual energy X-ray absorptiometry: From basics to
new tools. Quant. Imaging Med. Surg. 2020, 10, 1687–1698.
https://doi.org/10.21037/qims.2020.03.02.
52. Bartoli, E.; Fra, G.P.; Carnevale Schianca, G.P. The oral glucose tolerance test (OGTT)
revisited. Eur. J. Intern. Med. 2011, 22, 8–12. https://doi.org/10.1016/j.ejim.2010.07.008.
53. Maitre, L.; de Bont, J.; Casas, M.; Robinson, O.; Aasvang, G.M.; Agier, L.; Andrušaitytė, S.;
Ballester, F.; Basagaña, X.; Borràs, E.; et al. Human Early Life Exposome (HELIX) study: A
European population-based exposome cohort. BMJ Open 2018, 8, e021311.
https://doi.org/10.1136/bmjopen-2017-021311.
54. Smith-Warner, S.A.; Spiegelman, D.; Ritz, J.; Albanes, D.; Beeson, W.L.; Bernstein, L.;
Berrino, F.; van den Brandt, P.A.; Buring, J.E.; Cho, E.; et al. Methods for pooling results of

38
epidemiologic studies: The Pooling Project of Prospective Studies of Diet and Cancer. Am.
J. Epidemiol. 2006, 163, 1053–1064. https://doi.org/10.1093/aje/kwj127.
55. Machhi, J.; Herskovitz, J.; Senan, A.M.; Dutta, D.; Nath, B.; Oleynikov, M.D.; Blomberg,
W.R.; Meigs, D.D.; Hasan, M.; Patel, M.; et al. The Natural History, Pathobiology, and
Clinical Manifestations of SARS-CoV-2 Infections. J. Neuroimmune Pharmacol. 2020, 15,
359–386. https://doi.org/10.1007/s11481-020-09944-5.
56. Karatzi, K.; Poulia, K.A.; Papakonstantinou, E.; Zampelas, A. The Impact of Nutritional and
Lifestyle Changes on Body Weight, Body Composition and Cardiometabolic Risk Factors in
Children and Adolescents during the Pandemic of COVID-19: A Systematic Review.
Children (Basel) 2021, 8, 1130. https://doi.org/10.3390/children8121130.
57. Kreutz, R.; Dobrowolski, P.; Prejbisz, A.; Algharably, E.A.E.; Bilo, G.; Creutzig, F.; Grassi,
G.; Kotsis, V.; Lovic, D.; Lurbe, E.; et al. Lifestyle, psychological, socioeconomic and
environmental factors and their impact on hypertension during the coronavirus disease
2019 pandemic. J. Hypertens. 2021, 39, 1077–1089.
https://doi.org/10.1097/hjh.0000000000002770.
58. Millen, B.E.; Abrams, S.; Adams-Campbell, L.; Anderson, C.A.; Brenna, J.T.; Campbell,
W.W.; Clinton, S.; Hu, F.; Nelson, M.; Neuhouser, M.L.; et al. The 2015 Dietary Guidelines
Advisory Committee Scientific Report: Development and Major Conclusions. Adv. Nutr.
2016, 7, 438–444. https://doi.org/10.3945/an.116.012120.
59. Gardner, C.D.; Trepanowski, J.F.; Del Gobbo, L.C.; Hauser, M.E.; Rigdon, J.; Ioannidis,
J.P.A.; Desai, M.; King, A.C. Effect of Low-Fat vs Low-Carbohydrate Diet on 12-Month
Weight Loss in Overweight Adults and the Association With Genotype Pattern or Insulin
Secretion: The DIETFITS Randomized Clinical Trial. JAMA 2018, 319, 667–679.
https://doi.org/10.1001/jama.2018.0245.
60. Koliaki, C.; Spinos, T.; Spinou, Μ.; Brinia, Μ.-E.; Mitsopoulou, D.; Katsilambros, N. Defining
the Optimal Dietary Approach for Safe, Effective and Sustainable Weight Loss in
Overweight and Obese Adults. Healthcare (Basel, Switzerland) 2018, 6, 73.
https://doi.org/10.3390/healthcare6030073.

39
2.8. Supplemental Material
Figure S2.1. Flowchart for study recruitment.
*Includes those who refused for reasons related to COVID-19 and no-shows for scheduled
visits.
Follow up visit total
n = 89
Compete data for diet and at least
one outcome (n = 88)
Fasting glucose data
n = 87
Complete OGTT data
n = 82
Additional CHS participants
interested in the study
(n = 3)
Completed follow up visit
n = 86
Baseline visit
n = 158
Included in cross-sectional
analysis for the baseline visit
n = 155
At least one glucose outcome
n = 154
Full body composition data
n = 57
Invited for follow up visit
n = 140
Missing diet data
(n = 3)
Did not provide consent for
future use of samples
(n = 15)
Did not participate:
Refused (n = 30*)
Moved away or changed contact
information (n = 11)
Non-response (n = 13)
Missing diet data
(n = 1)

40
Table S2.1. Results (effects and 95% CIs) for sensitivity analyses at the baseline visit.
Outcome
HEI DASH MDS DII
Effect (95% CI) Effect (95% CI) Effect (95% CI) Effect (95% CI)
Fasting glucose
Model 1 -1.52 (-3.84, 0.80) -1.53 (-3.90, 0.85) -1.90 (-4.18, 0.39) 0.089 (-2.59, 2.77)
Model 2 -1.55 (-3.89, 0.79) -1.31 (-3.73, 1.11) -2.20 (-4.48, 0.093) -0.37 (-2.68, 1.93)
Model 3 -1.41 (-3.78, 0.96) -1.33 (-3.75, 1.10) -2.12 (-4.42, 0.18) -0.83 (-3.27, 1.61)
2-hour glucose
Model 1 -7.26 (-13.04, -1.49) -2.62 (-8.63, 3.39) -6.90 (-12.62, -1.18) 4.36 (-2.36, 11.08)
Model 2 -7.01 (-12.86, -1.16) -1.88 (-8.03, 4.28) -7.43 (-13.25, -1.61) 3.78 (-2.02, 9.58)
Model 3 -5.73 (-11.51, 0.052) -2.00 (-7.97, 3.97) -6.69 (-12.31, -1.08) 0.90 (-5.09, 6.90)
HbA1c
Model 1 -0.063 (-0.15, 0.021) -0.065 (-0.15, 0.021) -0.066 (-1.47, 0.016) -0.024 (-0.12, 0.072)
Model 2 -0.062 (-0.15, 0.022) -0.055 (-0.14, 0.033) -0.08 (-0.16, 0.004) -0.070 (-0.15, 0.013)
Model 3 -0.063 (-0.15, 0.023) -0.055 (-0.14, 0.033) -0.075 (-0.16, 0.007) -0.081 (-0.17, 0.007)
Glucose AUC
Model 1 -8.57 (-18.29, 1.15) -3.06 (-13.08, 6.96) -7.39 (-17.02, 2.25) 2.43 (-8.82, 13.69)
Model 2 -8.02 (-17.65, 1.61) -1.67 (-11.71, 8.37) -8.03 (-17.53, 1.46) 2.81 (-6.69, 12.31)
Model 3 -6.38 (-16.00, 3.24) -1.83 (-11.71, 8.05) -7.28 (-16.69, 2.12) -1.06 (-11.0, 8.85)
T2D/prediabetes
Model 1 0.83 (0.57, 1.19) 0.84 (0.56, 1.21) 0.64 (0.44, 0.93) 1.28 (0.84, 2.00)
Model 2 0.85 (0.58, 1.23) 0.85 (0.57, 1.25) 0.64 (0.43, 0.92) 1.17 (0.81, 1.72)
Model 3 0.89 (0.60, 1.30) 0.84 (0.56, 1.24) 0.65 (0.44, 0.95) 1.05 (0.70, 1.57)
Model 1 covariates: age, sex, ethnicity, parental education.
Model 2 covariates: Model 1 + physical activity
Model 3 covariates: Model 2 + body fat percent
MDS models additionally control for baseline energy intake and not for sex. All effects are scaled to 1 SD of
exposure.

41
Table S2.2. Results (effects and 95% CIs) for sensitivity analyses at the follow up visit.
Outcome
HEI DASH MDS DII
Effect (95% CI) Effect (95% CI) Effect (95% CI) Effect (95% CI)
Fasting glucose
Model 1 -0.95 (-4.57, 2.66) -1.94 (-5.47, 1.60) 2.52 (-1.05, 6.09) -0.87 (-4.61, 2.86)
Model 2 -1.23 (-4.95, 2.50) -2.29 (-5.94, 1.37) 2.99 (-0.79, 6.77) -0.68 (-4.53, 3.18)
Model 3 -0.21 (-3.85, 3.43) -0.69 (-4.40, 3.03) 3.75 (-0.17, 7.68) -2.08 (-5.85, 1.69)
2-hour glucose
Model 1 -9.87 (-17.35, -2.38) -9.37 (-16.78, -1.95) -1.03 (-9.23, 7.16) 5.74 (-2.31, 13.80)
Model 2 -8.64 (-16.16, -1.12) -8.25 (-15.71, -0.78) -2.82 (-11.33, 5.70) 4.39 (-3.72, 12.50)
Model 3 -5.60 (-12.39, 1.19) -3.72 (-10.07, 4.04) -1.01 (-8.78, 6.75) -0.22 (-7.60, 7.15)
Glucose AUC
Model 1 -12.74 (-22.39, -3.10) -12.37 (-22.00, -2.75) -3.93 (-14.21, 6.34) 4.37 (-6.50, 15.24)
Model 2 -11.34 (-20.84, -1.84) -10.99 (-20.44, -1.53) -6.67 (-17.17, 3.82) 3.08 (-7.57, 13.73)
Model 3 -7.51 (-16.21, 1.19) -5.61 (-14.54, 3.32) -4.24 (-14.22, 5.73) -1.84 (-11.49, 7.81)
HbA1c
Model 1 -0.085 (-0.20, 0.028) -0.098 (-0.21, 0.015) 0.040 (-0.070, 0.15) -0.038 (-0.16, 0.080)
Model 2 -0.086 (-0.20, 0.031) -0.10 (-0.21, 0.018) 0.041 (-0.078, 0.16) -0.042 (-0.16, 0.080)
Model 3 -0.059 (-0.18, 0.058) -0.055 (-0.17, 0.064) 0.050 (-0.073, 0.17) -0.075 (-0.20, 0.045)
T2D/prediabetes
Model 1 0.28 (0.12, 0.56) 0.31 (0.13, 0.61) 0.76 (0.47, 1.22) 1.31 (0.80, 2.23)
Model 2 0.29 (0.12, 0.57) 0.27 (0.11, 0.57) 0.81 (0.48, 1.35) 1.25 (0.75, 2.12)
Model 3 0.19 (0.059, 0.48) 0.26 (0.082, 0.67) 0.80 (0.45, 1.40) 1.01 (0.52, 2.00)
Model 1 covariates: age, sex, ethnicity, parental education
Model 2 covariates: Model 2 + physical activity
Model 3 covariates: Model 1 + body fat percent
MDS models additionally control for baseline energy intake and not for sex. All effects are scaled to 1 SD of exposure.

42
Table S2.3. Results (effects and 95% CIs) for sensitivity analyses for the effects of change in diet score between visits
Outcome
Δ HEI Δ DASH Δ MDS Δ DII
Effect (95% CI) Effect (95% CI) Effect (95% CI) Effect (95% CI)
Δ Fasting glucose
Model 1 -2.30 (-6.51, 1.90) -2.01 (-6.57, 2.55) 2.72 (-1.29, 6.73) 0.90 (-2.59, 4.38)
Model 2 -2.44 (-6.78, 1.89) -2.23 (-6.97, 2.52) 2.98 (-1.32, 7.29) 1.10 (-2.55, 4.76)
Model 3 -2.10 (-6.58, 2.38) -1.97 (-6.82, 2.89) 3.05 (-1.37, 7.46) 1.12 (-2.47, 4.88)
Δ 2-hour glucose
Model 1 -0.53 (-9.68, 8.63) -4.18 (-13.98, 5.62) 0.10 (-8.87, 9.06) 1.00 (-6.72, 8.72)
Model 2 0.18 (-9.15, 9.52) -2.96 (-13.07, 7.14) -1.22 (-10.87, 8.43) 0.26 (-7.72, 8.25)
Model 3 0.51 (-9.16, 10.18) -2.82 (-13.17, 7.53) -1.41 (-11.31, 8.49) 0.34 (-7.72, 8.41)
Δ Glucose AUC
Model 1 -4.20 (-15.85, 7.45) -6.75 (-19.30, 5.81) -2.57 (-13.97, 8.83) 7.09 (-2.68, 16.86)
Model 2 -3.40 (-15.23, 8.43) -5.25 (-18.10, 7.59) -5.30 (-17.44, 6.83) 7.42 (-2.61, 17.45)
Model 3 -2.78 (-15.08, 9.53) -4.82 (-17.93, 8.29) -4.55 (-16.97, 7.87) 7.78 (-2.34, 17.89)
Δ HbA1c
Model 1 -0.049 (-0.18, 0.081) -0.057 (-0.20, 0.083) 0.065 (-0.057, 0.19) 0.0002 (-0.11, 0.11)
Model 2 -0.046 (-0.18, 0.088) -0.052 (-0.20, 0.093) 0.072 (-0.060, 0.20) -0.0003 (-0.11, 0.11)
Model 3 -0.023 (-0.16, 0.11) -0.035 (-0.18, 0.11) 0.71 (-0.064, 0.21) 0.0051 (-0.11, 0.12)
T2D/prediabetes at the follow up visit
Model 1 0.82 (0.72, 0.92) 0.15 (0.035, 0.47) 0.79 (0.42, 1.45) 1.24 (0.74, 2.15)
Model 2 0.83 (0.72, 0.93) 0.14 (0.030, 0.46) 0.87 (0.45, 1.68) 1.13 (0.66, 1.98)
Model 3 0.85 (0.74, 0.95) 0.15 (0.033, 0.54) 1.00 (0.51, 1.97) 1.06 (0.60, 1.90)
Model 1 covariates: age, sex, ethnicity, parental education, baseline diet score
Model 2 covariates: Model 2 + physical activity at follow up
Model 3 covariates: Model 1 + body fat percent at baseline
MDS models additionally control for baseline energy intake and not for sex. All effects are scaled to 1 SD of exposure.

43
Table S2.4. Estimated effect size and 95% CI for the relationship between body composition
and risk for prediabetes/type 2 diabetes.
Exposure
Odds Ratio (95% CI)
Baseline1 Follow-up1 Change between visits2
BMI (kg/m2
) 1.11 (1.03, 1.20) 1.17 (1.07, 1.29) 1.19 (1.04, 1.40)
Body Fat (%) 1.067 (1.003, 1.139) 1.41 (1.21, 1.72) 1.26 (1.10, 1.50)
FFMI (kg/m2
) 1.26 (1.04, 1.53) 1.31 (1.06 1.65) 1.53 (1.06, 2.32)
Fat Mass:Height Ratio 1.28 (1.07, 1.59) 1.46 (1.22, 1.82) 2.45 (1.27, 6.57)
Android:Gynoid Ratio3 1.057 (1.005, 1.126) 1.10 (1.04, 1.16) 1.15 (0.98, 1.41)
Trunk:Leg Ratio3 1.05 (0.99, 1.12) 1.08 (1.03, 1.14) 1.04 (0.87, 1.23)
Trunk:Limb Ratio3 1.027 (0.99, 1.07) 1.04 (1.01, 1.07) 1.07 (0.98, 1.18)
VAT4
(in3
) 1.33 (1.07, 1.74) 1.54 (1.25, 2.00) 2.16 (1.11, 5.21)
Abbreviations: BMI: body mass index; FFMI: fat free mass index; VAT: visceral adipose tissue
volume
1 Model A: outcome ~ diet score + covariates
2 Model B: outcome ~ diet score + covariates
3 Effects are expressed per 0.01 unit increase in ratio.
4 Effects are expressed per 100 in3
increase in VAT.
Model A covariates: age, sex, ethnicity, parental education, energy intake, and physical
activity
Model B covariates: Model A covariates + baseline exposure.

44
Chapter 3: Proteomic and metabolomic signatures of diet quality
3.1. Abstract
Assessment of “omics” signatures may contribute to personalized medicine and precision
nutrition. However, existing literature is still limited in the homogeneity of participants’
characteristics and in limited assessments of integrated omics layers. Our objective was to use
metabolomics and proteomics to identify biological pathways and functions associated with
high-quality diet in a population of primarily Hispanic young adults. We conducted protein- and
metabolite-wide association studies and functional pathway analyses to assess the relationships
between the Healthy Eating Index-2015 (HEI) and Dietary Approaches to Stop Hypertension
(DASH) diets and proteins (n=334) and untargeted metabolites (n=23,173), using data from the
MetaAIR study (n=154, 61% Hispanic). Analyses were performed for each diet index separately,
adjusting for demographics and BMI. Five proteins (ACY1, ADH4, AGXT, GSTA1, F7), and six
metabolites (undecylenic acid, betaine, hyodeoxycholic acid, stearidonic acid, iprovalicarb,
pyracarbolid) were associated with both diets. These proteins are involved in lipid and amino
acid metabolism and in hemostasis, while significant metabolites included amino acid
derivatives, bile acids, fatty acids, and pesticides. Significantly enriched biological pathways
were involved in macronutrient metabolism, immune function, and oxidative stress. These
findings in young Hispanic adults contribute to efforts to develop precision nutrition and
medicine for diverse populations.
3.2. Introduction
Recent advances in analytic techniques and “omics” technologies have generated
interest in using biomarkers to characterize dietary habits or assess food intake. Omics data
include, though are not limited to, the genome, epigenome, transcriptome, proteome, and

45
metabolome, and aim to comprehensively characterize all molecules within these domains
using advanced high-throughput methods [1]. When assessed together, omics domains can
construct a holistic view of the mechanisms underlying disease development. These data have
been used to identify clinically relevant biomarkers of disease, understand disease progression,
and characterize disease risk [2]. In the developing field of precision nutrition, omics analyses
have been used to assess individual responses to dietary interventions and evaluate dietary
intake [3]. For instance, the metabolome and microbiome have been found to change in
response to the Mediterranean diet [4], and proteomic profiles may reflect weight change after a
diet intervention [5]. Despite rapid advances in omics technologies and their biomedical and
nutritional applications, existing research is limited with respect to the multi-omics integrations
and attention to diverse populations that is necessary to bring precision medicine and precision
nutrition into clinical practice [6].
Though precision medicine and precision nutrition approaches have appeared more
frequently in clinical practice and health research [7, 8], many populations remain
underrepresented. In the United States, Hispanic populations are at increased risk for
metabolic-dysfunction associated steatotic liver disease (MASLD) [9], type 2 diabetes (T2D)
[10], and other metabolic diseases relative to Non-Hispanic Whites, but are not often included in
population-level nutritional or precision medicine investigations [11, 12]. This population may
have different dietary habits than other racial or ethnic groups in the US that may result in
differences in circulating nutrition-related omics biomarkers, or may affect overall health and
disease risk. The diets of Mexican-American and other Hispanic communities in the US have
been shown to differ from those of other ethnic groups, with previous work suggesting the
Hispanic children and adults may consume different amounts of fast food, sweets, fruit juice,
and refined grains [13, 14] relative to other ethnic groups, though overall diet quality is similar to
that of non-Hispanic Whites [15]. Given that the diet composition may be different in Hispanics,

46
including this population in precision nutrition research is important to ensure equity in future
clinical applications.
Recent improvements in analytical techniques have allowed researchers to characterize
the omics phenotypes associated with consumption of certain foods or the phenotype
associated with adherence to healthy dietary patterns, usually measured using diet indices. Diet
indices are constructed from reported dietary intake to characterize overall dietary habits, and
represent common dietary patterns such as the Healthy Eating Index 2015 (HEI) [16] and the
Dietary Approaches to Stop Hypertension (DASH) diet [17]. Adherence to these healthy eating
patterns is associated with lower risk for diseases like T2D, cardiovascular disease, and other
related health outcomes [18-20], though the biological processes or mechanisms underlying
these effects are not well understood. Proteomics and metabolomics in particular are promising
methods that may elucidate the molecular foundations of the relationship between diet and
disease, and have been used to identify pathways of disease development or potential
molecular mediators of these associations [19, 21].
The proteome and metabolome are of particular interest for precision nutrition because
they are the two omics layers most closely related to physiological phenotypes and represent
different but complementary parts of cellular function [22]. The proteome reflects transcriptomic
activity, though it is also affected by other environmental and physiological factors, and consists
of all expressed proteins in a specific tissue [23, 24]. The metabolome is a reflection of cellular
activity and contains molecules produced during cellular metabolism [3]. In diet assessment,
metabolomics and proteomics methods are currently used to identify potential new biomarkers
of dietary intake [25], measure the effects of diet interventions [4], or improve upon traditional
diet assessments [26]. Despite growing interest in the use of omics technologies to evaluate
habitual diet, only a limited number of studies have applied either proteomic or metabolomic
assessments to dietary patterns, and, to our knowledge, only one has applied both [27]. Of
these studies, the majority included relatively homogenous populations; a recent review of

