Mechanisms of human skin cell migration and wound healing

PDF

Download

Open document

Flip pages

Copy asset link

Request this asset

Request accessible transcript

Transcript (if available)

Content
MECHANISMS OF HUMAN SKIN CELL MIGRATION AND WOUND HEALING

by

Chieh‐Fang Cheng

A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(PATHOLOGY)

August 2008

Copyright 2008 Chieh-Fang Cheng

ii
Table of Contents
List of Tables iii
List of Figures iv
Abstract vi
Chapter 1: Introduction 1
Chapter 2: Profiling Motility Signal‐Induced Genes in Human 14
Keratinocytes
Introduction 14
Materials and Methods 19
Results 25
Discussion 54
Chapter 3: TGF α‐Stimulated Secretion of HSP90 α: Using LRP‐1/CD91 59
Receptor To Promote Human Skin Cell Migration Against
TGF β‐Rich Environment In Wound Healing
Introduction 59
Materials and Methods 63
Results 73
Discussion 107
Chapter 4: Conclusion 115
References 124

iii
List of Tables
Table 2‐1: TGF α- and insulin-regulated total and common genes 39
Table 2‐2: Gene expression profile detected by microarray 42
Table 2‐3: Primers for QPCR 47
Table 2‐4: Gene expression profile verification by QRT‐PCR 48
Table 3‐1: Purification of HK pro‐motility activity from HKC‐CM 72

iv
List of Figures
Figure 2‐1: TGF α and insulin‐stimulated keratinocyte migration on 27
type I collagen
Figure 2‐2: TGF β blocked growth factor‐stimulated HK proliferation, 31
but not migration
Figure 2‐3: A schematic representation of the overall design of approach 32
Figure 2‐4: TGF β did not affect the early signaling by TGF α or insulin 35
Figure 2‐5: The DNA microarray gene profile 40
Figure 2‐6: Functional characterization of three secreted gene products 53
in HK migration
Figure 3‐1: Secretion of TGF α‐treated HKCs promotes skin cell migration, 76
but not proliferation
Figure 3‐2: Dose‐dependent pro‐motility effect of HKC‐CM 78
Figure 3‐3: Identification of hsp90 α from HKC‐CM 81
Figure 3‐4: Hsp90 α is secreted selectively by HKCs and has no mitogenic 86
effect
Figure 3‐5: TGF α stimulates membrane translocation and secretion of 87
endogenous hsp90 α selectively in HKCs
Figure 3‐6: Detailked Kinetics of TGF α stimulated membrane translocation 89
of hsp90 α in HKs
Figure 3‐7: ATPase activity is required for hsp90 α secretion 92
Figure 3‐8: The exosome pathway mediates TGF α‐induced membrane 94
translocation and secretion of hsp90 α in HKCs

v
Figure 3‐9: The extracellular hsp90 α promotes cell migration independently 97
of ATP binding or ATPase activity
Figure 3‐10: Hsp90 α promotes HKC migration mainly through its middle 98
domain
Figure 3‐11: CD91 receptor mediates hsp90 α signaling to promote cell 103
migration
Figure 3‐12: Extracellular hsp90 α bypasses the inhibitory effect of TGF β 106
on human dermal cell migration
Figure 3‐13: A schematic summary: the TGF α > hsp90 α secretion > skin 109
cell migration >wound healing model

vi
Abstract
Skin cell migration is essential for skin wound healing. Steps for cell migration are
often disrupted in non‐healing wounds, causing patient morbidity and even fatality.
Currently‐available treatments are unsatisfactory. To identify novel wound‐healing
targets, we took two approaches. First, we studied the migratory gene profiles in
human keratinocytes (HKs). Second, we investigated secreted molecules from TGF α‐
stimulated human keratinoytes, which contained a strong motogenic, but not
mitogenic, activity. In the first study, the main challenge is to separate genes that are
often simultaneously induced by pleiotropic signals of a given growth factor,
including migration, proliferation and metabolism. Therefore, we designed the
following steps. First, we took advantage of a unique response of HKs to TGF‐beta,
which inhibits proliferation but not migration of the cells, to suppress selectively the
proliferation signal‐responding genes. Second, we independently stimulated HKs
with TGF‐alpha or insulin to identify the commonly regulated genes and eliminate
TGF‐alpha‐ or insulin‐specific genes. Under these designs, we obtained the profiles
of early genes by microarray analyses, followed by QRT‐PCR validation and
subsequently functional characterizations by RNAi. The study suggested the
importance of secretory molecules in keratinocyte migration and provided the
theoretical basis for the second study. In the second study, we did protein

vii
purification for the conditioned medium from HKs and identified the heat shock
protein‐90alpha (hsp90 α) as the factor fully responsible for the motogenic activity in
keratinocyte secretion. TGF α causes a rapid membrane translocation and
subsequent secretion of hsp90 α via the unconventional exosome pathway in the cells.
The secreted hsp90 α promotes both epidermal and dermal cell migration through
their surface receptor LRP‐1/CD91. The pro‐motility activity resides in the middle
domain plus the charged sequence of hsp90 α, but independently of the ATPase
activity. Most intriguingly, unlike canonical growth factors, the hsp90 α signaling
overrides the inhibition of TGF β, an abundant inhibitor of dermal cell migration in
skin wounds. This finding provides a long‐sought answer for how dermal cells
migrate into the wound environment to build new connective tissues and blood
vessels. Thus, secreted hsp90 α is potentially a new agent for wound healing.

1
Chapter 1: introduction
It was reported that more than 1.25 million people had burns and 6.5 million had
chronic skin ulcers in the United States every year. The chronic skin ulcers are
exacerbated by pressure, venous stasis, or diabetes mellitus. In severe cases, loss of
integrity in large portions of the skin will lead to major disability or even death. It
was estimated that the cost for treating unhealed wounds is more than $9 billion
each year in the United States (Willnow et al. 1995). Therefore, treatments that can
achieve rapid wound closure and normal appearance after recovery are in great
demand.

There are three phases in wound healing process – inflammation, tissue formation,
and tissue remodeling. When skin is wounded, platelets secrete several mediators of
wound healing to attract and activate neutrophils and macrophages. Several hours
later, wound re‐epithelialization begins, in which the stationary epidermal human
keratinocytes (HKs) undergo significant phenotypic alternation and start to migrate.
Afterward, the migration of fibroblasts in the dermis is involved in the tissue
remodeling phase (Singer and Clark 1999). These stages outline the fundamental
role of cell migration in wound healing.

2
The re‐epithelialization process is the key step for wound healing and is carried out
by human keratinocytes (HKs), the major cell type in epidermis. Growth factors
(GFs) and cytokines regulate HKs migration along with extracellular matrices
(ECMs). It is believed that ECMs initiate HKs migration through integrin signaling
pathways, whereas GFs and cytokines further enhance HKs migration by binding to
specific receptors in the presence of ECMs. GFs and cytokines reported to promote
HKs migration include epidermal growth factor (EGF), insulin‐like growth factor‐1
(IGF‐1), keratinocyte growth factor (KGF), hepatocyte growth factor (HGF),
transforming growth factor‐α (TGF‐α), interleukin‐1 (IL‐1), and interleukin‐8 (IL‐8)
(Li et al. 2004a). On the other hand, different ECMs have been shown to
differentially affect HKs migration. Fibronectin and collagen are reported to
promote HKs migration, whereas laminin is reported to inhibit HKs migration
(Woodley et al. 1988).

GFs and ECMs at the wound are responsible for the transition of HKs from a
stationary state to a migratory state (Eliceiri 2001). In healthy unwounded human
skin, HKs are in contact with plasma. When human skin is wounded, HKs at the
wound edges can encounter serum for the first time. It is likely that certain GFs in
serum but not plasma can promote HKs migration. However, it is also possible that

3
HKs at the cut edges interact with new ECM components that cause the transition of
HKs.

There is only one GF, platelet‐derived growth factor‐BB (PDGF‐BB), approved by
FDA for treating chronic cutaneous foot ulcers in diabetic patients. Identifying
potential factors for wound treatment is strongly demanded. Therefore, we decided
to screen the novel factors which stimulate HKs migration from two different
approaches. First, we studied the early gene expression profile of migrating HKs.
Second, we focused on the extracellular molecules secreted by migrating HKs.

GFs stimulate cell migration by binding and activating specific cell surface receptors.
The stimulating signal is transmitted through cytoplasmic signaling networks to the
nucleus, leading to both up and down regulation of a variety of genes. These gene
products in turn act either intracellularly or extracellularly as the ultimate
“migration executers” to enhance cell migration. While a great deal has been studied
about the cytoplasmic signaling pathways and individual genes that are critical for
cell migration, there is a lack of systematic analysis of the gene expression profiles in
migratory cells. Since most GFs are pleiotropic, e.g. a mitogen and motogen, the

4
genes detected in the cells following GF stimulation are considered mixtures of
genes for multiple cellular functions. Therefore, a major technical difficulty is how
to distinguish the migration‐related genes from other migration unrelated genes.

DNA microarray technology is a powerful new tool for investigating the expression
of huge numbers of genes in a single cell type under different conditions, or multiple
cell types under the same condition. With the ability to screen the expression of
thousands of known gene at the same time, microarry assay has been widely used in
gene discovery, disease diagnosis, drug discovery, and toxicological research.
However, microarray assay requires complete experimental plan. Otherwise, the
result may become inconclusive.

In the past, several groups have tried to study the gene expression profiles for cell
migration by using microarray assays. Berens’ group studied the genes regulated by
ECM in human glioma cell lines. They compared the gene expression profiles of cells
cultured on ECM or plastic and found that three groups of gene were regulated
under the ECM treatment. The three groups of genes were related to cell motility,
apoptosis, and cell cycle (Mariani et al. 2001). Condeelis’ group used an in vivo
invasion assay to study the metastasis of carcinoma in rat model. They injected rats

5
with a needle containing EGF and matrigel to form an in vivo gradient to direct the
migration of carcinoma cells, and later drew the migrated carcinoma cells around
the needle, which were defined as highly invasive cells. They then isolated the
RNAs and subjected them for microarray analyses. They found a gene Zip‐code
binding protein (ZBP1) which is highly suppressed in invasive cells. In their
functional assay, they concluded that ZBP1 is important in polarizing the cells for
migration (Wang et al. 2004). Martin’s group used an in vivo mouse model to study
the gene expression profile in the cells surrounding the wound at different time
points. They created incisional wounds on the mice back skin, and dissected for
RNA extraction in the wound area at designed time points. They compared the gene
expression profiles in wild type mice and mutant mice without inflammatory
responses and found a total of 1,001 genes expressed differently in four different
time points in these two groups of mice. After categorizing these genes, they
concluded that inflammation is not a required process for wound healing in mice
(Cooper et al. 2005).

Based on the information above, our first approach in studying novel factors for
HKs migration is to study the early gene expression profile in migrating HKs
through using microarray. The detailed procedure is described in chapter 2.

6

The second approach we used to identify the “migration executors” was to study the
molecules secreted by migrating HKs in response to TGFα since TGFα is the most
potent GF in stimulating HK migration (Li et al. 2006a). We chose to study
extracellular factors based on the following two reasons: first, the microarray study
showed that extracellular factors were extensively induced in response to GFs in
migrating HKs, and second, for the purpose of clinical treatment, extracellular
factors are easier to handle for drug design.

Based on this idea, we primed HKs with TGFα and collected conditioned‐medium
(HK‐CM). We showed that HK‐CM contained certain active ingredient(s) to
stimulate HK migration. We further fractionated the HK‐CM by using FPLC (fast
protein liquid chromatography) system with four different ion exchange and gel
filtration columns. With this approach, we identified hsp90α as the active ingredient
in HK‐CM stimulating HK migration.

Traditionally, hsp90 is known as an abundant intracellular ATP‐dependent
chaperone protein, which composes 1~2% of the protein amount in cell under
nonstress conditions. There are two isoforms of hsp90, α and β. they have 86% of

7
their amino acid sequences in common. Hsp90 is essential for stress tolerance and
protein folding in the cells, and is necessary for the viability of the cells. Together
with a number of cochaperones, hsp90 acts as specific chaperones for a wide range
of client proteins such as: ErbB2, Src, Abl, Raf, Akt, cyclin‐dependent serine kinases,
nuclear hormone receptors, and hypoxia‐inducible factor‐1 (HIF1). The functions of
these client proteins include signal transduction, cell cycle regulation, and
transcription regulation. The major function of hsp90 is to mature the unfolded
protein to a functional protein.

The structure of hsp90 is divided into four regions: the amino‐terminus, the charged
sequence, the middle domain, and the carboxyl‐terminus. The amino‐terminus binds
ATP and is responsible for the ATPase activity of hsp90. ATP triggers a
conformational change in hsp90, which is required for cochaperones and client
proteins binding. Geldanamycin (GA) and 17‐allyaminogeldanamycin (17‐AAG),
two widely used inhibitors for hsp90, can also bind to the amino‐terminus. They
compete with ATP for the ATP binding site and block the ATP dependent chaperone
function of hsp90 (Prodromou et al. 1997). The carboxyl‐terminus is responsible for
the dimerization of hsp90, which is also required for intracellular chaperone
function of hsp90 (Richter et al. 2001b). Functions of the charged sequence and the

8
middle domain of hsp90 were suggested involved in client protein binding (Ali et al.
2006; Shiau et al. 2006).

The most well‐known clinical role of hsp90 is in cancer therapy. It is known that the
expression level of hsp90 is 2 to 10 fold higher in tumor cells compared to the
expression level found in normal cells. The ATPase activity of hsp90 in tumor cells is
higher than in normal cells. 17‐AAG selectively kills tumor cells by binding to hsp90
within tumor cells with a higher affinity than in normal cells (Kamal et al. 2003).
Based on these results, 17‐AAG is currently in phase II clinical trial for anti‐tumor
therapy.

Though extensive studies have been done on the intracellular role of hsp90, the
extracellular role of hsp90 is less studied. However, several groups reported
functional extracellular hsp90 in different fields. Hsp90α is secreted from vascular
smooth muscle cells in response to oxidative stress (Liao et al. 2000). Extracellular
hsp90α is involved in fibrosarcoma cell invasion (Eustace et al. 2004). In addition,
cell surface hsp90 is involved in neural cell migration and melanoma cell invasion
(Sidera et al. 2004); (Stellas et al. 2007).

9
Although many GFs stimulated skin cell migration in vitro, most of them failed to
show any motogenic activity in vivo. Our primary goal would not be fulfilled if the
newly identified hsp90α could not help wound healing in vivo. Therefore, before we
began to study the mechanism of the extracellular motogenic activity of hsp90α, we
tested the motogenic activity of hsp90α by an in vivo mouse model. Our results
showed that hsp90α is able to enhance wound healing process in mice (Li et al. 2007).
With the supporting evidence, we studied the mechanism of releasing hsp90α in
HKs and searched for downstream targets responsible for HK migration.

The finding that an intracellular chaperone protein stimulates HK migration
extracellularly raised more questions. How does hsp90α leave from HKs? Is it
passively released from dead cells or actively secreted by stimulated cells? Is it
hsp90α itself or the “cargo” it carries that stimulates HK migration? What is the
downstream mechanism for the motogenic effect of hsp90α? In chapter 3, we show
that 1) TGFα stimulates the secretion of exosome, which contains hsp90α, 2) purified
hsp90α stimulates HK migration independent of its ATPase and ATP binding
activities, and 3) secreted hsp90α binds to cell surface receptor CD91 to stimulate cell
migration.

10
Traditionally, the secretory proteins are synthesized with a small signal peptide, the
signal recognition particle (SRP). Proteins with SRP will enter ER by binding to SRP
receptor on ER. After entering the ER, the secretory proteins will translocate to the
Golgi, then condense in trans‐Golgi network (TGN), and then bud from TGN
enveloped in a lipid bilayer. These small vesicles are known as secretory granules.
The granules will mature, translocate to the cell membrane, and fuse with the cell
membrane to release their proteins. This is known as the conventional exocytotic
pathway.

Recent studies have shown that certain proteins without SRP can also be secreted
from cells by a non‐conventional pathway, known as exosome secretion. Exosomes,
also referred to as the secreted intraluminal vesicles (ILVs), are nanovesicles (30‐
90nm in diameter) that reside in the internal lumen of multivesicular bodies (MVBs).
These nanovesicles in MVBs are either fused to lysosomes for degradation or
secreted to extracellular space for communication with other cells. Although heat
shock proteins in exosomes have been found in conditioned‐media from B cells,
tumor cells, reticulocytes, dendritic cells, and enterocytes, the mechanism for
exosomal secretion remains unclear. (Clayton et al. 2005);(Mambula and

11
Calderwood 2006);(Thery et al. 2002). In chapter 3, we show for the first time that
secretion of exosomes can be driven by a growth factor TGFα.

CD91 is also known as low density lipoprotein receptor related protein 1 (LRP1),
which belongs to the low density lipoprotein (LDL) receptor family. In humans,
there are ten members in the LDL receptor family. Most of them are known to
participate in lipoprotein uptake, steroid hormone uptake, regulation of Ca
2+

homeostasis, and cell surface protease activity (Herz and Bock 2002). In addition,
very low density lipoprotein receptor (VLDLR) and apolipoprotein E receptor 2
(APOER2/LRP8) are involved in Reelin regulated neuronal migration (Trommsdorff
et al. 1999); LRP5 and LRP6 are involved in Wnt/β‐catenin signaling pathway
(Kikuchi et al. 2007).

CD91/LRP1 is first synthesized as a 600kDa protein. The protein translocates to the
ER and then to the Golgi apparatus as in conventional exocytotic pathway. Finally,
it is transported to the cell surface. During the process, CD91/LRP1 is cleaved into
two non‐covalently linked subunits: the 515kDa α subunit and the 85kDa β subunit
(Herz et al. 1988). The α subunit contains four extracellular ligand binding domains
and is reported to bind more than 30 ligands, including lipoprotein, coagulating

12
factors, matrix metalloproteinases (MMPs), complement components, bacterial
exotoxins, viral proteins, and heat shock proteins. The β subunit contains the
transmembrane domain and intracellular domain. There are two NPxY tyrosine
phosphorylation sites that can transmit intracellular signaling. In addition, several
adaptors and scaffold proteins, including Disabled‐1 (Dab1), FE65, Shc, and GULP,
were reported to interact with the intracellular domain of CD91. Although the
mechanism remains unclear, CD91/LRP1 is known to 1) help endocytotic process
that internalize the bound ligands into cells, and 2) activate downstream signaling
pathways following ligands binding.

CD91/LRP1, when bound to different extracellular ligands, was reported to function
differently in various tissues. CD91/LRP1 is involved in lipid metabolism in the liver
as a chylomicron remnant receptor (Willnow et al. 1995). It is reported to control cell
proliferation in smooth muscle cells (Boucher et al. 2003); It is involved in controlling
cell motility by binding to thrombospodin to signal focal adhesion (Orr et al. 2003).
CD91/LRP1 regulates blood brain barrier permeability and calcium signaling in
neurons (Yepes et al. 2003). It regulates inflammatory response in lungs (Gardai et al.
2003). It is responsible for amyloid‐β peptide (Aβ) uptake in neurons and related to
the pathogenesis of Alzheimer’s disease (Bu et al. 2006). It also elicits antigen specific

13
cellular response by binding to heat shock proteins, including hsp90 (Basu et al.
2001). Due to the multiple functions of CD91/LRP1 in different tissues, it is not
surprising that CD91 KO mouse is embryonic lethal (Herz et al. 1992).

The study presented here provides the gene expression profile for migrating HKs,
which shows several potential factors that were induced in migrating HKs. Further
detailed studies are required to reveal the roles of these factors in HK migration. In
addition, the study provides a working mechanism for a novel extracellular factor,
hsp90α, in stimulating migration of HKs and other skin cells. The study can serve as
a basis for further studies in detailed mechanism of the motogenic activity of hsp90α
as well as in clinical studies for treatment of patients with unhealed wounds.

