University of Southern California Dissertations and Theses

The mechanism of mammalian immunoglobulin class switch recombination: R-loop structures and activation-induced deaminase site preferences

(USC Thesis Other)

The mechanism of mammalian immunoglobulin class switch recombination: R-loop structures and activation-induced deaminase site preferences

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content THE MECHANISM OF MAMMALIAN IMMUNOGLOBULIN CLASS SWITCH
RECOMBINATION: R-LOOP STRUCTURES AND ACTIVATION-INDUCED
DEAMINASE SITE PREFERENCES

by

Feng-Ting Huang

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
(BIOCHEMISTRY AND MOLECULAR BIOLOGY)

December 2007

Copyright 2007 Feng-Ting Huang
ii
ACKNOWLEDGEMENTS
I would like to sincerely thank my mentor, Dr. Michael Lieber, for his support
and encouragement during these years. I am deeply influenced by his scientific
guidance and his passion to science. I am most fortunate to have him as my
mentor.
I would like to express my gratitude to Dr. Chih-Lin Hsieh. Her suggestions
and patient to me help me been through difficulties. I would like to give my
appreciation to Dr. Michael Stallcup, Dr. Ian Haworth, and Dr. Joe Hacia. Their
valuable advices broaden my view of my projects.
I would like to thank all lab members: Dr. Kefei Yu, Dr. Sathees Raghavan, Dr.
Yunmei Ma, Dr. Ryan Irvine, Dr. Shigeru Sasaki, Dr. Noriko Shimazaki, Dr.
Xiaoping Ciu, Haihui Lu, Deepankar Roy, Albert Tsai, Jiafeng Gu, and Go
Watanable. Their technical help and discussions are beneficial to my projects.
I would like to thank for my family and friends. Their mental supports are
inevitable for me to been through these years. Most gratefully, I would like to
dedicate this work to my sister and parents for their unconditional love.

Feng-Ting Huang
iii
TABLE OF CONTENTS
Acknowledgements……………………………………………………………………...ii

List of Tables…………………………………………………………………………….iv

List of Figures……………………………………………………………………………v

Abstract………………………………………………………………………………....viii

Chapter 1 Introduction…………………………………………………….1

Chapter 2 DNA substrate length and surrounding sequence
affect the activation-induced deaminase activity
at cytidine……………………………………………………..12

Chapter 3 Downstream boundary of chromosomal R-loops at
Murine switch regions……………………………………….31

Chapter 4 Sequence-dependence of chromosomal R-loops
at the immunoglobulin heavy chain Sµ class
switch region………………………………………………….70

Chapter 5 Conclusion remarks………………………………………..118

Bibliography…………………………………………………………………………..129
iv
List of Tables
Table 3-1. The frequency of WRC and AGCT sequence motifs………………......63

Table 4-1. Single-strandedness was destroyed after in vitro RNase H treatment.90

v
List of Figures
Fig. 2-1. AID protein and deamination assay…………………………………..16

Fig. 2-2. Time-dependent AID deamination…………………………………….18

Fig. 2-3. Substrate length effect on AID activity………………………………..21

Fig. 2-4. Site preference in the AID deamination………………………………24

Fig. 3-1. Fluorescence-activated cell sorter analysis of the fraction of
murine splenic surface positive IgG3 cells……………………………43

Fig. 3-2. Single-strandedness on the top strand of the murine Sγ3 region
In stimulated B cells……………………………………………………..45

Fig. 3-3. Single-strandedness on the bottom strand at the murine Sγ3 of
the stimulated B cells (with native primers)…………………………..48

Fig. 3-4. Single-strandedness on the top strand at the murine Sγ3 region
of stimulated B cells (with native primers)…………………………….49

Fig. 3-5. Single-strandedness on the top strand immediately downstream
of murine Sγ3 in stimulated B cells…………………………………….50

Fig. 3-6. The Sγ2b switch repeat alignment…………………………………….56

vi
Fig. 3-7. Single-strandedness on the top strand within and downstream
of the Sγ2b switch region in stimulated B cells……………………….57

Fig. 3-8. Plots of the downstream edge of the R-loops, the G density, and
the AGCT locations for murine Sγ3 through the initial portion of
the first constant exon…………………………………………………..60

Fig. 3-9. Plots of the downstream edge of the R-loops, the G density, and
the AGCT locations for murine Sγ2b through the initial portion of
the first constant exon…………………………………………………..61

Fig. 3-10. Thermodynamic prediction of nucleic acid stability of RNA:DNA
versus DNA:DNA within the murine Sγ3 and Cγ3 regions and
the Sγ2b and Cγ2b regions………………………………..……………67

Fig. 4-1. Diagrams of the immunoglobulin m locus in C57BL/6 and two
different Sµ deletion mice………………………………………………75

Fig. 4-2. Single-strandedness on the top strand at the murine core Sµ in
stimulated B cells in the C57BL/6 mouse……………………………..80

Fig. 4-3. R-loop molecules at the murine core Sµ in stimulated B cells in
the C57BL/6 mouse……………………………………………………..84

Fig. 4-4. Single-strandedness on the nontemplate strand upstream and
downstream of the murine core Sµ in stimulated B cells……………88

vii
Fig. 4-5. The R-loop molecules at the murine Iµ-Cµ region in
stimulated B cells from the core Sµ deletion (∆SµTR) mouse……...93

Fig. 4-6. The stem-loop structure of the loxP site in the middle of the
R-loop at the Iµ-Cµ region in the ∆SµTR mouse…………………….97

Fig. 4-7. DNA methylation analysis of the lox P sequence in the middle
of the Iµ-Cµ region of the ∆SµTR allele……………………..………..99

Fig. 4-8. Test for single-strandedness on the nontemplate strand at
murine Iµ-Cµ in the stimulated B cells from mice having
the larger deletion around Sµ..………………………………………102

Fig. 4-9. Plots of R-loop location, class switch recombination
breakpoints, AGCT sites, and G density for murine Sµ
through the initial portion of the first constant exon……………….105

Fig. 4-10. Plots of R-loop location, mutation frequency, AGCT sites,
and G density for the region upstream of the core Sµ……………107

Fig. 4-11. Two mechanisms by which endogenous RNase H activity
could generate single-stranded regions on the bottom
(C-rich) DNA strand in or around the switch regions……………...116

Fig. 5-1. The current model of the class switch recombination……………..128

viii
ABSTRACT
Class switch recombination (CSR) is the mechanism responsible for
changing constant domains of antibodies to produce different effector functions of
B cells. Somatic hypermutation (SHM) mutates the V(D)J segment to enhance
the antigen-binding affinity of antibodies. Activation induced deaminase (AID) is
required and responsible for targeting in both SHM and CSR. With DNA repair
systems, AID causes mutations in the V(D)J segment and double strand breaks in
switch regions for SHM and CSR, respectively. However, the targeting and
cleavage mechanism of AID is still not clear yet.
Here we report that the AID preference motif is 5’-WRCr-3’, which
corresponds to the SHM hotspot-5’WRC-3’. In addition, switch regions have
abundant 5’-WGCW-3’ sites, which are palindromic AID hotspots, and can lead to
DNA double strand breaks (DSB). For the DNA level, we study the chromosomal
R-loops at murine switch regions, Sγ3, Sγ2b, and Sµ. At day 2 of stimulation of
primary murine B cells, the frequency of R-loops is 1 out of 570 and 1 out of 25
alleles at murine Sγ3 and Sµ, respectively. The G density of the sequence is
crucial for the R-loop formation, because R-loops end in the region where
G-density declines. The GGGG cluster of the sequence may be important for
ix
R-loop initiation. R-loops are crucial for murine CSR efficiency. The larger Sµ
deletion mice have severely impaired CSR and, no R-loops are detected in these
mice. The core Sµ deletion mice, ∆SµTR mice, have reduced CSR, and
abundant R-loops are detected in these mice. Furthermore, R-loop initiation
sites are in the same zone where AID mutation starts to increase. In conclusion,
in this study, we provide the first structure-function analysis for the role of R-loops
in the targeting mechanism in CSR. Furthermore, we provide explanations for
special features of switch sequences: high G density on the non-template strand
(R-loop formation) and high density of WGCW sites (DSB sites).

1
CHAPTER 1: INTRODUCTION
The production of specific antibodies by mature B cells in mammals involves
three different types of Ig gene alterations-VDJ recombination, somatic
hypermutation (SHM), and class switch recombination (CSR). The first two
mechanisms are responsible for creating antibody diversity and specificity. The
third one-CSR produces different effector functions of immunoglobulins by the
rearrangement of constant region genes.
In contrast to VDJ recombination, mechanisms of SHM and CSR are not well
understood yet. They have more common characteristics with each other than
with VDJ recombination. Both mechanisms occur in antigen-stimulated B cells,
and require transcription through the entire region and require activation-induced
cytidine deaminase (AID).
Once a mammalian B cell is producing IgM immunoglobulin, it can undergo
isotype switching to produce IgG, IgA or IgE in a process that requires a DNA
recombination event called class switch recombination (Yu and Lieber 2003;
Chaudhuri and Alt 2004). The murine IgH constant region locus has eight
different constant genes. Each constant gene, except Cδ, is 5’ flanked by a
switch region composed of palindromes and repetitive sequences. (Igδ is
2
created by alternative splicing of a transcript that includes Igµ.) Class switch
recombination takes place between two switch regions, leading to the deletion of
the intervening DNA segment (Luby, Schrader et al. 2001; Stavnezer and
Amemiya 2004). There are several steps to complete CSR. First, Sµ and one
of the downstream switch regions are targeted by the CSR initiating enzyme, AID.
Double-strand breaks are generated in two switch regions and concurrently, the
switch region DNA ends synapse to each other, and are ligated. Later on, two
switch regions would rejoin together. The detailed molecular events of each step
will be described in the following paragraphs.
Targeting: region-specific recombination
Germline transcription is an important feature of CSR (Stavnezer, Radcliffe et
al. 1988; Chaudhuri and Alt 2004). Transcription through a particular S region is
required for B cells switching from IgM to the corresponding C
H
region (Coffman,
Lebman et al. 1993; Xu, Gorham et al. 1993; Bottaro, Lansford et al. 1994). In
mature B cells, the Sµ region is constitutively transcribed from the intronic
promoter, 5’ from the Sµ region; while the downstream S region is inductively
transcribed by the cytokine-responsive promoter. Differential cytokine
stimulation activates responding intronic promoters of different S regions.
3
Transcription starts from the I exon to the C
H
region, and the RNA products are
called germline transcripts. Splicing removes the intronic S region, resulting in
the rejoining of the I and C
H
exons. In the I-exon promoter deletion or
3’-enhancer element deletion mice, germline transcription of the switch region
and CSR to corresponding constant genes are abolished or greatly impaired
(Chaudhuri and Alt 2004).