47
metabolomics in nutrition found that virtually all took place in North America and Europe, with
limited racial or ethnic diversity in the selected populations [28]. Young people are also
underrepresented, and only one study to date has focused on younger adults [21]. Additional
research is still needed to assess whether findings may be generalized to all populations.
The purpose of this study is to use a multi-omics approach to examine the impact of diet
quality on biological processes in a population underrepresented in omics research: young
adults of primarily Hispanic ethnicity. Using both proteomics and metabolomics, we will evaluate
the effects of two healthy dietary patterns (the HEI, DASH diets) on alterations in the
metabolome and proteome to identify molecular signatures of high-quality diet and describe the
biological effects of diet using functional pathway analyses.
3.3. Methods
3.3.1. Cohort
Between 2014 and 2018, 155 participants (age 17-22) were recruited from the Children’s
Health Study in Southern California [29] for the MetaAIR study [30]. To be eligible for the
MetaAIR study, participants had a history of overweight or obesity in early high school and did
not have type 1 or type 2 diabetes. At the clinical visit, participants completed a diet
assessment, detailed questionnaires, and an oral glucose tolerance test (OGTT) where plasma
was collected. This study was approved by the University of Southern California Institutional
Review Board. Written informed consent/assent was obtained from participants, or by
participants and their guardians for those under age 18.
3.3.2. Diet Assessment
Participants completed two 24-hour dietary recalls on non-consecutive days, one
weekday and one weekend day, using the Nutritional Data System for Research software
version 2014 (University of Minnesota, Minneapolis, MN). Two diet indices, the HEI-2015 and

48
DASH, were calculated from the recall data as described previously [31]. The HEI-2015 has
thirteen components which are each given a score and summed to a final score between 0 and
100. The components include: total fruit, whole fruit, total vegetables, greens and beans, whole
grains, dairy, total protein foods, seafood and plant protein, fatty acids, refined grains, sodium,
added sugar, and saturated fats [16]. The DASH diet score was calculated from 8 target
nutrients (total fat, saturated fat, total protein, fiber, cholesterol, calcium, magnesium, and
potassium), which are each given a value of 0, 0.5, or 1, to generate a final score between 0
and 8 [17].
3.3.3. Metabolomics
Untargeted plasma metabolomics were measured in plasma samples collected at the
two-hour OGTT time point. Liquid chromatography and high-resolution mass spectrometry
methods (LC-HRMS) were used as described in Liu, et al. (32), with dual column and dual
polarity approaches and both positive and negative ionization. This resulted in four analytical
configurations: reverse phase (C18) positive, C18 negative, hydrophilic interaction (HILIC)
positive, and HILIC negative. Unique features were identified using mass-to-charge ratio (m/z),
retention time, and peak intensity. Features were adjusted for batch variation [33] and excluded
if they were detected in < 75% of samples or if there was a > 30% coefficient of variability of the
quality control samples after batch correction. After processing, there were 3,716 features from
the C18 negative model; 5,069 from the C18 positive mode; 7,444 from the HILIC negative
mode; and 6,944 from the HILIC positive mode, for a total of 23,173 features included in the
analyses. The raw intensity values from LC-HRMS were scaled to a standard normal distribution
and log2 transformed. Details of the analytical process have been described previously [34].
Comparing metabolite features to a database of standards that were analyzed using the
same analytical method identified 466 confirmed compounds. Metabolites’ identities were
assigned using known standards, and detected peaks were matched using accurate mass m/z

49
(<5ppm) and retention time (<15 sec). In instances where multiple annotations were possible
due to more than one molecule having retention times within the allowable error, the annotation
with the closest retention time to the known standard was chosen.
3.3.4. Proteomics
Proteins were measured in fasting plasma samples using the proximity extension array
(PEA) method from Olink Explore 384 Cardiometabolic panel [35]. This panel measures the
relative abundance of 369 proteins, reported as normalized protein expression (NPX) levels
after log2 transformation [36]. For 23 proteins, 50% of observations were below the limit of
detection (LOD) and were excluded from the analysis, leaving 346 proteins from the initial 369
offered after processing.
3.3.5. Covariates
Demographic information was collected through questionnaires. Age was calculated
from visit date and birthday. Participants self-reported race and ethnicity, and were categorized
as Non-Hispanic White, Hispanic, or Other. Participants’ sex and highest level of parental
education (less than high school, completed high school, more than high school, or don’t know)
were recorded. Body mass index (BMI, kg/m2
) was calculated from clinical measurements of
weight and height.
3.3.6. Statistical Analyses
Descriptive statistics were calculated for all diet indices and covariates, including mean
and standard deviation for continuous variables and frequency and percent for categorical
variables.
Omics-Wide Association Studies (OWAS). Protein-wide association studies (PWAS) and
metabolite-wide association studies (MWAS) were conducted between the diet indices and each

50
of the 346 proteins and 23,173 metabolite features individually using linear regression models.
To account for multiple comparisons, we also report p-values adjusted for a false discovery rate
(FDR) (q-values), calculated used the Benjamini-Hochberg method [37]. Q-values < 0.20 were
considered additional evidence for a true association between diet and that individual protein or
metabolite. Each linear model adjusted for age, sex, ethnicity, and BMI. All analyses were
performed using R, version 4.1.0 (R Foundation for Statistical Computing, Vienna, Austria).
Pathway Analysis. Proteomic pathway enrichment analysis was conducted using the
core analysis function of Qiagen Ingenuity Pathway Analysis (IPA) separately for each diet
index (QIAGEN Inc., https://digitalinsights.qiagen.com/IPA) [38]. To identify pathways
associated with either DASH or HEI, UniProt identifiers and test statistics (t-scores) for proteins
nominally significant in the PWAS analysis (p < 0.05) uploaded to IPA separately for each diet
index. Canonical pathways significantly associated (p < 0.05) with each diet index and which
included three or more significant protein hits were extracted.
A metabolomic pathway enrichment analysis was performed for each diet index using
the effect estimates and t-scores for each diet index-metabolite relationship. The analysis was
conducted using MetaboAnalyst (version 5.0), version 2.0 of the MS peaks to path module [39],
using the mixed ion mode, 5 ppm mass tolerance, and Human MFN pathway library, with a pvalue threshold of 0.05. Effect estimates (), t-scores, and p-values for each diet-metabolite
relationship in the MWAS analysis were uploaded to MetaboAnalyst. Only pathways with three
or more significant metabolite hits were extracted. Integrated Gene Set Enrichment Analysis
(GSEA) [40] and mummichog [41] algorithms were used to identify significantly enriched
pathways, based on the combined p-values from both methods [42].
3.4. Results
3.4.1 Study Population Characteristics

51
At the study visit, the average age of participants was 19.7 years (SD: 1.2), and 61% were of
Hispanic or Latino ethnicity (Table 3.1). Participants had a mean HEI score of 52.7 (SD: 13.0)
out of 100 and a mean DASH score of 2.3 (1.5) out of 8. HEI and DASH scores were
significantly positively correlated (r = 0.45, p < 0.001). Of the 155 MetaAIR participants, one was
missing proteomics data and 31 were missing metabolomics, leaving 154 and 124 with
complete data for proteomics and metabolomics analyses, respectively.
Table 3.1. Descriptive statistics for MetaAIR participant demographics and diet indices.
Visit Date
2014-2018
n = 155
Age (years), Mean (SD) 19.7 (1.2)
Sex, n (%)
Female
Male
71 (45.8)
84 (54.2)
Ethnicity, n (%)
Hispanic/Latino
Non-Hispanic White
Other
94 (60.6)
52 (33.5)
9 (5.8)
Parental Education, n (%)
Did not complete high school
Completed high school
More than high school
Don’t know
31 (20.0)
23 (14.8)
96 (61.9)
5 (3.2)
BMI (kg/m2
), Mean (SD) 29.9 (5.1)
HEI, Mean (SD) 52.7 (13.0)
DASH, Mean (SD) 2.26 (1.51)
Abbreviations: BMI, Body Mass Index; HEI,
Healthy Eating Index; DASH, Dietary
Approaches to Stop Hypertension; SD,
Standard Deviation
3.4.2 Diet Quality Was Associated with Proteins and Metabolites
PWAS identified 44 proteins associated with HEI and 25 associated with the DASH score at a
nominal p-value threshold (p < 0.05) (Figure 3.1A). This represented 64 (17.5%) unique
features significantly associated with at least one diet index. There were 5 features (ACY1,
ADH4, AGXT, F7, GSTA1) associated with both diet indices (Table S3.1). These significant
proteins were all inversely associated with HEI and DASH, and were involved in amino acid

52
metabolism, immunity, lipid metabolism, and hemostasis. After adjustment for multiple
comparisons (q < 0.2), 16 proteins (CDH2, CA5A, RARRES2, CCL15, COL18A1, F9, ADH4,
LILRB1, F7, HMOX1, PTPRS, SIGLEC7, TNFSF13B, TFPI, LILRB2, GSTA1) were significantly
associated with HEI. No proteins were significantly associated with DASH after adjustment for
multiple comparisons.
MWAS was performed on all 23,173 untargeted metabolomic features to examine their
associations with each diet index. We found that 1250 metabolomic features were associated
with the HEI and 1106 were associated with DASH scores. Of the 2101 (9.0%) unique features
significantly associated with either diet index at a nominal p-value threshold (p < 0.05), 38 had
confirmed annotations (Figure 3.1B, Table S3.2). These metabolites included compounds
belonging to groups such as amino acids and components of amino acid metabolism, fatty
acids, bile acids, and pesticides. Six annotated metabolites were associated with both diet
indices: undecylenic acid (a medium-chain fatty acid) and betaine (an amino acid derivative)
were positively associated with HEI and DASH, while iprovalicarb and pyracarbolid (fungicides),
stearidonic acid (an omega-3 fatty acid and derivative of lineolic acid), and hyodeoxychoic acid
(a bile acid) were inversely associated with HEI and DASH. After adjustment for multiple
comparisons (q < 0.2), 10 and 1 features were associated with HEI and DASH, respectively.
None of these features had confirmed annotations.

53
Figure 3.1. Chord diagram showing the significant (p < 0.05) associations between each diet
index and each of the proteins (A) and annotated metabolites (B). The width of each arrow
represents the magnitude of the association, and the color represents the direction. Proteins
(B)
(A)

54
and metabolites are grouped by function. Abbreviations: 5-HIAA, 5-hydroxyindoleacetic acid;
AICAR, 5-aminoimidazole-4-carboxamide ribonucleotide
3.4.3 Diet Quality Was Associated with Biological Pathways
The results from the proteomics core analysis in IPA identified 14 unique canonical pathways
associated with at least one diet index (Figure 3.2A, Table S3.3). HEI was associated with 12
pathways and DASH with 2 pathways.
Pathway analysis identified 12 unique metabolic pathways associated with at least one diet
index, with HEI associated with 8 significantly enriched pathways and DASH with 6 significantly
enriched pathways (Figure 3.2B, Table S3.3). Most pathways were related to amino acid
metabolism and metabolism of cofactors and vitamins. Two pathways (butanoate metabolism
and vitamin B6 metabolism) were significantly enriched for both diet indices.

55
Figure 3.2. Significantly enriched proteomics (A) and metabolomics (B) pathways. Enrichment
refers to the number of significant features within the pathway divided by the total number of
features in the pathway.
3.5. Discussion
We have identified metabolomic and proteomic signatures of high-quality diet, as
measured by two established dietary indices: the HEI-2015 and the DASH diet. We identified 5
proteins (F7, ACY1, ADH4, AGXT, and GSTA1) and six metabolites (betaine, iprovalicarb,

56
pyracarbolid, undecylenic acid, stearidonic acid, and hyodeoxycholic acid) with significant
inverse associations with both the HEI and DASH diets, and two metabolites (betaine and
undecylenic acid) positively associated with both diets. Pathway analysis indicated that
adherence to these diets was associated with liver function, immune response, hemostasis, and
multiple disease-specific pathways. Results from both proteomics and metabolomics analyses
revealed omics signatures involved in amino acid and lipid metabolism, as well as oxidative
stress and inflammation, consistent with previous evidence [27, 43]. Our findings provide further
evidence that use of these high-throughput analytical methods may be useful in identifying the
mechanisms by which adherence to a healthy diet may prevent metabolic and other diseases,
and in evaluating individuals for increased risk for disease.
Many of the proteomic and metabolomic features identified in our study have been
previously linked to dietary intake, metabolic disease, or both. Higher levels of the protein F7
(coagulation factor VII) has been linked to cardiovascular disease and obesity [44], while
previous research indicates that proteins ACY1 (aminocyclase 1) and GSTA1 (glutathione Stransferase alpha 1) are positively associated with T2D [45, 46]. Circulating levels of betaine, a
compound involved in homocysteine metabolism and carnitine production, has frequently been
positively associated with reduced risk for T2D [47], a healthy gut microbiome [48], and lower
risk for other metabolic and cardiovascular diseases [49]. Undecylenic acid has also been
identified as a possible indicator of gut microbiome health [50], and has been found to have antiinflammatory and antioxidative properties [51, 52]. Additionally, the functional pathways
identified in this analysis are consistent with the known antioxidant, anti-inflammatory, and
immune responses associated with healthy diets [53]. Both the HEI and the DASH diets are
characterized by high intake of fiber, whole grains, fruit and vegetables, and low intake of
sodium, added sugar, and saturated fats, all dietary components associated with beneficial
circulating proteins and metabolites [27].

57
Our analysis also identified some proteins and metabolites associated with diet that have
not previously been specifically linked to dietary intake or metabolic disease. For instance,
ADH4 (alpha dehydrogenase 4), a protein involved in alcohol oxidation [54], is not known to be
associated with diet, though it is involved in retinol metabolism [55]. AGXT (alanine-glyoxylate
aminotransferase) is a liver enzyme responsible for the oxidation of glyoxylate to oxalate, as
well as a catalyst for aldehyde-ketone and amino acid interconversions [56]. Though circulating
AGXT has not been previously reported as a biomarker of diet or metabolic disease, impaired
expression of the AGXT gene may contribute to the development of atherosclerosis [57], and
AGXT in liver tissue may be a biomarker for hepatocellular carcinoma [58].
Several unexpected metabolites were also identified. Stearidonic acid was inversely
associated with DASH and HEI in our analysis; though it may appear unusual that an omega-3
fatty acid is inversely associated with a healthy diet, because stearidonic acid is not usually a
major component of the diet and is rapidly converted to eicosapentaenoic acid (EPA), low levels
may indicate that more EPA is being generated [59, 60]. Hyodeoxycholic acid is a secondary
bile acid produced by gut microbiota and is involved in the regulation of lipid homeostasis [61].
High levels of hyodeoxycholic acid and other bile acids are a possible sign of diet-related liver
disease, though existing research is limited [62]. Additionally, HEI and DASH were positively
associated with some pesticide metabolites (tebufenozide, pymetrozine, and clothianidin) and
inversely associated with others (ethioencarb/methiocarb, fenuron, flutriafol, iprovalicarb, and
pyracarbolid). The human health effects, if any, of these specific compounds are not clear, and
there are few reports of their detection in human tissue.
Both the HEI and DASH were associated with proteins involved in amino acid and lipid
metabolism, immune system function, and hemostasis. Both diets were also associated with
fatty acid, amino acid, bile acid, and pesticide metabolites, and with similar metabolic pathways.
This is likely due to similarities between the diet indices, each of which rewards increased
consumption of fiber and unsaturated fatty acids, and penalizes consumption of sodium and

58
saturated fat. These nutrients have previously been linked to fatty acid metabolism [63], gut
microbial composition [64], and immune function [65]. However, each dietary pattern was also
distinguished by unique associations with specific omics features and functions. The HEI was
associated with B vitamin and acyl carnitine metabolites, proteins related to stress response and
cell adhesion, and proteomics pathways related to cellular immune response, cell and
organismal growth and development, and multiple disease pathways. In contrast, only DASH
was associated with proteins involved in angiogenesis and few proteomics pathways were
significantly enriched in response to the DASH diet. These differences are likely due to the
different components used to calculate each index. The HEI includes components from whole
foods, such as fruits, vegetables, whole grains, and protein foods. These components include
likely sources of B vitamins (beans, green vegetables, animal protein) [66], foods that contain
antioxidants or anti-inflammatories (fruits, vegetables) [67], and food groups that may reduce
risk for disease (fruit, vegetables, dairy, whole grains) [68]. Because our measure of the DASH
diet includes only nutrients and no whole foods, this score may not have captured these
additional biomarkers and pathways.
Recently, advances in analytic techniques have increased interest in the potential for
‘omics assessments to contribute to “precision medicine”, in which treatments can be tailored to
an individual’s specific genetic and biological state [2]. Proteomics and metabolomics analyses
have been used to evaluate the impact of dietary interventions and weight loss [5], examine the
potential for certain compounds to mediate the relationship between diet and health outcomes
[19], and assess adherence to a diet or consumption of specific food groups [21]. A
combination of both proteomics and metabolomics may provide more information about an
individual’s diet than each omics layer alone. For instance, circulating proteins may reflect long
term or habitual diet, while metabolomics may be more responsive to recent intake or to specific
foods [69, 70]. The molecules and biological pathways identified in this study are potential

59
candidates to monitor adherence to a healthy diet or to track beneficial changes in response to
a clinical treatment plan.
This study has multiple strengths, including the use of both untargeted and targeted
metabolomics and a large proteomics panel, which provided over 800 identified proteins and
metabolites. The use of two established diet indices allowed us to identify features consistent
across different measures of diet quality, while analyzing both proteomics and metabolomics
allowed us to assess biological processes and pathways impacted by both omics layers. This
analysis, conducted in a population of primarily Hispanic young adults, may also reflect dietary
intake or omics signatures specific to this age or ethnic group. Our study participants come
primarily from Hispanic and Latinx communities in Southern California and these communities
have been understudied in precision nutrition research. Limitations include the relatively small
sample size, which likely reduced our ability to detect features that remained statistically
significant after adjustment for multiple comparisons. However, our findings are biologically
plausible and, in many cases, replicate previous work. Additionally, this cross-sectional study
reports associations between diet and omics measurements collected at a single time point. It is
possible that this diet assessment did not completely capture habitual intake, which may
attenuate our results.
Our study identified proteins, metabolites, and biological pathways associated with a
healthy diet, as measured by the HEI and DASH diet indices, in Hispanic young adults. These
results, which included features and pathways linked to amino acid metabolism, the gut
microbiome, immune function, oxidative stress, and inflammation, suggest that there is potential
for proteomics and metabolomics measurements to play a role in clinical assessments and
interventions. Our work contributes to efforts to identify biomarkers of healthy diet that may be
involved in the development of metabolic disease.

60
3.6. References
1. Hasin Y, Seldin M, Lusis A. Multi-omics approaches to disease. Genome Biology
2017;18(1):83. doi: 10.1186/s13059-017-1215-1.
2. Olivier M, Asmis R, Hawkins GA, Howard TD, Cox LA. The Need for Multi-Omics
Biomarker Signatures in Precision Medicine. Int J Mol Sci 2019;20(19). doi:
10.3390/ijms20194781.
3. LeVatte M, Keshteli AH, Zarei P, Wishart DS. Applications of Metabolomics to Precision
Nutrition. Lifestyle Genom 2022;15(1):1-9. doi: 10.1159/000518489.
4. Meslier V, Laiola M, Roager HM, De Filippis F, Roume H, Quinquis B, Giacco R,
Mennella I, Ferracane R, Pons N, et al. Mediterranean diet intervention in overweight
and obese subjects lowers plasma cholesterol and causes changes in the gut
microbiome and metabolome independently of energy intake. Gut 2020;69(7):1258-68.
doi: 10.1136/gutjnl-2019-320438.
5. Figarska SM, Rigdon J, Ganna A, Elmståhl S, Lind L, Gardner CD, Ingelsson E.
Proteomic profiles before and during weight loss: Results from randomized trial of
dietary intervention. Scientific Reports 2020;10(1):7913. doi: 10.1038/s41598-020-
64636-7.
6. Denny JC, Collins FS. Precision medicine in 2030-seven ways to transform healthcare.
Cell 2021;184(6):1415-9. doi: 10.1016/j.cell.2021.01.015.
7. Dennis JM, Young KG, McGovern AP, Mateen BA, Vollmer SJ, Simpson MD, Henley
WE, Holman RR, Sattar N, Pearson ER, et al. Development of a treatment selection
algorithm for SGLT2 and DPP-4 inhibitor therapies in people with type 2 diabetes: a
retrospective cohort study. Lancet Digit Health 2022;4(12):e873-e83. doi:
10.1016/s2589-7500(22)00174-1.
8. Ravi R, Kesari HV. Novel Study Designs in Precision Medicine - Basket, Umbrella and
Platform Trials. Curr Rev Clin Exp Pharmacol 2022;17(2):114-21. doi:
10.2174/1574884716666210316114157.
9. Kardashian A, Serper M, Terrault N, Nephew LD. Health disparities in chronic liver
disease. Hepatology 2023;77(4):1382-403. doi: 10.1002/hep.32743.
10. Aguayo-Mazzucato C, Diaque P, Hernandez S, Rosas S, Kostic A, Caballero AE.
Understanding the growing epidemic of type 2 diabetes in the Hispanic population living
in the United States. Diabetes Metab Res Rev 2019;35(2):e3097. doi:
10.1002/dmrr.3097.
11. The Lancet Digital H. Equitable precision medicine for type 2 diabetes. Lancet Digit
Health 2022;4(12):e850. doi: 10.1016/s2589-7500(22)00217-5.
12. Dhillon J, Jacobs AG, Ortiz S, Diaz Rios LK. A Systematic Review of Literature on the
Representation of Racial and Ethnic Minority Groups in Clinical Nutrition Interventions.
Adv Nutr 2022;13(5):1505-28. doi: 10.1093/advances/nmac002.