14
Chapter 2: Profiling Motility Signal-Induced Genes in Human
Keratinocytes

Introduction
Cell migration is the result of repeated cycles of cytoskeletal‐mediated protrusion
and polarization, formation of adhesive contacts, cell contraction, and retraction at
the trailing edge (Lauffenburger and Horwitz 1996). These sequential events are
initiated and continuously driven by two extracellular environmental cues,
extracellular matrices (ECMs) and soluble growth factors (GFs) (Schwartz and
Ginsberg 2002);(Eliceiri 2001). In response to a motility signal, a cell on an ECM
substratum begins to show an initial protrusion of the plasma membrane at the cell’s
leading edge by polymerizing actin filaments into a membrane structure called
lamellepodia. Adhesive complexes , which are composed of clusters of integrin
receptors, actin filaments and associated proteins at the plasma membrane, are then
established. As the cell migrates, the adhesive complexes at the leading edge of the
cell develop into larger, more organized complexes, called focal adhesions
(Gumbiner 1996). The focal adhesions then serve as points of traction over which
body of the cell translocates toward the leading edge. Subsequent release from the
ECM substratum at the rear of the cell allows the cell body to be displaced and

15
consequently achieve a net step in the forward direction (Mitchison and Cramer
1996). Between the two distinct migration cues, i.e. ECMs versus GFs, Li and
colleagues showed that ECMs, but not GFs, initiate cell migration. In contrast, GFs
optimize the ECM‐initiated cell migration and provide moving directionality (Li et
al. 2004; Wang et al. 2004). These studies suggest that haptotaxis (ECM‐driven)
appears to be a prerequisite for chemotaxis (growth factor‐induced). Haptotaxis can
occur without chemotaxis, as previously reported (Carter 1967). However, the
reverse is not true. Chemotaxis cannot take place in the absence of ECMs .

These regulations of cell migration are extremely important in the process of healing
a human skin wound, since epidermal, dermal, as well as non‐resident bone
marrow‐derived, cells, must all migrate over or into the wound bed. In un‐wounded
skin, the resident skin cells are nourished by a filtrate of plasma from dermal blood
vessels. When skin is wounded, the resident cells in the wound encounter an acute
environmental transition from plasma to serum for the first time. Then, as the
wound heals, wound remodeling initiates new blood vessel formation and a
subsequent transition from serum back to plasma. Interestingly, the “plasma‐to‐
serum‐to‐plasma” transition coincides with the three classical phases of skin wound
healing as previously described (Singer and Clark 1999). We reported previously

16
that human serum, but not human plasma, promotes keratinocyte migration (Henry
et al. 2003).

Re‐epithelialization, the process by which HKs migrate over the wound bed and
resurface the wound, is a central event that resolves problems associated with skin
wounds, i.e. infection, water loss, metabolic disturbance, nutritional loss and pain.
Re‐epithelialization largely depends upon the migration of HKs across a wound bed,
in which the microenvironment (growth factors and connective tissue components)
is very different from that experienced by the same HKs in unwounded skin
(Woodley et al. 1993; Odland and Ross 1968). The physiological factors that induce
HKs to migrate from the wound margins are not fully defined. However, we
compared all reported HK pro‐motility factors and demonstrated that transforming
growth factor‐alpha (TGF α) and insulin are the most potent ʺmotogensʺ in human
serum that enhance ECM‐initiated HK migration (Li et al. 2004). HK migration
occurs well before cell division that occurs only after a 36‐48 hour lag phase (Li et al.
2004c). Within the wound bed, fibronectin and collagen are two extracellular matrix
components capable of initiating HK migration. Odland and Ross (Odland and Ross
1968) demonstrated by electron microscopy that dramatic morphological changes
occur when HKs transform from a stationary mode to a migratory mode. Stationary

17
cells are polar and cuboidal shaped and exhibit desmosomal and hemidesmosomal
connections. When the same cells are in a migratory mode, they flatten out, extend
lamellipodia with a ruffled plasma membrane, dramatically decrease surface
desmosomes and hemidesmosomes, retract their tonofilaments to a perinuclear
location, increase surface gap junctions and dramatically assembly actin cables that
form belt‐like structures related to the inner aspects of the plasma membrane
(Woodley et al. 1986).

GFs stimulate cell migration by signal transduction from their cell surface receptors
to cytoplasmic signaling networks and then to gene expression in the nucleus. The
products of the newly expressed genes in turn will act either intracellularly or
extracellularly to execute the motility signal initiated at the cell surface, resulting in
cell migration. While a great deal has been learned about individual cytoplasmic
signaling pathways or genes involved in the control of cell migration, there are
limited studies that systematically analyze cell motility‐specific gene profiles. The
lack of this type of systematic analysis is largely due to technical problems of sorting
out the inseparable pleiotropic effects of a given GF, such as mitogenic, motogenic
and factor‐specific effects. In this study, we have designed and used a unique

18
approach to separate migration signal‐regulated genes from genes unrelated to
migration and yet induced by the same GF.

19

Materials and Methods
Primary human keratinocytes (HKs) were purchased from Cascade Biologics
(Portland, OR) and cultured in Epilife supplemented with human keratinocyte
growth supplements (HKGS), according to the manufacturer’s instruction. HKs at
passage 4 to 5 were used in all experiments. Native rat‐tail type I collagen was from
BD Biosciences (Bedford, MA). Recombinant human TGF α was from R&D systems
Inc. (Minneapolis, MN). Recombinant human insulin was from GIBCO (Grand
Island, NY). Human Genome U133A 2.0 GeneChip Array was obtained from
Affymetrix (Santa Clara, CA). Colloidal gold (gold chloride, G4022) was from
SIGMA (St. Louis, MI). Anti‐phospho‐Smad2/3 antibody was from Cell Signaling
(Danvers, MA). Anti‐human MMP10 antibody was from Chemicon (Temecula, CA).
Anti‐human HB‐EGF antibody was from R&D systems. Anti‐human CXCL3
antibody was from Aviva Systems Biology (San Diego, CA). Restriction enzymes, T4
DNA ligase were from New England BioLabs (Beverly, MA). Plasmid Midi Kit was
from Promega (Madison WI). XL‐10 Gold Ultra competent cells (XL‐10 Gold) were
from STRATAGENE (Kingsport, TN).

20
Treatment of HKs with stimulators and inhibitors
HKs were seeded on collagen I pre‐coated tissue culture plates (150 mm) around
45% confluence (~ 2.5 X 10
6
cells) in complete medium and incubated overnight. The
next day, the cells (~ 4.5 X 10
6
cells) were deprived of GFs and incubated in serum‐
free medium for 16 hours to arrest the cells at G0/G1 phase. The cells were then
washed with pre‐warmed serum‐free medium and incubated in fresh serum‐free
medium containing 20 ng/ml TGF β for 20 min at 37
o
C to block the expression of
growth signal‐related genes. TGF α or insulin was added (in the continued presence
of TGF β) for 30 min, 60 min and 120 min. At the end of each stimulation time point,
the cells were washed with ice‐cold PBS to halt the stimulation and subjected to
RNA isolation, as described below. Duplicate plates (100 mm) under the same
conditions were subjected to immunoblotting analyses for phosphotyrosine and
phospho‐Smad proteins.

RNA extraction and microarray
Total RNA was extracted from HKs using the RNeasy mini kit (Qiagen) following
the manufacturer’s User Manual. After confirming the RNA’s quality by
260nm/280nm ratio of spectrometer readings and by RNA electrophoresis, the RNA
was used for microarray processing on a Human Genome U133A 2.0 Array

21
containing 22,000 oligonucleotide probe sets. This array represents 18,400 transcripts
and variants, including 14,500 well‐characterized human genes. Briefly, RNA was
hybridized with a T7‐(dT) promoter and reverse‐transcribed to cDNA by
Superscript II reverse transcriptase (200U/ µl). Second‐strand cDNA was synthesized
by DNA polymerase I. For in vitro reverse transcription, T7 RNA polymerase and
biotin‐labeled ribonucleotides were also added. The synthesized biotin‐labeled
cRNA was fragmented and then hybridized to probe the array at 45
o
C for 16 hr. The
data were analyzed by using DCHIP software (Li and Hung Wong 2001). The signal
for each probe was verified by PM/MM ratio and reported as present, marginal, or
absent. One to one comparisons were performed with samples from control versus
growth‐factor‐treated cells. The data were discarded if signals were absent in both
groups. The comparison of expression level of each gene between growth factor
treatment and control was analyzed by SScore, a program from Bioconductor
(Kennedy et al. 2006), and genes with p<0.05 were selected. The expression pattern
of selected genes were summarized by hierarchical clustering analysis (Eisen et al.
1998). The entire DNA array experiment was repeated from cell culture to data
analyses and the data of consensus among the experiments was presented.

22
Quantitative RT‐PCR
Extracted total RNA was first subjected to reverse‐transcription assays using
TaqMan Reverse Transcription Reagents with oligo d(T)16 as the primer. The
detailed procedure was as described in the User Manual (Applied Biosystems,
Branchburg, NJ). Real‐time PCR was performed using SYBR Green PCR Master Mix.
The reactions were performed in 384‐well clear optical reaction plates in the 7900HT
system and exported by the SDS 2.1 program. All the instruments and reagents
related to reverse‐transcription reaction and real‐time PCR were from Applied
Biosystems. Primers for real‐time PCR were designed using the on‐line program
PrimerQuest from Integrated DNA Technology, Inc. (Coralville IA).

RNAi design and delivery by lentiviral system FG12
The RNAi Selection Program, as described previously (Yuan et al. 2004), was used to
identify possible RNAi target sequences, which were then scored as reported
(Reynolds et al. 2004). The selected RNAi sequences against MMP10 and HB‐EGF,
two well‐known pro‐motility factors, and CXCL3, a less characterized gene in HK
migration, were synthesized and cloned into the lentiviral RNAi delivery vector, FG‐
12 (Qin et al. 2003). Transfection and virus stock production were as described
previously by us (Qin et al. 2003). The gene transduction efficiency was analyzed

23
under a fluorescent microscope for a co‐expressed GFP (green fluorescent protein)
marker gene on the same vector. The selected RNAi sequences (sense) against the
three selected genes were: for human MMP10 was previously reported (Meyer et al.
2005), for human HB‐EGF, GAAGUUGGGCAUGACUAAU and for human CXCL3,
GUCCGUGGUCACUGAACUG. Down‐regulation of the gene products was verified
by immunoblotting analyses with antibodies against MMP10, HB‐EGF or CXCL3.

Colloidal gold migration assay
The colloidal gold migration assay was initially described by Albrecht‐Buehler (1977)
and modified by Woodley et al (1988). In short, cells are plated in plastic plates pre‐
coated with colloidal gold particles and various ECMs, treated with different kinds
of GFs, incubated at 37
0
C, and fixed with 1% formaldehyde after migration. Cells
that migrate during the period of time would ingest the colloidal gold particle and
clear out a “track” on the path they migrated. The size of migration tracks are
analyzed by computer‐based analyzing system. The motility of cells is defined by
migration index, which is calculated from 20 migration tracks.

24
DNA synthesis assays
To measure DNA synthesis, indicated cells were plated in 24‐well tissue culture
plates pre‐coated with collagen in triplicates, starved in serum‐free media for 24
hours and incubated with either indicated growth factors or HKC‐CM for 12 hours.
Stimulated cells were incubated [
3
H]‐thymidine for 6 hours. Induced DNA synthesis
was measured by incorporation of [
3
H]‐thymidine into chromosomal DNA by using
liquid scintillation analyzer from Packard instrument company (Downers Grove, IL).

25

Results
Strategy #1 for identification of HK motility‐specific gene expression: dual pro‐
motility growth factor stimulation
A major challenge for identifying a cell migration‐specific gene profile is to separate
migration signal‐regulated genes from migration‐unrelated, but often co‐expressed,
genes in response to the same growth factor stimulation, such as proliferation‐
related genes or genes involved in factor‐specific responses (such as glucose uptake
genes specifically induced by insulin, but not by other growth factors). We have
previously shown that migration of primary human keratinocytes (HKs) could be
equally stimulated by more than one growth factors, specifically TGF α and insulin
(Li et al. 2006). This property would allow us to select for the commonly‐regulated
genes by both growth factors and, at the same time, to eliminate factor‐specific genes.
Therefore, we selected TGF α and insulin as two parallel migration‐promoting cues
to induce independently gene expression in HKs.

We first wanted to determine what time points of the growth factor stimulation
should be used and why. The time course of TGF α and insulin‐stimulated HK
migration, as shown in Fig. 2‐1A, revealed that, for the initial 60 minutes following

26
stimulation, there was little detectable enhancement of HK migration in response to
either TGF α (diamond‐filled bars) or insulin (solid black bars) over collagen control
(open bars). A significant stimulation of HK migration became evident 120 minutes
after TGF α or insulin stimulation (bars 11 and 12 vs. 10). These observations
suggested that, following the addition of a growth factor, quiescent HKs need at
least 60 minutes of ”cellular preparation” prior to exhibiting induced migration. To
test if de novo protein synthesis is required, we treated HKs with increasing
concentrations of the protein synthesis inhibitor cycloheximide, and subjected the
cells to migration assays in response to TGF α or insulin. As shown in Fig. 2‐1B,
cycloheximide inhibited both TGF α‐stimulated (bars 3, 4, 5) and insulin‐stimulated
(bars 7, 8, 9) HK migration in a dose‐dependent manner. The inhibitory effects were
not due to cellular cytotoxicity, because more than 94% of the cells survived in fresh
medium after removal of cycloheximide. We decided to focus on the gene
expression profiles following 30 minute (to detect immediately early genes, or IEGs),
60 minute (early genes, or EGs) and 120 minute (delayed early genes, or DEGs)
stimulation of HKs by TGF α and insulin . Growth factor‐induced IEGs, EGs and
DEGs have previously been well defined by Stiles and Herschman (Rollins and
Stiles 1989; Herschman 1991).

27
Fig. 2‐1. TGF α and insulin‐stimulated keratinocyte migration on type I collagen.
HKs were serum starved overnight and subjected to colloidal gold migration assays
in the presence or absence of TGF α and insulin. Only Migration Index (MI) is
presented. (A) The migration of keratinocytes was time‐dependent in response to
TGF α or insulin. HKs showed significant migration after 2 hours of growth factor
stimulation. (B) Cycloheximide was added as indicated concentrations to the cells 30
minutes prior to growth factor stimulation. Presence of cycloheximide inhibited HK
migration stimulated by TGF α or insulin in a dose‐dependent manner, in
comparison to control (lanes 1, 2, 6) (*p <0.001).

28
B.
A.
10
20
Migration Index (%)
*
*
*
*
15 min 30 min 60 min 120 min 240 min
Col.
Col. + TGF α
Col. + Insulin
0
10
20
30
40
Migration Index (%)
*
*
*
*
*
*
1 2 3 4 5 6 7 8 9
TGF α: - + + + + - - - -
Ins: - - - - - + + + +
CHX ( µg/ml) CHX ( µg/ml)
0 5 15 45
0 5 15 45
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

29
Strategy #2 for identification of migration‐specific gene expression: using TGF β to
suppress proliferation signal‐induced genes
As mentioned previously, many growth factors stimulate both proliferation and
migration in HKs, like TGF α and insulin. To separate these two major signaling
events and focus on the migration‐specific genes, we took advantage of our previous
observation that TGF β inhibits serum‐stimulated HK proliferation, but not
migration (Sarret et al. 1992). To verify if this is also true for TGF α‐ and insulin‐
stimulated proliferation versus migration in HKs, we carried out HK migration and
proliferation assays in the presence or absence of TGF β. In the same experiments,
human dermal fibroblasts (HDFs) were included as a control. As shown in Fig. 2‐2,
TGF β inhibited the proliferation of both TGF α‐stimulated HKs (panel a) and PDGF‐
BB‐stimulated HDFs (panel b) in a dose dependent fashion. As previously shown
(Badyopahdhay et al. 2006), TGF β also inhibited PDGF‐BB‐stimulated HDF
migration (panel c). However, TGF β was unable to block TGF α‐stimulated HK
migration (panel d). Similar results were obtained for insulin‐stimulated HK
migration (data not shown). Thus, in HKs, TGF α and insulin signaling to promote
proliferation and migration can be separated by the presence of TGF β. This unique
property of HKs became extremely useful for our purpose of identifying the
migration‐specific gene profiles in the cells.

30

Final design of approach
Based on the observations of i) dual GF stimulation and ii) selective inhibition of HK
proliferation by TGF β, the entire design of our approach is schematically shown in
Figure 2‐3. We carried out the following sequential experimental approaches, aiming
to identify migration‐specific gene profiles: 1) HKs were to be pre‐treated with TGF β
to block cell proliferation and, therefore, the expression of key growth‐related genes.
2) The cells, then, were to be stimulated with TGF α or insulin for 30, 60 and 120
minutes. 3) Total RNA was to be isolated from the cells and subjected to DNA
microarray analyses. 4) Commonly regulated genes and factor‐specific genes in
response to both TGF α and insulin were to be identified. Steps 1 to 4 were to be
repeated using fresh cell cultures, prior to step 5. 5) Twelve genes were to be
randomly selected and subjected to QRT‐PCR verification.

31

Fig. 2‐2. TGF β blocked growth factor‐stimulated HK proliferation, but not
migration. Colloidal gold migration assay and DNA synthesis assay were carried
out as described by us previously (Bandyopahdhay et al. 2006). (A‐D) TGF β blocked
growth factor‐induced proliferation of HKs and dermal fibroblasts (DFs) in a dose‐
dependent manner (panels A and B). However, TGF β selectively blocked growth
factor‐induced migration of DFs, but not HKs (panel C, D). The experiments were
repeated more than 3 times.

32
Fig. 2‐3. A schematic representation of the overall design of approach.
TGF α and insulin are two potent stimulators of keratinocyte migration. Therefore,
dual stimulation with both GFs were used to eliminate the factor‐specific (FS) genes,
such as insulin‐specific genes involved in glucose uptake. TGF β selectively inhibit
TGF α‐ or insulin‐stimulated proliferation, but not migration, of human keratinocyte,
preventing growth‐specific (GS) gene expression. After two layers of “gene
filtration”, motility‐specific (MS) gene profiles are anticipated.

33

To investigate whether the early signaling pathways (i.e. prior to transcription
activation) of TGF α and insulin are affected by the co‐presence of TGF β, the lysates
of cells, which were serum‐starved and GF‐stimulated in the absence or presence of
TGF β, were subjected to immunoblotting analyses with anti‐phosphotyrosine, anti‐
phospho‐Smad or a loading control antibody. As shown in Fig. 2‐4, TGF α
stimulation for 30 min, 60 min and 120 min caused increased tyrosine
phosphorylation of several previously known protein species, including the 170‐kDa
EGFR, the 140‐kDa PLC γ, the 62‐kDa Dok and the 42/44‐kDa ERK1/2 (panel a, lanes
2, 3, 4 vs. lane 1). Weak detection of the EGFR tyrosine phosphorylation was
expected, because it peaks around 5 min after TGF α stimulation and then declines
(data not shown). In addition, we used a physiological concentration of 20ng/ml of
TGF α (in contrast to the 200‐500ng/ml frequently used for short periods of
stimulation, see Li et al. 2006 and ref. therein). Co‐presence of TGF β did not affect
the pattern or intensity of the phosphotyrosine proteins (panel a, lanes 6, 7, 8 vs. lane
5). Similarly, 5 µg/ml insulin stimulation caused increased phosphorylation of the
well‐characterized 70‐kDa β subunit of the insulin receptor (panel b, lanes 2, 3, 4 vs.
lane 1). As expected, co‐presence of TGF β did not interfere with the insulin‐
stimulated receptor activation (panel b, lanes 6, 7, 8 vs. lane 5). These data suggest

34
that TGF β’s blocking point is likely at downstream transcription control levels. For
example, it has previously been shown that TGF β blocks the expression of the
immediate early gene, c‐myc (Pietenpol et al. 1990). To make sure that TGF β
signaling remains active during the entire 120 minute period of growth factor
stimulation, we tested the activation of Smad2/3, a down‐stream signaling molecule
for TGF β. As shown, TGF β‐induced phosphorylation of Smad2/3 remained detectable
for the entire 2 hours in both TGF α‐stimulated (panel c, lanes 5‐8) and insulin‐
stimulated (panel e, lanes 5‐8) cells. In contrast, neither TGF α nor insulin stimulation
alone caused any detectable phosphorylation of Smad2/3 (panels c and d, lanes 1‐4).
Anti‐p38 blotting of the same membrane was used as an equal sample loading
control (panels d and f).

To estimate how many TGF α‐regulated genes were blocked in TGF β pre‐treated
HKs, HKs were pre‐treated with or without TGF β prior to TGF α stimulation, which
promotes both migration and proliferation. Total RNA isolated from the cells was
analyzed by DNA microarray analysis. We found that TGF β pre‐treatment blocked
124 of the total 500 TGF α‐regulated (up‐ and down‐regulated) genes at the 60
minute time point (data not shown). These data indicate that the approach of using
TGF β to block cell‐proliferation‐related gene expression is feasible.