Mammalian switch regions vary substantially in primary sequence but all
have the following features: highly repetitive and G-rich on the non-template
strand. They are 1-12 kb long in length. Switch regions can be divided into two
sub-groups depending on the repeat sequence. One group including Sµ, Sα,
and Sε contains two short sub-repeats: GAGCT and GGGGT. The other group
including Sγ1, Sγ2a, Sγ2b, and Sγ3 has a 49 bp repeat unit and also contains
numerous AGCT sites and clusters of G's also (Dunnick, Hertz et al. 1993;
Dunnick, Hertz et al. 1993). Most CSR breakpoints are identified in the switch
regions; however, they are less frequently found within the several hundred base
pairs upstream or downstream (Dunnick, Hertz et al. 1993; Lee, Kondo et al.
1998). Although switch regions contain the switch repeat, there is no preferential
4
position of CSR breakpoints. Therefore, CSR is regionally-specific
recombination.
Cleavage: RNA or DNA deamination model
The generation of a looped-out circular DNA that is deleted during CSR
strongly suggests that CSR proceeds through DSB intermediates. Both
staggered and blunt DSBs are detected at the Sµ region in primary murine B cells
and human cells (Catalan, Selz et al. 2003) (Rush, Fugmann et al. 2004). AID is
believed to play a role in producing DSB either by itself as the initiating enzyme or
by editing the CSR recombinase via an RNA editing model. Two models are
proposed for the role of AID in CSR: the RNA deamination (RNA editing) and DNA
deamination model. Both are described in the following paragraphs.
The RNA deamination model
The RNA deamination model was proposed because AID has strong
sequence homology to the RNA-editing cytidine deaminase APOBEC1.
APOBEC1 is an RNA-editing deaminase. It deaminates the residue 666 of
mRNA of ApoB100 and generates shorter form of mRNA, encoding ApoB48.
AID and associated cofactors may also have a similar function: editing of an
mRNA of a putative protein which is a DNA endonuclease in SHM and CSR.
5
There are several similarities between AID and APOBEC1 other than sequence
homology. First, both proteins need a cofactor. APOBEC1 requires the
cofactor, ACF, to recognize the editing site on ApoB mRNA (Mehta, Kinter et al.
2000). From mutagenesis studies, AID needs SHM and CSR-specific cofactors
(Ta, Nagaoka et al. 2003) (Shinkura, Ito et al. 2004). Second, both proteins have
NLS and NES signals and they facilitate the shuttling of AID between the nucleus
and the cytoplasm. In addition, the inhibition of protein synthesis severely
impairs CSR (Doi, Kinoshita et al. 2003). Nevertheless, no direct evidence
supports the model so far (Chaudhuri et al., 2004).
The DNA deamination model
The other model is the DNA deamination model. In this model of CSR, AID
converts deoxycytidine to deoxyuridine in S regions. The uracil is processed by
two DNA repair pathways-base excision repair (BER) and mismatch repair (MMR).
The uracil DNA glycosylase (UNG) removes uracil bases and leaves abasic sites.
The apurinic/apyrimidine endonuclease 1 (APE1) nicks 5’ to the abasic sites.
Msh2/Msh6 recognizes the mismatch, and recruits other proteins to excise the
mismatch region, resulting in DNA double-strand breaks. In this model, nearby
nicks, on the two strands of switch regions, form DNA double-strand breaks.
6
Several lines of evidence indicate that AID acts directly on DNA as a DNA-cytidine
deaminase. First, over-expression of AID in E.coli causes the mutation
(Petersen-Mahrt, Harris et al. 2002). It is unlikely that bacteria share the same
putative edited mRNA as mammals, strongly arguing that AID acts directly on
DNA. The mutation is more prone at C/G sites than at A/T sites, and that is
correlated where AID deaminates C to U. Second, the mutation rate increases in
uracil DNA glycosylase (Ung) deficiency mice. Although in theory, two models
are possible, more and more evidence supports the DNA deamination model.
Rejoining: the role of the NHEJ pathway
Two major DNA double strand break (DSB) repair systems in mammalian
cells are homologous recombination and the NHEJ pathway. Because of the
lack of junctional sequence homology of S regions, the NHEJ pathway is believed
be responsible for the rejoining phase in CSR. The main proteins involved in the
NHEJ pathway are Ku, DNA-PK, Artemis and XLF/XRCC4/LigaseIV. Ku70 and
Ku80 are required for CSR. The role of DNA-PK in CSR is controversial. In
DNA-PK deficient mice, CSR is impaired in all types of Ig except IgGγ1 (Manis,
Dudley et al. 2002). However, the debate is in SCID mice (which have mutant
DNA-PKcs with the inactivated kinase domain). The Bosma lab demonstrated
7
that DNA-PK
CS
is dispensable in CSR (Bosma, Kim et al. 2002). Artemis is not
required for CSR (Conticello, Langlois et al. 2007). So far, no ligase is
demonstrated essential in CSR. In addition, ligase IV or Xrcc4 (X-ray repair
complementing defective repair in Chinese hamster cells 4) knockout mice are
embryonic lethal. Ligase IV/XRCC4 have no other cellular function other than in
the NHEJ pathway. Substantial CSR still can occur in the Xrcc4-deficient B cells
(Yan, Boboila et al. 2007). The result indicates that although the major rejoining
system in CSR is NHEJ pathway, other ligases also are involved in the process.
Activation-induced deaminase (AID)
AID is the only B cell-specific factor required in CSR and SHM (Okazaki,
Kinoshita et al. 2002; Yoshikawa, Okazaki et al. 2002). It deaminates
single-stranded cytidine to uridine. Other repair proteins and error-prone
polymerases will repair and incorporate different nucleotides at the positions.
Therefore, AID has a mutational effect on genome sequence. Cells need to
develop tight control systems for AID function. First, there is transcriptional
regulation, AID is exclusively expressed in germinal center B cells (Muramatsu,
Nagaoka et al. 2007). Over-expression of AID in NIH 3T3 murine fibroblast cells
causes a high frequency of mutation in an artificial GFP substrate (Yoshikawa,
8
Okazaki et al. 2002). In addition, AID overexpression causes several
lymphomas in mice (Okazaki, Hiai et al. 2003). Second, AID has a NLS
sequence at the N-terminal end, and a NES signal at the C-terminal end (Ito,
Nagaoka et al. 2004). Most of the time, AID is present in the cytoplasm, and is
shuttled back and forth between the nucleus and the cytoplasm (Ito, Nagaoka et
al. 2004). Since AID is highly mutagenic, it’s better to keep AID away from the
genome. Third, one lab has reported post-translational modification. AID can
act on single-stranded DNA in vitro (Bransteitter, Pham et al. 2003; Chaudhuri,
Tian et al. 2003; Dickerson, Market et al. 2003; Bransteitter, Pham et al. 2004; Yu,
Huang et al. 2004). That same laboratory reports that AID can also act on the
non-template strand DNA of transcription-coupled dsDNA upon the
phosphorylation of the S38 residue (Chaudhuri, Khuong et al. 2004; Basu,
Chaudhuri et al. 2005). According to this work, protein kinase A (PKA) alpha
regulatory subunit (PKAr1α) associates with AID in activated B cells and
phosphorylates the S38 residue of AID (Basu, Chaudhuri et al. 2005). The
phosphorylated AID interacts with the 32 kDa subunit of replication factor A (RPA)
(Chaudhuri, Khuong et al. 2004; Basu, Chaudhuri et al. 2005). Upon
phosphorylation and interaction with RPA, AID gains dsDNA deamination activity
9
of the in vitro SHM substrates. The S38A AID mutation can not interact with RPA,
and it can not deaminate the non-template strand DNA of the
transcription-coupled dsDNA (Basu, Chaudhuri et al. 2005).
To date, a few other proteins have been reported to interact with AID,
including RNA Pol II (Nambu, Sugai et al. 2003; Chaudhuri, Khuong et al. 2004;
Basu, Chaudhuri et al. 2005), MDM2, and DNA-PKcs. The interaction with RNA
Pol II has not been confirmed yet, but it’s proposed that RNA Pol II may help the
targeting process, since transcription is required in CSR. Although MDM2
(MacDuff, Neuberger et al. 2006) and DNA-PKcs (Wu, Geraldes et al. 2005) are
found binding to the C-terminal of AID by yeast two hybrid and pull-down assays,
the physiological significance of the interaction is not certain yet.
Other proteins involved in class switch recombination
Proteins in other DNA repair pathways, base-excision repair (BER) and
mismatch repair (MMR), are involved in CSR: UNG2 in BER, MSH2, Msh6, PMS2,
MLH1, and EXO1 in MMR. Furthermore, other DNA-repair proteins participate in
CSR as well, such as γ-H2AX, 53BP1, and ATM (Chaudhuri and Alt 2004). The
precise functions of most of these proteins in CSR are uncertain. Further
investigation will provide more details of CSR mechanism.
10
Significance
Fully understanding the mechanism of class switch recombination will be
helpful for preventing or devising cures for many CSR-related diseases:
hyper-IgM syndrome (HIGM), autoimmunity, and B-cell lymphomas. Defects in
CSR cause hyper-IgM syndromes (HIGM), and the frequency of these is about 1
in 100,000 births (Durandy, Taubenheim et al. 2007). HIGM patients are
susceptible to severe bacterial infections, and they usually have a short lifespan.
They have normal or elevated concentration of serum IgM, but low or absent of
serum IgA, IgG, and IgE. Half of these patients have the X-linked CD40 ligand
deficiency. To achieve the antibody maturation, the CD40 ligand on follicular
helper T cells needs to interact with CD40 on B cells. Without the interaction, the
B cells can not proliferate and switch to other isotypes in vivo. Also, there is no
germinal center formation in secondary lymphoid organs. However, these B
cells are intrinsically normal and can switch to other isotypes in vitro (Durandy,
Taubenheim et al. 2007).
The other half of HIMG patients are B cell intrinsic defects, caused by
mutation of different genes involved in the CSR process. Fifty-three percent of
B-cell-intrinsically-defect patients have a deficiency of activation induced
11
deaminase (AID). Since AID is required in both CSR and SHM. Most of these
patients not only have the defect in CSR, but also in SHM. About three percent
of B-cell-intrinsic-defect patients have a deficiency of uracil DNA glycosylase
(UNG). In B cells of these patients, they have normal frequency but different
pattern of SHM. In mutations caused by SHM, there is a slight increase in
targeting in G/C residues, and almost all mutations are transition mutations.
About half of B-cell-intrinsic-defective patients don’t have AID or UNG defects,
and molecular defects of them still remain unknown so far. In addition, these
patients have normal SHM.
Furthermore, CSR is also involved in some autoimmunity and B-cell
malignancies. The c-myc gene translocates to the heavy chain switch regions in
all sporadic Burkitt’s lymphomas. Over 90 % of multiple myelomas involve
translocations into their switch regions (Bergsagel, Chesi et al. 1996).

12
CHAPTER 2: DNA substrate length and surrounding
sequence affect the activation-induced deaminase
activity at cytidine

SUMMARY
Activation-induced deaminase (AID) is required for both immunoglobulin

class
switch recombination and somatic hypermutation. AID is

known to deaminate
cytidines in single-stranded DNA, but the

relationship of this step to the class
switch or somatic hypermutation

processes is not entirely clear. We have
studied the activity

of a recombinant form of the mouse AID protein that was
purified

from a baculovirus expression system. We find that the length

of the
single-stranded DNA target is critical to the action

of AID at the Cs positioned
anywhere along the length of the

DNA. The DNA sequence surrounding a given
C influences AID deamination

efficiency. AID preferentially deaminates Cs in the
WRC motif,

and additionally has a small but consistent preference for purine

at the
position after the WRC, thereby favoring WRCr (the lowercase

r corresponds to
the smaller impact on activity).