61
13. Guerrero AD, Chung PJ. Racial and Ethnic Disparities in Dietary Intake among California
Children. J Acad Nutr Diet 2016;116(3):439-48. doi: 10.1016/j.jand.2015.08.019.
14. Rehm CD, Peñalvo JL, Afshin A, Mozaffarian D. Dietary Intake Among US Adults, 1999-
2012. Jama 2016;315(23):2542-53. doi: 10.1001/jama.2016.7491.
15. Tao MH, Liu JL, Nguyen UDT. Trends in Diet Quality by Race/Ethnicity among Adults in
the United States for 2011-2018. Nutrients 2022;14(19). doi: 10.3390/nu14194178.
16. Krebs-Smith SM, Pannucci TE, Subar AF, Kirkpatrick SI, Lerman JL, Tooze JA, Wilson
MM, Reedy J. Update of the Healthy Eating Index: HEI-2015. J Acad Nutr Diet
2018;118(9):1591-602. doi: 10.1016/j.jand.2018.05.021.
17. Mellen PB, Gao SK, Vitolins MZ, Goff DC, Jr. Deteriorating Dietary Habits Among Adults
With Hypertension: DASH Dietary Accordance, NHANES 1988-1994 and 1999-2004.
Archives of Internal Medicine 2008;168(3):308-14. doi:
10.1001/archinternmed.2007.119.
18. Jannasch F, Kröger J, Schulze MB. Dietary Patterns and Type 2 Diabetes: A Systematic
Literature Review and Meta-Analysis of Prospective Studies. J Nutr 2017;147(6):1174-
82. doi: 10.3945/jn.116.242552.
19. Kim Y, Lu S, Ho JE, Hwang SJ, Yao C, Huan T, Levy D, Ma J. Proteins as Mediators of
the Association Between Diet Quality and Incident Cardiovascular Disease and AllCause Mortality: The Framingham Heart Study. J Am Heart Assoc
2021;10(18):e021245. doi: 10.1161/jaha.121.021245.
20. Chiavaroli L, Viguiliouk E, Nishi SK, Blanco Mejia S, Rahelić D, Kahleová H, SalasSalvadó J, Kendall CW, Sievenpiper JL. DASH Dietary Pattern and Cardiometabolic
Outcomes: An Umbrella Review of Systematic Reviews and Meta-Analyses. Nutrients
2019;11(2). doi: 10.3390/nu11020338.
21. Shah RV, Steffen LM, Nayor M, Reis JP, Jacobs DR, Jr., Allen NB, Lloyd-Jones D,
Meyer K, Cole J, Piaggi P, et al. Dietary metabolic signatures and cardiometabolic risk.
European Heart Journal 2022;44(7):557-69. doi: 10.1093/eurheartj/ehac446.
22. Chen ZZ, Gerszten RE. Metabolomics and Proteomics in Type 2 Diabetes. Circ Res
2020;126(11):1613-27. doi: 10.1161/circresaha.120.315898.
23. Manzoni C, Kia DA, Vandrovcova J, Hardy J, Wood NW, Lewis PA, Ferrari R. Genome,
transcriptome and proteome: the rise of omics data and their integration in biomedical
sciences. Briefings in Bioinformatics 2016;19(2):286-302. doi: 10.1093/bib/bbw114.
24. Aslam B, Basit M, Nisar MA, Khurshid M, Rasool MH. Proteomics: Technologies and
Their Applications. J Chromatogr Sci 2017;55(2):182-96. doi:
10.1093/chromsci/bmw167.
25. Gibbons H, Brennan L. Metabolomics as a tool in the identification of dietary biomarkers.
Proc Nutr Soc 2017;76(1):42-53. doi: 10.1017/s002966511600032x.

62
26. Guasch-Ferré M, Bhupathiraju SN, Hu FB. Use of Metabolomics in Improving
Assessment of Dietary Intake. Clinical Chemistry 2018;64(1):82-98. doi:
10.1373/clinchem.2017.272344.
27. Walker ME, Song RJ, Xu X, Gerszten RE, Ngo D, Clish CB, Corlin L, Ma J, Xanthakis V,
Jacques PF, et al. Proteomic and Metabolomic Correlates of Healthy Dietary Patterns:
The Framingham Heart Study. Nutrients 2020;12(5). doi: 10.3390/nu12051476.
28. Andraos S, Beck KL, Jones MB, Han T-L, Conlon CA, de Seymour JV. Characterizing
patterns of dietary exposure using metabolomic profiles of human biospecimens: a
systematic review. Nutrition Reviews 2022;80(4):699-708. doi: 10.1093/nutrit/nuab103.
29. McConnell R, Shen E, Gilliland Frank D, Jerrett M, Wolch J, Chang C-C, Lurmann F,
Berhane K. A Longitudinal Cohort Study of Body Mass Index and Childhood Exposure to
Secondhand Tobacco Smoke and Air Pollution: The Southern California Children’s
Health Study. Environmental Health Perspectives 2015;123(4):360-6. doi:
10.1289/ehp.1307031.
30. Kim JS, Chen Z, Alderete TL, Toledo-Corral C, Lurmann F, Berhane K, Gilliland FD.
Associations of air pollution, obesity and cardiometabolic health in young adults: The
Meta-AIR study. Environment International 2019;133:105180. doi:
https://doi.org/10.1016/j.envint.2019.105180.
31. Costello E, Goodrich J, Patterson WB, Rock S, Li Y, Baumert B, Gilliland F, Goran MI,
Chen Z, Alderete TL, et al. Diet Quality Is Associated with Glucose Regulation in a
Cohort of Young Adults. Nutrients 2022;14(18):3734.
32. Liu KH, Nellis M, Uppal K, Ma C, Tran V, Liang Y, Walker DI, Jones DP. Reference
Standardization for Quantification and Harmonization of Large-Scale Metabolomics.
Analytical Chemistry 2020;92(13):8836-44. doi: 10.1021/acs.analchem.0c00338.
33. Luan H, Ji F, Chen Y, Cai Z. statTarget: A streamlined tool for signal drift correction and
interpretations of quantitative mass spectrometry-based omics data. Anal Chim Acta
2018;1036:66-72. doi: 10.1016/j.aca.2018.08.002.
34. Goodrich JA, Walker DI, He J, Lin X, Baumert BO, Hu X, Alderete TL, Chen Z, Valvi D,
Fuentes ZC, et al. Metabolic Signatures of Youth Exposure to Mixtures of Per- and
Polyfluoroalkyl Substances: A Multi-Cohort Study. Environ Health Perspect
2023;131(2):27005. doi: 10.1289/ehp11372.
35. Assarsson E, Lundberg M, Holmquist G, Björkesten J, Thorsen SB, Ekman D, Eriksson
A, Rennel Dickens E, Ohlsson S, Edfeldt G, et al. Homogenous 96-plex PEA
immunoassay exhibiting high sensitivity, specificity, and excellent scalability. PLoS One
2014;9(4):e95192. doi: 10.1371/journal.pone.0095192.
36. Petrera A, von Toerne C, Behler J, Huth C, Thorand B, Hilgendorff A, Hauck SM.
Multiplatform Approach for Plasma Proteomics: Complementarity of Olink Proximity
Extension Assay Technology to Mass Spectrometry-Based Protein Profiling. Journal of
Proteome Research 2021;20(1):751-62. doi: 10.1021/acs.jproteome.0c00641.

63
37. Benjamini Y, Hochberg Y. Controlling the False Discovery Rate: A Practical and
Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society: Series B
(Methodological) 1995;57(1):289-300. doi: https://doi.org/10.1111/j.2517-
6161.1995.tb02031.x.
38. Krämer A, Green J, Pollard J, Jr, Tugendreich S. Causal analysis approaches in
Ingenuity Pathway Analysis. Bioinformatics 2013;30(4):523-30. doi:
10.1093/bioinformatics/btt703.
39. Pang Z, Chong J, Zhou G, de Lima Morais DA, Chang L, Barrette M, Gauthier C,
Jacques P-É, Li S, Xia J. MetaboAnalyst 5.0: narrowing the gap between raw spectra
and functional insights. Nucleic Acids Research 2021;49(W1):W388-W96. doi:
10.1093/nar/gkab382.
40. Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, Paulovich
A, Pomeroy SL, Golub TR, Lander ES, et al. Gene set enrichment analysis: A
knowledge-based approach for interpreting genome-wide expression profiles.
Proceedings of the National Academy of Sciences 2005;102(43):15545-50. doi:
doi:10.1073/pnas.0506580102.
41. Li S, Park Y, Duraisingham S, Strobel FH, Khan N, Soltow QA, Jones DP, Pulendran B.
Predicting network activity from high throughput metabolomics. PLoS Comput Biol
2013;9(7):e1003123-e. doi: 10.1371/journal.pcbi.1003123.
42. Chong J, Wishart DS, Xia J. Using MetaboAnalyst 4.0 for Comprehensive and
Integrative Metabolomics Data Analysis. Curr Protoc Bioinformatics 2019;68(1):e86. doi:
10.1002/cpbi.86.
43. Playdon MC, Moore SC, Derkach A, Reedy J, Subar AF, Sampson JN, Albanes D, Gu F,
Kontto J, Lassale C, et al. Identifying biomarkers of dietary patterns by using
metabolomics. Am J Clin Nutr 2017;105(2):450-65. doi: 10.3945/ajcn.116.144501.
44. Bernardi F, Mariani G. Biochemical, molecular and clinical aspects of coagulation factor
VII and its role in hemostasis and thrombosis. Haematologica 2021;106(2):351-62. doi:
10.3324/haematol.2020.248542.
45. Elhadad MA, Jonasson C, Huth C, Wilson R, Gieger C, Matias P, Grallert H, Graumann
J, Gailus-Durner V, Rathmann W, et al. Deciphering the Plasma Proteome of Type 2
Diabetes. Diabetes 2020;69(12):2766-78. doi: 10.2337/db20-0296.
46. Rooney MR, Chen J, Echouffo-Tcheugui JB, Walker KA, Schlosser P, Surapaneni A,
Tang O, Chen J, Ballantyne CM, Boerwinkle E, et al. Proteomic Predictors of Incident
Diabetes: Results From the Atherosclerosis Risk in Communities (ARIC) Study.
Diabetes Care 2023;46(4):733-41. doi: 10.2337/dc22-1830.
47. Morze J, Wittenbecher C, Schwingshackl L, Danielewicz A, Rynkiewicz A, Hu FB,
Guasch-Ferré M. Metabolomics and Type 2 Diabetes Risk: An Updated Systematic
Review and Meta-analysis of Prospective Cohort Studies. Diabetes Care
2022;45(4):1013-24. doi: 10.2337/dc21-1705.

64
48. Koistinen VM, Kärkkäinen O, Borewicz K, Zarei I, Jokkala J, Micard V, Rosa-Sibakov N,
Auriola S, Aura AM, Smidt H, et al. Contribution of gut microbiota to metabolism of
dietary glycine betaine in mice and in vitro colonic fermentation. Microbiome
2019;7(1):103. doi: 10.1186/s40168-019-0718-2.
49. Rosas-Rodríguez JA, Valenzuela-Soto EM. The glycine betaine role in
neurodegenerative, cardiovascular, hepatic, and renal diseases: Insights into disease
and dysfunction networks. Life Sci 2021;285:119943. doi: 10.1016/j.lfs.2021.119943.
50. Wilmanski T, Rappaport N, Earls JC, Magis AT, Manor O, Lovejoy J, Omenn GS, Hood
L, Gibbons SM, Price ND. Blood metabolome predicts gut microbiome α-diversity in
humans. Nat Biotechnol 2019;37(10):1217-28. doi: 10.1038/s41587-019-0233-9.
51. Henry GE, Momin RA, Nair MG, Dewitt DL. Antioxidant and cyclooxygenase activities of
fatty acids found in food. J Agric Food Chem 2002;50(8):2231-4. doi: 10.1021/jf0114381.
52. Jantas D, Piotrowski M, Lason W. An Involvement of PI3-K/Akt Activation and Inhibition
of AIF Translocation in Neuroprotective Effects of Undecylenic Acid (UDA) Against ProApoptotic Factors-Induced Cell Death in Human Neuroblastoma SH-SY5Y Cells. J Cell
Biochem 2015;116(12):2882-95. doi: 10.1002/jcb.25236.
53. Badimon L, Chagas P, Chiva-Blanch G. Diet and Cardiovascular Disease: Effects of
Foods and Nutrients in Classical and Emerging Cardiovascular Risk Factors. Curr Med
Chem 2019;26(19):3639-51. doi: 10.2174/0929867324666170428103206.
54. Luo X, Kranzler HR, Zuo L, Lappalainen J, Yang BZ, Gelernter J. ADH4 gene variation is
associated with alcohol dependence and drug dependence in European Americans:
results from HWD tests and case-control association studies.
Neuropsychopharmacology 2006;31(5):1085-95. doi: 10.1038/sj.npp.1300925.
55. Yin SJ, Chou CF, Lai CL, Lee SL, Han CL. Human class IV alcohol dehydrogenase:
kinetic mechanism, functional roles and medical relevance. Chem Biol Interact
2003;143-144:219-27. doi: 10.1016/s0009-2797(02)00167-9.
56. Danpure CJ. Variable peroxisomal and mitochondrial targeting of alanine: glyoxylate
aminotransferase in mammalian evolution and disease. Bioessays 1997;19(4):317-26.
doi: 10.1002/bies.950190409.
57. Liu Y, Zhao Y, Shukha Y, Lu H, Wang L, Liu Z, Liu C, Zhao Y, Wang H, Zhao G, et al.
Dysregulated oxalate metabolism is a driver and therapeutic target in atherosclerosis.
Cell Rep 2021;36(4):109420. doi: 10.1016/j.celrep.2021.109420.
58. Zhao CL, Hui Y, Wang LJ, Yang D, Yakirevich E, Mangray S, Huang CK, Lu S. Alanineglyoxylate aminotransferase 1 (AGXT1) is a novel marker for hepatocellular carcinomas.
Hum Pathol 2018;80:76-81. doi: 10.1016/j.humpath.2018.05.025.
59. Shahidi F, Ambigaipalan P. Omega-3 Polyunsaturated Fatty Acids and Their Health
Benefits. Annu Rev Food Sci Technol 2018;9:345-81. doi: 10.1146/annurev-food111317-095850.

65
60. Prasad P, Anjali P, Sreedhar RV. Plant-based stearidonic acid as sustainable source of
omega-3 fatty acid with functional outcomes on human health. Crit Rev Food Sci Nutr
2021;61(10):1725-37. doi: 10.1080/10408398.2020.1765137.
61. Song C, Hiipakka RA, Liao S. Selective activation of liver X receptor alpha by 6alphahydroxy bile acids and analogs. Steroids 2000;65(8):423-7. doi: 10.1016/s0039-
128x(00)00127-6.
62. Batta AK, Arora R, Salen G, Tint GS, Eskreis D, Katz S. Characterization of serum and
urinary bile acids in patients with primary biliary cirrhosis by gas-liquid chromatographymass spectrometry: effect of ursodeoxycholic acid treatment. J Lipid Res
1989;30(12):1953-62.
63. Rosqvist F, Iggman D, Kullberg J, Cedernaes J, Johansson HE, Larsson A, Johansson
L, Ahlström H, Arner P, Dahlman I, et al. Overfeeding polyunsaturated and saturated fat
causes distinct effects on liver and visceral fat accumulation in humans. Diabetes
2014;63(7):2356-68. doi: 10.2337/db13-1622.
64. Makki K, Deehan EC, Walter J, Bäckhed F. The Impact of Dietary Fiber on Gut
Microbiota in Host Health and Disease. Cell Host Microbe 2018;23(6):705-15. doi:
10.1016/j.chom.2018.05.012.
65. Fu Y, Wang Y, Gao H, Li D, Jiang R, Ge L, Tong C, Xu K. Associations among Dietary
Omega-3 Polyunsaturated Fatty Acids, the Gut Microbiota, and Intestinal Immunity.
Mediators Inflamm 2021;2021:8879227. doi: 10.1155/2021/8879227.
66. Hanna M, Jaqua E, Nguyen V, Clay J. B Vitamins: Functions and Uses in Medicine.
Perm J 2022;26(2):89-97. doi: 10.7812/tpp/21.204.
67. Serafini M, Peluso I. Functional Foods for Health: The Interrelated Antioxidant and AntiInflammatory Role of Fruits, Vegetables, Herbs, Spices and Cocoa in Humans. Curr
Pharm Des 2016;22(44):6701-15. doi: 10.2174/1381612823666161123094235.
68. Schwingshackl L, Hoffmann G, Lampousi AM, Knüppel S, Iqbal K, Schwedhelm C,
Bechthold A, Schlesinger S, Boeing H. Food groups and risk of type 2 diabetes mellitus:
a systematic review and meta-analysis of prospective studies. Eur J Epidemiol
2017;32(5):363-75. doi: 10.1007/s10654-017-0246-y.
69. Bourgonje AR, Bolte LA, Vranckx LLC, Spekhorst LM, Gacesa R, Hu S, van Dullemen
HM, Visschedijk MC, Festen EAM, Samsom JN, et al. Long-Term Dietary Patterns Are
Reflected in the Plasma Inflammatory Proteome of Patients with Inflammatory Bowel
Disease. Nutrients 2022;14(12). doi: 10.3390/nu14122522.
70. Lloyd AJ, Willis ND, Wilson T, Zubair H, Chambers E, Garcia-Perez I, Xie L, Tailliart K,
Beckmann M, Mathers JC, et al. Addressing the pitfalls when designing intervention
studies to discover and validate biomarkers of habitual dietary intake. Metabolomics
2019;15(5):72. doi: 10.1007/s11306-019-1532-3.

66
3.7. Supplemental Material
Table S3.1. Proteins significantly associated with at least one diet index.
Gene Name Protein Name UniProt Diet Beta P-Value Q-Value Protein Function
ACY1 Aminoacylase 1 Q03154
HEI;
DASH
-0.012;
-0.153
0.020;
0.007
0.28;
0.44
amino acid
metabolism
ADAMTS16
ADAM
metallopeptidase
with
thrombospondin
type 1 motif 16 Q8TE57 HEI
-0.008
0.020 0.28 hemostasis
ADH4
Alcohol
dehydrogenase 4 P08319
HEI;
DASH
-0.025;
-0.207
0.003;
0.026
0.14;
0.56 lipid metabolism
AGXT
Alanine--glyoxylate
aminotransferase P21549
HEI;
DASH
-0.012;
-0.147
0.043;
0.028
0.36;
0.56
amino acid
metabolism
ANG Angiogenin P03950 HEI -0.006 0.032 0.35 stress response
AZU1 Azurocidin 1 P20160 DASH 0.113 0.032 0.56 immunity
BPIFB1
BPI fold containing
family B member 1 Q8TDL5 DASH 0.135 0.036 0.56 immunity
CA4
Carbonic anhydrase
4 P22748 DASH 0.040 0.045 0.56 CO2 homeostasis
CA5A
Carbonic anhydrase
5A P35218 HEI -0.023 0.001 0.13 CO2 homeostasis
CANT1
Calcium activated
nucleotidase 1 Q8WVQ1 HEI -0.004 0.043 0.36 nucleotidase
CCL14
C-C motif
chemokine ligand
14 Q16627 HEI
-0.006
0.034 0.35 immunity
CCL15
C-C motif
chemokine ligand
15 Q16663 HEI
-0.012
0.003 0.14 immunity
CCL27
C-C motif
chemokine ligand
27 Q9Y4X3 HEI
-0.021
0.047 0.37 immunity
CDH2 Cadherin 2 P19022 HEI -0.011 0.0002 0.10 cell adhesion
CDHR5
Cadherin related
family member 5 Q9HBB8 HEI -0.006 0.042 0.36 cell adhesion
CES1 Carboxylesterase 1 P23141 HEI -0.014 0.015 0.23 lipid metabolism
CHIT1 Chitinase 1 Q13231 DASH 0.333 0.010 0.44
carbohydrate
metabolism
CLC
Charcot-Leyden
crystal galectin Q05315 DASH -0.084 0.043 0.56 immunity
COL18A1
Collagen type XVIII
alpha 1 chain P39060 HEI -0.007 0.002 0.14 cell adhesion
CPA1
Carboxypeptidase
A1 P15085 DASH 0.109 0.032 0.56
amino acid
metabolism
CPB1
Carboxypeptidase
B1 P15086 DASH 0.125 0.003 0.44
amino acid
metabolism
CST3 Cystatin C P01034 HEI -0.005 0.023 0.29 protease inhibitor
CTSZ Cathepsin Z Q9UBR2 HEI -0.004 0.030 0.35 peptidase
DPP7
Dipeptidyl peptidase
7 Q9UHL4 HEI -0.009 0.043 0.36 immunity
F7
Coagulation factor
VII P08709
HEI;
DASH -0.007 0.004;
0.022
0.14;
0.56 hemostasis
F9
Coagulation factor
IX P00740 HEI -0.062 0.003 0.14 hemostasis
FAP
Fibroblast activation
protein alpha Q12884 DASH -0.010 0.018 0.56 angiogenesis