35

Fig. 2‐4. TGF β did not affect the early signaling by TGF α or insulin.
Serum‐starved HKs were unstimulated or stimulated with TGF α (20 ng/ml) or
insulin (5 µg/ml) in the absence or presence of TGF β (panels A and B). Total lysates
of the cells were subjected to immunoblotting with anti‐phosphotyrosine antibodies.
TGF α or insulin stimulated tyrosine phosphorylation of the previously reported
proteins (lanes 1‐4). None of the phosphotyrosine proteins was affected by the
presence of TGF β (lanes 5‐8). The effectiveness of TGF β presence was indicated by
increased Smad phosphorylation in both TGF α‐ (panels C and D) and insulin‐
(panels E and F) treated cells. The experiments were repeated 3 times.

36

Commonly up‐ and down‐regulated IEG, EG and DEG profiles in migrating, but
not proliferating, HKs.
We used SScore to compare the expression levels of each gene before and after
growth factor stimulation at each of the three time points. SScore determines the
significance of the relatively expression level by probe‐to‐probe comparison
(Kennedy et al. 2006). The results are shown in Tab. 2‐1. At 30‐minute stimulation
(to detect IEGs), TGF α up‐regulated a total of 71 genes and down‐regulated total of
46 genes. At the same time point, insulin up‐regulated 75 genes and down‐regulated
28 genes. Among them, 25 up‐regulated genes and one down‐regulated genes were
detected in both TGF α‐ and insulin‐stimulated cells. Therefore, among the IEGs,
approximately 35% of the genes up‐regulated and 2% of the genes down‐regulated
by TGF α were shared by insulin stimulation; whereas 33% of the genes up‐regulated
and 4% of the genes down‐regulated by insulin stimulation were shared by TGF α
stimulation.

At 60‐minute stimulation (to detect EGs), TGF α caused up‐regulation of 174 genes
and down‐regulation of 41 genes. Insulin caused up‐regulation of 109 genes and
down‐regulation of 47 genes. Among them, 58 up‐regulated and 15 down‐regulated

37
genes were detected in both TGF α‐ and insulin‐stimulated cells. Among the EGs,
33% of the genes up‐regulated and 37% of the genes down‐regulated by TGF α were
shared by insulin stimulation. Whereas, 53% of the genes up‐regulated and 32% of
the genes down‐regulated by insulin were shared by TGF α stimulation.

At 120‐minute stimulation, TGF α up‐regulated 209 genes and down‐regulated 74
genes. Insulin up‐regulated 42 genes and down‐regulated 39 genes. Among them, 13
up‐regulated and 3 down‐regulated genes were detected in both TGF α‐ and insulin‐
stimulated cells. Among these DEGs, 6% of the up‐regulated and 4% of the down‐
regulated genes by TGF α were shared by insulin. 31% of up‐regulated and 8% of
down‐regulated genes by insulin were shared by TGF α stimulation. The genes that
were significantly regulated (p < 0.05) by both TGF α and insulin are presented in Fig.
2‐5. As shown, we classified the identified genes into 5 different categories based on
the cellular location of the genes’ products. The 5 categories are: extracellular factors,
cell surface molecules, cytosolic molecules, nuclear molecules, and less‐well
characterized proteins. We generated a dendrogram for the genes in each category
and, therefore, created a total of 5 dendrograms. These analyses revealed the kinetics
of the gene expression and comparison of their expression patterns. The up‐
regulated genes are shown in red, and the down‐regulated genes are shown in green.

38
The unchanged genes are shown in black. The degree of the changes is shown by an
increasing intensity of the color, as indicated by the scale bar.

To understand the expression profiles, we analyzed IEGs, EGs and DEGs according
to the cellular locations and functions of their encoded gene products (Tab. 2‐2).
Among the up‐regulated IEGs, ten genes were extracellular factors (40%), one gene
was cell surface protein (4%), five genes were cytosolic molecules (20%), eight genes
were nuclear molecules (32%) and one was less characterized (4%). Only one IEG, a
nuclear protein, was down‐regulated.

Among the up‐regulated EGs, 13 were secreted extracellular factors (23%), eight
were cell surface molecules (14%), 21 were cytosolic molecules (36%), 6 were nuclear
molecules (10%) and 10 were less characterized (17%). Among the down‐regulated
EGs, two were cell surface molecules (13%), four were cytosolic molecules (27%),
nine were nuclear molecules (60%). For the up‐regulated DEGs, 7 were extracellular
factors (54%), 5 were cytosolic molecules (38%), and one was nuclear molecules (8%).
For the down‐regulated DEGs, one was cell surface molecule (33%), one was
cytosolic molecule (33%) and one was nuclear molecule (33%).

39

Tab. 2-1. TGF α- and insulin-regulated total and common genes
The result of DNA microarray analysis is shown, based on significance of gene
expression (p<0.05) on the probe based analysis calculated by using SScore software.
The numbers shown from TGF α‐ or insulin‐stimulated cells for the indicated time
are the total numbers of up‐ or down‐regulated genes. Among them, the commonly
up‐ or down‐regulated genes are referred to as “shared genes”. The percentage of
shared genes over the total number of genes regulated under TGF α or insulin
treatment is shown as “%”.

Up‐regulated genes

Down‐regulated genes

Growth
Factor
30min 60min 120min 30min 60min 120min
TGFα 71 174 209 46 41 74
Insulin 75 109 42 28 47 39
Shared
Genes
  25 58 13 1 15 3
% 35(TGF α)
33 (ins.)
33(TGF α)
53 (ins.)
6(TGF α)
31(ins.)
2(TGF α)
4(ins.)
37(TGF α)
32 (ins.)
4(TGF α)
8 (ins.)

40
Fig. 2‐5. The DNA microarray gene profile.
The relative expression levels of the genes in HKs in response to TGF α or insulin
stimulation for 30, 60, and 120 minutes are shown by dendrograms. Only those
genes that were significantly regulated by both TGF α and insulin at least for one
time point were shown. These genes were classified into different categories based
on the locations of their protein products: (A) extracellular factors, (B) surface
molecules, (C) cytosolic molecules, (D) nuclear molecules, (E) less characterized
molecules. The genes that were up‐regulated at certain time points are shown in red,
and down‐regulated at certain time points are shown in green. Genes with similar
kinetic expression patterns in response to both TGF α and insulin were clustered
together. Similarity of the expression patterns are indicated by the hierarchy lines on
the left hand side. For certain genes, there were more than one probe sets that
detected the expression of the genes. The probe sets, which detected a significant
difference in response to the growth factor stimulation for at least one time point,
are presented here. The final presented data resulted from the consensus of DNA
array analysis of mRNA samples from two independent cultures for each time point.
A.
Insulin
30min
60min
120min
30min
60min
120min
TGF α
FSTL3 203592_s_at
MMP10 205680_at
MMP1 204475_at
IL11 206924_at
SERPINE1 202628_s_at
EDN1 218995_s_at
PI3 41469_at
PI3 203691_at
SLPI 203021_at
ARTN 216052_x_at
ARTN 207675_x_at
ARTN 210237_at
EREG 205767_at
IL6 205207_at
CXCL2 209774_x_at
CXCL3 207850_at
VEGF 210513_s_at
VEGF 212171_x_at
IL8 211506_s_at
IL8 202859_x_at
THBD 203888_at
THBD 203887_s_at
IL1B 205067_at
IL1B 39402_at
VEGF 210512_s_at
LIF 205266_at
HBEGF 203821_at
HBEGF 38037_at
IL1A 210118_s_at
IL1F9 220322_at
B.
ZYX 215706_x_at
ULBP2 221291_at
RHCG 219554_at
PTDSR 212723_at
EPHA2 203499_at
SPRY4 221489_s_at
YRDC 218647_s_at
PLAUR 211924_s_at
PLAUR 210845_s_at
CLDN4 201428_at
ERBB3 202454_s_at
EFNA1 202023_at
DCBLD2 213865_at
Insulin
30min
60min
120min
30min
60min
120min
TGF α

41
Fig.2‐5, continued

42
Tab. 2‐2. Gene expression profile detected by microarray**The genes that were
significantly regulated by both TGF α and insulin were shown here. These genes
were classified based on the time point that they were significantly regulated,
location of their protein product, and function of their protein product. The
significantly regulated genes and their fold of change in response to both growth
factors at 30 minutes, 60 minutes, and 120 minutes were shown in Table A to C,
A. 30min
up down
Category subcategory genes TGFα   ins. genes TGFα   ins.
Extracellular growth EREG 1.27 1.13
Factors factors VEGF 1.23 1.11
HBEGF 2.97 1.81
cytokines IL1B 1.49 1.43
IL6 1.57 1.40
IL1A 1.61 1.46
chemokines CXCL3 1.65 1.73
CXCL2 1.44 1.38
IL8 1.87 1.53
others EDN1 1.73 1.15
Cell surface ZYX 1.05 1.05
Molecules
Cytosolic enzyme DUSP1 2.46 1.24
Molecules PTGS2 1.58 1.18
DUSP6 1.76 1.31
cytoskeleton ARHGEF2 1.23 1.08
components
others PHLDA1 1.49 1.33
Nuclear transcription JUN 1.49 1.16 ELF1 ‐1.30 ‐1.42
Molecules regulators EGR1 11.08 2.46
FOS 7.65 1.63
CITED2 1.44 1.28
DDIT3 1.55 1.64
cell cycle DUSP4 1.34 1.13
regulators PPP1R15A 1.17 1.11
others TNFAIP3 1.37 1.23
Less
characterized C4orf10 1.29 1.44

43
Tab.2‐2 continued
B. 60min
up down
Category subcategory gene TGFα ins. gene TGFα ins.
Extracellular growth ARTN 2.36 2.13
Factors factors VEGF 3.09 1.58
HBEGF 9.47 2.61
cytokines IL1B 2.72 2.12
LIF 11.58 2.85
IL11 1.68 1.47
IL1A 2.57 1.77
IL1F9 2.29 1.85
others SERPINE1 1.43 1.27
FSTL3 1.28 1.37
SLPI 2.37 3.43
PI3 1.81 2.18
THBD 2.53 2.07
Cell surface CLDN4 1.51 1.31 EFNA1 ‐1.34 ‐1.22
Molecules EPHA2 2.44 1.84 DCBLD2 ‐1.42 ‐1.64
PLAUR 2.24 1.49
PTDSR 2.62 1.89
YRDC 2.29 1.85
RHCG 1.58 2.06
ULBP2 1.63 1.82

SPRY4 ///
LOC653170 5.69 2.14
Cytosolic signaling GNE 1.42 1.24 CDC42 ‐1.51 ‐1.50
Molecules molecules PLEKHG3 1.51 1.39 TRIO ‐1.16 ‐1.38
PTPRE 2.26 1.29
cytoskeleton
TUBB2A ///
TUBB2B 1.37 1.36
components PKP1 1.37 1.42
enzymes MAT2A 2.08 1.95
PTGS2 3.24 1.56
HS3ST1 2.03 1.37
FUT1 2.21 1.49
PLK3 3.01 1.90
FBXL18 2.15 1.87

44
Tab. 2‐2 continued
others SLC20A1 1.80 1.34 HSP90B1 ‐1.52 ‐2.15
ALAS1 1.43 1.34 IFI6 ‐1.23 ‐1.39
ISG20L2 1.53 1.61
SPRR2B 2.25 1.72
NALP1 1.44 1.43
GRPEL1 1.68 1.52
CAPN3 1.84 1.80
SPRR1A 1.39 2.16
IVL 2.42 3.06
PHLDA1 4.32 1.61
Nuclear transcription FOSL1 5.48 2.70 BCL6 ‐1.72 ‐1.55
Molecules regulators EGR3 32.55 2.95 KLF9 ‐2.75 ‐2.45
ID1 1.44 1.38 NFIL3 ‐1.96 ‐1.53
FOS 20.41 2.78 MAF ‐1.80 ‐1.32
GATA6 ‐1.30 ‐1.47
SOX4 ‐1.81 ‐1.55
cell cycle ZFP36L2 1.82 1.72 CCNG2 ‐1.89 ‐2.41
regulators CDKN1A 1.34 1.29 CDKN1B ‐2.42 ‐1.92
others SGK ‐1.39 ‐1.53
Less IER2 2.35 1.41
Characterized DDIT4 1.89 1.77
C7orf44 2.64 2.06
JMJD3 2.61 1.89
C1orf106 1.64 1.67
HSPC159 2.00 1.61
LRRC8E 1.39 1.31

C11orf17
/// NUAK2 2.60 1.88
C1orf56 2.54 1.97
KIAA0251 1.65 1.50

45
Tab. 2‐2 continued
C. 120min
up down
Category subcategory gene TGFα ins. gene TGFα ins.
Extracellular
growth
factors EREG 3.23 1.60
Factors cytokines IL1B 3.49 1.89
IL1F9 3.08 2.40
chemokines CXCL2 1.86 2.55
IL8 4.61 2.15
MMPs MMP1 2.30 2.19
MMP10 4.50 4.13
Cell surface ERBB3 ‐2.12 ‐1.70
molecules
Cytosolic signaling PTP4A1 1.54 1.28
molecules molecules
cytoskeleton EVPL ‐1.48 ‐1.38
components
enzymes MAT2A 1.61 1.41
CYP27B1 1.72 1.38
others SCG5 2.00 1.57
PHLDA1 7.50 2.05
Nuclear transcription ATF3 ‐1.54 ‐1.30
molecules regulators
others ARL4C 2.60 1.75

46

Verification of DNA microarray data by quantitative RT‐PCR
To verify the gene profiles of the microarray analyses, we randomly selected 12
genes that were either up‐ or down‐regulated at least at one of the three stimulation
time points by both TGF α and insulin. Expression of these genes in the original total
RNA samples was analyzed by quantitative RT‐PCR analyses. A total of 34 reactions
for both TGF α (17 reactions) and insulin (17 reactions) were carried out. The DNA
sequences of the 12 pairs of QPCR primers plus a pair of GAPDH primers are shown
in Tab. 2‐3. As shown in Tab. 2‐4, data of the QRT‐PCR analyses on all the genes
selected was qualitatively consistent with the data of DNA microarray analyses.
Quantitatively, however, among the 34 reactions, 28 (82%) exhibited greater fold
changes by QRT‐PCR than by microarray analyses. The remaining 6 reactions
showed smaller fold changes between QRT‐PCR and microarray analyses.

47
Tab. 2‐3. Primers for QPCR*
*Twelve genes were selected and the expression levels of these genes were verified
by QRT‐PCR. The primer sequences of these twelve genes and GAPDH for QPCR
were shown here.
Forward Reverse
CCNG2 TGCCTAGCCGAGTATTCTTCT
CCT
TCGTTGGCAGCTCAGGAACAC
TAT
CXCL2 AAGGAGGCCCTGCCCTTA GTGGCCTCTGCAGCTGTGT
CXCL3 GCATCCCCCATGGTTCAG TCAGTTGGTGCTCCCCTTGT
GAPDH GAAGGTGAAGGTCGGAGT GAAGATGGTGATGGGATTTC
HBEGF TGAAGTTGGGCATGACTAATT
CC
GTCCCCAGCCGATTCCTT
IL1α TAGCAACCAACGGGAAGGTT
CTGA
AAGGTGCTGACCTAGGCTTGA
TGA
IL1β GCACCTCTCAAGCAGAAAAC
ATG
CCTGGCCGCCTTTGGT
IL8 ACTGCGCCAACACAGAAATT TTCTCCACAACCCTCTGCAC
LIF TCTTCCAGAAGAAGAAGCTG
GGCT
CTCGGTTCACAGCACACTTCA
AGA
MMP1 AGTGACTGGGAAACCAGATG
CTGA
GCTCTTGGCAAATCTGGCGTGT
AA
MMP10 GACCTGGGCTTTATGGAGATA
TT
GCTTCAGTGTTGGCTGAGTGA
A
PHLDA1 AGGAAGTGGGACGAGCACAT
TTCT
TCCAAACTACTTGATCTGGTGC
GG
TUBB2 ACTCGGTGCTGGATGTTGTGA
GAA
ACGTGTTCATGATCCTGTCTGG
GT

48
Tab. 2‐4. Gene expression profile verification by QRT‐PCR Twelve genes were
randomly selected and their expression levels verified by QRT‐PCR at one or more
time points. The table here showed the expression level of each gene detected by
microarray and QRT‐PCR. The results of QRT‐PCR were repeated and averaged
from at least three independent experiments.

49

stimulus microarray real time
cyclin G2(60min) TGF α ‐1.89 ‐3.53 down‐
regulated cyclin G2(60min) insulin ‐2.41 ‐3.93
CXCL2(30min) TGF α 1.44 1.30
CXCL3(30min) TGF α 1.65 2.03
IL8(30min) TGF α 1.87 1.64
PHLDA1(30min) TGF α 1.49 2.30
CXCL2(30min) insulin 1.38 1.61
CXCL3(30min) insulin 1.73 1.60
IL8(30min) insulin 1.53 1.87
Up‐
regulated

PHLDA1(30min) insulin 1.33 2.30
HBEGF(60min) TGF α 9.47 17.39
IL1α(60min) TGF α 2.57 9.57
IL1β(60min) TGF α 2.72 6.46
LIF(60min) TGF α 11.58 12.28
PHLDA1(60min) TGF α 4.32 6.16
TUBB2(60min) TGF α 1.37 1.23
HBEGF(60min) insulin 2.61 3.86
IL1α(60min) insulin 1.77 3.30
IL1β(60min) insulin 2.12 3.73
LIF(60min) insulin 2.85 2.60
PHLDA1(60min) insulin 1.61 2.52
Up‐
regulated

TUBB2(60min) insulin 1.36 1.04
CXCL2(120min) TGF α 1.86 5.40
IL1β(120min) TGF α 3.49 11.15
IL8(120min) TGF α 4.61 7.78
MMP1(120min) TGF α 2.30 9.85
MMP10(120min) TGF α 4.50 22.46
PHLDA1(120min) TGF α 7.50 16.94
CXCL2(120min) insulin 2.55 3.43
IL1β(120min) insulin 1.89 3.92
IL8(120min) insulin 2.15 2.85
MMP1(120min) insulin 2.19 7.11
MMP10(120min) insulin 4.13 17.66
Up‐
regulated

PHLDA1(120min) insulin 2.05 4.26

50

Functional characterization of three secreted gene products
Finally, to investigate whether the “migration signal‐regulated” genes indeed play a
role in HK migration, we selected three genes that all encode secreted polypeptides:
hairpin‐binding EGF‐like growth factor (HB‐EGF), stromelysin‐2 (MMP‐10) and
chemokine (CXC family) ligand 3 (CXCL3). Both HB‐EGF and MMP‐10 have
previously been reported to promote HK migration and wound healing (Hashimoto
et al. 1994; Shirakata et al. 2005; Krampert et al. 2004). While little is known about the
role of CXCL3 in HK migration, mice that are deficient in CXCR2, a receptor for
CXC family ligands, also showed delayed wound healing (Devalaralia et al. 2000).
We designed small hairpin RNA (shRNA) against each of the three genes and cloned
them into the lentiviral siRNA delivery system, FG12 (Please see Methods). As
previously demonstrated by us, the FG12 system offers more than 95% RNAi
transduction efficiency and permanent gene knockdown in HKs (Fan et al. 2006;
Bandyopahdhay et al. 2006).

To verify the effectiveness of the RNAi constructs, total RNA was extracted from the
cells and subjected to QRT‐PCR analyses. As shown in Figure 2‐6A, in siHB‐EGF‐
infected HKs, the HB‐EGF mRNA level was reduced by 83% (bar 2), in comparison

51
to HKs infected with control siLac Z (bar 1). Similarly, the mRNA levels of CXCL3
and siMMP‐10 were reduced by more than 60‐70% (bars 3 and 4). At the protein
level, cell lysates or conditioned medium were collected from the gene knockdown
or control (vector alone or siLacZ) cells and subjected to immunoblotting assays
with antibodies against each of the three gene products. Consistently, the level of
HB‐EGF protein was reduced more than 95% (panel b, lane 3 vs. lanes 1 and 2). Anti‐
p38 antibody immunoblots were used as sample loading controls. The protein level
of CXCL3 in the cells was reduced by more than 70%, in comparison to the HKs
infected with empty vector or siLacZ (panel d, lane 3 vs. lanes 1 and 2).

Since we were unable to detect the intracellular protein level of MMP‐10 in HKs, we
collected TGF α stimulated‐conditioned media of HKs infected with either control
constructs or siMMP‐10 and subjected them to immunoblotting analysis with anti‐
MMP‐10 antibodies. As shown in Figure 2‐6, we detected a significant amount of
secreted MMP‐10 from conditioned media of the control cells (lanes 1 and 2).
However, MMP‐10 knockdown almost completely blocked the MMP‐10 secretion
(panel f, lane 3 vs. lanes 1 and 2). Therefore, our RNAi approach effectively down‐
regulated the genes of interest. As mentioned previously, these down‐regulations
were permanent throughout the lifespan of the cells in culture.