13
INTRODUCTION
Despite a recent series of reports on AID biochemistry, much

remains to be
learned concerning AID function. Different forms

of recombinant AID from
various laboratories have shown some

contradictory properties. One group has
provided data that the

AID deamination target is a WRC (W = A or T; R = A or G)
hotspot

motif (Pham, Bransteitter et al. 2003), which is similar to the
RGYW/WRCY hotspot identified

from in vivo studies (Shapiro, Aviszus et al. 2002)
(Rogozin, Pavlov et al. 2001). However, two other groups did

not observe such
a hotspot predilection in biochemical studies

of AID action (Dickerson, Market et
al. 2003) (Sohail, Klapacz et al. 2003). One group has shown that AID is
activated

by removal of RNA by RNase A (Pham, Bransteitter et al. 2003), but two
other groups reported

that RNase A had no effect (Dickerson, Market et al. 2003)
(Sohail, Klapacz et al. 2003), leaving the issue of RNase

A-activation unclear.

Here, we examined the activity of a recombinant form of the

mouse AID
protein purified using a baculovirus expression system.

We find that the length of
the single-stranded DNA target is

critical to the action of AID at the Cs positioned
anywhere

along the length of the DNA. Having established the minimum

optimal
length, we did further analyses on how the DNA sequence

context influences AID
14
activity. We have kept the sequence surrounding

each C identical so that the
sites are fully comparable. We

find that AID preferentially deaminates Cs in the
WRC motif

and additionally has a small but consistent preference for purine

at the
position after the WRC, thereby favoring WRCr, where

the lowercase r designates
the smaller impact.

MATERIALS AND METHODS
Oligonucleotide Substrate
All oligonucleotides were synthesized

by Qiagen/Operon Technologies
(Richmond, CA). Oligonuleotides

used to determine the suitable AID substrate
length are as follows:

KY333 (5'-TTTTTTTTTACGATTTTTTT-3', 20mer); KY334
(5'-TTTTTTTTTTTACGATTTTTTTTT-3',

24mer); KY335 (5'-TTTTTTTTTTTTTT
ACGATTTTTTTTTTTT-3', 30mer).

Uracil-DNA-glycosylase (UDG) substrates
are: KY414 (5'-TTTTTTTAUGATTTTT-3',

16 mer); KY415 (5'-TTTTTTTTTAU
GATTTTTTT-3', 20 mer); KY416 (5'-TTTTTTTTTTTAUGATTTTTTTTT-3',

24 mer).
All other oligonucleotides used are indicated in the

specified figures.

Recombination Protein
Mouse AID cDNA was kindly provided

by Dr. Tasuku Honjo. The glutathione
S-transferase (GST) coding

sequence was obtained from pAcG2T (Pharmingen).
15
A PCR reaction

was used to construct the GST-mAID DNA fragment with an
enterokinase

recognition site between the GST and mAID. The GST-mAID coding

region was fully sequenced. Recombinant baculovirus (vKY9) was

made using
the Bac-to-Bac baculovirus expression system from

Invitrogen according to the
manufacturer's instructions. SF21

cells (Invitrogen) were infected with vKY9 at a
multiplicity

of infection (MOI) of 10 and harvested 3 days after infection.

Cells
were resuspended in buffer A (50 mM Tris-Cl, pH 8.0, 250

mM NaCl, 10 mM NaF, 5
mM 2-mercaptoethanol) that contains a

mixture of proteinase inhibitors and
sonicated. The cell lysate

was ultracentrifuged at 100,000 x g for 1 h, and the
clear cell

extract was mixed with glutathione agarose (Sigma). The glutathione

agarose was washed 5 times with 10 volumes (relative to bed

volume) of buffer A
and GST-mAID recombinant protein was eluted

with an equal volume (relative to
bed volume) of buffer B (50

mM Tris-Cl, pH 8.0, 100 mM NaCl, 10 mM reduced
glutathione,

1 mM dithiothreitol). The AID protein on a Coomassie stained

SDS-polyacrylamide gel is shown in Fig. 2-1A.

16
Figure 2-1. AID protein and deamination assay. (A) purified AID protein.
Recombinant AID protein with an N-terminal GST fusion was purified, resolved on
a 10% SDS-polyacrylamide gel, and stained with Coomassie Brilliant Blue. The
identity of each band is indicated. (B) scheme of the AID activity assay. A
single-stranded oligonucleotide substrate is labeled at the 5'-end (asterisk).
Uracil generated by AID is removed from DNA by UNG, resulting in an abasic (AP)
site. DNA cleavage at AP sites is introduced by alkali treatment at elevated
temperatures. The cleaved product is separated from the substrate on a
denaturing polyacrylamide gel.

17
Deamination Assay
The AID activity assay is shown schematically

in Fig. 2-1B. Uracil generated by
AID can be removed from DNA

by UNG, resulting in an abasic (AP) site. DNA
cleavage at AP

sites can be introduced by alkali treatment at elevated
temperatures

(95 °C) (Abner, Lau et al. 2001). Therefore, DNA cleavage can
serve as a readout

for AID-mediated cytidine deamination. The reaction mixture

(10 µl) containing 200 fmol of radioisotope-labeled oligonucleotide

substrate, 50
ng (1 pmol) of GST-mAID, 100 ng of RNase A, and

0.1 unit (1 unit catalyzes the
release of 1 nmol of uracil in

one h at 37 °C) of UNG (Invitrogen) was incubated
for 10min

at 37 °C in a buffer containing 25 mM Tris-Cl, pH 8.0, 50

mM NaCl, and
5 mM EDTA. The reaction was stopped by the addition

of 1 µl of 2M NaOH and
heated for 5 min at 95 °C. Eleven microliters of formamide were then added and
samples

were heated at 100 °C for 5 min and plunged in an ice water

bath.
Reaction products were separated on 10% denaturing polyacrylamide

gels
running in 1x Tris borate-EDTA (TBE) buffer. Gels were

visualized by
autoradiography using a phosphor-imager FX (BioRad

Laboratories, Hercules,
CA) and quantified with Quantity One

software (version 4.2). A time course of a
typical reaction

for two different DNA substrates is shown in Fig. 2-2.
18
Figure 2-2. Time-dependent AID deamination. AID-mediated cytidine
deamination was measured every 5 min for the first 30 min. Both a high
efficiency (T
11
AGCAT
11
, open circles) and a low efficiency (T
11
GTCTT
11
, open
triangles) substrate were tested. The sequences surrounding the target C
residues are listed.

19
RESULTS
AID activity assay
Two different radioisotope-labeled oligonucleotides, T11AGCAT11 and
T11GTCTT11 were incubated with 5 fold of recombinant mouse AID with an
N-terminal GST fusion protein (GST-mAID) at different time period. In the assay,
excess uracil glycosylase (UDG) was added to assure the AID deamination is the
rate-limiting step in the reaction. The reaction reaches the plateaus in about 15
to 20 minutes (Fig.2-2). Therefore, the following experiments were done in 10
minutes to ensure the reaction still in the linear range.
DNA Length-dependence of the Action of AID at Cytidines
To optimize the deamination condition, the minimal length of radioisotope
-labeled oligonulcleotide was determined by using different lengths of
oligonucleotides. The design of the oligonucleotide is the same central
sequence and different lengths of surrounding sequence, tracts of Ts. The
shortest substrate is 20 mer, and the conversion is only 5.3 %. With the longer
length, the deamination rate is higher. The conversion rate of the 24 mer-
substrate is about 49 %. In addition, the conversion rate of the 30 mer-substrate
is about 62 % (Fig.2-3A). The deamination of several different lengths of
20
oligonucleotides indicated the deamination rate reached the plateaus at the
26-mer substrate (Fig. 2-3B). The length-dependent deamination is the intrinsic
feature of AID, but not UNG in the reaction. The conversion rate of different
lengths of UNG substrates was similar (Fig. 2-3C).

21
Figure 2-3. Substrate length effect on AID activity. (A) AID deamination assay.
Radioisotope-labeled oligonucleotide substrate (single-stranded) was incubated
with recombinant AID protein. Oligonucleotides were heat-cleaved in alkali
conditions at the deaminated position. The numbers listed on the right of the gel
indicate the length of each oligonucleotide. (B) variation of AID activity as a
function of substrate length. The x-axis indicates the single-stranded DNA
substrate length. The y-axis indicates the percentage of product resulting from
AID deamination (see "Experimental Procedures"). (C) UNG activity is
independent of substrate length. UNG removes uracil from synthetic
uracil-containing oligonucleotide substrates (equivalent to an AID deamination
product). The resulting AP site was then cleaved with alkali treatment. The
cleaved product was resolved on a denaturing polyacrylamide gel. The amount
of UNG used and the length of the substrates are indicated. Lanes 1, 4, and 7,
KY414; lanes 2, 5, and 8, KY415; lanes 3, 6, and 9, KY416.
22

23
Quantitation of the Sequence Preferences of AID on Single-stranded

DNA
To know the AID preference motif, the basic design of oligonucleotides is
maintaining only one C in the sequence, and changing three positions of
surrounding sequences of C. The olionucleotide is T
11
NNCNT
11
. For the
NNCN motif, the first N was named -2 position, the second N was -1, C was 0,
and the last N was +1. To avoid the complicity, no C was in the surrounding
three positions (N). The length of the oligonucleotide is 26 mer, which is
sufficient for AID to act. There are 27 different oligonucleotides. The
deamination reaction was done in 10 minutes, still in a linear range. The hotspot
sequence of AID deamination is 5’-WRCN-3’ motif (Fig. 2-4A). The conversion
rate of four groups of WRC-containing oligonucleotides is all over 50 %. For all
sets of nine groups, the +1 position only provided the small effect of the AID
preference. Either A or G at the +1 position gave a little bit higher deamination
conversion rate. Therefore, the AID hotspot motif is 5’-WRCr-3’ (r indicated the
small effect). The highest conversion rate is about 80 % (T11TGCNT11). The
lowest conversion rate is about 10 % (T11GTCNT11). All of substrates can be
deaminated by AID, despite the preference.

24
Figure 2-4. Site preference in the AID deamination. (A) deamination assay
with 27 C-containing substrates. Oligonucleotide substrates with sequence
variation around the C (N
-2
N
-1
C
0
N
+1
) were assayed for deamination efficiency.
The exact sequence of each substrate around the C is listed on the top of the gel.
A unique number was assigned to each substrate (bottom of the gel). S,
substrate; P, product. The histogram shows the average conversion efficiency
for each substrate based on three independent experiments. The four groups
that fit the WRC motif are emboldened above the histogram. (B) deamination
assay with substrates containing two or more Cs. The sequences surrounding
each C are listed on the top of the gel as described in panel A. The deamination
efficiencies are listed in the histogram in order (Product 1, 2, and 3) from the
longest (C that is furthest away from the radioisotope label) to the shortest
product.
25

26
To further test the AID preference motif, oligonucleotides containing more
then one C were used in AID deamination assay (Fig.2-4B). Since the +1
position only had the small effect on the preference, the +1 was kept constant by
T nucleotide. The drawback of the experiment was the first C cleavage product
masked the second C cleavage product, since the oligonucleotide was 5’ labeled.
Overall, the cleavage profile was fitted with the WRCr motif. However, there
were some exceptions in the result. For the second product of the substrate 28,
T11ACCTT11, the conversion rate is low although it fit within the WRC motif.
The interesting one is the first product of the substrate 30, it was a non-WRC motif
but had high deamination rate.
DISCUSSION
We determined the AID site preference using a comprehensive

set of fully
comparable oligonucleotide substrates. Our data

and that of others (Pham,
Bransteitter et al. 2003) suggest that the 5'RGYW3'/5'WRCY3' hotspot

motif found
in somatic hypermutation may be largely due to AID

enzyme site preference.
Our consensus is WRCr, given a small

but consistent preference at the +1
position for purines.