67
Gene Name Protein Name UniProt Diet Beta P-Value Q-Value Protein Function
FETUB Fetuin B Q9UGM5 DASH -0.069 0.040 0.56 fertilization
GH1 Growth hormone 1 P01241 DASH -0.069 0.006 0.44 growth
GSTA1
Glutathione Stransferase alpha 1 P08263
HEI;
DASH
-0.019;
-0.203
0.008;
0.010
0.18;
0.44 lipid metabolism
GUSB Glucuronidase beta P08236 DASH -0.115 0.023 0.56
carbohydrate
metabolism
HMOX1 Heme oxygenase 1 P09601 HEI 0.010 0.006 0.16 immunity
HYOU1
Hypoxia upregulated 1 Q9Y4L1 HEI -0.004 0.011 0.21 stress response
ICAM1
Intercellular
adhesion molecule
1 P05362 HEI
-0.005
0.029 0.35 immunity
IGFBP2
Insulin like growth
factor binding
protein 2 P18065 DASH
0.159
0.008 0.44 growth
IGSF8
Immunoglobulin
superfamily member
8 Q969P0 HEI
-0.005
0.040 0.36 immunity
LGALS1 Galectin 1 P09382 HEI -0.005 0.046 0.37 immunity
LILRA5
Leukocyte
immunoglobulin like
receptor A5 A6NI73 HEI
-0.007
0.012 0.21 immunity
LILRB1
Leukocyte
immunoglobulin like
receptor B1 Q8NHL6 HEI
-0.007
0.004 0.14 immunity
LILRB2
Leukocyte
immunoglobulin like
receptor B2 Q8N423 HEI
-0.006
0.008 0.17 immunity
MARCO
Macrophage
receptor with
collagenous
structure Q9UEW3 DASH
-0.050
0.038 0.56 immunity
MFAP3
Microfibril
associated protein 3 P55082 DASH -0.085 0.025 0.56
microfibril
component
MNDA
Myeloid cell nuclear
differentiation
antigen P41218 DASH
0.165
0.007 0.44 immunity
MSTN Myostatin O14793 DASH -0.129 0.026 0.56 growth
NRP1 Neuropilin 1 O14786 HEI 0.004 0.042 0.36 angiogenesis
PDCD6
Programmed cell
death 6 O75340 DASH -0.156 0.038 0.56 angiogenesis
PDGFRA
Platelet derived
growth factor
receptor alpha P16234 HEI
-0.005
0.040 0.36 immunity
PGLYRP1
Peptidoglycan
recognition protein 1 O75594 DASH 0.103 0.008 0.44 immunity
PILRB
Paired
immunoglobin like
type 2 receptor beta Q9UKJ0 HEI
-0.009
0.048 0.38 immunity
PRSS2 Serine protease 2 P07478 DASH 0.087 0.018 0.56 digestion
PTGDS
Prostaglandin D2
synthase P41222 HEI -0.004 0.038 0.36 lipid metabolism
PTPRS
Protein tyrosine
phosphatase
receptor type S Q13332 HEI
-0.005
0.006 0.16 cell adhesion

68
Gene Name Protein Name UniProt Diet Beta P-Value Q-Value Protein Function
RARRES2
Retinoic acid
receptor responder
2 Q99969 HEI
-0.013
0.001 0.13 inflammation
SEMA3F Semaphorin 3F Q13275 HEI -0.006 0.010 0.20 cell adhesion
SERPINA11
Serpin family A
member 11 Q86U17 HEI -0.009 0.021 0.28 protease inhibitor
SERPINB5
Serpin family B
member 5 P36952 HEI 0.011 0.0201 0.28 tumor suppressor
SIGLEC7
Sialic acid binding
Ig like lectin 7 Q9Y286 HEI -0.006 0.006 0.16 cell adhesion
SOST Sclerostin Q9BQB4 HEI -0.008 0.011 0.21 bone formation
ST6GAL1
ST6 betagalactoside alpha2,6-sialyltransferase
1 P15907 HEI
-0.005
0.022 0.28 transferase
TFPI
Tissue factor
pathway inhibitor P10646 HEI -0.006 0.007 0.17 hemostasis
THPO Thrombopoietin P40225 HEI -0.007 0.014 0.23 immunity
TNFSF13B
TNF superfamily
member 13b Q9Y275 HEI -0.005 0.005 0.16 immunity
UMOD Uromodulin P07911 DASH 0.096 0.036 0.56 immunity
VASN Vasorin Q6EMK4 HEI -0.005 0.032 0.35 signal inhibitor

69
Table S3.2. Annotated metabolites significantly associated with at least one diet index.
MZ_RT Metabolite Name Diet Beta
PValue
QValue
Compound
Class
118.086145_21.36496892
5-Aminovaleric
acid DASH 0.143 0.022 0.94
delta amino
acids and
derivatives
118.086230393501_242.60893
5893286 Betaine
DASH;
HEI
0.132;
0.016
0.035;
0.016
0.99;
0.84
alpha amino
acids
129.092076556291_69.054199
4234249 Heptanoate HEI 0.014 0.048 0.99
medium-chain
fatty acids
130.0862206_20.25140869 Pipecolic acid HEI 0.017 0.015 0.65
alpha amino
acids
131.0825382_22.89270939 Ornithine DASH -0.142 0.030 0.88
alpha amino
acids
132.0767511_22.76024081 Creatine DASH -0.115 0.043 0.94
alpha amino
acids and
derivatives
141.0714769_20.73123889 Dihydrobiopterin DASH 0.139 0.031 0.94
biopterins and
derivatives
165.102438823495_67.772780
1186609 Fenuron DASH -0.137 0.024 0.99 herbicide
166.014645408756_337.90179
8527843 Quinolinic acid HEI -0.014 0.034 0.99
pyridinecarbox
ylic acids
170.0395578_20.02746515 AICAR DASH 0.133 0.027 0.94
1-ribosylimidazolecarbo
xamides
170.0811096_29.25485743 Pyridoxine HEI -0.016 0.019 0.67 pyridoxines
173.045559280627_317.28170
9506299 Shikimate HEI 0.015 0.032 0.99
shikimic acids
and derivatives
177.0403687_21.45490336 L-Gulonolactone HEI 0.016 0.024 0.77
gamma
butyrolactones
182.046024882254_52.486036
6311606 4-Pyridoxate HEI 0.014 0.039 0.99
methylpyridine
s
184.0727804_21.31249393 5-Methylcytosine DASH 0.133 0.038 0.88
hydroxypyrimid
ines
185.1535534_244.9770271 Undecylenic acid
HEI;
DASH
0.015;
0.213
0.035;
0.001
0.73;
0.38
medium-chain
fatty acids
190.0508276_26.82145169 5-HIAL HEI 0.017 0.013 0.77 hydroxyindoles
212.00232296579_39.3293524
152959 Indoxyl Sulfate HEI -0.017 0.015 0.99 arylsulfates
238.083995324624_33.676866
8312743 Pyracarbolid
DASH;
HEI
-0.174;
-0.015
0.005;
0.022
0.74;
0.99 fungicide
245.0951691_61.59122947 Biotin HEI -0.014 0.038 0.75
biotin and
derivatives
246.1697704_42.0591879
2-
Methylbutyroylcar
nitine HEI -0.013 0.027 0.70 acyl carnitines
248.069957634648_79.880224
6237701
Ethiofencarb;
Methiocarb HEI -0.015 0.022 0.84 pesticide
248.1477719_38.97478419
Hydroxybutyrylcar
nitine HEI 0.017 0.016 0.65 acyl carnitines
250.014750590405_58.569574
80077 Clothianidin HEI 0.020 0.002 0.74 insecticide
255.232921924428_58.501132
6171589 Palmitate DASH -0.125 0.034 0.95
long-chain fatty
acids
275.201721113245_59.446303
1320155 Stearidonic acid
DASH;
HEI
-0.200;
-0.020
0.001;
0.002
0.58;
0.86
lineolic acids
and derivatives
281.1129803_31.69698361 Pymetrozine DASH 0.133 0.040 0.94 insecticide

70
MZ_RT Metabolite Name Diet Beta
PValue
QValue
Compound
Class
305.248512471191_57.211486
6743449
5Z,8Z,11ZEicosatrienoic
acid HEI -0.014 0.023 0.99
long-chain fatty
acids
343.137111429024_71.319310
0597706 Flutriafol DASH -0.138 0.034 0.99 pesticide
347.2213741_240.323714 Corticosterone DASH 0.131 0.043 0.94
21-
hydroxysteroid
s
362.2423655_268.1736072 Iprovalicarb
HEI;
DASH
-0.014;
-0.166
0.049;
0.011
0.77;
0.94 fungicide
370.980388110205_338.02043
6813762
2-
Phosphoglycerate DASH -0.134 0.039 0.95
sugar acids
and derivatives
387.184837897759_65.393913
8352356 Tebufenozide DASH 0.115 0.048 0.95 insecticide
389.271769092124_67.789315
9938276
7-Ketodeoxycholic
acid HEI -0.017 0.015 0.84
dihydroxy bile
acids, alcohols
and derivatives
391.2851729_236.6284427
Hyodeoxycholic
acid
HEI;
DASH
-0.018;
-0.150
0.012;
0.021
0.77;
0.87
dihydroxy bile
acids, alcohols
and derivatives
400.342176167492_181.73034
4383984 Palmitoylcarnitine HEI -0.012 0.032 0.88 acyl carnitines
401.3408332_420.5308696 7 alpha-3ox-C DASH -0.146 0.020 0.94
cholesterols
and derivatives
498.2894603_225.6807061
Taurochenodesox
ycholic acid DASH -0.134 0.042 0.91
taurinated bile
acids and
derivatives

71
Table S3.3. Proteomic and metabolomic pathways and their contributing features.
Omics
Layer Pathway
Top Level
Pathway Diet Enrichment P-value
Contributing
Features
Proteome
Agranulocyte
Adhesion and
Diapedesis
Cellular
Immune
Response
HEI 0.019 < 0.001 CCL14,CCL15,CCL27,
ICAM1
Proteome Coagulation
System
Cellular
Stress and
Injury
HEI;
DASH
0.0857;
0.0286
< 0.001;
0.036
F7,F9,TFPI (HEI); F7
(DASH)
Proteome
Granulocyte
Adhesion and
Diapedesis
Cellular
Immune
Response
HEI 0.0212 < 0.001 CCL14,CCL15,CCL27,
ICAM1
Proteome
Hepatic
Fibrosis /
Hepatic
Stellate Cell
Activation
Ingenuity
Toxicity List
Pathways;
DiseaseSpecific
Pathways
HEI 0.0155 0.006 COL18A1,ICAM1,PDG
FRA
Proteome
Hepatic
Fibrosis
Signaling
Pathway
Cellular
Growth,
Proliferation
and
Development
; Cellular
Stress and
Injury;
DiseaseSpecific
Pathway
HEI 0.00709 0.044 COL18A1,ICAM1,PDG
FRA
Proteome Natural Killer
Cell Signaling
Cancer;
Cellular
Immune
Response
HEI 0.0152 0.006 COL18A1,LILRB1,SIG
LEC7
Proteome
Neuroprotecti
ve Role of
THOP1 in
Alzheimer's
Disease
DiseaseSpecific
Pathways
DASH 0.0248 < 0.001 F7,FAP,PRSS2
Proteome
Pathogen
Induced
Cytokine
Storm
Signaling
Pathway
DiseaseSpecific
Pathways
HEI 0.0135 0.001 CCL14,CCL15,CCL27,
COL18A1,TNFSF13B
Proteome
Pulmonary
Fibrosis
Idiopathic
Signaling
Pathway
Cellular
Stress and
Injury;
DiseaseSpecific
Pathways
HEI 0.0092 0.022 CDH2,COL18A1,PDG
FRA
Proteome
Regulation Of
The Epithelial
Mesenchymal
Transition By
Growth
Factors
Pathway
Growth
Factor
Signaling;
Organismal
Growth and
Development
HEI 0.0156 0.005 CDH2,PDGFRA,TNFS
F13B
Proteome
Role of
Macrophages
, Fibroblasts
and
DiseaseSpecific
Pathways
HEI 0.00904 0.023 ICAM1,SOST,TNFSF1
3B

72
Omics
Layer Pathway
Top Level
Pathway Diet Enrichment
P
-value
Contributing
Features
Endothelial
Cells in
Rheumatoid
Arthritis
Proteome
Role Of
Osteoblasts
In
Rheumatoid
Arthritis
Signaling
Pathway
Disease
-
Specific
Pathways
HEI 0.0123 0.010 CTSZ,SOST,TNFSF13 B
Proteome
SPINK1
Pancreatic
Cancer
Pathway
Cancer;
Disease
-
Specific
Pathways
DASH 0.05 < 0.001 CPA1,CPB1,PRSS2
Proteome
Xenobiotic
Metabolism
Signaling
Xenobiotic
Metabolism;
Ingenuity
Toxicity List
Pathways
HEI 0.0103;
0.00685
0.017;
0.014 CES1,GSTA1,HMOX1
Metabolome
Arachidonic
acid
metabolism
Lipid
metabolism DASH 0.16 0.030
261.22078_66.5;
277.21716_59.2;
278.22062_59.3;
279.23201_63.8
(Gamma
-Linolenate);
280.23498_63.4;
293.21229_64.1;
297.24276_63.0;
277.21670_323.5;
278.22025_323.9;
297.24235_315.4;
337.23795_236.7;
348.07018_268.5;
606.28581_267.0;
646.27833_264.1;
153.12755_65.1;
269.22624_63.9;
287.23705_69.3;
303.23212_57.5;
304.23564_57.3;
305.24611_65.6;
319.22730_65.6;
327.22865_69.0;
339.20902_57.7;
349.23632_57.2;
363.25182_56.8;
285.22204_337.7;
303.23258_338.3
(Arachidonic acid);
304.23585_338.2;
339.20995_336.6;
311.25781_277.1;
377.22949_269.0;
377.22883_360.7;
299.20130_317.7;
397.13883_280.7
(Thyrotropin Releasing
Hormone);
167.10767_273.8;

73
Omics
Layer Pathway
Top Level
Pathway Diet Enrichment
P
-value
Contributing
Features
265.21591_248.2;
301.21701_259.0;
317.21181_272.2;
319.22634_267.0;
337.23626_269.3;
359.22082_266.7;
405.22327_258.6;
167.10768_202.5;
335.22235_234.5;
336.22572_235.9;
351.21778_229.8;
372.17280_224.1
Metabolome
Benzoate
degradation
via CoA
ligation
Xenobiotics
biodegradati
on and
metabolism
HEI
1 0.026
109.02955_34.8;
134.06114_26.5;
93.03453_40.7;
93.03464_41.5
Metabolome Biopterin
metabolism
Metabolism
of cofactors
and vitamins
DASH 0.267 0.021
110.05995_32.6;
110.06010_24.7;
120.08066_27.2
(Phenylethanolamine);
120.080870_220.7;
122.09633_23.3;
122.09650_218.6;
130.06506_33.3;
130.06512_214.2;
136.07559_25.4;
136.07581_79.3;
138.09143_219.3;
138.09147_76.0;
146.05999_35.0;
148.07562_208.6 (3
-
Methyl
-
2
-Oxindole);
148.07576_220.7;
149.05967_215.0;
149.05968_26.3;
149.05970_221.0;
154.08621_37.0
(Dopamine);
154.08640_76.0;
164.07051_71.7;
164.07070_73.1;
164.07127_26.5
(Epinephrine);
165.05459_22.4;
165.07491_26.5;
166.08618_27.4
(Phenylalanine);
166.08638_220.7;
167.08955_26.2;
167.09034_221.2;
180.06654_23.1;
182.08086_22.8;
183.08437_22.8;
201.04057_65.1;
222.09838_80.6;
236.07885_145.7;
240.08828_42.0
Metabolome Butanoate
metabolism
Carbohydrat
e metabolism
DASH;
HEI 0.0833 0.048;
0.023
101.02449_207.5;
101.05930_69.9 (5
-
Valeroalacetane);

74
Omics
Layer Pathway
Top Level
Pathway Diet Enrichment
P
-value
Contributing
Features
103.04008_20.2 (2
-
Hydroxybutyrate);
103.04017_214.6;
111.04412_66.6
(Pyrocatechol);
113.02443_22.0;
113.02454_309.9;
114.09144_75.9;
116.07055_22.1;
116.07062_246.3
(Proline);
116.07172_22.4
(Valine);
116.10705_119.9;
117.01934_18.0
(Succinate);
117.01942_205.2;
117.07506_22.4;
117.10227_33.1;
118.02271_17.7;
118.08615_21.4 (5
-
Aminovaleric acid);
118.08623_242.6
(Betaine);
119.08954_21.4;
119.08966_241.9;
124.07568_84.9;
124.07570_24.2;
127.04001_59.7;
127.04019_65.4;
129.05471_66.9;
131.03498_300.9;
131.03503_308.2
(Glutarate);
132.03842_313.5;
132.10182_24.1;
133.01426_18.3
(Malate);
140.06816_19.9;
141.01367_23.0;
142.08623_23.0;
142.08642_81.7;
143.01073_15.9;
143.07045_69.5;
143.11786_25.5;
144.06673_215.1;
144.06675_17.8;
144.10188_21.9;
144.10200_237.9;
145.05079_64.8;
145.13347_33.4;
147.02989_20.1
(Citramalate);
147.02999_207.6;
149.04547_20.4
(Arabinose; Xylose);
149.04558_221.2;
152.07058_37.5;
152.07070_237.6;

75
Omics
Layer Pathway
Top Level
Pathway Diet Enrichment
P
-value
Contributing
Features
156.04209_20.4;
158.04601_18.4;
160.09653_26.8;
160.096927_78.6;
160.09693_241.2;
161.12836_17.7;
162.07739_16.8;
163.02475_22.5 (24
-
Dihydroxypteridine);
163.06119_22.3
(Ramnose);
168.06676_17.5;
170.08111_29.3
(Pyridoxine);
171.00554_15.6;
172.09676_36.6
(Mevalolacetone);
176.09261_23.9;
177.04037_21.5 (L
-
Gulonolacetone);
179.05638_105.2;
186.07719_16.7;
188.09142_32.7
(Adipate);
188.09188_234.6;
189.12325_23.0;
191.05601_24.0;
191.05605_308.1;
191.05616_300.7
(Quinic acid);
227.07890_33.8;
227.10376_244.4;
348.07018_268.5;
99.00878_20.3;
Metabolome Carnitine
shuttle
Lipid
metabolism HEI 0.258 0.008
146.11752_21.3
(Deoxycarnitine);
146.11761_238.4;
172.13313_26.1;
172.13329_230.0;
174.14879_23.5;
174.14899_227.7;
182.11776_231.3;
190.14379_221.6;
200.12788_22.5;
200.12812_228.1;
218.13824_23.0
(Propanoylcarnitine);
218.13879_218.3;
219.14173_22.9;
219.14175_217.9;
288.25340_51.1;
328.32070_268.5;
356.35194_276.9;
372.31067_258.1;
372.34713_285.1;373.
31381_258.1;
386.17240_53.1;
386.32662_182.7;
394.33224_312.6;
398.32612_262.5;

76
Omics
Layer Pathway
Top Level
Pathway Diet Enrichment
P
-value
Contributing
Features
399.32952_262.6;
400.34181_274.1;
400.34218_181.7
(Palmitoylcarnitine);
401.34510_274.1;
401.34569_181.0;
414.35741_281.6;
430.31658_189.2;
446.32561_261.4;
450.35838_179.2;
451.36162_179.0;
454.32987_207.2;
465.28574_260.8;
468.30739_275.9;
480.28640_269.9;
494.32329_283.5;
494.32365_203.7;
496.33943_203.6;
496.33961_309.2;
512.46751_345.1;
512.46802_172.2;
513.47131_172.4;
524.37141_201.2;
554.26721_316.3;
580.29948_204.9
Metabolome Histidine
metabolism
Amino acid
metabolism DASH 0.0909 0.045
107.06013_29.6;
110.07119_16.9
110.07134_235.4;
111.05528_24.6;
111.05533_165.5;
112.08692_164.6;
123.05649_16.3;
125.07090_29.0;
128.08188_168.0;
128.08189_25.7;
128.08191_232.3;
136.05180_310.9;
136.05180_137.0;
137.03580_184.8;
138.06617_51.9;
138.06632_28.1;
138.06632_166.2;
139.05019_22.2
(Urocanate);
139.05146_19.0;
143.08150_27.4;
153.03068_18.3;
154.06216_20.4;
154.06241_310.9;
155.06548_27.7;
156.07676_16.7
(Histidine);
156.07685_234.3;
157.06074_26.9;
157.08008_17.4;
169.06201_20.1;
178.05860_17.2;
183.04111_17.9;
183.07758_19.0;
195.08875_164.0;

77
Omics
Layer Pathway
Top Level
Pathway Diet Enrichment
P
-value
Contributing
Features
197.05663_24.9;
200.06786_139.0;
214.08349_17.9;
214.08352_137.5;
348.07018_268.5
Metabolome Lysine
metabolism
Amino acid
metabolism HEI 0.211 0.003
110.05995_32.6;
110.06010_24.7;
110.06010_228.2 (2
-
Aminophenol);
110.06013_158.6;
111.04412_66.6
(Pyrocatechol);
116.07055_22.1;
116.07062_246.4
(Proline);
117.01934_18.0
(Succinate);
118.02271_17.7;
118.08615_21.4 (5
-
Aminovaleric acid);
118.08623_242.6
(Betaine);
126.05505_228.1;
128.07066_235.3;
128.07070_75.4;
128.07072_167.1;
128.07179_289.9;
129.05455_36.6;
129.10233_291.0;
130.08622_20.3
(Pipecolic acid);
130.08638_234.4;
131.08963_21.9;
133.01426_18.3
(Malate);
134.08116_268.6;
142.05087_26.4
(Aminoadipate);
142.05097_268.9;
142.05101_41.3;
143.15434_17.2;
144.06551_26.2 (2
-
Oxobutyric acid);
144.06561_248.7;
144.06566_80.1;
144.06675_17.8;
145.04942_22.1;
145.04962_235.1;
145.09819_23.2
(Lysine);
146.08111_25.4 (5
-
Hydroxypipecolic
acid);
146.08127_242.0;
146.08130_81.9;
147.11272_16.6
(Diethanolamine);
148.09674_22.4;
148.09688_235.6;
158.04601_18.4;