52

The above gene‐knockdown HKs were then subjected to cell migration assays in
response to TGF α and insulin. As shown in Fig. 2‐6 (panel G), as expected, both
TGF α and insulin strongly stimulated the migration of HKs infected with empty
vector lentivirus (bars 2 and 3 versus bars 1). However, the growth factor‐induced
migration of siCXCL3‐lentiviral infected HKs was decreased by 40%‐50% in
comparison to the control HKs (bars 5, 6 vs. bar 2, 3). These data were quantitatively
consistent with the immunoblotting results in which the cellular protein level of
CXCL3 was down‐regulated approximately 70% by the siCXCL3 (see panel c).
Therefore, CXCL3 is required for HK migration in response to the growth factors.
Surprisingly, the growth factor‐stimulated HK migration was not affected by HB‐
EGF knockdown (bars 11 and 12). On the other hand, MMP‐10 knockdown only
slightly, but significantly, lowered TGF α‐stimulated HK migration (bar 8 vs. bar 2,
but did not affect insulin‐induced HK migration (bar 9 vs. bar 3). These results
suggest that MMP‐10 and HB‐EGF, although secreted by HKs, are not essential for
HK migration. We postulate that when HK‐derived MMP10 and HB‐EGF are
secreted into the wound environment, they in turn influence migration of other skin
cell types such as dermal fibroblasts and dermal microvascular endothelial cells,
whose migration is equally critical for wound healing and re‐modeling.

53
Fig. 2‐6. Functional characterization of three secreted gene products in HK
migration
HKs, infected with FG12 system carrying vector alone, control siLac‐Z or siRNA
against each of the three candidate genes, were subjected to RT‐PCR,
immunoblotting and colloidal gold migration assays. (A) Total RNA from these cells
were isolated five days following infection and subjected to RT‐PCR. (B‐F) To detect
the protein expression levels, lysates of the HKs, starved in serum free medium
overnight, were extracted and immunoblotted with antibodies against HB‐EGF or
CXCL3. To detect secreted MMP10 proteins, infected HKs were starved in serum
free medium overnight. Conditioned medium (CM) was collected and concentrated
100 times. The CM was analyzed by Western blot with an anti‐MMP10 antibody
(panel F). (G) After confirming the effectiveness of the RNAi, colloidal gold
migration assay was carried out to functionally characterize the importance of these
genes in HK migration. The experiments were repeated 3 times (*p <0.05).
A.
B.
C.
D.
E.
F.
G.
Vector siLac Z siRNA
HB-EGF
p38
CXCL3
p38
MMP10 (in media)
1 2 3
*
0
10
20
30
siLa c Z siHBEGF siCXCL3 siMMP10
Migration Index(%)
unstimulated TGF α insulin
0
0.2
0.4
0.6
0.8
1
1.2
untreated siHB-EGF siCXCL3 siMMP10
relative mRNA level
1 2 3 4
*
*
1 2 3 4 5 6 7 8 9 10 11 12

54

Discussion
Cell migration and cell proliferation are two mutually exclusive events. Migrating
cells are not proliferating and, similarly, proliferating cells are not simultaneously
migrating. In cell migration assays, mitomycin C is often used to eliminate the
potential contribution of cell proliferation to the results of these assays and,
therefore, it might be a useful alternative to block proliferation‐related gene
expression. However, since the blocking point of mitomycin C in proliferating cells
lies in the very downstream G2/M transition phase of the cell cycle, it will not block
the growth signal‐induced early gene expression. Therefore, mitomycin C does not
help to identify migration signal‐specific genes. The central idea of the current study
is to make use of the fact that TGF β blocks HK proliferation but not migration
(Sarret et al. 1992). In addition, however, TGF β is known to block expression of both
IEGs such as c‐myc (Pietenpol et al. 1990) and Rb‐regulated DEGs (Koike et al. 1994).
Furthermore, to further ensure the “purity” of the migration‐related gene profiles,
we stimulated HKs with two distinct pro‐motility factors. We have shown that
TGF α and insulin are the main keratinocyte pro‐motility factors in human serum. (Li
et al. 2006).

55
It is well established that TGF β’s effects can be divided into primary and secondary
effects. Its primary effects on most cell types include anti‐proliferation and/or anti‐
motility. Its secondary and main effect is to induce expression of many genes
including ECMs and mitogenic/motogenic growth factors, cytokines and MMPs.
These newly produced factors could in turn act either in an autocrine or paracrine
fashion to affect various responses of the cells. Our migration assays, colloidal gold
and “scratch” assays, last 12‐16 hours. Within this period, the physiological
concentrations of TGF β did not cause any significant change in human keratinocyte
migration stimulated by TGF α, as we and others have previously reported, (Sarret et
al. 1992; Tsuboi et al. 1992; Garlick and Taichman 1994; Ashcroft et al. 1999).
However, as mentioned earlier, the ECMs and growth factors produced by the cells
in response to TGF β may later on act to stimulate cell migration (e.g. > 48 hours)
(Nickoloff et al. 1988; Hebda 1988).

Microarray analysis allows profiling large‐scale and comparative gene expression
under two different conditions. Therefore, this approach can provide a useful global
picture of the gene expression involved in a given activity taking place inside the
cells. In conjunction with the available knowledge on the kinetics and functions of a
given gene, the technique is helpful to reveal intracellular signaling networks that

56
execute a specific cellular response such as motility. Here, we have identified 82
commonly up‐regulated genes and 19 commonly down‐regulated genes by both
TGF α and insulin following stimulation for up to two hours. Among these genes, 31
of the up‐regulated genes (38%) have previously been reported to play a role in cell
migration, and 9 of the down‐regulated genes (47%) were reported for cell migration.
Among the 14,500 well‐characterized genes in the Affymetrix HU‐133A 2.0 array,
there are only 187 probes representing 72 genes (0.5%) that are known to relate to
cell migration. Therefore, our strategy “enriched” the pool of migration signal‐
specific genes more than 80 fold. In analyzing the DNA microarray data, we used
SScore program to select significantly regulated genes by both TGF α and insulin and
generated dendrograms based on this selection. In the dendrogram, genes with
more than one probe on the microarray chip were shown more than once, such as
HB‐EGF and PHLDA1. These different probe sets for the same genes tend to cluster
together, indicating that their expression patterns are similar. Results of this analysis
further confirmed the induction or repression of the genes by TGF α and insulin,
since multiple probe sets for the same genes showed similar expression patterns.

Our study focused on the migration‐specific gene profiles in primary human
keratinocytes in vitro. Are these profiles relevant to those in HKs in vivo?

57
Unfortunately, this kind of study using human skin wounds is technically
impossible. It is not possible to wound human skin, take biopsies at different time
points, isolate the RNA from them and obtain gene expression profiles. Even in mice,
it is not feasible to isolate the wound cells/tissues without contaminating the
specimens with unwounded tissue. Since the latter is not in the environment of the
wound, the gene expression profiles from unwounded tissue do not reflect wound
healing. Furthermore, gene profiles from mouse tissue may not be at all relevant to
the gene profiles of human skin wounds. Second, in a real skin wound, the
environment is a mixture of multiple factors (cytokines, growth factors, ECMs,
MMPs et al). It is not possible to distinguish what genes are inflammation‐related,
proliferation‐related, migration‐related, differentiation‐related, metabolism‐related
and so on. It is impossible to know which gene is induced by what factor.

While microarray has been widely used to determine gene expression patterns
under various environmental and cellular conditions, to our knowledge this is the
first report of gene expression profiles designed in response to a specific type of cell
surface signal. Several previous studies also used a DNA microarray approach to
identify gene profiles associated with wound healing. Fitsialos et al surveyed the
gene expression profiles during the course of an in vitro scratch assay using cultured

58
HKs. They used inhibitors of ERK, p38‐MAPK and phosphatidylinositol 3‐kinase
(PI3K) to narrow down the pathway‐specific gene expression profiles (Fitsialos et al.
2007). They identified several extracellular factors that were induced at early time
points (1~3 hours), such as HB‐EGF, VEGF, EREG, LIF, SERPINE‐1, and IL‐8, and
transcription factors, such as Fos, Jun and EGR1. These results are consistent with
our observations. Roy et al compared the gene expression profiles in wound‐derived
blood vessels and blood vessels from intact human skin (Roy et al. 2007). They
demonstrated that the extracellular proteinase MMP1 was up‐regulated, which was
also observed in our study. In conclusion, our study has provided comprehensive
profiles of the early gene expression induced by HK pro‐motility signals. These
profiles may serve as a foundation to gain new insights into the mechanism of the
re‐epithelialization process during wound healing.

59
Chapter 3: TGF α‐Stimulated Secretion of HSP90 α: Using LRP‐
1/CD91 Receptor To Promote Human Skin Cell Migration
Against TGF β‐Rich Environment In Wound Healing

Introduction:
Transforming growth factor‐alpha (TGF α) belongs to the EGF family, which has ten
members in humans. The mature form of TGF α is a 50‐amino acid polypeptide
originally isolated from conditioned medium of virally transformed cells and tumor
cells (de Larco and Todaro 1978; Roberts et al. 1980). Active TGF α shows a
heterogeneous molecular mass ranging from 5‐20 kDa in SDS‐PAGE, due to variable
degrees of N‐ and O‐linked glycosylation (Teixido and Massague 1988). Although
the amount of TGF α in human circulation (plasma) is low or undetectable, it is
widely expressed in developing embryos and a number of adult tissues. Thus, TGF α
is mainly an autocrine/paracrine growth factor. TGF α elicits all its biological effects
by binding to and activating a 170‐kDa cell surface tyrosine kinase receptor (EGFR
or c‐erbB1), a member of the EGFR subfamily. This receptor family also includes c‐
erbB2/HER2/Neu, c‐erb3/HER3, and c‐erbB4/HER4 (Fantl et al. 1993). A major well‐
characterized effect of TGF α on cells of the epidermal origin is enhancement of cell

60
proliferation and migration, which often cannot occur simultaneously. In vivo
studies showed that increased local concentrations of TGF α or overexpression of
EGFR play an important role in both physiological processes, such as wound healing
and hair follicle development, and pathological disorders, such as psoriasis
(hyperproliferation of keratinocytes) and skin tumorigenesis (Fuchs and Byrne 1994;
Nanney et al. 1996). However, ablation of TGF α or EGFR gene in mice only showed
some visible abnormalities in the outer root sheath and hair follicle architecture,
likely due to the multiple family members at both the ligand and receptor levels of
TGF α (Luetteke et al. 1993; Mann et al. 1993). While the EGFR > Grb2/Sos > Ras‐GTP
> Raf > Mek1 > ERK1/2 pathway is widely accepted as the mechanism to promote cell
proliferation, the motility signaling pathway remains poorly understood.

During human skin wound healing, a critical step is the initiation of the resident
epidermal and dermal cells at the wound edge to migrate into the wound bed
(Martin 1997; Singer and Clark 1999). Human keratinocytes (HKCs) laterally migrate
across the wound bed from the cut edge to eventually close the wound, the process
known as re‐epithelialization. The dermal cells, including dermal fibroblasts (DFs)
and dermal microvascular endothelial cells (HDMECs), start to move into the
wound following the keratinocyte migration, where these cells deposit matrix

61
proteins, contract and remodel the newly closed wound and subsequently build
new blood vessels. The HKC migration is largely driven by TGF α from human
serum (Li et al. 2006a) and unaffected by TGF β family cytokines co‐present in the
wound (Bandyopadhyay et al. 2006). In contrast, the presence of TGF β blocks both
DF and HDMEC migration even in the presence of their growth factors, PDGF‐BB
and VEGF (Ali et al. 2006). Therefore, while it is clear why HKC migration
jumpstarts ahead of the DF and HDMEC migration during wound healing, it has
remained as a long time puzzle how DFs and HDMECs have moved into the wound
bed in the presence of TGF β.

The heat shock protein (hsp) families include chaperon proteins that are either
constitutively expressed, such as the hsp90 family, or stress‐induced expression,
such as the hsp70 and hsp27 families. Historically, their function is to interact with
and facilitate proper folding and intracellular trafficking of target proteins to
maintain the cellular homeostasis and to promote cell survival (Gething and
Sambrook 1992). However, recent studies showed that hsp proteins could also be
secreted by the cells. The secreted hsp proteins then carry out important
extracellular functions, including stimulation of immunological cytokine production,
activation of antigen presenting cells (APCs) and anti‐cancer functions (Binder et al.

62
2004; Schmitt et al. 2007). We recently showed that hypoxia causes hsp90 secretion,
which in turn mediates hypoxia‐induced dermal fibroblast migration during wound
healing (Li et al. 2007). Since hsp proteins lack any signal sequences at the amino
terminus, these proteins cannot be secreted via the classical endoplasmic
reticulum/Golgi transport pathway. Instead, they are secreted via a discrete
population of nano‐vesicles (30‐90nm in diameter), called exosomes (Multhoff and
Hightower 1996; Hegmans et al. 2004; Clayton et al. 2005; Lancaster and Febbraio
2005; Yu et al. 2006). The exosome secretion mechanism has, therefore, constituted a
potential mode of intercellular communication and opens up new therapeutic and
diagnostic strategies (Mignot et al. 2006). In the present study, our findings indicated
that TGF α “pushes” hsp90 α out of HKCs via the exosome pathway, which in turn
promotes migration of both the epidermal and dermal cells through the cell surface
receptor LRP‐1/CD91. The physiological significance is also discussed.

63

Materials and Methods
Primary human neonatal HKCs, DFs, HDMECs and melanocytes (MCs) were
purchased from Clonetics (San Diego, CA). HKCs were cultured in the EpiLife
medium with added human keratinocyte growth supplements (HKGS). MCs were
maintained in Medium 154 supplemented with human melanocyte growth
supplement (HMGS) (Cascade Biologics, Portland, Oregon). DFs were cultured in
DMEM supplemented with 10% fetal bovine serum. HDMECs were cultured in
growth factor‐supplemented Medium 131 (Cascade Biologics). The 5‐7th passages of
the cultured cells were used in conditioned medium preparation and 3
rd
or 4th
passage in cell migration assays. Human sera, collected from a variety of donors,
were purchased from Sigma‐Aldrich (St. Louis, MI). Human recombinant TGF α and
TGF β3 were purchased from R&D System (Minneapolis, MN). Gel Filtration
Calibration Kit HMW (Code, 28‐4038‐42) was from GE Healthcare (Uppsala,
Sweden). Anti‐hsp90 α antibody (SPA‐840) for Western analysis, anti‐hsp90 α
neutralizing antibody (SPS‐771, blocking the target‐binding site of hsp90) were all
from Stressgen (Victoria, BC Canada). Recombinant hsp‐90 α was from Stressgen and
produced in our laboratory. The construct of GST‐human 14‐3‐3 σ was a gift from Dr.
Mong‐Hong Lee (University of Texas, Houston, Texas). Recombinant

64
AMF/phosphoglucose isomerase/neuroleukin, Brefeldin A (BFA) and dimethyl
amiloride (DMA) were purchased from purchased from Sigma‐Aldrich (St. Louis,
MO). Rhodamine‐conjugated Phalloidin were from Sigma. Rat type I collagen was
purchased from BD Biosciences (Bedford, MA). Anti‐ β actin antibody and anti‐
GAPDH antibody were from Cell Signaling Technology.

Immunostaining cells and normal human skin.
Skin cells or human skin were cultured on glass coverslip pre‐coated with different
ECMs in sterile 8‐well plate. Cells were starved overnight and stimulated with
growth factors. Stimulated cells or human skin were fix with 4% formaldehyde/PBS.
Cells were permiabilized with 0.1% Triton X‐100/PBS. Fixed Cells or human skin
were blocked with 1%BSA/PBS and incubated with proper primary and secondary
antibody sequentially. The images of the stained cells or human skin were visualized
under either NIKON‐TE‐200U microscope using the green (FITC), blue (DAPI), or
red (Phalloidin) filters, or Nikon Eclipse 80i confocal microscope. Usually, 80‐120
cells per condition were analyzed.

Immunodepletion of hsp90 α from HKC‐CM.
10x concentrated HKC‐CM was pre‐incubated with the optimized amount of anti‐

65
hsp90 α rabbit polyclonal antibody (7 µg IgG/2 ml of 10X HKC‐CM to remove all
hsp90 α) for 16 h at 4
o
C. The immune complexes were precipitated by incubation
and agitation with protein G‐sepharose (25 µl of packed beads, Pharmacia) for 2 h.
After centrifugation, the hsp90 α‐free supernatants were removed and used for cell
migration assays.

In Vitro wound Healing Assay
Twelve‐well plates were precoated with collagen (45μg/ml) in triplicates, followed
by further BSA blocking. A sufficient number of serum‐starved HKCs were plated,
so that they became confluent in the wells right after attachment (1~2 hours).
Scratches were then made with 200μl‐yellow tip. Floating cells were removed by
PBS wash. Control media or conditioned media were added to the wells and
incubated for additional 16 hours. Mitomycin C (10μg/ml) was always included in
the media to prevent cell proliferation. Five representative images of the scratched
areas under each condition were photographed. To estimate the relative migration
of the cells, the unclosed cell‐free areas from five prints under each condition were
excised and weighed on a scale (Mettler AE50). We used “average gap” (AG, %) to
quantify the data. The collagen alone at 0 h was considered 100% AG. The wells that
showed the most HKC migration had the least AG values.

66

Sub‐cloning, protein production and purification of hsp90 α and mutants
The cDNAs for wt hsp90 α and three mutants (E47D, E47A, and D93N) were kindly
provided by Dr. Ulrich Hartl (Max Planck Institute for Biochemistry, Germany). The
coding regions of these cDNAs were subcloned into the His‐tag pET15b vector
(EMD biosciences, Inc., San Diego, CA) at BamH1 using a PCR‐based cloning
technique. The primers for PCR were as followed: 5’‐
GGATCCGATGCCTGAGGAAACCCAG‐3’ and 5’‐
ACTGTCGGATCCTTAGTCTACTTCTTCCAT‐3’. The pET15b‐hsp90 α constructs
were transformed into BL21‐codonPlus (DE3)‐RP competent cells (Stratagene, La
Jolla, CA) following the manufacturer‐provided protocol. Protein synthesis was
induced by the addition of 0.25 mM IPTG to the bacterium culture (O.D. ≅ 0.8) and
incubation for five hours at 25
o
C. The his‐tag proteins were first purified by Ni‐NTA
column with the HisBind purification kit (EMD biosciences, Inc.) according to the
manufacturer’s procedure. The purified proteins were concentrated in Centricon
YM‐100 (Millipore, Billerica, MA) to 4 ml and loaded onto a Superdex 200 HiLoad
gel filtration column (GE healthcare, Piscataway, NJ) and separated by FPLC.
Proteins were eluted by DPBS buffer with a flow speed of 1.2 ml/min. The fractions
with hsp90 α were further concentrated in a Centricon YM‐100 to the final

67
concentration of 1mg/ml. Proteins were stored in 10% glycerol‐DPBS at ‐70
o
C.
Generally, 20mg of hsp90 α could be generated from 1L of the bacterium culture.

Five distinct domains (N’, M‐1, M‐2, C’‐1 and C’‐2) of hsp90 α were constructed into
the His‐tag pET15b as above. After Ni‐NTA column purification, each of the domain
proteins was concentrated in Centricon YM‐50 or YM‐10 to the volume of ~4 ml
depending on the size of the domains. These domains were then purified by FPLC in
a Superdex 75 HiLoad gel filtration column (GE healthcare, Piscataway, NJ) and
followed by concentrated in Centricon YM‐50 or YM‐10 to the final concentration of
1mg/ml.

GFP‐ hsp90 α and siRNA against LRP1/CD91.
The cDNA for wt hsp90 α was amplified using PCR and subcloned in frame into the
3’ end of a fluorescence protein (GFP) gene. This fusion gene was inserted into the
lentivirus‐derived vector, pRRLsinhCMV, at BamHI + EcoRI. To identify possible
target sequences for siRNA against CD91, we used the RNAi Selection Program (Fan
et al. 2006). Four potential sites were selected and synthesized. The effectiveness of
synthetic double strand siRNA in down‐regulation of CD91 was first measured in
293 cells by transfection and the cell lysates blotted with corresponding antibodies.