27
RGYW is generally regarded as the hotspot for somatic hypermutation

based
on the analysis of endogenous genomic immunoglobulin genes (Shapiro, Aviszus
et al. 2002). G is the most heavily mutated residue within that motif.

The
complementary strand of the 5'RGYW3' hotspot is 5'WRCY3'. It is easy to
understand the presence of G in the RGYW hotspot

because AID deaminates the
complementary C in WRCY. However,

we did not observe the preference for Y
in the WRCY (corresponding

to the R in RGYW) motif in our deamination assays.
Instead,

we found that it is slightly favored to have R at that position.

Considering
the smaller effect at this position, the hotspot

preference determined by our assay
is best described as a 5'WRCr3'

motif (and the complementary sequence would
be 5'yGYW 3'). We

also observed some deviations from the WRC preference.
For example,

the second C in CGCT, which does not conform to a WRC motif,

shows a moderate to high deamination efficiency. This illustrates

that although
AID activity is influenced by the surrounding

sequence, local structural factors
probably underlie this in

a manner that is not fully described by primary sequence.

Although a sequence preference in somatic hypermutation is not

surprising
given known hotspots, what might be the relevance

of a WRC preference in class
switch recombination? After completing

this study, we realized several
28
interesting sequence features

relevant to this point. First, WRC is common in the
switch regions

of all vertebrates. It is commonly present in the form of
5'-AGCT-3'

or 5'-AGCA-3' on the G-rich strand (non-template strand). This

is the
case in vertebrates as distant as Xenopus from human (Kitao, Arakawa et al.
2000). Second, AGCT is a palindrome; hence, the anti-parallel

strand reads
5'-AGCT-3' as well, thereby conforming to the WRC

motif. Although AGCA is not
a palindrome, the anti-parallel

strand is TGCT, which also conforms to the WRC
motif. AID may

act on the WRC motifs on the two strands concurrently, given

that the edges of the R-loop may have single-stranded character

on the C-rich
(template) DNA strand as well (Tian and Alt 2000). If this were

the case, then a
double-strand break would result, consistent

with the known features of class
switch recombination (Dunnick, Hertz et al. 1993). These

aspects may explain
the relevance of the WRC motif to class

switch recombination.

In this regard, it is interesting that recent work has shown

that a G-rich region,
when substituted for the Sγ1 switch region

was able to form R-loops in vitro
(Shinkura, Tian et al. 2003). However, its ability

to undergo class switch
recombination was about 13-fold lower

than the normal Sγ1, and the basis for this
lower efficiency

was unclear (Shinkura, Tian et al. 2003). In light of the WRC
29
motif observation noted

above, it is noteworthy that the G-rich region used in that

study had very few WRC motifs, and the few which were present

did not have one
nearby on the anti-parallel strand. Hence,

WRC motifs close together on the two
DNA strands may be quite

important for class switch recombination in a manner
that has

not been previously considered.

Unlike class switch regions, V(D)J regions are unlikely to form

RNA:DNA
hybrids upon transcription, and hence, the required

single-strandedness in the
V(D)J region could not be generated

by an extensively displaced non-template
strand. It has been

hypothesized that the non-template strand is single-stranded

transiently at the transcription bubble, which would then serve

as a target for AID.
Indeed, recent studies have shown that

AID preferentially acts on the
non-template strand when transcription

is carried out in vitro or in Escherichia coli
(Pham, Bransteitter et al. 2003; Ramiro, Stavropoulos et al. 2003; Sohail, Klapacz
et al. 2003).

It is interesting to note, however, that the RGYW motif found

at
hotspots for SHM is not strand-specific (Milstein, Neuberger et al. 1998) (Michael,
Martin et al. 2002). Therefore,

both strands of the V(D)J region should be
equally accessible

to AID. Reconciliation of the substrate specificity of AID with

30
the precise structural features of its endogenous substrates

awaits further
investigation.

A recent study used AID deamination to introduce mutations into

a
single-stranded LacZα genetic reporter located on an otherwise

double-stranded
circular substrate (Pham, Bransteitter et al. 2003). That study reached

a similar
conclusion regarding the sequence preference, namely

that AID prefers the WRC
motif. The agreement on the sequence

preference is stronger given that both a
biological readout

and a purely biochemical one have yielded similar results.

We observed a DNA substrate length dependence for AID action

on
single-stranded DNA. Given the small size of AID (24 kDa)

we anticipated that
16-20 nt would be sufficient. In fact, 24 nts is necessary to reach full or nearly
full

activity on single-stranded DNA. This surprising length dependence

may
reflect a larger than expected footprint size of AID, but

multimerization of AID is
another possibility (Chaudhuri, Tian et al. 2003) (Ta, Nagaoka et al. 2003).

31
Chapter 3: The Downstream Boundary of Chromosomal
R-loops at Murine Switch Regions: Implications for the
Mechanism of Class Switch Recombination

SUMMARY
R-loops form at Sγ3 and Sγ2b immunoglobulin class switch regions in the
chromosomes of stimulated murine primary B cells and are suspected to be a
general feature of mammalian class switch regions. The in vivo upstream
boundary of the R-loops is known to begin within the switch repeats. To
determine how precisely the R-loop structure conforms to the repetitive zone of
the murine Sγ3 and Sγ2b switch regions, a chemical probing method was used to
obtain structural information on the downstream boundary. We find that only 61
to 67% of the R-loops terminate within the Sγ3 and the Sγ2b repetitive zones, and
the remainder terminate downstream, usually within the first 600 bp immediately
downstream of the core switch repeats. Interestingly, the nontemplate strand G
density falls to the random level gradually through this same region. Hence, the
R-loops terminate as the G-richness of the nascent RNA strand falls. This
finding is consistent with thermodynamic predictions for RNA:DNA duplex
32
strength relative to that of DNA:DNA duplexes. This contrasts with the location
of known recombination breakpoints, which correlate not with G-richness and
R-loop location, but rather with AGCT density. The implications of these findings
are discussed in the context of models for the targeting of class switch
recombination.
INTRODUCTION
To date, the targeting mechanism of AID in CSR is still unclear. Many
hypotheses were proposed, and one of them is DNA secondary structure based
on the feature of the switch region. This model arises because of the
regional-specificity of CSR. Several DNA structures have been proposed to form
in the switch region. By computer modeling, a DNA stem-loop structure was
suggested in Xenopus and mammalian switch regions (Tashiro, Kinoshita et al.
2001) (Mussmann, Courtet et al. 1997). Guanine quartets (G4) have been
suggested because of the G-rich nature of the switch region (Sen and Gilbert
1988) (Dempsey, Sun et al. 1999) (Duquette, Handa et al. 2004). However,
neither hypothesis has in vivo data to support it.
Upon transcription, an RNA:DNA hybrid forms in the switch region sequence
in vitro (Reaban and Griffin 1990; Daniels and Lieber 1995; Tian and Alt 2000;
33
Shinkura, Tian et al. 2003). R-loops are found at the Sγ3 and Sγ2b regions in
vivo in LPS-stimulated B cells (Yu, Chedin et al. 2003). Upon antigen stimulation,
germline transcription is activated in specific switch regions. The transcribed
RNA pairs with the template strand DNA, leaving the non-template G-rich strand
DNA single-stranded. The single-strandedness within R-loops or variants of
them (such as collapsed R-loops) may enhance the efficiency of CSR relative to
its efficiency in organisms that do not have R-loop forming switch regions (Zarrin,
Alt et al. 2004).
The upstream boundary of the R-loops (referred to as the 5' boundary) was
recently established in the initial description of in vivo switch region R-loops at Sγ3
and Sγ2b (Yu, Chedin et al. 2003). From previous experiments, although
R-loops are heterogeneous, 95% begin somewhere within the core repeats of the
switch regions. The downstream boundary of the R-loops is of interest because
it provides information on how quickly the RNA polymerase, RNA, and template
interaction revert to normal after the switch sequences have been passed.
Here we find that, although 61-67% of the R-loops end within the switch
region, most of the remaining ones end within the 600 bp downstream of the Sγ3
or the Sγ2b regions. The G-richness, which is a suspected predictor of R-loop
34
formation (Zarrin, Alt et al. 2004), declines gradually over this same 600 bp region,
providing an important correlation between R-loop structure and nontemplate
strand G density. Interestingly, high AGCT density terminates abruptly within the
boundaries of the core switch repeats, and this is where over 90 to 95% of the
recombination is known to occur in vivo at Sγ2b and Sγ3, respectively (Dunnick,
Hertz et al. 1993). Hence, R-loop formation tracks with the G-richness feature of
switch regions, but switch region recombination points track with the AGCT
density. These findings, as well as the frequency of R-loop formation, support a
model in which mammalian switch regions have evolved two motifs: one for
R-loop formation (G clusters) to provide the single-strandedness required by AID
and one for AID deamination within the non-B DNA portions of that R-loop (the
AGCT sites).
MATERIALS AND METHODS
Enzymes and reagents
All restriction enzymes were from New England Biolabs (Beverly, MA).
RNase H was from Promega (Madison, WI). Sodium bisulfite and other
chemicals were from Sigma. The C57BL/6 mice were purchased from the
Jackson Laboratories (Bar Harbor, Maine).
35
Purification and culture mouse B cells
Mouse B cells were taken from spleens of 8-12 week old C57BL/6 mice. A
single-cell suspension was prepared from spleens and red blood cells were
removed. Naïve B cells were purified by magnetic cell sorting (Miltenyi, Auburn,
CA). A single-cell suspension was incubated with the anti-CD43 magnetic beads
for 15 minute at room temperature. An LS column in the magnet was placed on
the MACS Multistand and washed with PBS containing 0.5 % BSA. The cell
suspension and beads were loaded onto the LS column. The column was
washed three times with 3 ml PBS containing 0.5 % BSA. The flow through was
collected. The cell suspension was centrifuged 10 minutes at 1,000 rpm at 4 °C.
The naïve B cells were counted and cultured at density of 5x 10
5
cells /ml.
These B cells were cultured with LPS (lipopolysaccharide) for two days.
Extraction and purification of genomic DNA
Cell pellets were washed with PBS once and dissolved in 10 mM Tris, 1 mM
EDTA (pH8.0). SDS was added to 0.5 % and proteinase K to 0.2 mg/ml final
concentrations. Cells were incubated at 37 °C overnight. Phenol/chloroform
extraction and ethanol precipitation of the genomic DNA were done. Genomic
DNA was dissolved in 10 mM Tris, 1 mM EDTA (pH8.0) to 1mg/ ml. The
36
genomic DNA was digested with EcoRI at 37 degrees overnight.
Phenol/chloroform extraction and ethanol precipitation of DNA were repeated, and
the DNA was dissolved in TE pH 8.0.
RNase H was used to verify R-loops molecules in stimulated mouse B cells.
As a control, the genomic DNA from stimulated mouse B cells was treated with
3 units of E.coli RNase H before the bisulfite modification (Promega, Madison, WI)
mixed with the restriction enzyme at 37 °C overnight. The treated genomic DNA
was purified by phenol/chloroform extraction and ethanol precipitation. The DNA
was dissolved in TE, pH 8.0.