78
Omics
Layer Pathway
Top Level
Pathway Diet Enrichment
P
-value
Contributing
Features
159.02999_204.8;
160.06159_256.2;
162.07609_258.6 (N
-
Methylglutamate);
162.07612_22.5 (2
-
Aminoadipic acid);
162.07620_92.3;
163.02475_22.5 (24
-
Dihydroxypteridine);
165.01920_15.0 (34
-
Dihydroxymandelate);
165.01921_290.3;
165.04057_21.8;
166.04426_27.5;
166.07213_17.2;
166.07287_282.1;
172.13313_26.1;
175.02483_207.0;
177.04037_21.5 (L
-
Gulonolacetone);
180.08669_25.6
(Galactosamine);
180.08673_237.5;
180.08680_341.9 (D
-
Mannosamine);
181.03551_26.5;
181.03680_274.9;
186.07719_16.7;
188.05652_324.4 (N
-
Acetylglutamate);
188.09290_295.3;
189.15952_17.6
(Propamocarb);
190.07140_337.5;
198.03501_112.2;
200.05649_205.9;
201.08823_136.3;
206.06698_239.6;
206.06705_109.6;
212.03208_14.6;
225.02408_22.7;
247.09343_344.3;
257.14700_21.1;
348.07018_268.5;
99.00878_20.3
Metabolome Tryptophan
metabolism
Amino acid
metabolism
(Aromatic
Amino Acids)
HEI 0.178 0.004
106.02846_15.3
(Nicotinic acid);
106.06491_64.4;
110.05995_32.6;
110.06010_24.7;
110.06010_228.2;
118.06505_371.2;
118.06506_32.4;
118.06518_66.1;
118.06519_226.6;
120.04433_35.9;
121.02829_131.0;
121.02832_34.6;
122.02475_19.5;

79
Omics
Layer Pathway
Top Level
Pathway Diet Enrichment
P
-value
Contributing
Features
122.02475_18.8
(Nicotinate);
122.02481_209.6;
130.06506_33.3;
130.06512_214.2;
132.08072_34.1;
132.08084_230.4;
132.08090_65.5;
134.05999_115.4;
134.06013_80.8;
136.04048_109.7 (4
-
Aminobenzoate);
137.07095_275.9;
138.01976_15.2;
138.05493_118.7;
138.05505_240.4;
139.05836_240.9;
140.03542_205.7;
142.06524_230.4;
144.04429_374.6;
144.08075_33.8;
144.08087_226.1;
146.02472_100.0;
146.02489_16.7;
146.05999_35.0;
146.06014_230.4;
148.03867_246.4;
148.07562_208.6 (3
-
Methyl
-
2
-Oxindole);
148.07576_220.7;
149.02330_13.8;
149.02349_232.1;
152.03522_231.7;
152.03550_19.1;
156.08098_234.0;
159.02999_204.8;
159.09163_33.6;
159.09178_230.4;
159.09268_33.9;
160.04048_139.4;
160.07565_32.8;
160.07567_177.9;
162.09193_71.8
(Indole
-
3
-Ethanol);
164.03533_26.4 (4
-
Pyridoxolacetate);
164.03543_142.7;
164.07051_71.7;
164.07070_73.1;
166.01464_337.9
(Quinolinate);
166.01468_15.9;
170.06000_34.5;
170.06007_230.3;
172.03868_449.8;
172.07567_220.7;
174.05465_44.1 (2
-
Quinolinecarboxylate);
174.05587_30.4;
174.05607_139.7;

80
Omics
Layer Pathway
Top Level
Pathway Diet Enrichment
P
-value
Contributing
Features
174.05608_80.0
(Indole
-
3
-Acetate);
175.02483_207.0;
175.08653_115.4
(Indole
-
3
-Acetamide);
175.08669_79.9;
176.07052_138.7;
176.07057_209.2;
177.07388_203.9;
177.10212_22.2;
178.08620_170.6;
178.08634_238.8;
178.08643_68.6;
179.05593_22.6
(Psicose);
180.06553_64.3
(Hippurate);
180.06563_73.8;
181.06064_49.1;
182.04466_24.0;
184.02517_337.9;
184.06031_25.1 (4
-
Pyridoxic acid);
186.05609_244.9;
188.03537_68.5;
188.07051_34.2;
188.07069_230.3;
188.07159_56.4;
188.07173_111.1;
189.07491_56.5;
190.04981_309.3;
190.04993_74.4
(Kynurenate);
190.05083_26.8 (5
-
Hydroxyindoleacetate);
190.05103_283.8;
190.05109_136.2;
190.08584_26.3;
190.08615_223.7
(Indole
-
3
-propionic
acid);
190.08631_220.8;
190.08641_68.1
(Indole
-
3
-Methyl
Acetate);
191.08948_223.6;
192.06538_26.2;
192.06562_225.4;
192.10198_231.2;
194.04591_72.4;
194.08110_194.9;
194.08111_107.0 (2
-
Hydroxyphenylacetic
acid);
195.05069_19.2;
195.11288_224.7;
197.06838_115.5;
200.01039_146.9;
200.01200_147.2;

81
Omics
Layer Pathway
Top Level
Pathway Diet Enrichment
P
-value
Contributing
Features
200.05649_205.9;
202.05034_211.0;
203.05232_20.8;
203.08138_27.8;
203.08249_34.2;
203.08268_283.7;
204.06599_31.4;
204.06665_174.7
(Cinnamoylglycine
-
trans);
204.085705_34.0;
204.08588_283.7;
205.06187_138.1;
205.06977_32.0
(Methylglutarate);
205.07003_175.9;
205.09709_230.8;
205.09713_33.9
(Tryptophan);
206.08112_177.1
(Indolelactic acid);
206.08138_75.8;
206.10049_33.8;
206.10077_230.4;
207.07751_284.0;
(Kynurenine);
207.08450_177.2;
208.09675_124.8;
208.09680_74.3 (N
-
Acetylphenylalanine);
209.09208_225.4;
210.09552_225.4;
217.06182_225.3
221.09192_26.5;
222.09838_80.6;
223.10739_27.5;
227.07890_33.8;
233.05719_49.5;
234.07714_109.7;
243.05282_32.4;
246.01584_147.0;
247.10748_66.5;
247.10777_79.3 (N
-
Acetyltryptophan);
248.09152_66.6;
248.09190_79.3;
249.08770_33.4;
249.08784_283.6;
253.08251_227.2;
260.97936_20.2;
263.10325_26.8;
263.10351_283.8;
263.10356_232.5;
264.10663_27.1;
264.10682_231.7;
265.11809_66.5;
265.11857_77.3;
266.12154_66.6;
266.12163_79.4;

82
Omics
Layer Pathway
Top Level
Pathway Diet Enrichment
P
-value
Contributing
Features
267.99842_29.3
(Pyridoxal
-Phosphate);
285.99142_249.6;
287.09989_66.5;
287.10013_79.4 (N
-
acetyl
-tyrosine);
299.08025_230.3;
301.07722_230.3;
303.07375_66.8;
303.07417_79.5;
348.07018_268.5;
89.02441_19.8
(Glyceraldehyde);
89.04031_21.9;
99.53025_15.8
Metabolome
Vitamin A
(retinol)
metabolism
Metabolism
of cofactors
and vitamins
DASH 0.107 0.040
135.11693_64.7;
269.22624_63.9;
270.22980_63.7;
271.24194_69.5;
285.22204_337.7;
287.23705_69.3;
289.21583_262.0
(Testosterone);
299.20035_262.7;
299.20130_317.7;
300.20441_311.5;
301.21677_58.3;
301.21697_341.4
(Eicosapentaenoate);
301.21702_259.0;
303.23212_57.5;
305.24611_65.6;
312.17194_41.4;
315.19629_255.7;
317.21084_267.8;
318.21489_266.6;
319.22634_267.0;
327.23261_57.9
(Docosahexaenoate);
343.22734_273.2;
345.20744_310.7;
345.24340_65.5;
347.25902_63.5;
348.07018_268.5;
357.23835_69.6;
363.25243_277.8;
372.19307_43.8
Metabolome
Vitamin B6
(pyridoxine)
metabolism
Metabolism
of cofactors
and vitamins
HEI;
DASH 0.3; 0.1 0.003;
0.043
123.09171_217.4;
124.07568_84.9;
124.07570_24.2;
124.07581_219.6;
125.10728_28.7;
125.10741_209.1;
126.09127_29.5;
126.09138_94.0;
132.04433_115.0;
134.05999_115.4;
134.06013_80.8;
141.10216_31.2;

83
Omics
Layer Pathway
Top Level
Pathway Diet Enrichment
P
-value
Contributing
Features
142.08623_23.0;
142.08638_224.1;
142.08642_81.7;
150.05594_39.7
(Noradrenaline);
151.08669_221.4;
152.07058_37.5;
152.07070_237.6;
152.07076_74.5;
153.05480_68.2;
164.03533_26.4 (4
-
Pyridoxolacetone);
166.05101_120.0
(Pyridoxal);
168.06676_17.5;
168.08998_221.3;
168.09061_26.0;
169.09702_23.5
(Pyridoxamine);
170.08111_29.3
(Pyridoxine);
170.08135_79.9;
182.04602_52.5 (4
-
Pyridoxate);
183.04916_26.5;
183.04964_46.9;
183.07758_19.0;
184.06031_25.1 (4
-
Pyridoxic acid);
188.05653_215.5;
188.09142_32.7
(Adipate);
188.09188_234.6;
189.04050_117.0;
189.04052_226.1;
214.07212_125.1;
249.00738_45.7;
96.04327_24.5
Metabolome Vitamin E
metabolism
Metabolism
of cofactors
and vitamins
HEI 0.194 0.023
119.05027_35.4;
121.06471_24.73
(Phenylacetaldehyde);
135.04515_33.0;
143.56203_242.5;
165.05557_26.9 (3
-(4
-
hydroxyphenyl)propan
oic acid);
179.07144_44.8;
192.11357_246.1;
198.97542_21.3;
198.97569_23.2;
264.13546_237.8;
348.07018_268.5;
391.35701_63.0;
427.35704_369.1;
427.35736_63.0;
429.37376_49.7;
430.37901_64.4;
443.35324_50.3;
445.36748_376.3;
445.36784_63.0;

84
Omics
Layer Pathway
Top Level
Pathway Diet Enrichment P-value
Contributing
Features
445.36886_50.4;
446.37232_50.6;
461.36351_386.5;
461.36401_49.3 (25-
Hydroxycholesterol);
462.36731_49.8;
463.37768_383.1;
475.37924_57.4;
485.35958_370.0;
489.39519_49.6;
505.39035_49.7;
521.38511_49.5;
531.36379_371.0
Metabolome
Vitamin H
(biotin)
metabolism
Metabolism
of cofactors
and vitamins
HEI 0.363 0.038
245.09517_61.6
(Biotin);
348.07018_268.5;
129.10233_291.0;
166.07287_282.1;
130.08622_20.3
(Pipecolic acid);
145.09819_23.2
(Lysine);
147.11272_16.6
(Diethanolamine);
166.07213_17.2;
225.02408_22.7

85
Chapter 4: Metabolites and proteins mediate the relationship between diet and insulin
sensitivity
4.1. Abstract
Background. Poor diet quality is a known risk factor for type 2 diabetes and related outcomes,
including insulin sensitivity. Biological changes that occur in response to diet quality are likely to
explain this relationship.
Methods. High dimensional mediation analyses (HIMA) were performed to identity metabolites,
proteins, and miRNA that might mediate the relationship between the Healthy Eating Index2015 (HEI) and insulin sensitivity (Matsuda Index) after four years of follow up in participants
from the MetaAIR cohort (n = 77, 52% female, 57% Hispanic). HIMA analyses were performed
using both an “early” omics integration and a “late” integration approach. Features identified
using both HIMA approaches were further assessed using causal mediation analyses, and the
indirect effects and proportion mediated by selected features were calculated for each feature
independently and together.
Results. Early integration identified four potential mediators: three metabolites (5Z, 8Z, 11Zeicosatrienoic acid, pipecolic acid, and biotin), and one protein (F9). Late integration identified
one metabolite (5Z, 8Z, 11Z-eicosatrienoic acid) and five proteins (ACY1, CDH2, CST3, F9, and
PTGDS) as potential mediators. The two features identified by both approaches, 5Z, 8Z, 11Zeicosatrienoic acid (a polyunsaturated fatty acid) and F9 (coagulation factor IX), each explained
41.7% (95% CI: 6.1, 198) and 43.5% (95% -9.0, 149.6) of the total effect of HEI on insulin
sensitivity, respectively. Together, these two features were found to mediate 69% (95% CI: -
12.9, 249.3%) of the total effect.
Conclusions. F9 and 5Z, 8Z, 11Z-eicosatrienoic acid are both linked to inflammation and were
previously reported as possible biomarkers for type 2 diabetes. These findings indicate that

86
metabolites and proteins may mediate the association between diet and diabetes-related
outcomes like insulin sensitivity, likely through pathways related to inflammation. F9 and 5Z, 8Z,
11Z-eicosatrienoic acid may be potential targets for monitoring efforts for diet adherence or
disease prevention.

87
4.2. Introduction
Young-onset type 2 diabetes (T2D) is a health condition of increasing concern, as the
number of adolescents and young adults at risk for this condition grows [1]. Because health
outcomes for youth with T2D are more severe than those diagnosed in later adulthood [2], there
is a need for greater understanding of the complex mechanisms underlying T2D disease
progression and development. Insulin insensitivity is a metabolic disruption leading to glucose
intolerance and greatly increases the risk for T2D, and often can be detected well in advance of
a T2D diagnosis [3-5]. Understanding and targeting this condition before progression to T2D
may prevent later disease and disability.
The pathology of both young-onset and adult-onset T2D is characterized by pancreatic
beta cell dysfunction and insulin resistance in the liver and peripheral tissues [6]. In a healthy
person, the pancreas will secrete insulin after a meal, glucose uptake will occur in hepatic and
muscle cells, and glucose production in the liver will be suppressed [7]. Failure or impairment of
any of these processes will lead to metabolic disturbance and eventually T2D, and can occur in
two main ways: insulin insufficiency caused by impaired beta cell function, and hepatic and
muscle tissue insensitivity to insulin [3, 7]. As a result, fasting glucose and insulin, and post-oral
glucose tolerance test (OGTT) glucose and insulin rise, leading to prediabetes and T2D. In
youth, this hyperinsulinemia appears to be more rapid than in older adults, and the progression
to diabetes follows more quickly [8]. When insulin resistance or altered glucose homeostasis are
detected, initial treatment usually involves lifestyle changes, particularly to the diet.
Poor diet is a significant risk factor for T2D and its precursor conditions. Consuming a
diet high in fruits, vegetables, and fiber, and low in added sugars, saturated fat, and refined
grains, is recommended by the United States Department of Agriculture (USDA) Dietary
Guidelines [9], and has been found to be protective against T2D in both young people and in
middle aged and older adults [10-12]. Recently, advances in “omics” technologies have made it

88
possible to evaluate the direct impact of dietary intake on biological processes, including those
that might lead to the development of T2D and other metabolic diseases [13-15]. It is
hypothesized that some or all of the effects of dietary intake on metabolic disease may be
mediated by omics features, such as those from the metabolome, proteome, or others [16-18].
A mediation analytical approach may provide insight into the molecular mechanisms
underlying disease development and the biological effects of an exposure on disease risk. In a
traditional causal mediation framework, the total effect of an exposure on an outcome is
decomposed into the “indirect” effect of a single intermediate on the causal pathway, and the
remaining “direct” effect of the exposure [19]. High dimensional mediation analysis, with
potentially thousands of omics features to act as mediators, also allows us to determine the
extent to which the effects of a given exposure are explained by a collection of biological
features along the path between an exposure (lifestyle factors, treatments, or other
interventions) and a health outcome (insulin resistance, prediabetes, or T2D) [20]. In this way,
omics data can be used to determine the mechanism (or mechanisms) by which diet affects risk
for young-onset T2D.
The purpose of this study is to determine the extent to which the relationship between
diet and markers of insulin sensitivity is mediated by underlying biological conditions. We have
previously shown that high-quality diet is protective against prediabetes and is associated with
glucose homeostasis in a cohort of young adults [10]. Here, we assess the longitudinal
relationship between diet and markers of insulin sensitivity in the same cohort, along with
potential mediators from three omics layers: miRNA, proteomics, and metabolomics.
4.3. Methods
4.3.1. Study Population

89
Between 2014 and 2018, 155 participants (age 17-22) were recruited from the Children’s
Health Study in Southern California [21] for the MetaAIR study [22]. Participants were enrolled if
they had a history of overweight or obesity in early adolescence, did not have type 1 or type 2
diabetes, had no other serious medical conditions, and were not taking any medications known
to influence glucose metabolism. Between January 2020 and March 2022, 140 MetaAIR
participants were invited to participate in a follow up visit, and 85 participants completed the
second visit [10]. Both study visits took place at the Diabetes and Obesity Research Institute at
the University of Southern California (USC). This study was approved by the USC Institutional
Review Board and written informed consent was obtained from participants (and their
guardians, if under age 18) at both visits.
4.3.2. Diet Assessment
At baseline, participants completed two 24-hour dietary recalls on two nonconsecutive
days, one weekday and one weekend day. Recalls were completed by trained interviewers
using the Nutritional Data System for Research (NDSR) software version 14 (University of
Minnesota, Minneapolis, MN). These recalls were used to calculate the Healthy Eating Index
2015 (HEI), which measures adherence to the USDA 2015 Dietary Guidelines. This index
consists of thirteen components that, when summed, produce a final score between 0 and 100.
The components of the HEI are: total fruit (range: 0-5), whole fruit (0-5), total vegetables (0-5),
greens and beans (0-5), whole grains (0-10), dairy (0-10), total protein foods (0-5), seafood and
plant proteins (0-5), fatty acids (0-10), refined grains (0-10), sodium (0-10), added sugar (0-10),
and saturated fats (0-10) [9].
4.3.3. Outcome Assessment
A two-hour oral glucose tolerance test (OGTT) was performed at the baseline and
follow-up visits, using a glucose load of 75g of glucose per kg of body mass (max 75g). Blood

90
samples were taken before the OGTT and at 30-, 60-, 90-, and 120 minutes post-glucose
challenge. Glucose and insulin concentrations at each time point were measured in plasma. The
OGTT was not completed if participants had a fasting glucose value greater than 126 mg/dL,
measured by glucometer.
HOMA- and HOMA-IR were calculated to assess beta cell function and insulin resistance,
respectively, using the following formulas [23]:
HOMA-IR =

22.5
HOMA- =
20
−3.5
Where units of insulin are IU/mL and units of glucose are mmol/mL.
To measure whole-body insulin sensitivity, the Matsuda index (MI) was calculated as follows,
where higher values reflect increased insulin sensitivity [24]:
MI =
10,000
√( ) ( )
4.3.4. miRNA
Circulating miRNA were measured in baseline serum samples using NanoString [25].
The NanoString platform uses fluorescent barcodes bound to a capture probe to count 798
possible miRNA. Details of the analytical process have been described previously [26]. After

91
data processing and restricting the analysis to miRNA detected in more than 90% of
participants, 142 miRNA remained for analysis.
4.3.5. Proteomics
Proteins were measured at baseline in fasting plasma using the Olink Explore 384
Cardiometabolic panel [27]. The Cardiometabolic panel measures the relative abundance of 369
proteins, 23 of which had 50% of observations below the limit of detection. These 23 proteins
were excluded from the analysis, leaving 346 proteins for analysis. Proteins are reported as
normalized protein expression levels after log2 transformation [28].
4.3.6. Metabolomics
Untargeted metabolomics were measured at baseline in plasma samples collected at the
two-hour OGTT time point using liquid chromatography and high-resolution mass spectrometry
(LCMS) methods as described previously [29, 30]. Unique features were identified using massto-charge ratio (m/z), retention time, and peak intensity. Features were adjusted for batch
variation [31] and excluded if they were detected in < 75% of samples or if there was a > 30%
coefficient of variability of the quality control samples after batch correction. After processing,
the resulting 23,173 features were scaled to a standard normal distribution and log2
transformed. Of these 23,173 features, 466 confirmed compounds were identified using known
standards. Metabolites’ identities were assigned by matching mass m/z (<5ppm) and retention
time (<15sec). Where more than one molecule had retention times within the allowable error,
the annotation with the closest retention time to the known standard was chosen.
4.3.7. Covariates
Demographic characteristics were collected through questionnaires administered at
baseline. Age was calculated from visit date and birthday. Participants self-reported their race

92
and ethnicity, which were categorized as Non-Hispanic White, Hispanic, and Other, their sex,
and if they exercised (yes or no). Participants also reported their parents’ highest level of
education (less than high school, completed high school, more than high school, or don’t know).
Body mass index (BMI, kg/m2
) was calculated from clinical measurements of weight and height.
4.3.8. Statistical Analysis
Omics-wide association studies (OWAS) were performed to identify miRNA, proteins,
and metabolites nominally associated with HEI (p < 0.05) after adjusting for participant age,
ethnicity, sex, and BMI. The proteomic and metabolomic OWAS results have been reported
previously (see Chapter 3). The miRNA OWAS was performed to identify miRNA nominally
associated with HEI (p < 0.05), after adjusting for participant age, ethnicity, sex, and BMI. Two
miRNA, miR-30a-5p and miR-377-3p, were associated with HEI and retained for the mediation
analysis (Table S1). Of the 346 measured proteins, 44 were previously found to be associated
with HEI and included in the mediation analysis. Nineteen annotated metabolites previously
found to be nominally associated with HEI (p < 0.05) were considered as possible mediators,
excluding exogenous chemicals.
Descriptive statistics were calculated for exposure, outcomes, and covariates. Linear
regression was used to determine the relationship between baseline HEI and all outcome
variables at the follow up visit, adjusting for age, sex, ethnicity, parental education, and
exercise. Mediation analyses were conducted for those relationships with statistically significant
associations. miRNA, proteomic, and metabolomic features were considered as possible
mediators (Table 4.1).