68
The two most effective RNAis were then cloned into the lentiviral RNAi delivery
vector, FG‐12 (Fan et al. 2006). The selected RNAi sequence (sense) against human
CD91 for FG12 cloning was GACCAGTGCTCTCTGAATA (RNAi‐1) and
GGAGTGGTATTCTGGTATA (RNAi‐2). These constructs were used to transfect
293T cells together with two packaging vectors, pCMV ∆R8.2 and pMDG, to produce
virus stocks.

ATPase assay
The ATPase assay was based on a regenerating coupled enzyme assay (Ali et al.
1993), in which the phosphorylation of ADP by pyruvate kinase (PK) at the expense
of phosphoenol pyruvate is coupled to the reduction of the resulting pyruvate by
lactate dehydrogenase (LDH) at the expense of NADH. Oxidation of NADH to
NAD
+
produces a loss of optical density at the NADH absorbance maximum of 340
nm, in direct stoichiometry to the amount of ADP phosphorylated. The ATPase
activity was expressed as the turnover of NADH per minute per microgram of
protein. Each 400 µl assay contained 50 mM HEPES‐KOH, pH 8.0, 5 mM MgSO4, 0.6
mg ATP (Sigma), 30 µg NADH (Sigma), 2 mM phosphoenol pyruvate (Sigma), 1.4
units of PK, 2 units of L‐LDH (PK/LDH mix from Sigma), and 200 µg of wild type or
mutant hsp90 α proteins.

69

Protein purification
A total of 10 liters of HK‐CM were collected in 500 ml batches, each of which tested
positive for pro‐motility activity on HKs using colloidal gold migration assays. The
volume of this starting material was concentrated sequentially to ~5ml using 20‐ml
Centricons with a 10 kDa cutoff for proteins and equilibrated to 10 ml with the Start
Buffer for SP Sepharose column (50mM MES pH 6.0). The sample was passed
through the SP Sepharose column in the FPLC System (Pharmacia LKB) at 2.5
ml/min. Unbound proteins were collected in the flow‐through. The column was
washed with a start buffer and, then, bound proteins were eluted stepwise with
increasing concentrations of NaCl (.1M, .2M, .3M, .4M, .6M and 1.0M). Fractions
containing the protein peaks from each NaCl‐concentration of elution were pooled
and concentrated to a final volume of ~2 ml in Centricons. During the concentrating
process, the NaCl concentrations were equalized to ~150 mM with MES ± 1M NaCl.
The materials were further equilibrated by washing and centrifuging twice in 1:10‐
fold dilutions with DMEM and concentrated to the original volume (2 ml). They
were then tested for stimulation of HK migration using colloidal gold migration
assays, in comparison to the original 10x HK‐CM in equal amount of total proteins.
The peak activity was observed in the 0.2M fractions. This “positive” material was

70
concentrated to 0.5ml using a 2ml‐Centricon 10 and loaded into a Superdex 75
HiLoad gel filtration column (Pharmacia). This seizing column was run with PBS at
1 ml/min and filtrates collected in 5ml/tube fractions. The fractions containing a
protein peak and to be used for further analysis were washed in a Centricon 10 with
2x in DMEM and concentrated to 2ml. The HK pro‐motility activity of the samples
was tested as previously described with the amount of total protein per sample
equalized to that in 10x HK‐CM. The fractions with distinct protein peaks were also
tested in combinations and we did not detect any synergistic effects in any of the
combined fractions. The migration assay revealed that significant activity was
contained in fractions 18 and 19. Fractions 18 and 19 were combined and
concentrated to 2 ml with a 20ml‐centricon. The sample was then diluted to 10 ml
(1:5) with equilibration buffer (20mM Tris, pH 8.0) for the Q Sepharose column
(Pharmacia). The diluted sample was loaded into the Q Sepharose column at 2.5
ml/min and washed in the same equilibration buffer. The bound proteins were
eluted with increasing NaCl concentrations (.1M, .2M, .3M, .4M, .5M, .6M, and 1.0M
NaCl). The fractions containing a protein peak at each elution of a given NaCl
concentration were pooled individually. The NaCl in these fractions were equalized
to 150mM with Tris, pH 8.0 +/‐ 1M NaCl using Centricons and the volumes were
reduced to 2 ml. The samples were washed twice (1:10) with DMEM in 20 ml‐

71
Centricons and concentrated to equal volumes. Each sample was tested for HK pro‐
motility activity with equal amounts of total protein and the 0.3M NaCl fraction
showed the strongest activity. This fraction was concentrated to 500 µl and loaded
onto a Superdex 200 HiLoad gel filtration column (Pharmacia) using PBS as the
running solution at 0.5 ml/min. 500 µl/tube fractions were collected and HK pro‐
motility activity was tested for each of the fractions with equal sample volumes as
opposed to equal amounts of protein as used previously. Half of each fraction was
used for the migration assay (250 µl). The recovered protein amount and activity
after purified with each column is as shown in Tab. 3‐1.

72

Tab. 3‐1. Purification of HK pro‐motility activity from HKC‐CM Ten liter
conditioned medium was collected, concentrated, and purified sequentially through
four different columns. The amount of protein recovered and the activity remained
after each purification are shown.
Steps Volume (ml) Protein (mg) Specific Activity(MI) Total Activity Recovery
HKC-CM 500 (20 x) ~ 370 25-30/0.41 mg 100%
SP-Sepharose 6.0 ~ 36 25-30/122 µg~37%
Q-Sepharose 4.5 ~1.44 25-30/23.6 µg~7.9%
Superdex 75 2.5 ~9.5 25-30/48.5 µg~18.5%
Superdex 200 4.0 ~ 0.07 25-30/16.3 µg ~6.7%

73

RESULTS
TGF α stimulates HKCs to export stimulators of motogenic, but not mitogenic,
activity.
TGF α is a potent mitogen and motogen in human serum for epidermal HKCs (Li et
al. 2006a). Migrating cells are not proliferating and vice versa. We wanted to identify
downstream and motility signal‐specific targets that satisfy the following criteria: 1)
a primary target of TGF α signaling; 2) a secreted protein and 3) a factor that
selectively promotes migration, but not proliferation. We focused on serum‐free
conditioned media of TGF α‐primed HKCs, according to the step‐by‐step procedure
depicted schematically in Fig. 3‐1A. The starting control medium (Con) and the
conditioned medium (HKC‐CM) were subjected independently to migration and
DNA synthesis assays. As shown in Fig. 3‐1B, 10x HKC‐CM (panel d), but not 10x
control medium (panel c), was able to duplicate the pro‐motility effect of TGF α
(panel b vs. panel a) in the colloidal gold migration assays, quantitated as migration
index (MI) (Li et al. 2004c). This observation was confirmed by using the in vitro
“scratch” assay in the presence of a DNA synthesis inhibitor, mitomycin C. As
shown in Fig. 3‐1C, the control medium only slightly closed the “wounded” area
following overnight incubation (panel c’ versus panel a’), likely due to coated

74
collagen‐driven cell migration (Li et al. 2004a). Incubation with the HKC‐CM,
however, caused complete closure of the wounded area (panel d’), similar to the
TGF α‐stimulated migration (panel b’). The data was quantitated as “average gap”
(AG) of the unclosed space (Li et al. 2004c). 10x HKC‐CM was chosen, because it was
the lowest concentrated HKC‐CM that reached a plateau pro‐motility activity (Fig.
3‐2).

Interestingly, unlike the dual effects of TGF α on cell migration and growth, HKC‐
CM showed little mitogenic effects on either epidermal or dermal cells. As shown in
Fig. 3‐1D, TGF α strongly stimulated DNA synthesis in HKCs, as expected (bar 4 vs.
bar 1). In comparison, HKC‐CM showed little stimulation of DNA synthesis (bar 7).
Similar effects of HKC‐CM on HDMECs (bar 8 vs. bar 2) and DFs (bar 9 vs. bar 3)
were observed. In comparisons, PDGF‐BB and FBS strongly stimulated DNA
synthesis in these cells (bars 5 and 6), as expected. In contrast, we found that HKC‐
CM was also able to stimulate migration of DFs and HDMECs. As shown in Fig. 3‐
1E, in comparison to the pro‐motility effects of PDGF‐BB and FBS, respectively (bars
5 and 6), HKC‐CM showed an equivalent level of stimulation of HDMEC (bar 8) and
DF (bar 9) migration. Therefore, this yet‐to‐be identified pro‐motility activity in the
HKC‐CM has so far satisfied two of the three proposed criteria, i.e. 1) a secreted

75
factor(s) and 2) with motogenic but not mitogenic effect on all the major types of
human skin cells involved in wound healing.

Exported hsp90 α is fully responsible for motogenic activity in HKC‐CM.
What is the factor(s) in HKC‐CM that promotes skin cell migration? An SDS‐PAGE
slab gel and silver staining of HKC‐CM, as shown in Fig. 3‐3A, revealed multiple
proteins with a wide range of molecular masses (left column). To identify which of
these proteins is responsible for the pro‐motility activity found in HKC‐CM, we
started to purify HKC‐CM. The detailed purification procedures are discussed in
Materials & Methods section.

76
Fig. 3‐1 Secretion of TGF α‐treated HKCs promotes skin cell migration, but not
proliferation. (A) The outline for preparing serum‐free conditioned medium of
HKCs (HKC‐CM) and control medium (Con). (B) Comparison of HKC‐CM with
TGF α in stimulation of HKC migration using the colloidal gold migration assay
(Methods) n=4, p < 0.05. An average size of the migration tracks under each
condition was marked with an open circle. (C) Comparison of HKC‐CM with TGF α
on HKC migration in the “scratch” assay. AG, average gap (Methods), n=3, p < 0.05.
(D) Unlike serum growth factors, HKC‐CM did not cause DNA synthesis in major
human skin cell types, HKCs, DFs (dermal fibroblasts) and HDMECs (human
dermal microvascular endothelial cells). n=4. Bars with * are statistically significant
over serum‐free controls, p < 0.01. (E) Comparison of HKC‐CM on migration of
HKCs, DFs and HDMECs with serum growth factors using the colloidal gold
migration assay. Only Migration Index (MI) of the experiments is shown, n=3, Bars
with * are statistically significant over their serum‐free controls, p < 0.05.

77
HKs on collagen in
medium w/ TGF α
Remove medium
and rinse 5x w/ PBS
Add serum-free medium,
rinse and save as control (Con.)
Add serum-free medium
& incubate overnight
Collect conditioned medium
(HKC-CM)
Concentrating &
migration assays
A.
SF+ TGF α Con.
a
b
c
d
MI (%):
7 ± 2.6
25 ± 3.1
8 ± 1.7
29 ± 4.2
B.
a’
d’
AG (%):
100 ± 13
23 ± 14
78 ± 9.6
19 ± 16
SF/ 0 hr
C
.
D.
b’
Con./16 hr
c’
[
3
H-Thymidine] Incorp. (fold increase)
1
2
3
serum-free HKC-CM
(10 x)
1 2 3 4 5 6 7 8 9
4
HDMEC HK
DF
Migration Index (%)
10
20
HK
30
serum-free
HKC-CM
(10x)
E.
1 2 3 4 5 6 7 8 9
HDMEC DF
*
*
*
*
*
*
*
*
*

78
Fig. 3‐2. Dose‐dependent pro‐motility effect of HKC‐CM. HKC‐CM, collected from
human keratinocyte culture, is defined as 1x. The protein contents of this medium
was then concentrated to 5x, 10x, 20x and 40x and tested for their ability to promote
HK migration in comparison to the effect of TGF α. 10x HKC‐CM gave rise to a
TGF α‐equivalent stimulation of the cell migration. Quatitation and statistic analyses
are detailed in Methods.

10
20
Migration Index (%)
30
Con.
HKC-CM
1 2 3 4 5 6 7 8 9 10
TGF α (ng/ml)
5 10 15 20
1x 5x 10x 20x 40x

79

The protein contents of ten liters of HKC‐CM were concentrated 100 fold and
subjected to FPLC (Fast Protein Liquid Chromatography), as diagrammed
schematically in Fig. 3‐3A (middle column, see a detailed protein purification
procedure in Table 1). Fractions 17‐18 of the final chromatography (Superdex 200),
which contained a peak of pro‐motility activity, were pooled, concentrated and
resolved by SDS‐PAGE. Silver stain of the SDS gel revealed four major polypeptides.
Mass spectrometry analyses of the excised individual bands unveiled four
previously known gene products. From high to low molecular masses they were:
heat shock protein‐90 α (hsp90 α), autocrine motility factor (AMF)/neuroleukin,
fibronectin isoform 6 preproprotein and 14‐3‐3 σ. To narrow down if any of the
proteins is responsible for the motogenic activity in HKC‐CM, the recombinant
forms of these four proteins were individually tested. As shown in Fig. 3‐3B, we
surprisingly discovered that recombinant hsp90 α stimulated HKC migration in a
dose‐dependent manner, which was equivalent to TGF α stimulation (bars 3‐6 vs. bar
2). In contrast, neither AMF nor 14‐3‐3 σ showed any significant pro‐motility effects
on HKCs (bars 8‐11 or bars 12‐15 vs. bar 1). We reported previously that fibronectin
minimally stimulates HKC migration (Li et al. 2004c).

80

To prove the physical presence of hsp90 α protein in HKC‐CM and estimate its
working concentration in comparison to recombinant hsp90 α, 10x HKC‐CM,
together with a series of known amounts of recombinant hsp90 α proteins, was
subjected to anti‐hsp90 α Western blotting analyses. As shown in Fig. 3‐3C, based on
a standard curve generated from the increasing amounts of recombinant hsp90 α
(lanes 1, 2, 3), 10x HKC‐CM contained 0.06‐0.08 µM of hsp90 α, which is close to the
optimal pro‐motility effect of 0.1 µM of recombinant hsp90 α (see Figure 3B, bar 5).
To verify that hsp90 α is the pro‐motility factor in HKC‐CM, we depleted hsp90 α
from HKC‐CM by anti‐hsp90 α antibody immunoprecipitation and then tested the
hsp90 α‐free HKC‐CM on HKC migration. As shown in Fig. 3‐3D, anti‐hsp90 α
antibody depletion eliminated the pro‐motility activity of HKC‐CM in antibody
dose‐dependent manner (bars 6‐10). In contrast, neither a control IgG nor
neutralizing antibodies against AMF or 14‐3‐3 σ showed any inhibitory effect on
HKC‐CM‐stimulated HKC migration (bars 3, 4 5 vs. bar 2). These results clearly
indicated that the pro‐motility activity in HKC‐CM is hsp90 α.

81
Fig. 3‐3 Identification of hsp90 α from HKC‐CM. (A) 50 µl of 20x HKC‐CM was
analyzed by SDS‐PAGE and silver stain (left column). Migration assays were used to
identify the “positive” fraction(s) of purification that contained the peak of pro‐
motility activity. Pooled and concentrated materials of the #17‐18 fractions from the
final Superdex 200 chromatography were resolved in SDS gel and silver stained
(right column). The four identifiable bands were excised and subjected to mass
spectrometry analyses and their identities are shown (arrows). (B) Recombinant
proteins of human hsp90 α, AMF and 14‐3‐3 σ were individually tested for
stimulation of HKC migration, in comparison to the stimulation of TGF α. The
results of the colloidal gold migration assays are shown as Migration Index (MI), n=3,
Bars with * are statistically significant over the control, p < 0.05. (C) Samples with
known amounts of recombinant hsp90 α, as indicated, were loaded side‐by‐side with
increasing volumes of a 10x HKC‐CM on an SDS gel and subjected to
immunoblotting analysis with an anti‐hsp90 α antibody. A standard curve was
established based on densitometry scanning of the control bands and was used to
estimate the amount of hsp90 α in HKC‐CM. n = 3. (D) Neutralizing antibodies
against human hsp90 α (bars 6‐10), AMF (bar 4) or 14‐3‐3 σ (bar 5) or control IgG (bar
3) were used to “deplete” each of the corresponding antigens in HKC‐CM by
immunoprecipitations. The treated HKC‐CM were tested for their stimulation of

82
HKC migration in colloidal gold migration assays. Only MI of the experiments is
shown. n=4, Bars with * are statistically decreases over the positive controls, p < 0.01.

83

We were curious whether secretion of hsp90 α is a general event in other types of
skin cells stimulated with growth factors, i.e. DFs, HDMECs and melanocytes (MC).
As shown in Fig. 3‐4A, we found that hsp90 α was undetectable in CM of PDGF‐BB‐
stimulated DFs (lane 1), VEGF‐stimulated HDMECs (lane 4) or FBS‐stimulated MCs
(lane 3). However, a 90‐kDa protein was clearly detected in TGF α‐stimulated HKC‐
CM (lane 2) that co‐migrated with the recombinant hsp90 α control (lane 5). We then
tested whether hsp90 α also satisfies our third criterion, i.e. not a mitogen. As shown
in Fig. 3‐4B, consistent with the lack of mitogenic effect in HKC‐CM, recombinant
hsp90 α was unable to cause any significant increase in DNA synthesis in the three
primary human skin cell types studied (bars 7‐9).

As far as the intracellular hsp90 α levels were concerned, we observed that HKCs
and DFs had similar levels and HDMECs showed a slightly lower level of
intracellular hsp90 α proteins (Fig. 3‐4C, panels a and b). Hsp90 β was included as a
specificity control. HKCs showed a lower expression and DFs and HDMECs showed
similar levels of hsp90 β (Fig. 3‐4C, panels c and d). To examine the expression of
hsp90 α versus hsp90 β (control) in vivo with monoclonal antibodies specifically
against either hsp90 α or hsp90 β, as shown in Fig. 3‐4D, we found that hsp90 α

84
staining is significantly stronger in the epidermis than in the dermis (panel b vs.
panel a). In reverse, hsp90 β staining was strongly detected in the dermis, but weakly
in the epidermis (panel e vs. panel d).

TGF α stimulates hsp90 α translocation and secretion selectively in HKCs via the
exosome pathway.
Based on the above findings, we speculated that TGF α should stimulate membrane
translocation and secretion of intracellular hsp90 α in HKCs, but not in dermal cells.
To test this hypothesis, we undertook two independent approaches, 1) direct
immunostaining the endogenous hsp90 α with a monoclonal anti‐hsp90 α antibody
and 2) detection of secreted hsp90 α from the conditioned media. As shown in Fig. 3‐
5A, TGF α stimulation caused a rapid and robust membrane relocation and cell
surface clustering of hsp90 α in a time‐dependent manner in HKCs (left column,
panels b to d vs. panel a). The accumulation of hsp90 α, particularly between 15 min
to 60 min following TGF α stimulation, was so overwhelming that it formed
“budding structures” toward outside the cell membrane (panels c and d, see
inserted enlargement). In fact, this translocation could occur as rapidly as two
minutes following TGF α stimulation (Fig. 3‐6). The presence of cyclohexamide, a
common protein synthesis inhibitor, showed no inhibitory effect on the translocation

85
(data not shown). Interestingly, however, no detectable hsp90 α relocation occurred
in DFs in response to PDGF‐BB, a potent DF pro‐motility factor (right column,
panels a’ to d’). More surprisingly, as shown in Fig. 3‐5B, although TGF α can also
stimulate DF migration, it was unable to induce any detectable membrane relocation
of hsp90 α in DFs (panel d), as it did in HKCs in a side‐by‐side experiment (panel b).
The reason remains unknown.

To verify directly that TGF α stimulation causes further secretion of hsp90 α from
HKCs, two identical dishes of serum‐starved HKCs were either untreated or treated
with TGF α for four hours. Both conditioned media and total cell lysates were
comparatively analyzed by Western blot with an anti‐hsp90 α specific antibody. As
shown in Fig. 3‐5C, TGF α stimulation resulted in a detectable decrease in the
intracellular hsp90 α level (panel a, lane 2 vs. lane 1). Anti‐ β‐actin blot was used as
the loading control (panel b). Consistently, TGF α stimulated a concomitant increase
in secreted hsp90 α in the conditioned medium over a basal level (panel c, lane 2 vs.
lane 1). Thus, these data provided direct evidence that TGF α stimulates secretion of
hsp90 α from its pre‐existing intracellular pool. Estimation of the secreted hsp90 α
versus the total intracellular hsp90 α at both four hour and 16‐hour (overnight) time
points revealed 16‐20% secretion from the intracellular hsp90 α pool.