Bisulfite modification assay
For the bisulfite treatment, 20 µg of digested genomic DNA in 30 µ l was
mixed with 12.5 µl of 20 mM hydroquinone and 457.5 µl of 2.5 M sodium bisulfite
(pH 5.2). The mixture was covered with mineral oil in a microcentrifuge tube in
the dark at 37 °C overnight. The bisulfite-treated-DNA was purified by the Wizard
DNA clean-up system (Promega, Madison, WI), according to the manufacturer’s
protocol. The purified DNA was desulfonated with 0.3 M NaOH at 37 °C for 15
minutes and recovered by ethanol precipitation. The purified DNA was
37
resuspended in 30 µl of 10 mM Tris, 1 mM EDTA (pH 8.0) and stored at –20
degrees.
Two kinds of primers were used in the PCR. In normal PCR, two primers
were used which have the usual sequence complementarity to the top and bottom
strands of the template. We call these 'native' primers. When we enrich for
R-loops, one primer is always native, but the other (called a converted primer) has
a sequence with 2 to 8 (out of 20 to 24 nt) changes that anticipate the bisulfite
conversions [C to T for primers directed downstream (forward primers) and G to A
for primers directed upstream(reverse primers)] at a region of
single-strandedness. The annealing temperature used was standardized
according to the primer used in each specific PCR reaction. The PCR reaction
was carried out for 30 cycles and resolved on an agarose gel, and the
correct-sized fragments were recovered using a Gene Clean kit (BIO101, CA).
Purified PCR products were cloned using the TOPO-TA cloning kit (Invitrogen,
CA). Plasmid DNA from each clone of interest was purified using the Gene Elute
Plasmid Mini Prep kit (Sigma, St. Louis, MO). Sequencing reactions were
carried out using the SequiTherm Excel II sequencing kit (Epigenetics, Madison,
WI) and MWG thermal cycler model Primus 96 Plus (MWG Biotech, High Point,
38
NC). Automated sequencing was carried out using the Li-Cor DNA Analyzer
model 4200 (Li-Cor, Lincoln, NE).
Oligonucleotides
Oligomers were from Operon (Richmond, CA). The following primers were
used in the PCR. FTH5 (converted primer), 5’-GGGGTTATTAGATTATA
GGGTTTT-3’; FTH7 (native primer), 5’-CCATAGTTCCATTTTACAGTTACC-3’;
FTH13 (converted primer), 5’-GGGGATTAGGTTGAGTAGTTTTT-3’; FTH16
(converted primer), 5’-TAATCAACTCAATCTTACTAACTAAA-3’; KY245 (native
primer), 5’- TAAGGAGTCTGACCAAGCAACCATA-3'; KY246 (native primer), 5'-
AGTCCACTTTTCCACACTCCTATCT-3'; KY320 (native primer), 5’-ACTTTCCC
TTTCTCTTTCTTCATC-3’; KY484 (biotin tagged-KY320); KY505 (native primer),
5’-ATGATCCTGGGGGATTATGGAAA-3’; FTH99 (converted primer), 5’- ACTCCC
CAAAACTACCCAATCTAA-3’; FTH100 (converted primer), 5’- ACTCCCTAAA
AACTACTCAACCTAA-3’; FTH101 (converted primer), 5’- AAACTCCCCC
AAAACTACCCAACCTA-3’; FTH71 (converted primer), 5’- GGTGGGAGTATT
AGGGATTAATTTT-3’; FTH73 (native primer), 5’- CTTACACTCCCATTTATC
ACATTTC-3’; FTH88 (native primer), 5’- CCTGTACATATCTTCCACAATCCCC-3’.
39
Calculating the thermodynamic stability
Predicted nucleic acid hybridization thermodynamics was calculated by the
HyTher
TM
program on the website http://ozone2.chem.wayne.edu/. The program
provides the calculation for four forms of nucleic acid: DNA duplex, RNA/DNA,
DNA/RNA, and RNA/RNA. The hybridization conditions were based on 0.14 M
NaCl, 0.002 M MgCl and at 37 °C mimicking the physiological condition.

Determination of the frequency of R-loops
In the first phase of the method, the top strands of R-loops were primer
extended. For this, the biotin-tagged native primer (KY484) was used in the
primer extension of stimulated B cell genomic DNA that had already been treated
with bisulfite. The biotin-primer extension product was purified using
streptavidin-magnetic beads (Sigma, St. Louis). The purified primer extension
products were used as templates in the PCR with two native primers, KY505 and
KY246. The PCR reaction was carried out for 30 cycles and resolved on an
agarose gel, and the correct-sized fragments were recovered using a Gene Clean
Kit (BIO101, CA). Purified PCR products were cloned using the TOPO-TA
cloning kit (Invitrogen, CA). Each clone was picked and restreaked as a line on
40
the surface of a new ampicillin agar plate. Each plate contained about 110
different clones.
The second phase of the method involved colony lifts onto nylon membranes.
The nylon membrane was pressed against the agar plate (110 clones) for 2
minutes. The membrane was then transferred to denaturing solution (0.5 M
NaOH, 1.5 M NaCl) for 15 minutes. The bacterial side was up. The membrane
was then transferred to 1 M Tris, pH 7.5 for 15 minutes, followed by 1 M Tris, pH
7.5, 1.5 M NaCl for 15 minutes, followed by a rinse with 2xSSC. Kimwipes were
used to wipe the bacterial debris off of the surface of the membrane, while it was
under the 2xSSC. The DNA was then fixed on the membrane by UV crosslinking.
The membrane was then rinsed with 2xSSC and put into a plastic hybridization
bag.
In the third phase, the three converted primers, FTH99, FTH100, and
FTH101, were individually 5'-labeled with polynucleotide kinase and then mixed to
generate the probe for hybridization. The membrane was pre-hybridized at 66°C
for 15 minutes. Three different oligonucleotide probes (5 pmol each) was
5'-labeled using polynucleotide kinase and then added, and the membrane was
hybridized at 66 °C overnight. The membrane was washed with 2xSSC, 0.5 %
41
SDS for 10 minutes twice, and 0.1XSSC, 0.5 % SDS for 10 minutes twice. The
membrane was exposed overnight.
RESULTS
The downstream boundary of the R-loops at Sγ3
The sodium bisulfite method was used to obtain structural information on the
3’ boundary of Sγ3 R-loops (Yu, Chedin et al. 2003). The bisulfite anion carries out
a nucleophilic attack on the C5-6 bond of cytosine only when it is not protected by
stacking, which occurs when it is single-stranded. Reactivity of all of the C’s in a
region indicates a stable zone of single-strandedness.
Only a small proportion of B cells undergo CSR at any given time.
Approximately 3.3% of naïve murine B cells become IgG3
+
cells after 4 days of
LPS stimulation, and here we stimulate for only two days (Fig. 3-1). To study the
R-loop structure, an enrichment method was used to assess R-loop molecules
after the bisulfite modification treatment. The strategy relies on one normal
(native) primer and one 'converted' primer. The converted primer has all C’s
replaced with T’s in order to preferentially anneal with a strand where all of the C’s
have been converted with bisulfite to U (which is replaced with T during the PCR
step). Using this approach, any PCR products arise predominantly from R-loops
42
rather than from duplex DNA. Based on the design of converted primers, the
result can provide structural information specifically for the top strand or the
bottom strand of Sγ3 or Sγ2b (Yu, Chedin et al. 2003).
Here, genomic DNA was extracted from LPS-stimulated B cells from the
spleens of C57BL/6 mice. The entire murine Ig Sγ3 region is 1,801 bp, and the
region between the switch region and the constant region is 2,080 bp. For PCR,
one primer was the forward converted primer, and the other one was the native
reverse primer. The converted primer, FTH13, was located 400bp downstream
of the start of the murine Ig Sγ3 and contained 7 C’s all converted to T’s. This is
a more strongly enriching converted primer than we have used previously (Yu,
Chedin et al. 2003). Because FTH13 is a converted primer, it only anneals to the
form of the template strand that has the C to T conversions; and these C to T
conversions are only in the template strand after the first round of PCR when the
nontemplate strand has been copied. The native primer, KY320, was 674 bp
downstream of the end of the core repeats of the switch region.

43
Figure 3-1. Fluorescence-activated cell sorter analysis of the fraction of
murine splenic surface positive IgG3 cells. The purified naïve mouse B cells
were stimulated with 20 µg/mL LPS for 4 days. After the stimulation, B cells
were stained with FITC-conjugated rat anti-mouse IgG3 monoclonal antibody.
The X-axis shows the staining of IgG3. The demarcated right portion of the
distribution represents IgG3-producing B cells. After 4 days of stimulation with
20 µg/mL LPS, 3.3 % of the B cells expressed surface IgG3.

44
After 30 cycles of PCR, the 2,084 bp PCR product was obtained. A total of 34
clones were sequenced and found to represent distinct molecules. Eleven out of
34 sequenced molecules showed sporadic conversion, where conversion had
occurred at some or all of the primer site C’s but at only occasional intermittent
C’s downstream of the primer. Twenty-three molecules revealed R-loop
molecules that had long stretches of conversion (Fig. 3-2). Out of these 23
molecules, 14 R-loop molecules had their endpoints within the switch region (the
endpoints of the regions of bisulfite reactivity indicate the boundaries of the
R-loops). Nine molecules had long stretches of conversion which extended
outside of the switch region. Four of the 9 molecules had long stretches of
conversion which extended more than 600 bp downstream of the switch region (to
the primer KY320 and indicated by an asterisk on the right side of the sequence
line in Fig. 3-2A). The longest R-loop was 2,084 nucleotides, which covered the
entire length of the region analyzed. Interestingly, three R-loop molecules had
gaps within these long stretches of conversion (Fig. 3-2A, indicated by 'gap' on
the right side of the sequence line).

45
Figure 3-2. Single-strandedness on the top strand of the murine Sγ3 region
in stimulated B cells. (A) Top strand sequence without RNase H treatment prior
to bisulfite treatment. Primary splenic B cells were stimulated in culture for 2
days with LPS. Genomic DNA was prepared and then treated with sodium
bisulfite as described in the Methods. A single round of PCR (30 cycles) was
done using one regular (native sequence) primer and one converted primer
whose sequence is complementary to the top strand where the C's have been
converted to U's. FTH13, the converted primer containing 7 C's converted to 7
T's, is located 400 bp downstream of the beginning of the murine Ig Sγ3 region.
KY320, the native primer, is located 674 bp downstream of the end of the Sγ3
region. Each long line represents an independent molecular clone. The small
vertical bar on each line indicates a C on the sequence converted to a T. The
asterisk indicates that the region of conversion extends to the KY320 site. The
word, gap, indicates that the clone has gaps in the long stretches of C to T
conversion (The gaps have more than two consecutive unconverted C's). In the
diagram at the top, the long ellipse represents the Sγ3 region, which is 1801 bp
long. The region between the Sγ3 and the Cγ3 (constant region of the Igγ3) is
2,080 bp. (B) Top strand sequence with RNase H treatment prior to bisulfite
46
treatment. The experiment was done as in Figure 3-2A, except that an RNase H
incubation was done prior to bisulfite treatment. The bottom line displays every C
residue on the top strand of the PCR product.