93
Table 4.1. Omics features associated with HEI and included in mediation analyses.
miRNA Proteins Metabolites
miR-30a-5p ACY1 Betaine
miR-377-3p ADAMTS16 Heptanoate
ADH4 Pipecolic Acid
AGXT Quinolinic Acid
ANG Pyridoxine
CA5A Shikimate
CANT1 L-Gulonolactone
CCL14 4-Pyridoxate
CCL15 Undecylenic Acid
CCL27 5-HIAA
CDH2 Indoxyl Sulfate
CDHR5 Biotin
CES1 2-Methylbutyroylcarnitine
COL18A1 Hydroxybutyrylcarnitine
CST3 Stearidonic Acid
CTSZ 5Z, 8Z, 11Z-Eicosatrienoic Acid
DPP7 7-Ketodeoxycholic Acid
F7 Hyodeoxycholic Acid
F9 Palmitoylcarnitine
GSTA1
HMOX1
HYOU1
ICAM1
IGSF8
IGALS1
LILRA5
LILRB1
LILRB2
NRP1
PDGFRA
PTPRS
RARRES2
SEMA3F
SERPINA11
SERPINB5
SIGLEC7
SOST
ST6GAL1
TFPI
THPO
TNFSF13B

94
VASN
Mediation analyses were conducted using the hima package [32] in R version 4.3.1.
High dimensional mediation analysis (HIMA) is a three-step variable-selection approach to
mediation with large numbers of potential variables (greater than or near sample size). This
approach first excludes any variables not strongly associated with the outcome, then conducts
further variable selection using a penalty (here, the minmax concave penalty). Finally joint
significance testing is used to test for the mediation effect [32].
High dimensional mediation was conducted using two approaches: “early” integration
and “late” integration. In the early integration approach, all three omics layers (miRNA,
proteomics, and metabolomics) were concatenated before analysis, while in late integration
each omics layer was analyzed in an independent model. These two approaches offer different
methods to account for any correlation or biological relationships between each omics layer,
where early integration assumes little correlation between each layer and later integration
accounts for stronger relationships between each omics layer by assessing them separately. Pvalues were adjusted for a Bonferroni false discovery rate (q-values), and features were
considered to significantly mediate the relationship between exposure and outcome if q < 0.05.
All omics features were normalized prior to inclusion in the mediation models. Each mediation
model adjusted for age, sex, ethnicity, parental education, and exercise.
Traditional causal mediation analyses were used to validate features selected by HIMA.
Estimates and 95% CIs were calculated for the indirect (mediation) effect, direct effect, total
effect, and proportion mediated using the “mediation” (v4.5.0, single mediators) [33] and
“CMAverse” (multiple mediators) [34] R packages. To determine whether the selected features
from each mediation approach improved the fit of the linear regression model for HEI on the
outcome, the selected features were added as additional covariates and the resulting R2s
compared to the base model with no additional features.

95
4.4. Results
4.4.1. Study Population Characteristics
Of the 85 participants who completed both study visits, 8 were missing metabolomics
data, leaving 77 for analysis. At baseline, the average age of participants was 20 years old (SD
= 1.2). The second visit took place approximately four years after the first (SD = 1.1 years).
Participants had a mean HEI score of 54.8 (SD = 13.2) out of a possible 100, were majority
Hispanic, and about half were female (Table 4.2). Due to incomplete OGTT measurements, 3
participants were missing 2-hour glucose and 2-hour insulin, and 5 participants were missing MI
scores.
Table 4.2. Descriptive statistics for participant demographics,
exposure, and outcomes.
Variable n = 77
Baseline Age (years), Mean (SD) 20.0 (1.2)
Follow Up Time (years), Mean (SD) 4.0 (1.1)
Sex, n (%)
Female
Male
40 (51.9%)
37 (48.1%)
Ethnicity, n (%)
Hispanic/Latino
Non-Hispanic White
Other
44 (57.1%)
28 (36.4%)
5 (6.5%)
Parental Education, n (%)
Did not complete high school
Completed high school
More than high school
Don’t know
13 (16.9%)
10 (13.0%)
52 (67.5%)
2 (2.6%)
Exercise, n (%)
Yes
No
59 (76.6%)
18 (23.4%)
HEI, Mean (SD) 54.8 (13.2)
Fasting Glucose, Mean (SD) 95.4 (17.2)
2-Hour Glucose, Mean (SD)
Missing, %
120 (35.8)
3 (3.9%)

96
Fasting Insulin, Mean (SD) 13.0 (11.5)
2-Hour Insulin, Mean (SD)
Missing, n (%)
73.5 (70.5)
3 (3.9%)
HOMA-, Mean (SD) 146 (123)
HOMA-IR, Mean (SD) 3.28 (3.86)
MI, Mean (SD)
Missing, n (%)
4.53 (2.85)
5 (6.5%)
4.4.2. Diet-Outcome Associations
HEI was negatively, but not significantly, associated with 2-hour glucose, fasting insulin,
2-hour insulin, HOMA-IR, and HOMA- (Table 4.3). We also observed positive but nonsignificant associations between HEI and fasting glucose. However, HEI was significantly
positively associated with MI: one point increase in HEI was associated with a 0.05 (95% CI:
0.01, 0.09) unit increase in MI, indicating that higher HEI scores at baseline were linked to
higher insulin sensitivity at the follow up visit.
As this was the only relationship to achieve statistical significance, the association
between HEI and MI was the only one further investigated for mediation by omics features.
Table 4.3. Associations between baseline HEI and insulin- and glucose-related
outcomes.
Outcome n Beta SE P-Value
Fasting glucose 77 0.10 0.15 0.51
2-hour glucose 74 -0.21 0.32 0.51
Fasting insulin 77 -0.09 0.09 0.34
2-Hour Insulin 74 -0.61 0.59 0.31
HOMA-B 77 -1.09 1.00 0.28
HOMA-IR 77 -0.004 0.03 0.89
Matsuda Index 72 0.05 0.02 0.04
4.4.3. High-Dimensional Mediation and Feature Selection
For the early integration mediation approach, features from all three omics layers (2
miRNA, 44 proteins, and 19 metabolites) were considered in the same model. This resulted in

97
four selected features, one protein and three metabolites: F9, pipecolic acid, biotin, and 5Z, 8Z,
11Z-eicosatrienoic acid. Pipecolic acid was positively associated with both HEI and MI, while the
other three features were negatively associated with both HEI and MI. F9, or protein coagulation
factor IX, contributed the most (50.9%) to the total effect of HEI on MI (Figure 4.1A).
In the late integration approach, features were assessed separately by omics layer for
mediation effects. This approach selected six features, five proteins and one metabolite: ACY1,
CDH2, CST3, F9, PTGDS, and 5Z, 8Z, 11Z-eicosatrienoic acid. ACY1, F9, PTGDS, and 5Z, 8Z,
11Z-eicosatrienoic acid were negatively associated with both HEI and MI, and were associated
with an overall positive indirect effect. CST3 and CDH2 were negatively associated with HEI
and positively associated with MI, and were associated with an overall negative indirect effect.
ACY1 (aminocyclase 1) and CST3 (cystatin-C) contributed the most to the total effect of HEI on
MI, (62.5% and -62.2%, respectively) (Figure 4.1B).
Neither the early nor late integration approaches selected any miRNA.

98
Figure 4.1. Features selected from the early integration (A) and late integration (B) approaches.
Alpha is the estimate for the association of the HEI on each feature (M ~ X + covariates), beta is
the estimate for the association of each feature on MI adjusted for HEI (Y ~ M + X + covariate),
and TE(%) is the percent of the total effect mediated by all selected features.
4.4.4. Causal Mediation
Traditional causal mediation analyses were conducted with the two features (5Z, 8Z,
11Z-eicosatrienoic acid and F9) that were selected by both the early and late integration
approaches. The indirect effects of each feature were positive, and the overall indirect effect of

99
the two mediators together was also positive (Figure 4.2, Table S4.2). F9 and 5Z, 8Z, 11Zeicosatrienoic acid together mediated 69% (95% CI: -12.9, 249.3%) of the total effect of HEI on
MI.
In the base linear regression model for the relationship between HEI and MI, adjusting
for age, ethnicity, sex, parental education, and exercise, the adjusted R2 was 0.18. After adding
F9 and 5Z, 8Z, 11Z-eicosatrienoic acid to the model, the adjusted R2
increased to 0.37 (Table
4.4). A likelihood ratio test indicated that the addition of these two variables significantly
improved the model fit (p < 0.0001). The direct effect of HEI on MI, after adjusting for covariates
and the two mediators, is 0.016 (95% CI: -0.029, 0.060), suggesting that most of the effect of
HEI on insulin sensitivity is mediated by F9 and 5Z, 8Z, 11Z-eicosatrienoic acid.
Figure 4.2. Effect estimates and 95% confidence intervals from causal mediation analysis for
HIMA-selected features and their joint mediation effects. ACME is the average causal mediation
effect, for each one unit increase in HEI.
Table 4.4. Comparisons between the base and mediator models.
Beta (HEI) 95% CI DF R
2 Chi2 LRT p-value
Model 1 0.051 0.0035, 0.098 65 0.18 - -
Model 2 0.028 -0.017, 0.073 64 0.31 11.5 0.0007
Model 3 0.029 -0.017, 0.075 64 0.29 13.2 0.0003
Model 4 0.016 -0.029, 0.060 63 0.37 20.5 < 0.0001
Model 1: MI ~ HEI + age + sex + parental education + exercise + ethnicity
Model 2: Model 1 + F9
Model 3: Model 1 + 5z,8z,11z-eicosatrienoic acid
Model 4: Model 1 + 5z,8z,11z-eicosatrienoic acid + F9
Abbreviations: HEI: Healthy Eating Index 2015; SE: Standard error; DF: Degrees of freedom;
LRT: Likelihood ratio test

100
4.5. Discussion
In this study, we assessed miRNA, proteomics features, and metabolomics features for
potential mediation effects using both early and late integration approaches in high dimensional
mediation. This analysis identified two omics features, one metabolite and one protein, that
significantly mediate the relationship between diet quality and insulin sensitivity in a cohort of
young adults. We found that 5z,8z,11z-eicosatrienoic acid, a polyunsaturated fatty acid, and F9,
a protein involved in blood clotting, mediated more than half of the total effect between HEI
scores measured at the baseline visit and the Matsuda index assessed four years later.
Accounting for these two biomarkers in a linear model with HEI and other risk factors
(demographics, exercise) substantially improved the model fit and explained much more of the
variable in the Matsuda index than HEI alone. This novel analysis is one of the first, to our
knowledge, to examine multiple omics layers for mediation between diet and insulin sensitivity.
Despite interest in the application of omics technologies to nutritional epidemiology and
nutrition assessment [14, 35, 36], few studies have considered the mediating effects of omics
features in established relationships between diet and disease. The mechanisms underlying
these relationships are likely numerous and not yet completely understood, though our results
provide evidence that metabolic and proteomic alterations resulting from dietary intake may be
responsible for the biological changes that lead to insulin insensitivity, glucose intolerance, and
eventually T2D. The two mediators identified in this analysis, F9 and 5z,8z,11z-eicosatrienoic
acid, have previously been linked to T2D and may play an important role in the development of
metabolic disease.
Both F9 and 5z,8z,11z-eicosatrienoic acid (also known as mead acid) were inversely
associated with HEI and insulin sensitivity, and appear to be elevated in plasma in response to
inflammation. F9 is a critical protein involved in hemostasis and has been identified in previous
proteomics analyses as a possible biomarker for T2D and related conditions [37-39].

101
Coagulation cascade pathways, which include F9, are associated with beta cell function [37]
and are activated in response to increased inflammation [40]. Mead acid is an endogenous n-9
polyunsaturated fatty acid, is detected in high levels in the presence of essential fatty acid (EFA)
deficiency [41], and previously has been found to be associated with increased risk for T2D [42].
Poor diet quality may contribute to insufficient intake of EFAs or disrupt the balance of omega-3
and omega-6 fatty acids [43], which may result in increased production of mead acid. This may
increase inflammation, which then leads to insulin resistance in muscle and hepatic cells [44].
This study has many strengths. The longitudinal study design means that these findings
are not the result of reverse causation, and the use of the Matsuda index calculated from five
OGTT time points provides an indicator of combined hepatic and peripheral tissue sensitivity to
insulin [24]. The Matsuda index may thus capture more information on overall insulin resistance,
regardless of an individual’s unique pathology or disease progression. Additionally, this study
was conducted in an ethnically diverse population. As insulin sensitivity may differ across racial
and ethnic groups [45], this study provides evidence for mediation by omics biomarkers in a
primarily Hispanic cohort. Agreement between our mediation results and previous work
identifying biomarkers of T2D suggests that these results may be generalizable across
ethnicities and age of T2D onset, though additional validation is needed.
This study also has several limitations. We have assumed that the diet and omics
measures reported here are indicative of long-term diet and long-term biological processes,
though this is a common assumption in both nutritional and molecular epidemiology. We also
recognize the limitations of mediation analysis, especially in its relatively new application to
high-dimensional data. In this analysis, we have assumed that there is no unmeasured
confounding between our exposure and mediators, the mediators and outcome, or between our
exposure and outcome. Because many factors may influence omics measurements, including
environmental factors, pharmaceuticals, or genetics, this may not be a valid assumption. An
additional consideration for high dimensional mediation is the assumption that there is no

102
influence by one mediator on another [46]; while we have no specific biological evidence for a
relationship between F9 and 5z,8z,11z-eicosatrienoic acid, biological systems are complex and
we cannot fully exclude the possibility. Finally, our relatively small sample size limited the
number of statistically significant relationships that could be assessed for mediation, and may
have prevented us from identifying additional mediators between diet and insulin sensitivity.
However, we applied a strict FDR correction for mediator selection to limit spurious findings, and
our selected mediators are biologically plausible.
4.6. Conclusion
Two omics features, F9 and 5z,8z,11z-eicosatrienoic acid, were found to be significant
mediators of the relationship between diet quality and insulin sensitivity, and to explain the
majority of the direct effect between this exposure and outcome. These mediators have been
previously associated with T2D in adults, and may be involved in the inflammatory responses
that are though to underly the development of insulin resistance. While the sample size in the
study is small, and these findings should be considered preliminary, these results identify
possible biomarkers for insulin sensitivity that could be assessed over the course of disease
development to monitor diet adherence or other preventive interventions for young-onset T2D.
4.7. References
1. Divers J, Mayer-Davis EJ, Lawrence JM, Isom S, Dabelea D, Dolan L, Imperatore G,
Marcovina S, Pettitt DJ, Pihoker C, et al. Trends in Incidence of Type 1 and Type 2
Diabetes Among Youths - Selected Counties and Indian Reservations, United States,
2002-2015. MMWR Morb Mortal Wkly Rep 2020;69(6):161-5. doi:
10.15585/mmwr.mm6906a3.
2. Magliano DJ, Sacre JW, Harding JL, Gregg EW, Zimmet PZ, Shaw JE. Young-onset
type 2 diabetes mellitus - implications for morbidity and mortality. Nat Rev Endocrinol
2020;16(6):321-31. doi: 10.1038/s41574-020-0334-z.

103
3. Ferrannini E, Gastaldelli A, Iozzo P. Pathophysiology of prediabetes. Med Clin North Am
2011;95(2):327-39, vii-viii. doi: 10.1016/j.mcna.2010.11.005.
4. Thomas DD, Corkey BE, Istfan NW, Apovian CM. Hyperinsulinemia: An Early Indicator
of Metabolic Dysfunction. J Endocr Soc 2019;3(9):1727-47. doi: 10.1210/js.2019-00065.
5. Song Y, Manson JE, Tinker L, Howard BV, Kuller LH, Nathan L, Rifai N, Liu S. Insulin
Sensitivity and Insulin Secretion Determined by Homeostasis Model Assessment and
Risk of Diabetes in a Multiethnic Cohort of Women: The Women's Health Initiative
Observational Study. Diabetes Care 2007;30(7):1747-52. doi: 10.2337/dc07-0358.
6. Boden G. Pathogenesis of type 2 diabetes. Insulin resistance. Endocrinol Metab Clin
North Am 2001;30(4):801-15, v. doi: 10.1016/s0889-8529(05)70216-4.
7. DeFronzo RA. The Triumvirate: β-Cell, Muscle, Liver: A Collusion Responsible for
NIDDM. Diabetes 1988;37(6):667-87. doi: 10.2337/diab.37.6.667.
8. Savic Hitt TA, Katz LEL. Pediatric Type 2 Diabetes: Not a Mini Version of Adult Type 2
Diabetes. Endocrinol Metab Clin North Am 2020;49(4):679-93. doi:
10.1016/j.ecl.2020.08.003.
9. Krebs-Smith SM, Pannucci TE, Subar AF, Kirkpatrick SI, Lerman JL, Tooze JA, Wilson
MM, Reedy J. Update of the Healthy Eating Index: HEI-2015. J Acad Nutr Diet
2018;118(9):1591-602. doi: 10.1016/j.jand.2018.05.021.
10. Costello E, Goodrich J, Patterson WB, Rock S, Li Y, Baumert B, Gilliland F, Goran MI,
Chen Z, Alderete TL, et al. Diet Quality Is Associated with Glucose Regulation in a
Cohort of Young Adults. Nutrients 2022;14(18):3734.
11. Qian F, Liu G, Hu FB, Bhupathiraju SN, Sun Q. Association Between Plant-Based
Dietary Patterns and Risk of Type 2 Diabetes: A Systematic Review and Meta-analysis.
JAMA Internal Medicine 2019;179(10):1335-44. doi: 10.1001/jamainternmed.2019.2195.
12. Neuenschwander M, Ballon A, Weber KS, Norat T, Aune D, Schwingshackl L,
Schlesinger S. Role of diet in type 2 diabetes incidence: umbrella review of metaanalyses of prospective observational studies. BMJ 2019;366:l2368. doi:
10.1136/bmj.l2368.
13. Meslier V, Laiola M, Roager HM, De Filippis F, Roume H, Quinquis B, Giacco R,
Mennella I, Ferracane R, Pons N, et al. Mediterranean diet intervention in overweight
and obese subjects lowers plasma cholesterol and causes changes in the gut
microbiome and metabolome independently of energy intake. Gut 2020;69(7):1258-68.
doi: 10.1136/gutjnl-2019-320438.
14. LeVatte M, Keshteli AH, Zarei P, Wishart DS. Applications of Metabolomics to Precision
Nutrition. Lifestyle Genom 2022;15(1):1-9. doi: 10.1159/000518489.
15. Walker ME, Song RJ, Xu X, Gerszten RE, Ngo D, Clish CB, Corlin L, Ma J, Xanthakis V,
Jacques PF, et al. Proteomic and Metabolomic Correlates of Healthy Dietary Patterns:
The Framingham Heart Study. Nutrients 2020;12(5). doi: 10.3390/nu12051476.