86
Fig. 3‐4 Hsp90 α is secreted selectively by HKCs and has no mitogenic effect. (A)
Equal volumes of serum‐free CM from four major types of primary human skin cells,
HKCs, DFs, HDMECs and MCs (melanocytes), were analyzed by Western
immunoblotting with anti‐hsp90 α antibodies. Commercial human recombinant
hsp90 α protein was included as the positive control. (B) Serum‐starved HKCs, DFs
and HDMECs were untreated or treated with indicated stimuli in [
3
H]‐thymidine
incorporation assays. The data was presented as fold increases, based on triplicates
per condition. Arrows point to the effect of hsp90 α (bars 7‐9), in comparison to the
expected mitogenic effect of GF stimulation (bars 4‐6), n=3. Bars with * are
statistically significant over serum‐free controls, p < 0.01. (C) Comparison of the total
cellular levels of hsp90 α (panels a, b) versus hsp90 β (panels c, d) among epidermal
HKCs, dermal DFs and dermal HDMECs in Western blot analyses with anti‐hsp90 α
antibodies. These experiments were repeated four times and similar observations
were made. (D) Normal human skin sections were subjected to indirect
immunofluorescence staining with the antibodies against hsp90 α (panel b), hsp90 β
(panel d) or corresponding IgG controls (panels a and c). The images show skin
tissue distribution of hsp90 α (b), hsp90 β (d). Dotted lines refer to the basement
membrane zone of skin. Derm, dermis; Epi, epidermis.

87
Fig. 3‐5 TGF α stimulates membrane translocation and secretion of endogenous
hsp90 α selectively in HKCs. (A) HKCs, cultured on collagen‐coated coverslips and
serum‐starved overnight, were either untreated or treated with an optimal
concentration of TGF α (20 ng/ml) for the indicated periods of time (also see detailed
kinetics in Fig. 6). In the same experiments, PDGF‐BB‐treated DFs were included as
a cell type‐specificity control. Both cell types were fixed and immunostained with a
monoclonal antibody specifically against human hsp90 α. The results were
visualized by fluorescence‐conjugated secondary antibody. 80‐120 cells per
condition were randomly selected and analyzed. Under each condition, a
representative image and the percentage of the cells that shared the image over the
total number of the cells examined is shown. (B) HKCs and DFs were
simultaneously cultured, serum starved and treated with TGF α. As shown, TGF α
stimulates hsp90 α membrane translocation selectively in HKCs, but not in DFs. (C)
Serum‐starved HKCs were untreated or treated with 20ng/ml TGF α for four hours.
Both total lysates and equal volumes of serum‐free conditioned media (concentrated
from 4ml to 50 µl prior to gel analysis) were subjected to Western blot with an anti‐
hsp90 α antibody. TGF α‐induced decrease in the intracellular hsp α versus the
untreated cells was averaged based on results of four independent experiments. n=4,
p < 0.05.

88
Cell % w/
the image
91
79
84
77
TGF α
-
+15’
+30’
+60’
Cell % w/
the image
86
85
89
76
PDGF
-
+15’
+30’
+60’
a
b
e
a’
b’
c’
d’
e’
A.
d
c
HK DF
-
DFs HKs
C d
a b
B.
a
TGF α (30’) TGF α (30’)
-
Cell %
w/ the image: 96 ± 4.5 81 ± 9.7 90 ± 6.4 88 ± 5.3
C.
TGF α: - +
(equal volum e)
TGF α: - +
extracellular intracellular
1 2
hsp90 α hsp90 α
β-actin
c.
a.
b.
%: 100 72

89
Fig. 3‐6. Detailed Kinetics of TGF α stimulated membrane translocation of hsp90 α
in HKs. HKs, cultured on collagen‐coated coverslips and serum‐starved overnight,
were either untreated or treated with an optimal concentration of TGF α (20 ng/ml)
for the indicated periods of time. Cells were fixed and immunostained with a
monoclonal antibody against human hsp90 α. The results were visualized by
fluorescence‐conjugated secondary antibody. 80‐120 cells per condition were
randomly selected and analyzed. Under each condition, the representative image
plus the percentage of the cells that shared the similar image over the total number
of the cells examined are shown.
a b cd
e
f
g
h
TGF α: - 2’ 5’ 15’
TGF α: 30’ 60’ 90’ 120’

90

We confirmed the observation by constructing and introducing GFP (green
fluorescent protein)‐linked wt and ATP‐binding and ATPase mutants of human
hsp90 α fusion genes. Results of the experiments clearly showed that TGF α
stimulation increased the amount of extracellular GFP‐hsp90 α. Mutations in ATP‐
binding and ATPase domains of hsp90 α decreased the efficiency of membrane
translocation in response to TGF α (Fig. 3‐7A). Furthermore, TGF α stimulation
caused a time‐dependent secretion of the GFP‐hsp90 α fusion protein (Fig. 3‐7B).

Finally, to investigate the mechanism by which hsp90 α was exported, we asked the
question of whether secretion of hsp90 α is mediated by the classical endoplasmic
reticulum/Golgi transport pathway or via the non‐conventional “exosome traffic
pathway”. We took advantage of two inhibitors, Brefeldin A (BFA) and dimethyl
amiloride (DMA), which block classical‐ and exosome‐mediated protein secretion,
respectively (Savina et al. 2003; Lancaster and Febbraio 2005). In addition, confocal
microscopy was used to analyze hsp90 α membrane translocation at a specific layer
of the cells. As shown in Fig. 3‐8A, TGF α stimulation caused clear membrane
translocation of hsp90 α (panel f vs. a). DMA inhibited the membrane translocation
in a dose‐dependent fashion (panels g to i vs. panel f). In contrast, BFA even at its

91
reported high inhibitory concentration showed little effect (panel j vs. panel f).
Vehicle alone treated cells were included as positive controls (panel f vs. panel a).
Furthermore, similar inhibition of TGF α‐stimulated secretion of hsp90 α by of DMA
was observed. As shown in Fig. 3‐8B, DMA (lane 6), but not BFA (lane 4) almost
completely blocked hsp90 α secretion into the extracellular environment by HKCs in
response to TGF α (panel a, lane 6 vs. lanes 2 and 4). As expected, as shown in Fig. 3‐
8B, DMA blocked secretion of hsp70, another well‐known exosome‐mediated
secretion (panel b, lane 6) (Clayton et al. 2005; Lancaster and Febbraio 2005), and
BFA blocked the classical endoplasmic reticulum/Golgi transport pathway‐mediated
secretion of MMP10 (panel c, lane 4)). These data indicated that TGF α stimulates
secretion of hsp90 α in HKCs via the exosome pathway.

92
Fig. 3‐7. ATPase activity is required for hsp90 α secretion. Four GFP hsp90 α fusion
genes, including GFP‐hsp‐90 α‐wt, GFP‐hsp90 α‐E47A, GFP‐hsp90 α‐E47D and GFP‐
hsp90 α‐D93N, were constructed into lentiviral vector pRRLsin‐CMV and were used
to individually infect HKCs (Fan et al, 2005). (A) In the absence of TGF α stimulation,
both wt and the mutants of GFP‐hsp90 α spread the entire cells (panels a, e, i and m).
TGF α stimulation caused cell surface relocation of GFP‐hsp90 α‐wt in a time‐
dependent fashion (panels b‐d, pointed by arrows). TGF α also caused a clear
relocation of the GFP‐D93N mutant, but the percentage of the cells showing the
relocation, was decreased to one half (panels f, g, h). The percentage of the cells,
which showed TGF α‐stimulated GFP‐E47D relocation, was 70‐80% (panels j, k, l, but,
this number was reduced to 20‐30% for the cells expressing E47A mutant (panels n,
o, p). These data suggested that the ATP‐binding function and ATPase activity of
hsp90 α are required for efficiency of its membrane translocation in response to
TGF α. (B) The conditioned media (CM) of un‐stimulated or TGF α‐stimulated cells
expressing GFP‐vector alone or GFP‐hsp90 α‐wt were immunoblotted with an anti‐
GFP antibody. No secretion of the control GFP proteins were detected (lanes 1 and
2). A small amount of basal level of GFP‐hsp90 α was detected (lanes 3 and 5).
However, TGF α stimulation significantly increased the amount of extracellular GFP‐
hsp90 α (lanes 4 and 6).

93
TGF α: - 15’ 30’ 60’
D93N
E47D
wt
a b
c
d
e f
g
h
i j
k l
91%
86% 92% 90%
97% 44%
53% 67%
76% 92% 71% 80%
o p
E47A
31% 99% 27% 23%
m n
o
p
A.
B.
TGF α: - 8h - 2h - 8h
hsp90 α
1 2 3 4 5 6
GFP GFP-hsp90-wt
Blot w/
Anti-GFP

94
Fig. 3‐8 The exosome pathway mediates TGF α‐induced membrane translocation
and secretion of hsp90 α in HKCs. (A) HKCs, were either untreated or treated with
TGF α (20 ng/ml) in the absence or presence of the indicated concentrations of DMA
or BFA for 15 min. The experiments and data quantitation were carried out as
described for Figure 5A. (B) Conditioned media (concentrated from 4ml to 50 µl
prior to gel analysis) of the cells, either untreated or treated with 20ng/ml TGF α in
the absence or presence of DMA or BFA and for four hours, were subjected to
Western blot with anti‐hsp90 α antibodies. Equal loading of the samples is controlled
at three levels: 1) equal numbers of the cells in different dishes of culture, 2) equal
volumes of the media added to the dishes and 3) further calibration of the volumes
of loading samples against the cell numbers at the end of the incubation.
Control 0.6 nM 3 nM 15nM
10 µg/ml
96.6% 94.6% 95.3% 92.6%
94%
76.2% 71.2% 58.8% 73.8%
89%
-
+TGF
DMA BFA
A.
TGF: - + - + - +
Hsp90 α
(secreted)
DMA BFA
control
1 2 3 4 5 6
B.
f g h i j
a b c d e
Hsp70
(secreted)
MMP10
(secreted)
a. b.
c.
1 2 3 4 5 6 1 2 3 4 5 6
TGF: - + - + - +
DMA BFA
control
TGF: - + - + - +
DMA BFA
control

95

The ATP‐binding, ATPase and functional domains of Hsp90 α in promoting cell
migration. Hsp90 α is an ATP‐binding protein and has an intrinsic ATPase activity,
both of which are required for its intracellular chaperon function in the so‐called
hsp90 chaperone cycle (Obermann et al. 1998; Young et al. 2001; Richter and Buchner
2001a). Therefore, we asked whether these activities are also required for hsp90 α’s
extracellular function to promote cell migration. His‐tagged recombinant proteins of
hsp90 α‐wt, hsp90 α‐E47A, hsp90 α‐E47D and hsp90 α‐D93N proteins were produced
in bacteria, isolated by Ni‐NTA column and FPLC‐purified (Fig. 3‐9A). To verify the
effects of the mutations, we subjected the purified proteins to the in vitro ATPase
assay. As shown in Fig. 3‐9B, hsp90 α‐wt showed an expected ATPase activity (bar 2)
over BSA control (bar 1). Hsp90 α‐E47D showed about a half of the ATPase activity
of the hsp90 α‐wt (bar 3). Neither hsp90 α‐E47A nor hsp90 α‐D93N had any
significant ATPase activity (bars 4 and 5) over the BSA control (bar 1). When these
proteins were tested for their effects on HKC migration, however, we surprisingly
found that none of the mutations affected the proteins’ ability to stimulate HKC
migration. As shown in Fig. 3‐9C, the three mutant hsp90 α proteins (bars 6‐14)
showed a dose‐dependent stimulation of HKC migration, just like the wild type
hsp90 α (bars 3‐5) or TGF α (bar 2) stimulation. These results indicate that, unlike its

96
intracellular action as an ATP‐dependent chaperon, extracellular hsp90 α promotes
cell motility without the need to bind any co‐factors or ATP.

What part(s) of Hsp90 α promotes migration, then? Hsp90 α is composed of an N’‐
terminal domain, a charged sequence, a middle domain and a C’‐terminal domain
(Fig. 3‐10). To narrow down the domain that carries out the pro‐motility function of
hsp90 α, we constructed each of the individual domains, expressed it in bacteria as
His‐tagged proteins and purified them by FPLC (Methods). Equal molarities of the
proteins were then tested for pro‐motility effects on HKCs. As shown in Fig. 3‐10,
the full‐length (WT) hsp90 α showed a remarkable pro‐motility activity, in
comparison to the control medium. The middle domain plus the charged sequence
(M‐1) show a similar degree of activity as the WT hsp90 α. However, the middle
domain lacking the charged sequence showed a significantly decreased activity (M‐
2), although the charged sequence plus the entire N’‐terminal domain (N’) showed
no stimulating activity. The two C’‐terminal domains (C’‐1 and C’‐2) both showed a
moderate pro‐motility activity. Therefore, hsp90 α promotes HKC migration mainly
through its middle plus the charged sequence, consistent with previous observation
that middle domain plus the charged sequence is exposed on the surface of hsp90 α
molecule (Nemoto et al. 1997).

97
Fig. 3‐9 The extracellular hsp90 α promotes cell migration independently of ATP
binding or ATPase activity. (A) Purified wild type and mutant human recombinant
hsp90 α proteins (wt, E47D, E47A and D93N) were subjected to an SDS‐PAGE
analysis together with a commercial hsp90 α (1 µg) and a series known amounts (1‐
10 µg) of BSA. The proteins were visualized by Coomassie blue staining. (B) The
recombinant proteins were subjected to in vitro ATPase assays (Methods), using BSA
as the baseline control. (C) The wt, E47D, E47A and D93N mutant hsp90 α proteins
showed similar dose‐dependent pro‐motility effect on HKCs. Only the migration
indices were shown, n=5, p < 0.05.

98
Fig. 3‐10 Hsp90 α promotes HKC migration mainly through its middle domain.
Similar to the procedures in Figure 9, purified wild type and the indicated fragments
of human hsp90 α proteins were: (A) verified in an SDS‐PAGE and Comassie blue
staining and (B) subjected to colloidal gold migration assays (0.1 µM each), in
comparison to control serum‐free medium. Only the migration index (MI) is shown.
n=3.
WT
N’
M-1
C’-1
1 272
236
629
272 629
732 617
732 635
MI(%)
19.26 ± 2.74
5.79 ± 1.32
19.73 ± 3.14
13.81 ± 3.12
12.23 ± 2.94
13.42 ± 2.52
P < 0.01
P = 0.18
P < 0.01
P < 0.01
P < 0.01
P < 0.01
Hsp90 α
M-2
Control Medium
C’-2
4.95 ± 1.46
236 272 629 1 723 ATP-binding
Charged
sequence
Middle domain C’ domain
WT N’ M-1 C’-1 M-2 C’-2
BSA ( µg)
5 10 20
118
100
54
38
29
20
7
kDa
1 2 3 4 5 6 7 8 9
Comassie blue stain
A.
B.

99

LRP1/CD91 is the receptor for extracellular hsp90 α to promote skin cell migration.
To investigate how hsp90 α promotes skin cell migration from outside of the cells,
we focused on a reported common receptor for heat shock proteins, LRP‐1 (LDL
receptor‐related protein)/ α2‐macroglobulin receptor/CD91 (Basu et al. 2001). CD91
consists of a 515‐kDa extracellular subunit and a membrane‐anchoring 85‐kDa
subunit, resulting from proteolytic products of a common 600‐kDa precursor (Herz
et al. 1990). We undertook two independent approaches to study the role of CD91.
First, we designed and delivered two shRNAs targeted against different sites on the
CD91 mRNA by lentiviral infection to HKCs. As shown in Fig. 3‐11A, CD91‐RNAi‐1
and CD91‐RNAi‐2 down‐regulated almost completely the endogenous CD91 protein
(lane 2 and 3 vs. lane 1). When cells infected with lentivirus encoding a control non‐
specific shRNA were subjected to migration assays, as shown in Fig. 3‐11D, hsp90 α
strongly stimulated HKC migration (bar 2). However, down‐regulation of CD91
completely blocked HKC cell migration in response to hsp90 α (bars 4 and 6 vs. bar
2). Second, we used a monoclonal neutralizing antibody to block the cell surface
CD91, as schematically shown in Fig. 3‐11E, and similar results were obtained from
the experiments. As shown in Fig. 3‐11F, hsp90 α strongly stimulated HKC
migration (bars 2 vs. bar 1). Addition of a control IgG showed little effect (bar 3).

100
However, addition of even 3 µg/ml of anti‐CD91 antibodies completely blocked
hsp90 α‐induced HKC migration in a dose‐dependent manner (bars 4‐6 vs. bar 2).

We confirmed the importance of CD91 in hsp90 α signaling using other human skin
cell types. In comparison to HKCs, DFs and HDMECs are not only CD91‐positive
but express relatively even higher CD91 (Fig. 3‐11B, lanes 3 and 4 vs. lane 2).
Dendritic cells were included as a positive control (lane 1). When CD91 in DFs was
down‐regulated by siRNA (Fig. 3‐11C) and tested in migration assays, we found
that hsp90 α was no longer able to stimulate migration of DFs (Fig. 3‐11G, bar 4). In
contrast, hsp90 α still stimulated migration of the vector alone‐infected DFs (bar 2).

Finally, we studied whether hsp90 α is able to bind CD91. Using GST‐hsp90 α pull‐
down assays, as shown in Fig. 3‐11H, GST alone was unable to bind any CD91 from
the lysates of HKCs (lane 1). However, GST‐hsp90 α pulled down CD91 in a dose‐
dependent manner (lanes 2‐5). Using His‐tagged domains of hsp90 α on beads, as
shown in Fig. 3‐11I, we observed that, among the N’‐terminal ATP binding and
ATPase domain plus charged sequence, the middle domain plus charged sequence
and the C’‐terminal domain, the middle domain plus charged sequence of hsp90 α
(lane 3) bound strongest to CD91 (lanes 4 and 2). Full‐length hsp90α (lane 5), a

101
known CD91‐binding protein, and empty beads (lane 1) were included as positive
and negative controls, respectively. Taken together, these findings indicated that
CD91 is a receptor for secreted hsp90 α, whose extracellular function is to promote
migration of all CD91
+
skin cell types during wound healing.

Extracellular hsp90 α overrides TGF β inhibition to promote dermal cell migration.
In human skin, the main body of cells in epidermis is HKCs (>90%) and the two
main types of cells in the dermis are DFs and HDMECs. HKC migration is initiated
immediately following wounding. The dermal cells, however, move into the wound
approximately four days after the injury, where they deposit ECMs and remodel the
wound after keratinocytes close the wound (Singer and Clark 1999). It has long
remained as a puzzle how the dermal cells could have migrated into the wound,
since the high concentration of TGF β (30‐50ng/ml) present in human serum (the
main soluble environment in the wound) would completely block their migration
(Roberts et al. 1980). In contrast, HKC migration is insensitive to TGF β, due to lower
expression level of the type II TGF β receptor in the cells (Bandyopadhyay et al. 2006).
Thus, a clinically relevant question was whether or not the hsp90 α‐induced
migration of dermal cells is sensitive to TGF β. Consistent with our previous
obseravtion, as shown in Fig. 3‐12A, HKC‐CM and recombinant hsp90 α promoted

102
migration of all three cell types (bars 7 to 12), similarly to their growth factor
stimulation (bars 4 to 6). Based on these data, we conclude that hsp90 α is a general
pro‐motility factor for all the major skin cell types involved in wound healing. As
shown in Fig. 3‐12B, human plasma (HP, containing little TGF β) stimulated DF
migration (bar 2 vs. bar 1), whereas human serum (HS, with high concentrations of
TGF β) completely blocked DF migration (bar 3), consistent with our previous
findings (Bandyopadhyay et al. 2006). As expected, PDGF‐BB and hsp90 α equally
stimulated DF migration (bar 4 vs. bar 5). However, in the co‐presence of TGF β, HP‐
and PDGF‐BB‐induced DF migration was completely shut down (bars 6 and 8).
Intriguingly, TGF β was unable to inhibit hsp90 α‐stimulated DF migration (bar 9 vs.
bar 5). The reverse was also true. When purified hsp90 α was used to “rescue” the
inhibition of DF migration from HS or TGF β inhibition, the addition of hsp90 α
overrode the blockage of HS or TGF β3 on PDGF‐BB‐induced DF migration (bars 11
and 12 vs. bars 3 and 7). Therefore, the secreted hsp90 α by HKCs may subsequently
serve as a major pro‐motility factor in the wound bed.