47
The structural information from the bottom strand was examined as well.
First, converted primers specifically annealing to the converted bottom strand
were used in the PCR. No PCR product was obtained, indicating that no
stretches of single-strandedness could be detected on the bottom strand with or
without RNase H. Second, we used native primers outside of the Sγ3 region for
the PCR, reasoning that though large regions of single-strandedness may not be
detectable, native primers outside of the switch regions might permit us to access
any single-strandedness within the switch region. KY245, which is 350 bp
upstream of Sγ3, was used with KY246, which is 200 bp downstream of Sγ3.
The expected 2,405 bp PCR product was obtained. Without in vitro RNase H
treatment, a total of 11 molecules were sequenced, and all showed only sporadic
conversion (Fig. 3-3, lower portion). With in vitro RNase H treatment, eleven
molecules were sequenced and all showed only sporadic conversion (Fig. 3-3,
upper portion). In conclusion, before and after in vitro RNase H treatment, the
bottom-strand molecules had no long stretches of conversion. In order to
complete the analysis, we examined the top strand with native primers outside of
the switch region. Without (Fig. 3-4, lower portion) and with (Fig. 3-4, upper
portion) in vitro RNase H treatment, molecules only showed sporadic conversion.
48
Fig. 3-3. Single-strandedness on the bottom strand at the murine Sγ3 of
the stimulated B cells (with native primers). The upper part of the figure
represents clones with In vitro RNase H treatment. The lower part of the figure
represents clones without in vitro RNase H treatment. Two primers were used in
the PCR; KY245, the native primer, was located 350 bp upstream of the upstream
edge of the murine Ig Sγ3. KY246, the native primer, was located 200 bp
downstream of the downstream boundary of Sγ3. Each long line represents an
independent clone. The small bar on each line indicates that the C on the
sequence was converted to a T. The bottom line of the diagram displays every C
residue on the bottom strand of the PCR product.

49
Fig. 3-4. Single-strandedness on the top strand at the murine Sγ3 region of
stimulated B cells (with native primers). The experiment was done as in
Figure 3-3 except that the molecules shown here are from the top strand.

50
Figure 3-5. Single-strandedness on the top strand immediately
downstream of murine Sγ3 in stimulated B cells. Two primers were used in
an initial round of PCR (30 cycles) and yielded a faint band. To generate more
DNA for cloning, this band was cut out and amplified with the same primers with
an additional round of 30 cycles. FTH5, the converted primer containing 7 C's
converted to 7 T's, is located immediately within the downstream edge of the
murine Ig Sγ3 region. FTH7, a native primer, is located 130 bp downstream of
the beginning of the first Cγ3 exon. Each long line represents an independent
molecular clone. The small vertical bar on each line indicates a C on the
sequence converted to T. The most bottom line displays every C residue on the
top strand of the PCR product.

51

52
To further investigate the ends of R-loops, a more downstream primer set,
FTH5 paired with FTH7, was used to define the 3’ boundary of R-loops (Fig. 3-5).
The converted forward primer, FTH5, was located at the border of the repetitive
zone of the switch region and contained 7 C’s converted to T’s. The native
reverse primer, FTH7, was 130 bp downstream of the start of the constant region
of the Sγ3. After two rounds of 30 cycles of PCR, the 2.2 kb PCR product was
cloned, and 50 clones were sequenced. Seven molecules showed only sporadic
conversion, whereas 43 clones revealed long regions of conversion consistent
with an R-loop conformation (Fig. 3-5). Forty of these 43 molecules had 3'
boundaries located within 600 bp downstream of the end of the switch region.
Two molecules ended only slightly further downstream, passing the position of
primer KY320. The conversion region for one of the 43 molecules was 1,696 nts
long, ending immediately before a 100 bp region composed entirely of A’s and G’s
and 390-488 bp upstream of the constant region. No R-loop molecules had long
stretches of conversion that extended into the constant region or near it. To
confirm these results, another PCR primer set was designed to see if R-loops
existed in the constant region. The converted reverse primer, FTH16, primes in
the constant region and contains 8 G’s all converted to A’s (see Materials and
53
Methods), which has more enrichment potential than other primers that we have
used (the partner primer for this PCR is FTH5). After two rounds of 30 cycles of
PCR, no PCR product was obtained. Therefore, no R-loops were detectable in
the constant region even with primers with extremely high enrichment potential.
Structure of the Sγ3 R-loops after in vitro RNase H treatment
To confirm that the above molecules contained actual R-loops, the same
genomic DNA was treated with E. coli RNase H1 before treating with bisulfite.
RNase H only digests RNA which is annealed with DNA. The genomic DNA was
treated in vitro with 3 units of E.coli RNase H1 for 16 hours. It is difficult to
calculate the fold excess of RNase H because the reaction contains an unknown
amount of RNA:DNA hybrid in the total genome (Li and Manley 2005) and
because some of the E. coli RNase H1 may be bound nonproductively by the vast
excess of genomic DNA. After one round of 30 PCR cycles using the converted
primer FTH13 and the native primer KY320, the 2,084 bp product was still
obtained, although the yield was slightly lower than for genomic DNA without in
vitro RNase H treatment (Fig. 3-2B). Twenty-two clones were sequenced. Five
out of the 22 sequenced molecules showed sporadic conversion on the top strand.
The remaining 17 molecules indicated long stretches of single-strandedness on
54
the top strand, consistent with R-loops. Interestingly, none of these extended
outside of the Sγ3 region, which contrasts with the analysis without RNase H,
where 39% of the molecules extended downstream of Sγ3. In addition, eight of
the 17 molecules had gaps in the long stretches of conversion, which is a higher
frequency of gaps than when no RNase H is used (Fig3-2). These features
indicate that this initial RNase H incubation partially digested the R-loops.
We wanted to more stringently test whether the R-loops downstream of the
core Sγ3 repeats had been destroyed by RNase H. The primer set, FTH5 paired
with FTH7, was used in a PCR analysis of the RNase H-treated material just
described. No PCR product was obtained, even after three rounds of PCR (30
cycles each round), indicating that no R-loops were detectable downstream of the
Sγ3 region, after one incubation with RNase H. This is consistent with the
possibility that the portion of the R-loop immediately downstream of Sγ3 is more
readily destroyed than the portion of the R-loop within the Sγ3 region.
Next, the genomic DNA was treated with a second incubation of 3 units of E.
coli RNase H1 for 16 hours. Using the FTH13 converted primer within the switch
region and the native primer downstream (KY320), no PCR product could be
amplified any longer. (Amplification of the DNA template using native primers
55
was quantitatively stable throughout these incubations, indicating that the DNA
was not affected.) Hence, the portion of the R-loop downstream of the Sγ3 can
be destroyed in the first incubation with RNase H, but the portion within the Sγ3
requires a second incubation. These findings suggest that long in vivo R-loops
are less optimal substrates than pure RNA:DNA duplexes and that the R-loops
within the repetitive core region of Sγ3 are somewhat more resistant than the
portion of the R-loop downstream of the core repeat region.
Downstream boundaries of R-Loops at Sγ2b
We were interested in determining the nature of the downstream R-loop
boundary at a second switch region, Sγ2b (Fig. 3-6). The entire murine Sγ2b is
3.8 kb, and the distance between the Sγ2b and Cγ2b is 1,837 bp (Fig. 3-7). To
find the downstream boundary of R-loops at murine Sγ2b, the same enrichment
strategy was used to detect the R-loops. The converted primer, FTH71,
containing 5 C’s all converted to T’s, is positioned 400 bp upstream of the end of
Sγ2b. Two native reverse primers, FTH73 and FTH88, were chosen at 100 and
800 bp downstream of the end of Sγ2b. The converted primer and one of the
native primers were used in the PCR with the same stimulated B cell genomic
DNA as in the study of Sγ3.
56
Figure 3-6. The Sγ2b switch repeat alignment. The consensus sequence of
Sγ2b is shown at the top of the figure. The start of the Sγ2b is base A, shown in
bold. The end of the Sγ2b is base G, shown in bold. The dash means the
sequence is the same as the consensus sequence. The murine Sγ2b has 68
switch repeats and 5 degenerate repeats. Portions of switch repeats were found
as much as 250 to 300 bp downstream of the last degenerate repeat.

57
Figure 3-7. Single-strandedness on the top strand within and
downstream of the Sγ2b switch region in stimulated B cells. (A) The
experiment was done as in Figure 3-2A. In the diagram at the top, the half
ellipse represents the downstream part of Sγ2b region. The region between the
Sγ2b and the Cγ2b (constant region of the Igγ2b) is 1,837 bp. FTH71, a
converted primer, is located 400bp upstream of the end of Sγ2b. FTH73 is a
native primer that is located 100 bp downstream of Sγ2b. (B) The same DNA as
in Fig. 3-7A was analyzed using a different downstream native primer. FTH88,
a native primer, is located 800bp downstream of Sγ2b. The bottom line displays
every C residue on the top strand of the PCR product.

58
When the FTH71-FTH73 set was used, 16 independent molecules were
identified based on sequencing (Fig. 3-7A). Five molecules had long stretches
of conversion. Of that, one had a long stretch of conversion to FTH73. In the
FTH71-FTH88 set, 18 independent molecules were identified based on
sequencing. Four molecules had long stretches of conversion. Of that, one
molecule had long stretches of conversion to the FTH88 position. Hence, the
majority of R-loops at Sγ2b terminate within the switch repeats, and the others
terminate downstream.
Frequency of R-loop formation at switch regions in stimulated splenic B cells
Use of a converted primer on one side to detect R-loops makes it impossible
to quantitate their absolute frequency because it enriches for the R-loop
molecules at the expense of the non-R-loop molecules. We have developed a
method which now permits us to determine the frequency of the R-loops in
genomic DNA from primary B cells. Murine splenic B cells are stimulated for 2
days, genomic DNA is harvested, and subjected to bisulfite treatment (see
Materials and Methods). We then use two native primers, one on each side of
the switch region, to amplify the entire switch region. These PCR products are
ligated into a vector and transformed into E. coli. Colony-lift hybridization is done
59
using a 'converted' oligonucleotide probe which anneals to any bisulfite-reacted
form of the top strand in any R-loops. This method provides a minimal estimate
of the actual frequency of R-loop formation because only R-loops located at the
site of the converted oligonucleotide probe will be detected. R-loops at other
locations in the same switch region will be overlooked. Nevertheless, this
method provides an unbiased minimal assessment of the actual R-loop frequency.
Any clones that are detected are confirmed by sequencing. Using this method,
we have determined that that the minimum frequency of R-loops at Sγ3 is one per
570 alleles in splenic B cells at day 2 of stimulation. This determination is based
on detection of three R-loops of length 1.4 to 2 kb among 1700 alleles examined.