104
16. Kim Y, Lu S, Ho JE, Hwang SJ, Yao C, Huan T, Levy D, Ma J. Proteins as Mediators of
the Association Between Diet Quality and Incident Cardiovascular Disease and All‐
Cause Mortality: The Framingham Heart Study. Journal of the American Heart
Association 2021;10(18):e021245. doi: doi:10.1161/JAHA.121.021245.
17. Yu EY, Ren Z, Mehrkanoon S, Stehouwer CDA, van Greevenbroek MMJ, Eussen S,
Zeegers MP, Wesselius A. Plasma metabolomic profiling of dietary patterns associated
with glucose metabolism status: The Maastricht Study. BMC Med 2022;20(1):450. doi:
10.1186/s12916-022-02653-1.
18. Lu Q, Chen J, Li R, Wang Y, Tu Z, Geng T, Liu L, Pan A, Liu G. Healthy lifestyle, plasma
metabolites, and risk of cardiovascular disease among individuals with diabetes.
Atherosclerosis 2023;367:48-55. doi: 10.1016/j.atherosclerosis.2022.12.008.
19. VanderWeele TJ. Mediation Analysis: A Practitioner's Guide. Annual Review of Public
Health 2016;37(1):17-32. doi: 10.1146/annurev-publhealth-032315-021402.
20. Zhang H, Hou L, Liu L. A Review of High-Dimensional Mediation Analyses in DNA
Methylation Studies. Methods Mol Biol 2022;2432:123-35. doi: 10.1007/978-1-0716-
1994-0_10.
21. McConnell R, Shen E, Gilliland Frank D, Jerrett M, Wolch J, Chang C-C, Lurmann F,
Berhane K. A Longitudinal Cohort Study of Body Mass Index and Childhood Exposure to
Secondhand Tobacco Smoke and Air Pollution: The Southern California Children’s
Health Study. Environmental Health Perspectives 2015;123(4):360-6. doi:
10.1289/ehp.1307031.
22. Kim JS, Chen Z, Alderete TL, Toledo-Corral C, Lurmann F, Berhane K, Gilliland FD.
Associations of air pollution, obesity and cardiometabolic health in young adults: The
Meta-AIR study. Environment International 2019;133:105180. doi:
https://doi.org/10.1016/j.envint.2019.105180.
23. Matthews DR, Hosker JP, Rudenski AS, Naylor BA, Treacher DF, Turner RC.
Homeostasis model assessment: insulin resistance and beta-cell function from fasting
plasma glucose and insulin concentrations in man. Diabetologia 1985;28(7):412-9. doi:
10.1007/bf00280883.
24. Matsuda M, DeFronzo RA. Insulin sensitivity indices obtained from oral glucose
tolerance testing: comparison with the euglycemic insulin clamp. Diabetes Care
1999;22(9):1462-70. doi: 10.2337/diacare.22.9.1462.
25. Goytain A, Ng T. NanoString nCounter Technology: High-Throughput RNA Validation.
Methods Mol Biol 2020;2079:125-39. doi: 10.1007/978-1-4939-9904-0_10.
26. Patterson WB, Holzhausen E, Chalifour B, Goodrich J, Costello E, Lurmann F, Conti DV,
Chen Z, Chatzi L, Alderete TL. Exposure to ambient air pollutants, serum miRNA
networks, lipid metabolism, and non-alcoholic fatty liver disease in young adults.
Ecotoxicol Environ Saf 2023;264:115486. doi: 10.1016/j.ecoenv.2023.115486.
27. Assarsson E, Lundberg M, Holmquist G, Björkesten J, Thorsen SB, Ekman D, Eriksson
A, Rennel Dickens E, Ohlsson S, Edfeldt G, et al. Homogenous 96-plex PEA

105
immunoassay exhibiting high sensitivity, specificity, and excellent scalability. PLoS One
2014;9(4):e95192. doi: 10.1371/journal.pone.0095192.
28. Petrera A, von Toerne C, Behler J, Huth C, Thorand B, Hilgendorff A, Hauck SM.
Multiplatform Approach for Plasma Proteomics: Complementarity of Olink Proximity
Extension Assay Technology to Mass Spectrometry-Based Protein Profiling. Journal of
Proteome Research 2021;20(1):751-62. doi: 10.1021/acs.jproteome.0c00641.
29. Liu KH, Nellis M, Uppal K, Ma C, Tran V, Liang Y, Walker DI, Jones DP. Reference
Standardization for Quantification and Harmonization of Large-Scale Metabolomics.
Analytical Chemistry 2020;92(13):8836-44. doi: 10.1021/acs.analchem.0c00338.
30. Goodrich JA, Walker DI, He J, Lin X, Baumert BO, Hu X, Alderete TL, Chen Z, Valvi D,
Fuentes ZC, et al. Metabolic Signatures of Youth Exposure to Mixtures of Per- and
Polyfluoroalkyl Substances: A Multi-Cohort Study. Environ Health Perspect
2023;131(2):27005. doi: 10.1289/ehp11372.
31. Luan H, Ji F, Chen Y, Cai Z. statTarget: A streamlined tool for signal drift correction and
interpretations of quantitative mass spectrometry-based omics data. Anal Chim Acta
2018;1036:66-72. doi: 10.1016/j.aca.2018.08.002.
32. Zhang H, Zheng Y, Zhang Z, Gao T, Joyce B, Yoon G, Zhang W, Schwartz J, Just A,
Colicino E, et al. Estimating and testing high-dimensional mediation effects in epigenetic
studies. Bioinformatics 2016;32(20):3150-4. doi: 10.1093/bioinformatics/btw351.
33. Tingley D, Yamamoto T, Hirose K, Keele L, Imai K. mediation: R Package for Causal
Mediation Analysis. Journal of Statistical Software 2014;59(5):1 - 38. doi:
10.18637/jss.v059.i05.
34. Shi B, Choirat C, Coull BA, VanderWeele TJ, Valeri L. CMAverse: A Suite of Functions
for Reproducible Causal Mediation Analyses. Epidemiology 2021;32(5):e20-e2. doi:
10.1097/ede.0000000000001378.
35. Guasch-Ferré M, Bhupathiraju SN, Hu FB. Use of Metabolomics in Improving
Assessment of Dietary Intake. Clinical Chemistry 2018;64(1):82-98. doi:
10.1373/clinchem.2017.272344.
36. Picó C, Serra F, Rodríguez AM, Keijer J, Palou A. Biomarkers of Nutrition and Health:
New Tools for New Approaches. Nutrients 2019;11(5):1092. doi: 10.3390/nu11051092.
37. Curran AM, Scott-Boyer MP, Kaput J, Ryan MF, Drummond E, Gibney ER, Gibney MJ,
Roche HM, Brennan L. A proteomic signature that reflects pancreatic beta-cell function.
PLOS ONE 2018;13(8):e0202727. doi: 10.1371/journal.pone.0202727.
38. Calimlioglu B, Karagoz K, Sevimoglu T, Kilic E, Gov E, Arga KY. Tissue-Specific
Molecular Biomarker Signatures of Type 2 Diabetes: An Integrative Analysis of
Transcriptomics and Protein–Protein Interaction Data. OMICS: A Journal of Integrative
Biology 2015;19(9):563-73. doi: 10.1089/omi.2015.0088.
39. Zhao D, Shen L, Wei Y, Xie J, Chen S, Liang Y, Chen Y, Wu H. Identification of
candidate biomarkers for the prediction of gestational diabetes mellitus in the early

106
stages of pregnancy using iTRAQ quantitative proteomics. PROTEOMICS – Clinical
Applications 2017;11(7-8):1600152. doi: https://doi.org/10.1002/prca.201600152.
40. Oikonomopoulou K, Ricklin D, Ward PA, Lambris JD. Interactions between coagulation
and complement--their role in inflammation. Semin Immunopathol 2012;34(1):151-65.
doi: 10.1007/s00281-011-0280-x.
41. Kawashima H, Yoshizawa K. The physiological and pathological properties of Mead
acid, an endogenous multifunctional n-9 polyunsaturated fatty acid. Lipids in Health and
Disease 2023;22(1):172. doi: 10.1186/s12944-023-01937-6.
42. Lu Y, Wang Y, Zou L, Liang X, Ong CN, Tavintharan S, Yuan J-M, Koh W-P, Pan A.
Serum Lipids in Association With Type 2 Diabetes Risk and Prevalence in a Chinese
Population. The Journal of Clinical Endocrinology & Metabolism 2017;103(2):671-80.
doi: 10.1210/jc.2017-02176.
43. Simopoulos AP. The importance of the ratio of omega-6/omega-3 essential fatty acids.
Biomed Pharmacother 2002;56(8):365-79. doi: 10.1016/s0753-3322(02)00253-6.
44. Sears B, Perry M. The role of fatty acids in insulin resistance. Lipids in Health and
Disease 2015;14(1):121. doi: 10.1186/s12944-015-0123-1.
45. Raygor V, Abbasi F, Lazzeroni LC, Kim S, Ingelsson E, Reaven GM, Knowles JW.
Impact of race/ethnicity on insulin resistance and hypertriglyceridaemia. Diabetes and
Vascular Disease Research 2019;16(2):153-9. doi: 10.1177/1479164118813890.
46. Blum MGB, Valeri L, François O, Cadiou S, Siroux V, Lepeule J, Slama R. Challenges
Raised by Mediation Analysis in a High-Dimension Setting. Environmental Health
Perspectives 2020;128(5):055001. doi: doi:10.1289/EHP6240.

107
4.8. Supplemental Material
Table S4.1. Results from an omics-wide association study between miRNA and the Healthy
Eating Index 2015.
miRNA Beta SE P-value
hsa-let-7a-5p -0.007 0.004 0.106
hsa-let-7b-5p 0.007 0.005 0.157
hsa-let-7c-5p -0.006 0.006 0.282
hsa-let-7d-5p -0.003 0.006 0.543
hsa-let-7e-5p -0.018 0.009 0.053
hsa-let-7f-5p -0.004 0.008 0.621
hsa-let-7g-5p 0.000 0.003 0.976
hsa-let-7i-5p -0.004 0.019 0.831
hsa-miR-106a-5phsa-miR-17-5p 0.003 0.004 0.472
hsa-miR-106b-5p 0.005 0.006 0.418
hsa-miR-107 -0.010 0.010 0.318
hsa-miR-1183 -0.001 0.011 0.941
hsa-miR-122-5p -0.011 0.012 0.382
hsa-miR-1246 -0.018 0.013 0.157
hsa-miR-1255a 0.006 0.014 0.649
hsa-miR-125a-5p -0.012 0.008 0.125
hsa-miR-125b-5p 0.004 0.009 0.689
hsa-miR-126-3p -0.006 0.004 0.101
hsa-miR-1260b 0.012 0.008 0.131
hsa-miR-127-3p 0.007 0.009 0.462
hsa-miR-128-3p 0.009 0.008 0.294
hsa-miR-1285-3p -0.003 0.012 0.827
hsa-miR-1299 0.002 0.009 0.841
hsa-miR-130a-3p -0.004 0.005 0.448
hsa-miR-132-3p 0.008 0.010 0.437
hsa-miR-136-5p -0.003 0.010 0.789
hsa-miR-140-5p 0.009 0.007 0.246
hsa-miR-142-3p -0.006 0.006 0.269
hsa-miR-144-3p 0.013 0.009 0.150
hsa-miR-145-5p 0.004 0.009 0.681
hsa-miR-146a-5p -0.004 0.007 0.547
hsa-miR-148a-3p -0.010 0.005 0.057
hsa-miR-148b-3p -0.012 0.007 0.096
hsa-miR-150-5p 0.007 0.006 0.280
hsa-miR-151a-3p 0.006 0.007 0.378
hsa-miR-151a-5p 0.005 0.008 0.493
hsa-miR-1537-3p 0.001 0.008 0.944

108
miRNA Beta SE
P
-value
hsa
-miR
-15a
-5p
-0.005 0.005 0.331
hsa
-miR
-15b
-5p
-0.002 0.004 0.567
hsa
-miR
-16
-5p 0.007 0.006 0.260
hsa
-miR
-181a
-5p
-0.007 0.006 0.242
hsa
-miR
-185
-5p
-0.003 0.006 0.659
hsa
-miR
-186
-5p 0.007 0.008 0.426
hsa
-miR
-18a
-5p
-0.002 0.007 0.751
hsa
-miR
-190a
-5p 0.008 0.010 0.402
hsa
-miR
-191
-5p
-0.001 0.004 0.898
hsa
-miR
-1910
-5p 0.009 0.008 0.271
hsa
-miR
-193b
-3p 0.016 0.011 0.147
hsa
-miR
-196a
-5p 0.007 0.011 0.533
hsa
-miR
-197
-3p
-0.014 0.008 0.076
hsa
-miR
-199a
-3p
-
hsa
-miR
-199b
-3p
-0.009 0.005 0.104
hsa
-miR
-199a
-5p
-0.012 0.006 0.060
hsa
-miR
-199b
-5p 0.002 0.011 0.843
hsa
-miR
-19a
-3p 0.003 0.007 0.698
hsa
-miR
-19b
-3p 0.008 0.005 0.108
hsa
-miR
-20a
-5p
-
hsa
-miR
-20b
-5p 0.006 0.005 0.195
hsa
-miR
-21
-5p
-0.002 0.006 0.741
hsa
-miR
-22
-3p 0.002 0.004 0.577
hsa
-miR
-221
-3p
-0.011 0.007 0.114
hsa
-miR
-222
-3p 0.001 0.005 0.814
hsa
-miR
-223
-3p
-0.002 0.004 0.538
hsa
-miR
-23a
-3p 0.002 0.004 0.669
hsa
-miR
-23b
-3p 0.004 0.008 0.656
hsa
-miR
-24
-3p 0.003 0.007 0.669
hsa
-miR
-25
-3p 0.008 0.006 0.162
hsa
-miR
-26a
-5p
-0.010 0.008 0.189
hsa
-miR
-26b
-5p 0.006 0.006 0.318
hsa
-miR
-27a
-3p
-0.006 0.010 0.520
hsa
-miR
-27b
-3p
-0.010 0.006 0.117
hsa
-miR
-28
-5p
-0.011 0.008 0.174
hsa
-miR
-299
-3p 0.005 0.011 0.679
hsa
-miR
-299
-5p 0.006 0.011 0.571
hsa
-miR
-29a
-3p 0.005 0.009 0.605
hsa
-miR
-29b
-3p
-0.010 0.006 0.108
hsa
-miR
-29c
-3p
-0.001 0.006 0.894
hsa
-miR
-301a
-3p
-0.011 0.009 0.217
hsa
-miR
-302d
-3p 0.007 0.017 0.687

109
miRNA Beta SE
P
-value
hsa
-miR
-30a
-5p 0.019 0.008 0.020
hsa
-miR
-30b
-5p
-0.004 0.009 0.692
hsa
-miR
-30c
-5p 0.003 0.008 0.697
hsa
-miR
-30d
-5p 0.000 0.007 0.975
hsa
-miR
-30e
-3p 0.003 0.009 0.748
hsa
-miR
-30e
-5p 0.003 0.008 0.751
hsa
-miR
-32
-5p 0.002 0.007 0.811
hsa
-miR
-320e
-0.001 0.010 0.899
hsa
-miR
-323a
-3p 0.000 0.009 0.983
hsa
-miR
-323b
-3p 0.003 0.011 0.745
hsa
-miR
-335
-5p 0.010 0.009 0.270
hsa
-miR
-337
-3p
-0.006 0.010 0.550
hsa
-miR
-337
-5p 0.000 0.010 0.962
hsa
-miR
-340
-5p
-0.001 0.008 0.902
hsa
-miR
-342
-3p 0.000 0.006 0.934
hsa
-miR
-361
-3p
-0.003 0.008 0.746
hsa
-miR
-361
-5p
-0.001 0.008 0.887
hsa
-miR
-363
-3p 0.007 0.009 0.469
hsa
-miR
-367
-3p 0.008 0.011 0.484
hsa
-miR
-374a
-5p
-0.006 0.004 0.157
hsa
-miR
-374b
-5p
-0.014 0.007 0.060
hsa
-miR
-376a
-3p 0.006 0.008 0.433
hsa
-miR
-376c
-3p 0.002 0.009 0.813
hsa
-miR
-377
-3p 0.017 0.009 0.045
hsa
-miR
-378g
-0.010 0.011 0.343
hsa
-miR
-378i
-0.006 0.011 0.566
hsa
-miR
-379
-5p 0.009 0.010 0.391
hsa
-miR
-382
-5p 0.000 0.009 0.999
hsa
-miR
-409
-3p 0.014 0.008 0.076
hsa
-miR
-421 0.021 0.011 0.053
hsa
-miR
-423
-3p
-0.006 0.009 0.477
hsa
-miR
-423
-5p 0.000 0.004 0.954
hsa
-miR
-424
-5p 0.001 0.009 0.893
hsa
-miR
-425
-5p 0.010 0.007 0.140
hsa
-miR
-432
-5p
-0.005 0.010 0.618
hsa
-miR
-4421 0.009 0.011 0.379
hsa
-miR
-4454
-
hsa
-miR
-7975
-0.016 0.009 0.081
hsa
-miR
-450a
-5p 0.003 0.009 0.767
hsa
-miR
-451a 0.009 0.008 0.264
hsa
-miR
-454
-3p 0.005 0.009 0.538
hsa
-miR
-485
-3p
-0.006 0.009 0.518

110
miRNA Beta SE P-value
hsa-miR-486-3p -0.007 0.009 0.426
hsa-miR-487b-3p 0.011 0.009 0.217
hsa-miR-491-5p 0.000 0.010 0.986
hsa-miR-495-3p -0.006 0.011 0.577
hsa-miR-503-5p 0.008 0.009 0.383
hsa-miR-506-3p 0.000 0.010 0.998
hsa-miR-513b-5p 0.006 0.009 0.486
hsa-miR-514a-5p 0.013 0.011 0.232
hsa-miR-518b 0.004 0.008 0.661
hsa-miR-526ahsa-miR-518c-5phsa-miR-518d-5p -0.006 0.011 0.595
hsa-miR-543 -0.009 0.010 0.359
hsa-miR-548d-5p 0.005 0.010 0.623
hsa-miR-590-5p 0.016 0.009 0.075
hsa-miR-597-5p -0.003 0.012 0.839
hsa-miR-598-3p 0.010 0.010 0.363
hsa-miR-612 0.009 0.011 0.400
hsa-miR-652-3p 0.007 0.007 0.306
hsa-miR-660-5p 0.006 0.009 0.537
hsa-miR-664a-3p 0.005 0.009 0.577
hsa-miR-7-5p 0.012 0.011 0.281
hsa-miR-92a-3p 0.010 0.007 0.143
hsa-miR-93-5p 0.006 0.004 0.097
hsa-miR-98-5p -0.012 0.008 0.151
hsa-miR-99b-5p 0.003 0.009 0.722
Notes: All analyses adjusted for age, sex, ethnicity, and
BMI
Table S4.2. Results from causal mediation analyses for HIMA-selected features and their joint
mediation effects.
Effect (95% CI)
F9
5z,8z,11zEicosatrienoic Acid Joint
ACME 0.023 (0.002, 0.048) 0.022 (0.004, 0.049) 0.035 (0.003, 0.065)
ADE 0.028 (-0.016, 0.074) 0.029 (-0.016, 0.071) -
Prop. Mediated 0.44 (-0.098, 1.71) 0.42 (0.052, 2.201) 0.691 (-0.129, 2.493)
Abbreviations: ACME, average causal mediation effect; ADE, average direct effect

111
Chapter 5: Summary
The prevalence of T2D in young people has increased dramatically in the past decade.
Prediabetes, the intermediate hyperglycemic state prior to T2D, is now present in 20-25% of
adolescents and young adults [1]. Should these young people progress to T2D, they will face
increased risks for nephropathy, neuropathy, retinopathy, cardiovascular disease, and
pregnancy complications, as well as a higher mortality rate than older adults with T2D [2, 3]. In
those with young-onset T2D, serious complications are expected to appear well before age 50,
severely decreasing life expectancy and quality of life [4]. To improve health outcomes for those
with T2D and prevent it in those at risk, a better understanding of the risk factors and
mechanisms responsible for T2D development is needed.
Diet and lifestyle interventions are first-line treatments in youth at risk for T2D. Healthy
dietary patterns, with high intake of fruits, vegetables, and fiber, and low intake of sodium,
added sugar, and saturated fat, reduce the risk for T2D in virtually all populations [5, 6]. Dietary
interventions for T2D that emphasize these components of healthy diets can improve insulin
sensitivity and glycemic control and prevent T2D and prediabetes [7, 8], though the
mechanisms behind this are not well understood. Diet may be particularly important for youth,
since failure rates for pharmaceutical interventions are high and disease progression is faster
and more severe than in older adults [3].
The mechanisms by which diet impacts T2D risk are still not well understood. Highthroughput technologies, like multi-omics analyses, have the potential to unravel these
uncertainties and improve T2D treatment and prevention [9]. Omics measurements can be
combined with typical methods of diet assessment, like food frequency questionnaires and 24-
hour recalls, to examine the impact of dietary intake on biological processes and determine how
these biological alterations might contribute to insulin resistance, beta cell function, or other
states of glucose dysregulation. Mediation methods, usually used in epidemiology to
characterize pathways by which an exposure, behavior, or circumstance results in disease, can

112
also be applied to omics analyses to elucidate the molecular mechanisms along the causal
pathways between diet and T2D. This dissertation represents an effort to identify some of the
biological mechanisms and pathways related to diet quality that may contribute to T2D in young
adults.
In Chapter 2, I demonstrated that high-quality diet was significantly associated with
reduced risk for prediabetes. Among participants in the MetaAIR-MetaCHEM cohort, diet quality
decreased, on average, from the baseline visit in 2014-2018 to the follow up visit in 2020-2022.
However, increases in diet quality as measured by the HEI and DASH diet indices were highly
protective against having prediabetes at follow up: a one-point change in DASH score between
visits reduced prediabetes risk by 64%, while a one-point change in HEI between visits reduced
prediabetes risk by 9%. In a cross-sectional analysis at the follow up visit, HEI and DASH were
also significantly associated with lower 2-hour glucose and glucose AUC, and the DASH diet
was also inversely associated with multiple measures of body composition, including BMI, body
fat percent, and visceral adipose tissue. These findings confirmed that diet was a strong risk
factor for prediabetes in this population and age group, with effect sizes that exceeded many
previously findings in older adults.
As the HEI and DASH diets had the strongest associations with prediabetes in Chapter
2, I choose these two dietary patterns for further analysis in Chapter 3. Here, I performed
proteome-wide and metabolome-wide association studies and functional pathway analyses to
identify omics features characteristic of a healthy diet and describe their biological functions.
This analysis was cross-sectional, as proteomics and metabolomics data were only available for
the baseline visit. There were 44 proteins associated with HEI and 25 with DASH, and 23 and
21 annotated metabolites associated with HEI and DASH, respectively. Proteins associated with
diet had many biological functions associated with metabolic disease: cell growth and adhesion;
amino acid, carbohydrate, and lipid metabolism; immune response; and stress response.
Metabolites of B-vitamins, amino acids, bile, acids, fatty acids, and acyl carnitines were

113
associated with both diets. Pesticides, including fenuron (an herbicide), pyracarbolid (a
fungicide), and clothianidin (an insecticide) were also significantly associated with diet. Five
proteins (ACY1, ADH4, AGXT, F7, GSTA1) and six metabolites (undecylenic acid, betaine,
iprovalicarb, pyracarbolid, stearidonic acid, hyodeoxychoic acid) were associated with both HEI
and DASH. Previous research had shown that many of these 11 overlapping features were
linked to metabolic disease, inflammation, or gut microbiome activity.
In Chapter 4, I performed a longitudinal analysis to investigate the mediation effects of
omics signatures on insulin sensitivity. This analysis focused on the relationship between
baseline HEI and Matsuda Index at the follow up visit, as this was the only relationship that
retained statistical significance with the reduced sample size. All proteomic and endogenous
metabolomic features found to be associated with HEI in Chapter 3 were included as possible
mediators, as well as miRNA associated with HEI. High-dimensional mediation analysis
selected one protein, F9, and one metabolite, mead acid, as significant mediators. Traditional
causal mediation analyses then revealed that these two features mediated approximately 69%
of the total effect between HEI and Matsuda Index. Moreover, adding these two molecules to a
linear regression model for the effect of HEI on Matsuda Index increased the R2
from 0.18 to
0.37, demonstrating a marked improvement in model fit. The fully adjusted model, with both F9
and mead acid, explained 37% of the total variation in the relationship between HEI and insulin
sensitivity. F9 and mead acid were linked to T2D in previous literature and are involved in
inflammation and beta cell function: thus, these findings support a possible mechanism for
insulin resistance that involves an inflammatory response and resulting damage to beta cells.
5.1. Implications and Next Steps
These findings identified proteins and metabolites associated with healthy dietary
patterns and provided evidence to support their involvement in mechanisms underlying the
effects of diet quality on risk for T2D. In particular, mechanisms involving an inflammatory