103
Fig. 3‐11 CD91 receptor mediates hsp90 α signaling to promote cell migration. (A)
Lysates of HKCs infected with lentivirus encoding either a control shRNA (LacZ‐
siRNA) or two shRNAs against CD91 (CD91‐RNAi‐1 and CD91‐RNAi‐2) were
analyzed by Western blot with an anti‐CD91 antibody. (B) CD91 expression in HKCs,
DFs and HDMECs, in comparison to dendritic cells, was shown by Western blots. (C)
Lysates of DFs infected with lentivirus encoding either a control shRNA (LacZ‐
shRNA) or the shRNAs against CD91 were analyzed by Western blot with an anti‐
CD91 antibody. (D) The effects of down‐regulation of CD91 by two distinct shRNAs
on HKC migration in response to hsp90 α (10 µg/ml) were analyzed in colloidal gold
migration assays. Bars with * are statistically significant over serum‐free controls, n
=4, p < 0.03. (E) A schematic presentation of how neutralizing anti‐CD91 antibodies
block the cell surface CD91 to prevent hsp90 α binding. (F) Effect of anti‐CD91
blocking antibody on HKC migration in response to hsp90 α. HKCs on collagen in
colloidal gold migration assays were pre‐incubated with increasing concentrations
of an anti‐CD91 neutralizing antibody for 30 min (and continued presence
throughout the assays) or control IgG prior to addition of hsp90 α. (G) The effects of
RNAi down‐regulation of CD91 on DF migration in response to hsp90 α (10 µg/ml)
were analyzed in colloidal gold migration assays. Bars with * are statistically
significant over serum‐free controls, n=3, p < 0.05. (H) Lysates of HKCs were

104
incubated with indicated amounts of GST‐hsp90 α fusion proteins or GST alone on
beads. Beads‐bound proteins were dissociated and analyzed by Western blot with
an anti‐CD91 antibody. (I) Lysates of HKCs were incubated with 20 µg each of three
His‐tagged domains of hsp90 α on Ni
+
beads. Full‐length hsp90α (FL) was used as
positive control. The bound proteins were analyzed by Western blot with an anti‐
CD91 antibody. n=3, p < 0.05.

105

106
Fig. 3‐12 Extracellular hsp90 α bypasses the inhibitory effect of TGF β on human
dermal cell migration. Primary human dermal fibroblasts (DFs) were serum‐starved
overnight and subjected to colloidal gold migration assays under either untreated or
treated with the various reagents as indicated: human plasma (HP, 3%), human
serum (HS, 3%), PDGF‐BB (15ng/ml) or hsp90 α (0.1 µM) in the absence or presence
of TGF β3 (0.4 ng/ml), the critical inhibitor of dermal cell migration present in human
serum (2). Only the migration index (MI) is shown, n=3, p < 0.05. Bars with * are
statistically significant over serum‐free control.
10
20
30
1 2 3 4 5 6 7 8 9 10 11 12
*
*
*
Migration Index (%)
*
+TGF β
*
*
*
hsp90 α
Migration Index (%)
10
20
HDMEC HK DF
30
serum-free
HK-CM
(10x)
1 2 3 4 5 6 7 8 9 10 11 12
hsp90 α
(0.1 µM)
B. A.
DFs

107

DISCUSSION
In the current study, we provide evidence that TGF α triggers HKCs to secrete
hsp90 α and the extracellularly‐located hsp90 α in turn acts as a pro‐motility factor on
all major human skin cell types. Among the major cell types in skin, only HKCs
secrete hsp90 α. Mechanistically, TGF α triggers a rapid membrane translocation of
hsp90 α and its secretion to extracellular environment via the exosome pathway.
Among the four domains of hsp90 α, only the middle domain plus the charged
sequence is able to fully duplicate the pro‐motility effect of the full‐length hsp90 α,
independently of its ATP‐binding and ATPase functions. Furthermore, extracellular
hsp90 α promotes migration of both epidermal and dermal cells though the cell
surface receptor LRP‐1/CD91. Physiologically significant, this study showed that the
extracellular hsp90 α‐induced skin cell migration could no longer be inhibited by
TGF β, which is abundantly present in skin wounds. TGF β potently inhibits
canonical growth factor‐stimulated migration of DFs and HDMECs, and, therefore,
our finding provides an explanation for the first time how the dermal cells move
into the wound bed in the presence of TGF β. A schematic representation of these
findings is shown in Fig. 3‐13. In support of this model in vivo, we recently reported
that topical application of purified hsp90 α enhanced skin wound healing in nude

108
mice (Li et al. 2007). Taken together, we propose that following wounding increased
TGF α from the serum stimulates HKCs to secrete hsp90 α. After reaching its
threshold, the extracellular hsp90 α not only promotes HKC migration for re‐
epithelialization, but also recruits dermal cells to migrate into the wound for new
connective tissue formation, new blood vessel formation and wound remodeling.
Moreover, the hsp90 α’s TGF β‐insensitive signaling has a clear advantage over the
canonical growth factors to promote cell migration under the TGF β‐rich micro‐
environment of skin wounds. Considering the fact that TGF β potently inhibits
dermal cell migration, such as DFs and HDMECs (Bandyopadhyay et al. 2006),
extracellular hsp90 α may represent a bona fide physiological driving force for dermal
cell migration into the wound bed to participate in wound healing.

109
Fig. 3‐13 A schematic summary: the TGF α > hsp90 α secretion > skin cell migration
>wound healing model. Following skin injury, paracrine‐ or autocrine‐released
TGF α stimulates membrane translocation and secretion of the pre‐existing hsp90 α
proteins in HKCs. Secreted hsp90 α jumpstarts HKC migration, a critical event of the
re‐epithelialization process, by binding to the LRP‐1/CD91 receptor on the cell
surface. After the extracellular hsp90 α also defuses into and reached certain
concentration in the wound bed, it induces migration DFs and HDMECs from the
cut edge into the wound bed even under “hazard” conditions: no adequate
concentrations of ATP and the presence of pontent motility inhibitors, such as TGF β.
Thus, extracelluar hsp90 α may be a useful target for skin wound healing.

Integrin
ECM
hsp90 α
hsp90 α
TGF α
EGFR
hsp90 α
keratinocyte
LRP-1/
CD91
?
CD91
+
Dermal Cell Migration
Unknown
mechanism?
skin wound repair
Re-
epithelailization
Migration
Remodeling &
Neovascurization

110

Historically, the hsp90 family chaperons regulate folding, transport, maturation, and
degradation of a diverse, but selective, set of client proteins, in particular signaling
molecules (Richter and Buchner 2001a; Neckers and Ivy 2003). Recently,
extracellularly‐located hsp90 proteins have also been reported. Liao and colleagues
showed that oxidative stress causes sustained release of hsp90 α, which in turn
stimulates activation of ERK1/2 in rat vascular smooth muscle cells (Liao et al. 2000).
Jay and his colleagues identified hsp90 α, but not hsp90 β, in the conditioned media
of tumor cells (Eustace et al. 2004). Moreover, two groups showed that heat stress
causes increased secretion of hsp90 and hsp70 (Clayton et al. 2005; Lancaster and
Febbraio 2005). Yu and colleagues showed that γ radiation stimulates secretion of
hsp90 β into conditioned medium in a p53‐dependent manner (Yu et al. 2006). As
previously mentioned, these proteins have to undergo the exosome‐mediated
exocytosis that exports cellular proteins that lack a signal sequence. In contrast, these
proteins cannot be secreted via the conventional endoplasmic reticulum/Golgi
transport pathway. Exosomes, also called ‘intraluminal vesicles’ (ILVs), are
contained within the multivesicular bodies (MVBs), whose known function is to act
as an intermediate in the degradation of proteins internalized from the cell surface
or sorted from the trans Golgi organelle (Denzer et al. 2000; Thery et al. 2002).

111
However, MVBs and their exosomes can also fuse with the plasma membrane to
release the cargo proteins into the extracellular space (Stoorvogel et al. 2002; Fevrier
and Raposo 2004). In fact, all of the proteins that have been identified in exosomes
are located in the cell cytosol or endosomal compartments, never in endoplasmic
reticulum, Golgi apparatus, mitochondria or the nucleus (Stoorvogel et al. 2002).
Hsp90 α has been identified in exosomes (Hegmans et al. 2004; Clayton et al. 2005).
In this paper, our data links EGFR activation to the exosome pathway, leading to
membrane translocation and secretion of hsp90 α. How EGFR communicates with
the MVB protein trafficking pathway remains an unanswered question.

How does extracellular hsp90 α promote cell migration? Jay and his colleagues
showed that extracellular hsp90 α, but not hsp90 β, interacts with and acts as an
activator of the matrix metalloproteinase 2 (MMP2), and functional inhibition of
extracellular hsp90 α inhibited tumor invasion (Eustace et al. 2004). However, unlike
hsp90 α, which is able to duplicate the full pro‐motility effect of TGF α on HKCs,
there is no evidence that addition of any recombinant MMP alone can promote cell
motility. In addition, presence of MMP inhibitors, such as GM6001 or MMP
Inhibitor III, had little inhibitory effect on hsp90 α‐stimulated HKC migration
(unpublished results). Instead, we provided evidence that CD91 is the cell surface

112
receptor that mediates hsp90 α signaling to promote cell migration. CD91 is found in
monocytes, hepatocytes, fibroblasts and keratinocytes (Herz et al. 1990; Kristensen et
al. 1990; Strickland et al. 1990; Vandivier et al. 2002). We showed that the three major
skin cell types, HKC, DFs and HDMECs, were all CD91‐positive and all responded
to the pro‐motility effect of hsp90 α. Basu and colleagues have shown that released
hsp90 α binds to CD91, triggering macrophages to secrete cytokines and causing
dendritic cells to express antigen‐presenting and co‐stimulatory molecules (Basu et
al. 2000; Basu et al. 2001). Besides hsp90 α, other extracellular heat shock proteins
that bind CD91 include gp96, hsp60, hsp70 and calreticulin (CRT) (Basu et al. 2001;
Ogden et al. 2001; Habich et al. 2002; Vandivier et al. 2002). However, while CD91 is
a common receptor for heat shock proteins, not every CD91‐binding heat shock
protein was able to stimulate cell migration. In our hands, for instance, recombinant
CRT showed no detectable pro‐motility effect on HKCs, in comparison to hsp90 α or
TGF α (unpublished data). Therefore, the outcomes of CD91 activation may depend
upon binding to its 515‐kDa extracellular domain by specific heat shock proteins,
such as hsp70 for antigen presentation (Binder et al. 2004) and hsp90 α for cell
migration. Interestingly, RNAi down‐regulation of CD91 did not block TGF α‐
stimulated HKC migration in vitro (Cheng, CF and Li, W unpublished data),

113
supporting again our hypothesis that the HKC‐secreted hsp90 α is physiologically
important for recruiting the dermal cells into the wound.

Hsp90 α is composed of an N’‐terminal domain, a charged sequence (linker), a
middle domain and a C’‐terminal domain. Monoclonal antibody screening analyses
suggest that the middle and C’‐terminal domains are exposed at the surface of
hsp90 α protein (Nemoto et al. 1997). Crystal structures of the yeast and E. Coli hsp90
revealed that the middle domain forms a highly conserved surface loop, suggesting
a common role as a potential client protein binding site (Ali et al. 2006; Shiau et al.
2006). Our study showed that the middle domain plus the changed sequence
strongly binds CD91 and contains the full pro‐motility effect compared with the full‐
length hsp90 α. Confirmation of these findings in in vivo wound healing models
could further reveal the therapeutic potential of hsp90 α.

Finally, the observation that TGF α selectively stimulates secretion of hsp90 α in
HKCs, but not in DFs, was unexpected, especially since TGF α can promote DF
migration as well (Li et al. 2004a). This finding suggests a previously unrecognized
signaling mechanism: a single growth factor can use different mechanisms in two
geographically close skin cell types to execute the same cellular response, i.e.

114
motility. While the physiological implication of such TGF α signaling specificity
remains to be studied, it is conceivable that a possible “linker” for communication
between TGF α‐bound EGFR and hsp90 α‐containing exosomes is selectively present
in HKCs, but absent in the dermal cell types. The nature of such a linker could be
one of several possibilities, by which intracellular signaling molecules interpret
upstream signals. First, the signaling molecules can be differentially organized by
scaffolding proteins, resulting in a variety of combinations. Second, distinct
locations of the same signaling molecules can dictate specific biological responses.
Third, the strength and duration of a given signal/pathway can affect the outcomes
of the responses. The choice of these mechanisms and their responses may vary from
cell type to cell type (Pawson 2004). The bottom line is that differences in the cellular
context of HKCs and DFs are responsible for determining the signaling specificity of
TGF α.

115
Chapter 4: Conclusion
In a wound, there are several growth factors, cytokines, chemokines, MMPs, and
other secretory molecules. Certain molecules in the messy environment are capable
of stimulating skin cell migration and facilitating wound healing. Identifying such
factors would be valuable in treating unhealed wounds. In order to better
understand the molecular mechanism of wound healing, we decided to focus on the
rate limiting step of wound healing—re‐epithelialization, which is directly achieved
by lateral migration of keratinocytes around the wound area.

Previous findings in our lab showed that TGFα and insulin are the two most potent
growth factors present in human serum that stimulate HK migration (Li et al. 2006a).
In addition, TGFβ blocks HK proliferation without affecting migration
(Bandyopadhyay et al. 2006). We studied the early gene expression profile of
migrating HK by 1) adding TGFβ to block proliferation related genes and 2)
selecting common genes regulated by both TGFα and insulin to eliminate factor
specific genes. We identified 101 genes using these strategies. These genes are
potential HK migration related genes.

116
Among the genes identified, we selected three secretory factors MMP10, HB‐EGF,
and CXCL3. The significance of these three genes in an in vivo wound study was
discussed in chapter 2. These three factors also represent three types of important
molecules in wounds: MMPs, growth factors, and chemokines. MMPs are known to
degrade ECMs and are involved in initiating cell migration. Growth factors and
chemokines provide chemotactic signaling to stimulate cell migration in the
presence of ECM signaling. Therefore, studying these factors’ role in cell migration
is relevant to wound healing clinically.

Although we did not provide detailed mechanism of the importance of these
identified genes in HK migration, the results confirmed the importance of secreted
molecules in HK migration. This provides theoretical basis for the study in chapter 3.

In chapter 3, we discussed the secretion of hsp90α by TGFα stimulated HK through
exosome secretion. Several heat shock proteins are secreted through exosome
secretion by different cell types, but the mechanisms are yet clear (Clayton et al.
2005). To our best knowledge, we report for the first time that exosome secretion can
be stimulated by a growth factor. What is the mechanism for the TGFα stimulated
exosome secretion? Although there is no clear answer now, there are some clues.

117
Our data showed that the secretion happened within two minutes after TGFα
stimulation. Neither de novo protein synthesis nor gene transcription could happen
within such a short period of time. The process is more likely to be activated by
intracellular signaling pathways directly. Furthermore, the secretion was observed
only in keratinocyte but not fibroblast, which suggested that the process is cell type
specific. This phenomenon and the ability of hsp90α overriding TGFβ in stimulating
dermal cell migration support the role of hsp90α in wound, as shown in our model
in chapter 3.

Another important question is: how does hsp90α enter ILVs? During the biogenesis
of exosomes, there are four mechanisms for different proteins to enter exosome. First,
cytosolic proteins can be wrapped into ILVs by non specific engulfment. Second, the
proteins can be sorted into ILVs through the ESCRT machinery (Babst 2005). Third,
certain proteins anchor lipid raft‐like domains or interact with proteins that anchor
lipid raft‐like domains can be co‐sorted into ILVs by lipid affinity (de Gassart et al.
2003). Finally, chaperone proteins, such as hsp70 and hsp90, can be driven into ILVs
by co‐sorting with their client proteins (Geminard et al. 2004). In our study, we
showed that the hsp90α mutant without ATPase activity fail to translocate to cell

118
membrane in response to TGFα. Since ATPase activity is required for the chaperone
functions of hsp70 and hsp90, our results are consistent with this model.

CD91/LRP1 is known to bind to hsp90 (Basu et al. 2001). In our study, we further
showed that CD91/LRP1 is the functional receptor for the motogenic effect of hsp90α.
However, due to the huge size (600KDa) and multiple ligand binding domains, it is
very hard to study CD91 by “rescue” experiments. To overcome this problem,
several groups dissect the ligand binding domains of CD91/LRP1. Herz’s group
initiated the strategy by generating a chimeric construct that contains the SRP, the
region 2 or region 4 of CD91/LRP α subunits, and the β subunit of CD91/LRP1.They
showed that tPA/PAI‐1 binds to region 2 of CD91/LRP1, whereas RAP binds both
(Willnow et al. 1994). Bu’s group constructed simplified “mini‐receptors” with the
same strategy for all four ligand binding domains of CD91/LRP1. They removed the
EGF repeats in the mini‐receptors to reduce the size of these constructs but still
maintained their functional activities (Obermoeller‐McCormick et al. 2001).
Therefore, these mini‐receptors are easier to handle and are potentially helpful in
studying the detailed mechanism of hsp90α induced skin cell migration.

119
CD91/LRP1 is essential for hsp90α stimulated skin cell migration and is physically
bound to hsp90α, but it is not clear whether CD91/LRP1 alone is enough to transmit
the motogenic signal of hsp90α. Two members in LRP family, LRP5 and LRP6, form
co‐receptors with Frizzled (Fz) in Wnt signaling pathway. CD91/LRP1 itself can also
interact with calreticulin (CRT) in several different events. CRT can exist in
extracellular space as ligand for CD91/LRP1 or form co‐receptor with CD91/LRP1 to
bind other ligands to regulate inflammatory response, help remove apoptotic cells,
and stimulate dermal cells migration (Basu et al. 2001);(Gardai et al. 2003);(Ogden et
al. 2001; Orr et al. 2003). Furthermore, surface CRT on apoptotic cells can trans‐
activate CD91/LRP1 on engulfing cells for apoptotic cells removal (Gardai et al.
2005).

CRT is a multiple functional protein and broadly located in the cells. It is found in
ER and identified as a Ca
2+
binding protein that regulates Ca
2+
storage in ER
(Ostwald and MacLennan 1974);(Baksh and Michalak 1991). It is known as an
intracellular chaperone protein responsible for proper folding and quality control of
newly synthesized glycoproteins (Spiro et al. 1996). In addition to the intracellular
roles, CRT is also found on cell surface. The mechanism for the translocation of CRT
to cell surface is yet clear. Cell surface CRT is known as receptor for C1q, which is

120
involved in classical complement pathway. When expressed on cell surface, CRT
does not have a transmembrane domain. Therefore, CD91/LRP1 was suggested and
reported as a co‐receptor with CRT (Ogden et al. 2001).

Is CRT important for skin cell migration stimulated by hsp90α? Does CRT form co‐
receptor with CD91/LRP1 and mediate the motogenic activity of hsp90α? Our data
showed that extracellular CRT alone has no effect on skin cell migration. Since CRT
is a multiple functional protein both in the cell and on cell surface, it is not feasible to
knock down the expression of endogenous CRT and study the role of CRT in skin
cell migration. Currently, there is no direct evidence to prove the involvement of
CRT in hsp90α stimulated skin cell migration.

The pro‐motility activity of hsp90α is not universal. Our preliminary data showed
that some cells, such as melanocyte and melanoma cells, express CD91/LRP1 but do
not migrate in response to hsp90α. In addition, these cells expressed similar amount
of CRT on cell surface. What is the mechanism for this cell type specific response?
The involvement of endocytotic process might provide an explanation for this
phenomenon. When bound to certain ligands, CD91/LRP1 can internalize these
ligands. Urokinase‐type plasminogen activator (uPA) is one of these ligands.

121
Another member in the LDLR family, LRP1B, can bind to uPA as efficient as
CD91/LRP1 but internalize the ligand much less efficiently. Consequently, the
function of uPA is determined by the relative expression level of CD91/LRP1 and
LRP1B (Li et al. 2002). The result suggests that LRP1B can block uPA signaling by
competing with CD91/LRP1. The binding ability of LRP1B to other CD91/LRP1
ligands is unclear. However, LRP1B and CD91/LRP1 show 59% similarity in cDNA
sequences. It is possible that LRP1B is another cell surface molecule that is involved
in controling hsp90α signaling. Comparing the expression level of LRP1B in HK, FB,
MC, and M24 cells would provide some hint for the involvement of LRP1B in
hsp90α stimulated skin cell migration.

Cell migration is a repeated cycle of cytoskeletal‐mediated process, which can be
divided into the following three stages: membrane protrusion, cell body
translocation, and rear detachment. The protrusion in the front of the cell forms
filopodia and lamellopodia. The process requires actin filament polymerization and
structural organization. The translocation of the cell body is driven by the
interaction between myosin and actin filament. The rear detachment that release
adhesions at the rear of the cell depends on actin filament contraction
(Lauffenburger and Horwitz 1996). Therefore, actin filament is strongly involved in

122
cell migration. In our preliminary data, we observed actin filament re‐organization
in HK and DF after hsp90α stimulation. The result further confirmed the role of
hsp90α in skin cell migration.