60
Figure 3-8. Plots of the downstream edge of the R-loops, the G density,
and the AGCT locations for murine Sγ3 through the initial portion of the first
constant exon. (A) The relative position of the 482 bp Iγ3 exon, the 805 bp
Iγ3-Sg3 intervening region, the 1,801 bp Sγ3, the 2,080 bp Sγ3-Cγ3 intervening
region, and the first 1,801bp of the Cγ3 region (exon/intron boundaries not
specified) are shown. R-loop downstream boundaries based on Figures 3-2 and
3-3 are shown as dots above or below the line, respectively. (B) The AGCT plot
of the (A) sequence. The short vertical bars on the plot indicate the AGCT sites.
(C) The G-density plot of the (A).
61
Figure 3-9. Plots of the downstream edge of the R-loops, the G density,
and the AGCT locations for murine Sγ2b through the initial portion of the
first constant exon. (A) The relative position of the 3807 bp Sγ2b, the 1837 bp
Sγ2b-Cγ2b intervening region, the 356 bp Ig2b-Sγ2b intervening region, the 400
bp Iγ2b exon, and 3807 bp of the Cγ2b constant exons/introns (boundaries not
shown). R-loop downstream boundaries based on Figure 3-6 are shown as dots
above (from Fig. 3-7A) or below the line (from Fig. 3-7B), respectively. (B) The
AGCT plot of the (A) sequence. The short vertical bars on the plot indicate the
AGCT sites. (C) The G-density plot of the (A).

62
DISCUSSIONS
R-loop endpoints correlate with G-richness
In previous work, we showed that the R-loops at Sγ3 and Sγ2b begin either
near the first repeat or somewhere within the switch region (Yu, Chedin et al.
2003). We initiated the current study to determine how quickly the RNA
polymerase, RNA, and template interactions revert to normal after the core switch
repeats have been passed at the downstream boundaries of the Sγ3 and Sγ2b
switch regions. Though 61-67% of the R-loops terminate within the core Sγ3 and
Sγ2b switch repeat regions, the others typically terminate within the 600 bp
immediately downstream of the core switch repeats (Fig. 3-5). This observation
is interesting when it is compared to the profile of G-richness across the Sγ3 and
Sγ2b regions and downstream toward the first constant exons (Figs. 3-8 & 3-9).
The G-richness drops to the random sequence level only gradually over the 600
bp downstream of Sγ3 and Sγ2b. The propensity for R-loop formation is thought
to be based on the G-richness of the non-template strand for reasons that are
discussed below. Our observation that the R-loops persist in the first few
hundred base pairs downstream of the switch region is probably related to this
elevated G-richness.
63
Table 3-1. The frequency of WRC and AGCT sequence motifs. Iγ3 is 805 bp
upsream of Sγ3. Cγ3 is 2,080 bp downstream of Sγ3. W represents A or T. R
represents A or G .

64
The class switch breakpoints correlate with AGCT density rather than WRC (W=A
or T, R=A or G) density
Of 23 breakpoints at Sγ3 in murine B cells, all were within the core switch
repeats (Dunnick, Hertz et al. 1993). However, at Sγ2b, two (out of 29
breakpoints) breakpoints in murine B cells were identified 150 bp and 600 bp
downstream of the core repeats (Dunnick, Hertz et al. 1993; Dunnick, Hertz et al.
1993). In addition, recombination breakpoints are found in the 5’ and 3’ flanking
regions of the Sµ core repeats (Lee, Kondo et al. 1998). One possible
explanation for the paucity of class switch recombination breakpoints downstream
could have been that the density of WRC sites (on the bottom strand) is much
lower in the region downstream of the switch region. However, upon
examination, the region downstream of the Sγ3 switch region is not significantly
reduced in WRC motifs (Table 3-1). In contrast, we find that the AGCT density
does correlate with the location of the downstream edge of the CSR breakpoints
(Fig. 3-8). AGCT is a palindromic form of the WRC motif, and AGCT is a distinct
motif in CSR sequences of all vertebrates that utilize class switch recombination.
The AGCT motif may be particularly important in achieving symmetrical action by
AID in order to generate double-strand breaks, as we have pointed out previously
65
(Yu, Roy et al. 2005; Yu, Roy et al. 2005). In contrast, somatic hypermutation
does not require this and relies more on nonpalindromic WRC motifs (Zarrin, Alt et
al. 2004) (Yu, Roy et al. 2005). This difference between AGCT sites, which are
palindromic and permit action by AID on both strands at the same position, and
nonpalindromic WRC sites, which permit action on only one strand, may be
fundamental to the difference between CSR and SHM. The sharp decrease in
AGCT at the end of the repetitive zone of the switch region may account for the
paucity of CSR sites downstream of the core repeat of the Sγ3 switch region, and
similarly for Sγ2b.
The Iγ3-Sγ3 and the Iγ2b-Sγ2b intervening regions are somewhat reduced for
AGCT density, but not as reduced as the intervening regions between the switch
regions and the constant regions. The G-density is also considerably reduced
upstream of the core switch repeats, but not as reduced as the constant regions.
The R-loops begin at or around the beginning of the core repeats (Yu, Chedin et al.
2003). Therefore, the onset of the R-loops requires the high level of G-density
intrinsic to the core switch repeats. The lack of recombination events upstream
of these two switch regions is likely because of the lack of R-loops here.
Additional factors, such as chromatin structure or more complex methods of CSR
66
targeting, may also contribute both upstream and downstream of the core switch
repeats.
Thermodynamic basis for the formation and stability of R-loops
Why and how do the R-loops form at the switch regions, and why are they
stable? The strength of RNA:DNA duplexes relative to the corresponding
DNA:DNA duplexes can be calculated (SantaLucia 1998). RNA:DNA duplexes
are always more stable (Fig. 3-10). The basis for this is not certain but may be
for several reasons. First, the RNA:DNA duplex is thought to adopt a
configuration that is closer to A-form DNA than B-form DNA. The A-form has a 3’
endo pucker of the sugar ring in contrast to the 2’ endo pucker in B-form DNA, and
this contributes to increasing stacking. Second, the RNA strand can form an
extra hydrogen bond between the 2’OH of the sugar at nucleotide i and the O4’ of
the sugar at the nucleotide at position i + 1. R-loops with a G-rich RNA, and
hence a G-rich nontemplate DNA strand, are by calculation particularly stable (Fig.
3-10). These features explain why the R-loop forms, but they do not explain how
it forms.

67
Fig. 3-10. Thermodynamic prediction of nucleic acid stability of RNA:DNA
versus DNA:DNA within the murine Sγ3 and Cγ3 regions and the Sγ2b and
Cγ2b regions. The figure represents the free energy for annealing of DNA:DNA
(D/D), DNA:RNA (D/R), and RNA:DNA (R/D) at the murine Sγ3 and Sγ2b regions.
More negative values of ∆G indicate greater stability of the nucleic acid form.
D/R represents R-loops formed upon transcription in the non-physiological
direction. R/D represents R-loops formed upon transcription in the physiological
direction (G-rich RNA). The length of Cγ3 used was the same as the length of
Sγ3 (1801 bp), and the length of the Cγ2b was the same as the length of the Sγ2b
(3807 bp). The difference in length accounts for why the energy scale for the γ2b
is larger than that for γ3.

68
The R-loops could conceivably form by either of two mechanisms. One
possibility is an extended RNA:DNA hybrid as is seen on single-stranded DNA
templates. In this model, the RNA, because of its G-richness and nearly A-form
structure, might not pass through the exit pore of the RNA polymerase (Gopal,
Brieba et al. 1999; Westover, Bushnell et al. 2004), but would remain annealed
with the template DNA. In the second model, the RNA would pass out of the exit
pore and then compete with the non-template DNA strand for annealing to the
template DNA strand upstream of the RNA polymerase (Drolet, Broccoli et al.
2003). There is no definitive evidence for or against either of these potential
mechanisms.
Frequency of R-loops at switch regions in stimulated B cells
We have been able to document chromosomal R-loops using normal PCR
primers, and this permits determination of the actual frequency of the R-loops
among alleles within stimulated B cells. This had not been possible previously
using pairs of primers in which one of the primers was 'converted.' The
frequency of one R-loop per 570 alleles at Sγ3 after 2 days of stimulation
compares to actual isotype switch (and surface positive for IgG3) of 3.3% of
murine B cells after 4 days of LPS stimulation. We only stimulate for 2 days in
69
the study here, and the probes for the R-loops cover only a fraction of the Sγ3
switch region. In addition, the R-loop is likely to be present for only a small
fraction of the time of stimulation and during only a fraction of the cell cycle.
Hence, this minimal estimate is well within a range that corresponds to switching
to Sγ3.