114
response as a result of a pro-inflammatory diet, diet-related obesity, or other external stimuli are
likely to contribute to loss of insulin sensitivity and later development of T2D. Inflammation is
associated with other risk factors for T2D and comorbidities commonly diagnosed in patients
with T2D, including obesity, NAFLD, and cardiovascular disease [10]. Nutritional interventions to
reduce inflammation, either through specific foods or supplements, or adherence to an antiinflammatory dietary pattern like those described by the DII [11] or GrAID [12], may be
especially valuable as treatment or prevention approaches.
Pharmaceuticals that target inflammation, either primarily or incidentally, may help
preserve insulin sensitivity or reduce insulin resistance. The two mediators identified in Chapter
4, mead acid and F9, may be potential biomarkers that could be used to monitor insulin
sensitivity or response to treatment. It is possible they may be potential drug targets
themselves, though additional validation studies are needed to confirm these findings and
explore future possibilities. As this analysis was performed in a relatively small cohort and we do
not yet have the ability to measure every possible protein or metabolite and confidently identify
them, additional mediators may emerge from larger cohorts with different omics analytical
techniques. Additionally, while the results presented in this dissertation largely agree with
previous work in older populations with T2D, a life course approach to T2D prevention is
needed to validate these finding and identify additional biomarkers specific to young-onset T2D.
In this dissertation, I used one high dimensional mediation analysis method (HIMA) to
evaluate the relationship between diet, omics, and T2D. However, there are many analytical
possibilities and none have emerged as a gold standard, nor is there consensus on how to
account for the correlation structures within and between omics layers. Mediation approaches
using latent factors, such as those identified using methods like JIVE [13], can be used to
identify possible mechanism through alternations in molecular pathways involving multiple
omics layers [14]. Approaches using latent clusters, like LUCID [15], can distinguish between

115
those at higher or lower risk of disease. Each method may contribute to the development of
precision health approaches for risk assessment, monitoring, and drug development.
5.2. Future Directions: Precision Health Approaches
The precision medicine approaches described previously are not limited to traditional
risk factors for T2D, like diet and genetics. Evidence is emerging that environmental exposures
may also increase the risk T2D and related diseases, and integrated data analysis approaches
are capable of expansion to include multiple risk factors and their joint effects on disease.
Interaction between environmental exposures and the genome, proteome, metabolome, and
microbiome can cause disease directly, increase susceptibility for disease, and affect treatment
response, and must be considered in a true precision health approach [16].
Several environmental factors have been associated with elevated risk for T2D,
including air pollution [17, 18], pesticides [19], perfluoroalkyl substances (PFAS) [20], and other
endocrine disrupting chemicals [21]. However, these associations are not as well described in
young people, and it is not clear if environmental exposures contribute to the increasing
epidemic of young-onset T2D. Some studies have not found any relationship between chemical
exposures and young-onset T2D [22, 23]. In children, evidence is mixed for the effects of
pesticide exposure on T2D, though some chemicals have been associated with other markers
of metabolic function [24, 25]. Additional research has shown that PFAS exposure may affect
glucose metabolism in adolescents and young adults [26, 27], and that exposure throughout
childhood and adolescence is associated with impaired glucose homeostasis in early adulthood
[28]. New developments in exposure and risk assessment may offer more information on the
exact nature of these relationships, as well as the relative contributions of both traditional and
emerging risk factors to young-onset T2D.
Risk factors are often closely linked to one other as well; diet, for instance, is associated
both with risk for T2D and can be a source of chemical exposures. Fish is a common source of

116
environmental contaminants, like PCBs, DDE, or methyl mercury, and the association of fish
intake with T2D risk may appear inconsistent as a result [29, 30]. Fruit and vegetable intake
may also mitigate PCB-associated risk for T2D [31], and diet has been proposed as a mediator
between PFAS and T2D risk [32]. Consideration for both diet and chemical exposures and the
correlation between them may provide a better estimate of disease risk than assessments of
either exposure alone. As shown here, dietary intake affects multiple omics layers in ways that
affect disease risk; previous work has also demonstrated that chemical exposures may impact
circulating miRNA [33], oxidative stress biomarkers [34], and the metabolome [35]. For precision
medicine to be realized, integration of the patients’ unique biological state, environment,
behavior, and knowledge of the mechanisms linking them together is required.
5.3. Conclusion
Despite increased interest in precision nutrition and precision medicine for diabetes,
current research integrating both diet and omics assessments is limited. In this dissertation, I
have performed an analysis of the effect of diet quality on risk for glucose dysregulation and
prediabetes using “traditional” diet assessment methods and analytical techniques. Then, to
examine the biological effects of adherence to a healthy diet, I identified proteomic and
metabolomic signatures of diet quality as measured by the HEI and DASH dietary patterns.
Finally, I explored the potential for omics to mediate the relationship between diet quality and
insulin resistance and identified a metabolite and protein that together mediate a large portion of
this effect. This work contributes to the literature, both by determining the association between
diet and T2D risk in cohort of young adults, a population frequently overlooked in T2D research,
and by establishing a framework to investigate the mediating effects of omics biomarkers in dietrelated T2D risk.

117
5.4. References
1. Andes LJ, Cheng YJ, Rolka DB, Gregg EW, Imperatore G. Prevalence of Prediabetes
Among Adolescents and Young Adults in the United States, 2005-2016. JAMA Pediatr
2020;174(2):e194498. doi: 10.1001/jamapediatrics.2019.4498.
2. Wilmot E, Idris I. Early onset type 2 diabetes: risk factors, clinical impact and
management. Therapeutic Advances in Chronic Disease 2014;5(6):234-44. doi:
10.1177/2040622314548679.
3. Savic Hitt TA, Katz LEL. Pediatric Type 2 Diabetes: Not a Mini Version of Adult Type 2
Diabetes. Endocrinology and Metabolism Clinics of North America 2020;49(4):679-93.
doi: https://doi.org/10.1016/j.ecl.2020.08.003.
4. Rhodes ET, Prosser LA, Hoerger TJ, Lieu T, Ludwig DS, Laffel LM. Estimated morbidity
and mortality in adolescents and young adults diagnosed with Type 2 diabetes mellitus.
Diabetic Medicine 2012;29(4):453-63. doi: https://doi.org/10.1111/j.1464-
5491.2011.03542.x.
5. Schwingshackl L, Hoffmann G, Lampousi AM, Knüppel S, Iqbal K, Schwedhelm C,
Bechthold A, Schlesinger S, Boeing H. Food groups and risk of type 2 diabetes mellitus:
a systematic review and meta-analysis of prospective studies. Eur J Epidemiol
2017;32(5):363-75. doi: 10.1007/s10654-017-0246-y.
6. Neuenschwander M, Ballon A, Weber KS, Norat T, Aune D, Schwingshackl L,
Schlesinger S. Role of diet in type 2 diabetes incidence: umbrella review of metaanalyses of prospective observational studies. BMJ 2019;366:l2368. doi:
10.1136/bmj.l2368.
7. Esposito K, Maiorino MI, Bellastella G, Chiodini P, Panagiotakos D, Giugliano D. A
journey into a Mediterranean diet and type 2 diabetes: a systematic review with metaanalyses. BMJ Open 2015;5(8):e008222. doi: 10.1136/bmjopen-2015-008222.
8. Campbell AP. DASH Eating Plan: An Eating Pattern for Diabetes Management. Diabetes
Spectr 2017;30(2):76-81. doi: 10.2337/ds16-0084.
9. Guo Y, Huang Z, Sang D, Gao Q, Li Q. The Role of Nutrition in the Prevention and
Intervention of Type 2 Diabetes. Frontiers in Bioengineering and Biotechnology 2020;8.
doi: 10.3389/fbioe.2020.575442.
10. Rohm TV, Meier DT, Olefsky JM, Donath MY. Inflammation in obesity, diabetes, and
related disorders. Immunity 2022;55(1):31-55. doi: 10.1016/j.immuni.2021.12.013.
11. Shivappa N, Steck SE, Hurley TG, Hussey JR, Hébert JR. Designing and developing a
literature-derived, population-based dietary inflammatory index. Public Health Nutrition
2014;17(8):1689-96. doi: 10.1017/S1368980013002115.
12. Campmans-Kuijpers MJE, Dijkstra G. Food and Food Groups in Inflammatory Bowel
Disease (IBD): The Design of the Groningen Anti-Inflammatory Diet (GrAID). Nutrients
2021;13(4). doi: 10.3390/nu13041067.

118
13. Lock EF, Hoadley KA, Marron JS, Nobel AB. JOINT AND INDIVIDUAL VARIATION
EXPLAINED (JIVE) FOR INTEGRATED ANALYSIS OF MULTIPLE DATA TYPES. Ann
Appl Stat 2013;7(1):523-42. doi: 10.1214/12-aoas597.
14. O’Connell MJ, Lock EF. R.JIVE for exploration of multi-source molecular data.
Bioinformatics 2016;32(18):2877-9. doi: 10.1093/bioinformatics/btw324.
15. Peng C, Wang J, Asante I, Louie S, Jin R, Chatzi L, Casey G, Thomas DC, Conti DV. A
latent unknown clustering integrating multi-omics data (LUCID) with phenotypic traits.
Bioinformatics 2020;36(3):842-50. doi: 10.1093/bioinformatics/btz667.
16. Baccarelli A, Dolinoy DC, Walker CL. A precision environmental health approach to
prevention of human disease. Nature Communications 2023;14(1):2449. doi:
10.1038/s41467-023-37626-2.
17. Wu Y, Fu R, Lei C, Deng Y, Lou W, Wang L, Zheng Y, Deng X, Yang S, Wang M, et al.
Estimates of Type 2 Diabetes Mellitus Burden Attributable to Particulate Matter Pollution
and Its 30-Year Change Patterns: A Systematic Analysis of Data From the Global
Burden of Disease Study 2019. Front Endocrinol (Lausanne) 2021;12:689079. doi:
10.3389/fendo.2021.689079.
18. Yang BY, Fan S, Thiering E, Seissler J, Nowak D, Dong GH, Heinrich J. Ambient air
pollution and diabetes: A systematic review and meta-analysis. Environ Res
2020;180:108817. doi: 10.1016/j.envres.2019.108817.
19. Evangelou E, Ntritsos G, Chondrogiorgi M, Kavvoura FK, Hernández AF, Ntzani EE,
Tzoulaki I. Exposure to pesticides and diabetes: A systematic review and meta-analysis.
Environ Int 2016;91:60-8. doi: 10.1016/j.envint.2016.02.013.
20. Zeeshan M, Zhang YT, Yu S, Huang WZ, Zhou Y, Vinothkumar R, Chu C, Li QQ, Wu
QZ, Ye WL, et al. Exposure to isomers of per- and polyfluoroalkyl substances increases
the risk of diabetes and impairs glucose-homeostasis in Chinese adults: Isomers of C8
health project. Chemosphere 2021;278:130486. doi:
10.1016/j.chemosphere.2021.130486.
21. Lind PM, Lind L. Endocrine-disrupting chemicals and risk of diabetes: an evidencebased review. Diabetologia 2018;61(7):1495-502. doi: 10.1007/s00125-018-4621-3.
22. Suarez-Lopez JR, Lee DH, Porta M, Steffes MW, Jacobs DR, Jr. Persistent organic
pollutants in young adults and changes in glucose related metabolism over a 23-year
follow-up. Environ Res 2015;137:485-94. doi: 10.1016/j.envres.2014.11.001.
23. Jensen TK, Timmermann AG, Rossing LI, Ried-Larsen M, Grøntved A, Andersen LB,
Dalgaard C, Hansen OH, Scheike T, Nielsen F, et al. Polychlorinated biphenyl exposure
and glucose metabolism in 9-year-old Danish children. J Clin Endocrinol Metab
2014;99(12):E2643-51. doi: 10.1210/jc.2014-1683.
24. Lee HA, Park SH, Hong YS, Ha EH, Park H. The Effect of Exposure to Persistent
Organic Pollutants on Metabolic Health among KOREAN Children during a 1-Year
Follow-Up. Int J Environ Res Public Health 2016;13(3). doi: 10.3390/ijerph13030270.

119
25. Burns JS, Williams PL, Korrick SA, Hauser R, Sergeyev O, Revich B, Lam T, Lee MM.
Association between chlorinated pesticides in the serum of prepubertal Russian boys
and longitudinal biomarkers of metabolic function. Am J Epidemiol 2014;180(9):909-19.
doi: 10.1093/aje/kwu212.
26. Goodrich JA, Alderete TL, Baumert BO, Berhane K, Chen Z, Gilliland FD, Goran MI, Hu
X, Jones DP, Margetaki K, et al. Exposure to Perfluoroalkyl Substances and Glucose
Homeostasis in Youth. Environmental Health Perspectives 2021;129(9):097002. doi:
doi:10.1289/EHP9200.
27. Alderete TL, Jin R, Walker DI, Valvi D, Chen Z, Jones DP, Peng C, Gilliland FD,
Berhane K, Conti DV, et al. Perfluoroalkyl substances, metabolomic profiling, and
alterations in glucose homeostasis among overweight and obese Hispanic children: A
proof-of-concept analysis. Environ Int 2019;126:445-53. doi:
10.1016/j.envint.2019.02.047.
28. Valvi D, Højlund K, Coull BA, Nielsen F, Weihe P, Grandjean P. Life-course Exposure to
Perfluoroalkyl Substances in Relation to Markers of Glucose Homeostasis in Early
Adulthood. J Clin Endocrinol Metab 2021;106(8):2495-504. doi:
10.1210/clinem/dgab267.
29. Marushka L, Hu X, Batal M, Tikhonov C, Sadik T, Schwartz H, Ing A, Fediuk K, Chan
HM. The relationship between dietary exposure to persistent organic pollutants from fish
consumption and type 2 diabetes among First Nations in Canada. Can J Public Health
2021;112(Suppl 1):168-82. doi: 10.17269/s41997-021-00484-w.
30. Wallin A, Di Giuseppe D, Orsini N, Åkesson A, Forouhi NG, Wolk A. Fish consumption
and frying of fish in relation to type 2 diabetes incidence: a prospective cohort study of
Swedish men. Eur J Nutr 2017;56(2):843-52. doi: 10.1007/s00394-015-1132-6.
31. Hofe CR, Feng L, Zephyr D, Stromberg AJ, Hennig B, Gaetke LM. Fruit and vegetable
intake, as reflected by serum carotenoid concentrations, predicts reduced probability of
polychlorinated biphenyl-associated risk for type 2 diabetes: National Health and
Nutrition Examination Survey 2003-2004. Nutr Res 2014;34(4):285-93. doi:
10.1016/j.nutres.2014.02.001.
32. Roth K, Imran Z, Liu W, Petriello MC. Diet as an Exposure Source and Mediator of Perand Polyfluoroalkyl Substance (PFAS) Toxicity. Frontiers in Toxicology 2020;2. doi:
10.3389/ftox.2020.601149.
33. Xu Y, Jurkovic-Mlakar S, Li Y, Wahlberg K, Scott K, Pineda D, Lindh CH, Jakobsson K,
Engström K. Association between serum concentrations of perfluoroalkyl substances
(PFAS) and expression of serum microRNAs in a cohort highly exposed to PFAS from
drinking water. Environ Int 2020;136:105446. doi: 10.1016/j.envint.2019.105446.
34. Chen JC, Baumert BO, Li Y, Li Y, Pan S, Robinson S, Rubbo B, Costello E, He J,
Hampson H, et al. Associations of per- and polyfluoroalkyl substances, polychlorinated
biphenyls, organochlorine pesticides, and polybrominated diphenyl ethers with oxidative
stress markers: A systematic review and meta-analysis. Environ Res 2023;239(Pt
1):117308. doi: 10.1016/j.envres.2023.117308.

120
35. Goodrich JA, Walker DI, He J, Lin X, Baumert BO, Hu X, Alderete TL, Chen Z, Valvi D,
Fuentes ZC, et al. Metabolic Signatures of Youth Exposure to Mixtures of Per- and
Polyfluoroalkyl Substances: A Multi-Cohort Study. Environmental Health Perspectives
2023;131(2):027005. doi: doi:10.1289/EHP11372.

Abstract (if available)

Abstract Type 2 diabetes (T2D), traditionally a disease of adults in late middle age or older, is becoming more prevalent in younger age groups. Young people who develop T2D tend to have more severe disease that progresses more quickly than older age groups, and T2D is also associated with more complication at younger ages. High quality diet can reduce the risk for T2D, prediabetes, and other early signs of insulin sensitivity and glucose dysregulation in all age groups. However, the biological mechanisms responsible for the protective effects of healthy diet are not well understood. High-throughput technologies provide an opportunity to investigate these mechanisms and include omics layer like genomics, transcriptomics, miRNA, proteomics, and metabolomics. Omics methods have the potential to help track disease progression or response to interventions and can be used to elucidate mechanisms underlying the relationship between diet and disease. This dissertation aims to investigate the relationship between diet quality and risk for prediabetes and insulin resistance, two metabolic states that increase the risk for T2D, in a cohort of primarily Hispanic young adults in Southern California. I use both traditional diet assessment methods and omics analyses to evaluate the protective effects of healthy dietary patterns over four years of follow up, and perform an integrated analysis with metabolomics, proteomics, and miRNA to investigate potential mechanisms responsible for the effects of diet on insulin sensitivity.
Here I present results showing that 1) improvements in diet quality are associated with reduced risk for prediabetes; 2) proteomics and metabolomics can be used to identify molecular signatures of high-quality diet; and 3) proteins and metabolites may mediate more than half the total effect of healthy diet on later insulin sensitivity. Diet quality, measured using the Healthy Eating Index-2015 and the Dietary Approaches to Stop Hypertension (DASH) diet, were highly protective against prediabetes and were linked to proteins, metabolites, and biological pathways related to nutrient metabolism, oxidative stress, and inflammation. One coagulation protein and one polyunsaturated fatty acid were identified as mediators of the relationship between diet and insulin sensitivity; both molecules have been previously linked to T2D development in older adults and are involved in inflammation, suggesting that this is a primary pathway by which insulin sensitivity is influenced by dietary intake. These findings confirm that healthy dietary patterns reduce the risk for T2D in young adults and identify molecular features and pathways that may be responsible for this effect. These pathways may be additional targets for precision medicine approaches to design targeted interventions, monitor disease progression, or predict future T2D. Future research directions include additional integration analyses for precision health, including consideration of environmental chemical exposures, particularly those also associated with dietary intake.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

The associations between ultra-processed food consumption and type 2 diabetes and obesity among young adults

PDF

Phenotypic and multi-omic characterization of novel C. elegans models of Alzheimer's disease

PDF

Ectopic fat and adipose tissue inflammation in overweight and obese African Americans and Hispanics

PDF

Diet quality and pancreatic cancer incidence in the multiethnic cohort

PDF

System biology approaches to cancer and diabetes metabolism

PDF

Investigating a physiological pathway for the effect of guided imagery on insulin resistance

PDF

Targeted metabolic signatures and diet in associations with obesity and insulin resistance in young adults

PDF

Genetic studies of inflammation and cardiovascular disease

PDF

Dietary carcinogens and genetic variation in their metabolism: epidemiological studies on the risk of selected cancers

PDF

Examining the relationship between common genetic variation, type 2 diabetes and prostate cancer risk in the multiethnic cohort

PDF

Association of single nucleotide polymorphisms in GCK, GCKR and PNPLA3 with type 2 diabetes related quantitative traits in Mexican-American population

PDF

Genes and environment in prostate cancer risk and prognosis

PDF

Quantity versus quality: how adipose tissue accumulation and immune cell profile associate with risk for type 2 diabetes in minority children and adults

PDF

The development of type 2 diabetes mellitus (type 2 DM) in California twins

PDF

Vision epidemiology and the impact of vision loss on vision-specific quality of life

PDF

Pharmacogenetic association studies and the impact of population substructure in the women's interagency HIV study

PDF

Identifying genetic, environmental, and lifestyle determinants of ethnic variation in risk of pancreatic cancer

PDF

Investigating the modulatory mechanisms of ILC2s as a therapeutic strategy in the context of health and disease

PDF

Multi-omic data mining to elucidate molecular adaptation mechanisms of filamentous fungi exposed to space environment

PDF

Association of comorbidity with prostate cancer tumor characteristics in African American men

Asset Metadata

Creator Costello, Elizabeth (author)

Core Title Molecular mechanisms of young-onset type 2 diabetes: integration of diet and multi-omics biomarkers

School Keck School of Medicine

Degree Doctor of Philosophy

Degree Program Epidemiology

Degree Conferral Date 2023-12

Publication Date 12/15/2024

Defense Date 12/07/2023

Publisher Los Angeles, California (original), University of Southern California (original), University of Southern California. Libraries (digital)

Tag diet,diet quality,mediation,metabolomics,multi-omics,OAI-PMH Harvest,prediabetes,proteomics,Young adults,young-onset type 2 diabetes

Format theses (aat)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Chatzi, Lida (committee chair), Alderete, Tanya (committee member), Chen, Zhanghua (committee member), Conti, David (committee member), Goran, Michael (committee member)

Creator Email e.e.costell@gmail.com,eecostel@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC113797217

Unique identifier UC113797217

Identifier etd-CostelloEl-12567.pdf (filename)

Legacy Identifier etd-CostelloEl-12567

Document Type Dissertation

Format theses (aat)

Rights Costello, Elizabeth

Internet Media Type application/pdf

Type texts

Source 20231218-usctheses-batch-1116 (batch), University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright.

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA

Repository Email cisadmin@lib.usc.edu