Wound healing is a significant problem for more than 6 million people in the United
States. The mechanism, skin cell migration, relies on the responses of individual cells
to the microenvironment. There are a lot of factors in the wound area which
contribute to the wound healing processes by different mechanisms. We identified
hsp90α secreted by TGFα primed HKs through exosome secretion. In addition,
hsp90α binds to downstream receptor CD91/LRP1 as a novel mechanism in
stimulating skin cell migration. Most importantly, dermal cell migration stimulated
by hsp90α overrides the inhibition effect of TGFβ, which was previously shown to
inhibit growth factor stimulated dermal cell migration. Topical application of
hsp90α on wounds in mice model provides a potential novel treatment for unhealed
wounds (Li et al. 2007).

Cell migration is an important process not only for skin wound repair, but for
developmental biology, neuron biology, tumor metastasis, angiogenesis, and many
other fields studied by basic and clinical scientists. Although a lot of questions

123
remain to be answered to clarify the role of extracellular hsp90α in other events,
identifying the motogenic activity of extracellular hsp90α in skin cells potentially
provides clues for the mechanisms of cell migration in other cell types.

124
References
Ali, J. A., A. P. Jackson, et al. (1993). "The 43-kilodalton N-terminal fragment of the DNA
gyrase B protein hydrolyzes ATP and binds coumarin drugs." Biochemistry 32(10):
2717-24.

Ali, M. M., S. M. Roe, et al. (2006). "Crystal structure of an Hsp90-nucleotide-p23/Sba1
closed chaperone complex." Nature 440(7087): 1013-7.

Babst, M. (2005). "A protein's final ESCRT." Traffic 6(1): 2-9.

Baksh, S. and M. Michalak (1991). "Expression of calreticulin in Escherichia coli and
identification of its Ca2+ binding domains." J Biol Chem 266(32): 21458-65.

Bandyopadhyay, B., J. Fan, et al. (2006). "A "traffic control" role for TGFbeta3:
orchestrating dermal and epidermal cell motility during wound healing." J Cell Biol
172(7): 1093-105.

Basu, S., R. J. Binder, et al. (2001). "CD91 is a common receptor for heat shock proteins
gp96, hsp90, hsp70, and calreticulin." Immunity 14(3): 303-13.

Basu, S., R. J. Binder, et al. (2000). "Necrotic but not apoptotic cell death releases heat shock
proteins, which deliver a partial maturation signal to dendritic cells and activate the
NF-kappa B pathway." Int Immunol 12(11): 1539-46.

Binder, R. J., R. Vatner, et al. (2004). "The heat-shock protein receptors: some answers and
more questions." Tissue Antigens 64(4): 442-51.

Boucher, P., M. Gotthardt, et al. (2003). "LRP: role in vascular wall integrity and protection
from atherosclerosis." Science 300(5617): 329-32.

Bu, G., J. Cam, et al. (2006). "LRP in amyloid-beta production and metabolism." Ann N Y
Acad Sci 1086: 35-53.

Carter, S. B. (1967). "Haptotaxis and the mechanism of cell motility." Nature 213(5073):
256-60.

Clayton, A., A. Turkes, et al. (2005). "Induction of heat shock proteins in B-cell exosomes." J
Cell Sci 118(Pt 16): 3631-8.

Cooper, L., C. Johnson, et al. (2005). "Wound healing and inflammation genes revealed by
array analysis of 'macrophageless' PU.1 null mice." Genome Biol 6(1): R5.

125
de Gassart, A., C. Geminard, et al. (2003). "Lipid raft-associated protein sorting in
exosomes." Blood 102(13): 4336-44.

de Larco, J. E. and G. J. Todaro (1978). "Growth factors from murine sarcoma virus-
transformed cells." Proc Natl Acad Sci U S A 75(8): 4001-5.

Denzer, K., M. J. Kleijmeer, et al. (2000). "Exosome: from internal vesicle of the
multivesicular body to intercellular signaling device." J Cell Sci 113 Pt 19: 3365-74.

Eisen, M. B., P. T. Spellman, et al. (1998). "Cluster analysis and display of genome-wide
expression patterns." Proc Natl Acad Sci U S A 95(25): 14863-8.

Eliceiri, B. P. (2001). "Integrin and growth factor receptor crosstalk." Circ Res 89(12): 1104-
10.

Eustace, B. K., T. Sakurai, et al. (2004). "Functional proteomic screens reveal an essential
extracellular role for hsp90 alpha in cancer cell invasiveness." Nat Cell Biol 6(6):
507-14.

Fan, J., S. Guan, et al. (2006). "PKCdelta clustering at the leading edge and mediating growth
factor-enhanced, but not ecm-initiated, dermal fibroblast migration." J Invest
Dermatol 126(6): 1233-43.

Fantl, W. J., D. E. Johnson, et al. (1993). "Signalling by receptor tyrosine kinases." Annu
Rev Biochem 62: 453-81.

Fevrier, B. and G. Raposo (2004). "Exosomes: endosomal-derived vesicles shipping
extracellular messages." Curr Opin Cell Biol 16(4): 415-21.

Fuchs, E. and C. Byrne (1994). "The epidermis: rising to the surface." Curr Opin Genet Dev
4(5): 725-36.

Gardai, S. J., K. A. McPhillips, et al. (2005). "Cell-surface calreticulin initiates clearance of
viable or apoptotic cells through trans-activation of LRP on the phagocyte." Cell
123(2): 321-34.

Gardai, S. J., Y. Q. Xiao, et al. (2003). "By binding SIRPalpha or calreticulin/CD91, lung
collectins act as dual function surveillance molecules to suppress or enhance
inflammation." Cell 115(1): 13-23.

Geminard, C., A. De Gassart, et al. (2004). "Degradation of AP2 during reticulocyte
maturation enhances binding of hsc70 and Alix to a common site on TFR for sorting
into exosomes." Traffic 5(3): 181-93.

126
Gething, M. J. and J. Sambrook (1992). "Protein folding in the cell." Nature 355(6355): 33-
45.

Gumbiner, B. M. (1996). "Cell adhesion: the molecular basis of tissue architecture and
morphogenesis." Cell 84(3): 345-57.

Habich, C., K. Baumgart, et al. (2002). "The receptor for heat shock protein 60 on
macrophages is saturable, specific, and distinct from receptors for other heat shock
proteins." J Immunol 168(2): 569-76.

Hegmans, J. P., M. P. Bard, et al. (2004). "Proteomic analysis of exosomes secreted by
human mesothelioma cells." Am J Pathol 164(5): 1807-15.

Herz, J. and H. H. Bock (2002). "Lipoprotein receptors in the nervous system." Annu Rev
Biochem 71: 405-34.

Herz, J., D. E. Clouthier, et al. (1992). "LDL receptor-related protein internalizes and
degrades uPA-PAI-1 complexes and is essential for embryo implantation." Cell 71(3):
411-21.

Herz, J., U. Hamann, et al. (1988). "Surface location and high affinity for calcium of a 500-
kd liver membrane protein closely related to the LDL-receptor suggest a
physiological role as lipoprotein receptor." Embo J 7(13): 4119-27.

Herz, J., R. C. Kowal, et al. (1990). "Proteolytic processing of the 600 kd low density
lipoprotein receptor-related protein (LRP) occurs in a trans-Golgi compartment."
Embo J 9(6): 1769-76.

Kamal, A., L. Thao, et al. (2003). "A high-affinity conformation of Hsp90 confers tumour
selectivity on Hsp90 inhibitors." Nature 425(6956): 407-10.

Kennedy, R. E., R. T. Kerns, et al. (2006). "SScore: an R package for detecting differential
gene expression without gene expression summaries." Bioinformatics 22(10): 1272-4.

Kikuchi, A., H. Yamamoto, et al. (2007). "Multiplicity of the interactions of Wnt proteins
and their receptors." Cell Signal 19(4): 659-71.

Kristensen, T., S. K. Moestrup, et al. (1990). "Evidence that the newly cloned low-density-
lipoprotein receptor related protein (LRP) is the alpha 2-macroglobulin receptor."
FEBS Lett 276(1-2): 151-5.

Lancaster, G. I. and M. A. Febbraio (2005). "Exosome-dependent trafficking of HSP70: a
novel secretory pathway for cellular stress proteins." J Biol Chem 280(24): 23349-55.

127
Lauffenburger, D. A. and A. F. Horwitz (1996). "Cell migration: a physically integrated
molecular process." Cell 84(3): 359-69.

Li, C. and W. Hung Wong (2001). "Model-based analysis of oligonucleotide arrays: model
validation, design issues and standard error application." Genome Biol 2(8):
RESEARCH0032.

Li, W., J. Fan, et al. (2004). "Mechanisms of human skin cell motility." Histol Histopathol
19(4): 1311-24.

Li, W., J. Fan, et al. (2004a). "Mechanisms of human skin cell motility." Histol Histopathol
19(4): 1311-24.

Li, W., G. Henry, et al. (2004). "Signals that initiate, augment, and provide directionality for
human keratinocyte motility." J Invest Dermatol 123(4): 622-33.

Li, W., G. Henry, et al. (2004c). "Signals that initiate, augment, and provide directionality for
human keratinocyte motility." J Invest Dermatol 123(4): 622-33.

Li, W., Y. Li, et al. (2007). "Extracellular heat shock protein-90alpha: linking hypoxia to skin
cell motility and wound healing." Embo J 26(5): 1221-33.

Li, Y., J. Fan, et al. (2006a). "Transforming growth factor-alpha: a major human serum factor
that promotes human keratinocyte migration." J Invest Dermatol 126(9): 2096-105.

Li, Y., J. M. Knisely, et al. (2002). "Low density lipoprotein (LDL) receptor-related protein
1B impairs urokinase receptor regeneration on the cell surface and inhibits cell
migration." J Biol Chem 277(44): 42366-71.

Liao, D. F., Z. G. Jin, et al. (2000). "Purification and identification of secreted oxidative
stress-induced factors from vascular smooth muscle cells." J Biol Chem 275(1): 189-
96.

Luetteke, N. C., T. H. Qiu, et al. (1993). "TGF alpha deficiency results in hair follicle and
eye abnormalities in targeted and waved-1 mice." Cell 73(2): 263-78.

Mambula, S. S. and S. K. Calderwood (2006). "Heat shock protein 70 is secreted from tumor
cells by a nonclassical pathway involving lysosomal endosomes." J Immunol 177(11):
7849-57.

Mann, G. B., K. J. Fowler, et al. (1993). "Mice with a null mutation of the TGF alpha gene
have abnormal skin architecture, wavy hair, and curly whiskers and often develop
corneal inflammation." Cell 73(2): 249-61.

128
Mariani, L., C. Beaudry, et al. (2001). "Glioma cell motility is associated with reduced
transcription of proapoptotic and proliferation genes: a cDNA microarray analysis." J
Neurooncol 53(2): 161-76.

Martin, P. (1997). "Wound healing--aiming for perfect skin regeneration." Science 276(5309):
75-81.

Meyer, E., J. Y. Vollmer, et al. (2005). "Matrix metalloproteinases 9 and 10 inhibit protein
kinase C-potentiated, p53-mediated apoptosis." Cancer Res 65(10): 4261-72.

Mignot, G., S. Roux, et al. (2006). "Prospects for exosomes in immunotherapy of cancer." J
Cell Mol Med 10(2): 376-88.

Mitchison, T. J. and L. P. Cramer (1996). "Actin-based cell motility and cell locomotion."
Cell 84(3): 371-9.

Multhoff, G. and L. E. Hightower (1996). "Cell surface expression of heat shock proteins and
the immune response." Cell Stress Chaperones 1(3): 167-76.

Nanney, L. B., J. P. Sundberg, et al. (1996). "Increased epidermal growth factor receptor in
fsn/fsn mice." J Invest Dermatol 106(6): 1169-74.

Neckers, L. and S. P. Ivy (2003). "Heat shock protein 90." Curr Opin Oncol 15(6): 419-24.

Nemoto, T., N. Sato, et al. (1997). "Domain structures and immunogenic regions of the 90-
kDa heat-shock protein (HSP90). Probing with a library of anti-HSP90 monoclonal
antibodies and limited proteolysis." J Biol Chem 272(42): 26179-87.

Obermann, W. M., H. Sondermann, et al. (1998). "In vivo function of Hsp90 is dependent on
ATP binding and ATP hydrolysis." J Cell Biol 143(4): 901-10.

Obermoeller-McCormick, L. M., Y. Li, et al. (2001). "Dissection of receptor folding and
ligand-binding property with functional minireceptors of LDL receptor-related
protein." J Cell Sci 114(Pt 5): 899-908.

Ogden, C. A., A. deCathelineau, et al. (2001). "C1q and mannose binding lectin engagement
of cell surface calreticulin and CD91 initiates macropinocytosis and uptake of
apoptotic cells." J Exp Med 194(6): 781-95.

Orr, A. W., C. E. Pedraza, et al. (2003). "Low density lipoprotein receptor-related protein is a
calreticulin coreceptor that signals focal adhesion disassembly." J Cell Biol 161(6):
1179-89.

Ostwald, T. J. and D. H. MacLennan (1974). "Isolation of a high affinity calcium-binding
protein from sarcoplasmic reticulum." J Biol Chem 249(3): 974-9.

129
Pawson, T. (2004). "Specificity in signal transduction: from phosphotyrosine-SH2 domain
interactions to complex cellular systems." Cell 116(2): 191-203.

Prodromou, C., S. M. Roe, et al. (1997). "Identification and structural characterization of the
ATP/ADP-binding site in the Hsp90 molecular chaperone." Cell 90(1): 65-75.

Qin, X. F., D. S. An, et al. (2003). "Inhibiting HIV-1 infection in human T cells by lentiviral-
mediated delivery of small interfering RNA against CCR5." Proc Natl Acad Sci U S
A 100(1): 183-8.

Reynolds, A., D. Leake, et al. (2004). "Rational siRNA design for RNA interference." Nat
Biotechnol 22(3): 326-30.

Richter, K. and J. Buchner (2001a). "Hsp90: chaperoning signal transduction." J Cell Physiol
188(3): 281-90.

Richter, K., P. Muschler, et al. (2001b). "Coordinated ATP hydrolysis by the Hsp90 dimer." J
Biol Chem 276(36): 33689-96.

Roberts, A. B., L. C. Lamb, et al. (1980). "Transforming growth factors: isolation of
polypeptides from virally and chemically transformed cells by acid/ethanol
extraction." Proc Natl Acad Sci U S A 77(6): 3494-8.

Sarret, Y., D. T. Woodley, et al. (1992). "Human keratinocyte locomotion: the effect of
selected cytokines." J Invest Dermatol 98(1): 12-6.

Savina, A., M. Furlan, et al. (2003). "Exosome release is regulated by a calcium-dependent
mechanism in K562 cells." J Biol Chem 278(22): 20083-90.

Schmitt, E., M. Gehrmann, et al. (2007). "Intracellular and extracellular functions of heat
shock proteins: repercussions in cancer therapy." J Leukoc Biol 81(1): 15-27.

Schwartz, M. A. and M. H. Ginsberg (2002). "Networks and crosstalk: integrin signalling
spreads." Nat Cell Biol 4(4): E65-8.

Shiau, A. K., S. F. Harris, et al. (2006). "Structural Analysis of E. coli hsp90 reveals dramatic
nucleotide-dependent conformational rearrangements." Cell 127(2): 329-40.

Sidera, K., M. Samiotaki, et al. (2004). "Involvement of cell surface HSP90 in cell migration
reveals a novel role in the developing nervous system." J Biol Chem 279(44): 45379-
88.

Singer, A. J. and R. A. Clark (1999). "Cutaneous wound healing." N Engl J Med 341(10):
738-46.

130
Spiro, R. G., Q. Zhu, et al. (1996). "Definition of the lectin-like properties of the molecular
chaperone, calreticulin, and demonstration of its copurification with
endomannosidase from rat liver Golgi." J Biol Chem 271(19): 11588-94.

Stellas, D., A. Karameris, et al. (2007). "Monoclonal antibody 4C5 immunostains human
melanomas and inhibits melanoma cell invasion and metastasis." Clin Cancer Res
13(6): 1831-8.

Stoorvogel, W., M. J. Kleijmeer, et al. (2002). "The biogenesis and functions of exosomes."
Traffic 3(5): 321-30.

Strickland, D. K., J. D. Ashcom, et al. (1990). "Sequence identity between the alpha 2-
macroglobulin receptor and low density lipoprotein receptor-related protein suggests
that this molecule is a multifunctional receptor." J Biol Chem 265(29): 17401-4.

Teixido, J. and J. Massague (1988). "Structural properties of a soluble bioactive precursor for
transforming growth factor-alpha." J Biol Chem 263(8): 3924-9.

Thery, C., L. Zitvogel, et al. (2002). "Exosomes: composition, biogenesis and function." Nat
Rev Immunol 2(8): 569-79.

Trommsdorff, M., M. Gotthardt, et al. (1999). "Reeler/Disabled-like disruption of neuronal
migration in knockout mice lacking the VLDL receptor and ApoE receptor 2." Cell
97(6): 689-701.

Vandivier, R. W., C. A. Ogden, et al. (2002). "Role of surfactant proteins A, D, and C1q in
the clearance of apoptotic cells in vivo and in vitro: calreticulin and CD91 as a
common collectin receptor complex." J Immunol 169(7): 3978-86.

Wang, W., S. Goswami, et al. (2004). "Identification and testing of a gene expression
signature of invasive carcinoma cells within primary mammary tumors." Cancer Res
64(23): 8585-94.

Willnow, T. E., S. A. Armstrong, et al. (1995). "Functional expression of low density
lipoprotein receptor-related protein is controlled by receptor-associated protein in
vivo." Proc Natl Acad Sci U S A 92(10): 4537-41.

Willnow, T. E., K. Orth, et al. (1994). "Molecular dissection of ligand binding sites on the
low density lipoprotein receptor-related protein." J Biol Chem 269(22): 15827-32.

Woodley, D. T., P. M. Bachmann, et al. (1988). "Laminin inhibits human keratinocyte
migration." J Cell Physiol 136(1): 140-6.

131
Yepes, M., M. Sandkvist, et al. (2003). "Tissue-type plasminogen activator induces opening
of the blood-brain barrier via the LDL receptor-related protein." J Clin Invest 112(10):
1533-40.

Young, J. C., I. Moarefi, et al. (2001). "Hsp90: a specialized but essential protein-folding
tool." J Cell Biol 154(2): 267-73.

Yu, X., S. L. Harris, et al. (2006). "The regulation of exosome secretion: a novel function of
the p53 protein." Cancer Res 66(9): 4795-801.

Yuan, B., R. Latek, et al. (2004). "siRNA Selection Server: an automated siRNA
oligonucleotide prediction server." Nucleic Acids Res 32(Web Server issue): W130-4.

Asset Metadata

Creator Cheng, Chieh-Fang (author)

Core Title Mechanisms of human skin cell migration and wound healing

School Keck School of Medicine

Degree Doctor of Philosophy

Degree Program Pathobiology

Publication Date 07/24/2008

Defense Date 06/19/2008

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

Tag OAI-PMH Harvest,skin cell migration,wound healing

Language English

Advisor Hofman, Florence M. (committee chair), Li, Wei (committee chair), Roy-Burman, Pradip (committee member), Stallcup, Michael R. (committee member), Stiles, Bangyan L. (committee member)

Creator Email chiehfac@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m1398

Unique identifier UC1144124

Identifier etd-Cheng-20080724 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-195316 (legacy record id),usctheses-m1398 (legacy record id)

Legacy Identifier etd-Cheng-20080724.pdf

Dmrecord 195316

Document Type Dissertation

Rights Cheng, Chieh-Fang

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email uscdl@usc.edu

Abstract (if available)

Abstract Skin cell migration is essential for skin wound healing. Steps for cell migration are often disrupted in non‐healing wounds, causing patient morbidity and even fatality. Currently‐available treatments are unsatisfactory. To identify novel wound‐healing targets, we took two approaches. First, we studied the migratory gene profiles in human keratinocytes (HKs). Second, we investigated secreted molecules from TGFalpha‐stimulated human keratinoytes, which contained a strong motogenic, but not mitogenic, activity. In the first study, the main challenge is to separate genes that are often simultaneously induced by pleiotropic signals of a given growth factor, including migration, proliferation and metabolism. Therefore, we designed the following steps. First, we took advantage of a unique response of HKs to TGF‐beta, which inhibits proliferation but not migration of the cells, to suppress selectively the proliferation signal‐responding genes. Second, we independently stimulated HKs with TGF‐alpha or insulin to identify the commonly regulated genes and eliminate TGF‐alpha‐ or insulin‐specific genes.