70
Chapter 4: Sequence-Dependence of Chromosomal
R-loops at the Immunoglobulin Heavy Chain Sµ Class
Switch Region

SUMMARY
The mechanism by which the cytidine deaminase, AID, acts at immunoglobulin (Ig)
heavy chain class switch regions during mammalian class switch recombination
(CSR) remains unclear. R-loops have been proposed as a basis for this
targeting. Here we show that the difference between various forms of the Sµ
locus that can or cannot undergo CSR correlates well with the location and
detectability of R-loops. The Sµ R-loops can initiate hundreds of base pairs
upstream of the core repeat switch regions, and this is where the AID mutation
frequency begins to rise, despite a constant density of WRC sites in this region.
The frequency of R-loops is one in 25 alleles, regardless of the presence of the
core Sµ repeats, again consistent with the initiation of most R-loops upstream of
the core repeats. These findings explain the surprisingly high level of residual
CSR in B cells from mice lacking the core Sµ repeats, but the marked reduction in
CSR in mice with deletions of the region upstream of the core Sµ repeats. These
71
studies also provide the first analysis of how R-loop formation in the eukaryotic
chromosome depends on DNA sequence.
INTRODUCTION
The Honjo laboratory discovered the key lymphoid-specific enzyme-AID for
both CSR and SHM (Muramatsu, Kinoshita et al. 2000) (Muramatsu,
Sankaranand et al. 1999). AID is a 24 kDa protein which deaminates C in DNA
(Di Noia and Neuberger 2002) (Petersen, Casellas et al. 2001), but only when
that DNA is single-stranded (Bransteitter, Pham et al. 2003) (Pham, Bransteitter et
al. 2003; Yu, Huang et al. 2004; Yu, Roy et al. 2005). A key question in CSR and
SHM concerns how the DNA becomes single-stranded. Because transcription
appears to be required for both CSR and SHM, one could propose that the 9 bp
bubble in the DNA (due to the RNA/DNA hybrid) created by the RNA polymerase
is sufficient to serve this purpose. This simple explanation alone is not adequate
because all transcribed genes in the genome would be mutated (as in SHM) or be
recombined (as in CSR). Therefore, there must be another explanation.
Moreover, the switch regions evolved hundreds of million of years after the
presence of AID because SHM is more ancient than CSR (Barreto,
Pan-Hammarstrom et al. 2005) (Ichikawa, Sowden et al. 2006) (Wakae, Magor et
72
al. 2006). Therefore, the unusual features of mammalian switch regions likely
target these regions for recombination rather than mutation.
For CSR, we have proposed that R-loop formation can improve the
availability of single-straneded regions in which AID can act (Yu, Chedin et al.
2003; Yu and Lieber 2003). Several groups, including ours, have demonstrated
R-loop formation at switch regions in vitro, when these are transcribed by
prokaryotic polymerases (Daniels and Lieber 1995) (Leung and Maizels 1994)
(Reaban and Griffin 1990) (Tian and Alt 2000). In 2003, we demonstrated
kilobase-length chromosomal R-loops in vivo at the murine Sγ3 and Sγ2b switch
regions (Yu, Chedin et al. 2003), and this has been confirmed (Ronai,
Iglesias-Ussel et al. 2007). We note that R-loops are not the only mechanism by
which AID can gain access to C's in duplex DNA. During SHM, transcription and
single-stranded binding proteins (RPA in particular) may function to liberate
lengths of single-strandedness (Chaudhuri and Alt 2004). In addition, CSR in
Xenpus occurs in a region that is not G-rich and does so at an efficiency that is
only about 4-fold lower than for the mouse Sγ1 region (Zarrin, Alt et al. 2004). It
is possible that the G-rich repeats of mammalian switch regions evolved to
improve the generation of single-stranded regions by 4-fold at each switch region.
73
For a donor and an acceptor switch region, this improvement may be
multiplicative (16-fold).
Previously, we were unable to carry out PCR across the Sµ region, and other
groups have had similar difficulty (Xue, Rada et al. 2006). Hence, Sµ, arguably
the most important switch region, has remained unexamined up to this point.
Mice with deletion of the core Sµ region (designated ∆SµTR) retain a surprisingly
high level of CSR (11-63%, depending on which acceptor switch region is utilized)
(Luby, Schrader et al. 2001). This finding posed a major challenge for any model
of CSR targeting because the vast majority of the AGCT sites, the clusters of G's
within the Sµ repeats, and the overall G-richness were all deleted. In contrast,
the more extensive deletion in the Iµ-Cµ deletion mice causes a reduction in CSR
to about 50-fold below wild type, demonstrating that some major difference
between the core Sµ and the Iµ-Cµ deletion mice must account for the efficiency
of CSR (Khamlichi, Glaudet et al. 2004), even though the core Sµ repeats are
missing in both mice. The difference between these two deletion mice (the
difference between locations labeled 2 versus 3 in Fig. 4-1A and between mouse
alleles 2 and 3 in Fig. 4-1B) must have some role in maintaining the efficiency of
CSR. What could the difference between these remaining sequences be?
74
Here we find that the less severely-deleted allele still permits R-loop formation,
but the larger deletion results in a marked reduction in R-loop formation.
Therefore, a major fraction of R-loop formation can begin upstream of the core Sµ
repeat region, and the large majoritiy of these R-loops initiate in a 50 bp region
that is 50% G-rich on the nontemplate strand. These observations help
reconcile many aspects of CSR targeting at Sµ and begin to establish DNA
sequence rules for where chromosomal R-loops initiate.
MATERIALS AND METHODS
Methods
All methods are described in Chapter three.
Oligonucleotides
The following primers were used in the PCR. For the amplification of the core Sµ
in the C57BL/6 mouse. FTH84, the 3 C’s converted primer, 5’-TGAGTTGGGG
TAAGTTGGGATGAGT-3’; KY293, the native primer, 5'–AACTCTACTGCCTACA
CTGGAC-3'; KY294, the native primer, 5’-CAGCACAATCTGGCTCACTT-3’. For
the amplification of the upstream and downstream regions of the core Sµ.
FTH111, 3 C’s converted primer, 5’-AGATAAGTTAGGTTGAGTAGGGTT-3’;
FTH119, 3 C’s converted primer, 5’-CTATTCTTTCTCAATTCTATACAACTA-3’;
75
Figure 4-1. Diagrams of the immunoglobulin µ locus in C57BL/6 and two
different Sµ deletion mice. The top diagram shows the map of the Eµ-Cµ
region in the C57BL/6 mouse. The vertical bar labeled as 2 indicates the
boundary of gene replacement in the core Sµ deletion mouse (∆SµTR) (Luby,
Schrader et al. 2001). The vertical bar labeled as 3 indicates the location of the
boundary of the gene replacement in the Iµ-Cµ deletion mouse (Khamlichi,
Glaudet et al. 2004). Eµ, intronic enhancer; Iµ, I µ exon; Cµ, µ constant region
exon. The arrows below the diagram indicate the locations of class switch
recombination breakpoints in the mouse. If the class switch recombination
efficiency in the C57BL/6 mouse is defined as 100%, then the efficiency in the
∆SµTR mouse is 35% (the efficiency varies for differenct acceptor switch regions,
ranging from 11 to 63 %), and the efficiency in the Iµ-Cµ mouse ranges from1-2
%.

76

77
FTH50, the native primer, 5’-TTGAAGGAACAATTCCACACAAA-3’; FTH94, the
native primer, 5’- CTGGGAGAACTATTCTCATCCCAAA-3’. For the amplification
of the Iµ-Cµ in the larger deletion of Sµ mouse. FTH52, 6 C’s converted primer,
5’-AATGGTAAGTTAGAGGTAGTTAT-3’; FTH51, the native primer, 5’-
CCCATGGCCACCAGATTCTTATC3’; FTH48, the native primer, 5’-
TCTCCATTCAATTCTTTTCCAATA-3’. For the probe of the hybridization in the
core Sµ in the C57BL/6 mouse. FTH98, 5’- ACTCAACTCAACTCAAC
TCAACTCAA-3’; FTH102, 5’- AATTCTAACCAACCAACTCTACTCA-3’; FTH104,
5’- ACTCAACTCAACTCAACCCAACTCAA-3’, FTH105, 5’- ACCCAACTCAAC
CCAACTCAACCCAA-3’. For the probe of the hybridization in the ∆SµTR mouse.
FTH130, 5’- ATACAACTATAACCTTCCTTCTACAT-3’; FTH131, 5’- CACATTAA
ATTATAAATCAAAAATATAATAA-3’; FTH132, 5’- CATCAACCAACCC
AATTAAATCCAA-3’; FTH134, 5’- ACTCAACCCAATTCATAATCCCAAT-3’;
FTH135, 5’- ATATAAATAACCCAAACAACAATACTC-3’. For the probe of the
hybridization in the Iµ-Cµ deletion mouse. FTH178, 5’- ACATATAAACT
AACTTAAAAACCCTTC-3’; FTH179, 5’- AAAAAACCCAAAATCCAAACCTAC
C-3’; FTH180, 5’- CTTTAAAAACAACAACCACAACTATAA-3’.

78
RESULTS
R-loops are detected in the core Sµ in the wild type (C57Bl/6) mouse
Previously, the R-loops were detected at murine chromosomal Sγ3 and Sγ2b
regions in stimulated B cells. Therefore, we believe that R-loops could be a
general feature in all switch regions. To further investigate the role of R-loops in
the class switch recombination, we tried to detect any R-loop in the Sµ region.
The Sµ region is the donor switch region, and hence, is the most important switch
region.
To detect the R-loops at Sµ region, the bisulfite modification assay was used
to study the DNA structure. First, long single-strandedness was detected by the
enrichment method, the converted / native PCR primers (Fig. 4-2). The 3Cs
converted primer was designed in the core Sµ region, and the native primer is
about 200 bp downstream of the end of the core Sµ. The PCR product is 1,253
bp. Because of the repetitive nature of the core Sµ, the converted primer can
prime at multiple locations. Therefore, the products of the PCR include multiple
species, most of which are shorter than the full-length Sµ; hence, a single product
band was not obvious at the full-length position. The region surrounding the
anticipated full-length DNA size was cut out, cloned, and transformed into E. coli.
79
Upon sequencing, molecules having different lengths of long stretches of
single-strandedness were observed. The longest was nearly 1,100 nts in length.
The majority of R-loops ended within the core Sµ. Six of 34 molecules extended
downstream of the core Sµ, and their endpoints fell within the region containing
several degenerate Sµ repeats. The sequenced molecules had different sizes of
deletions. In some cases, the deleted product was several hundred base pairs
shorter than the expected product. The reason for the deletions is that the core
Sµ region consists of uniform direct repeats, and the converted primer can anneal
at many regions within the core Sµ, not simply at the designated locations. In
addition, the Sµ sequence may be unstable in bacteria. Other groups have
previously reported deletions within the Sµ sequence.

80
Figure 4-2. Single-strandedness on the top strand at the murine core Sµ in
stimulated B cells in the C57BL/6 mouse. After bisulfite treatment, the region
was amplified using one native and one converted primer. The converted primer,
FTH84 (starts at nt 137,610 of Genbank accession no. AC073553), containing 3
C’s converted to 3’Ts, is located 250 bp downstream of the start of the core Sµ.
The native primer, KY294, is located 200 bp downstream of the end of the core Sµ.
The expected PCR product is 1,253 bp, but the PCR products are distributed from
this point to smaller sizes. The region of the gel at approximately 1,253 was cut
out and cloned. Thirty-three clones showed different lengths of R-loops in the
core Sµ. All clones had small deletions and mutations. Among all clones,
twenty-five clones had large deletions, and these are shown as shorter lines. All
large deletions occurred in the core Sµ region. The naive B cells are stimulated
with 20 µg/mL LPS for two days.

81

82
RNase H treatment is used to assess whether the long stretches of
single-strandedness are due to R-loop formation. The RNase H destroys the
RNA in R-loops, thereby permitting Sµ repeats in the 'top' (nontemplate) strand to
anneal to repeats in the 'bottom' (template) strand. If this occurs, then we expect
to lose the long stretches of bisulfite reactivity and the enriched primer can no
longer anneal to a site that is not converted. That is, the destroyed R-loops are
not amplified by the enriching primer after the RNase H treatment. One gapped
molecule was detected, which had discontinuous stretches of
single-strandedness. It consisted of several alternations between duplex DNA
and single-stranded DNA.
The frequency of R-loops at murine core Sµ
To determine the frequency of R-loops at Sµ in DNA from splenic B cells that
were stimulated with LPS+IL-4, the core Sµ was first amplified using a pair of
native primers, which avoids any enrichment. The native primers were close to
the core Sµ. The expected PCR product size is 1,518 bp, but the PCR products
were distributed over a range around1,518 bp, and this region was used for
cloning and transformation into E.coli. Colony-lift hybridization was done using a
'converted' oligonucleotide probe which anneals to any bisulfite-reacted form of
83
the top strand in any R-loops. The probe is actually a mixture of four
oligonucleotide probes, based on four different locations within the Sµ core region,
so that R-loops covering a greater fraction of the switch region can be detected.
After testing 921 colonies (see Methods), 296 were determined to provide
information about the top (nontempate) strand, whereas the remainder reflected
the bottom (template) strand. Colony-lift hybridization showed that 12 of these
molecules had long stretches of conversion consistent with R-loops (Fig. 4-3;
duplicate molecules are not shown). Hence, the R-loop frequency is
approximately 4% of alleles (12 out of 296). Of course, this can be an
underestimate of the actual R-loop frequency because the probes only cover a
small fraction of the region. Therefore, R-loops could exist between the probes,
or upstream or downstream of them. More importantly, R-loops that extend to
the site where the native primer is located will not be amplified because the native
primers will not anneal at priming sites that have had their single-stranded C's
converted by bisulfite. In other words, efficient native primer amplification rests
on the assumption that the native primer site is in a duplex DNA conformation.
For these and other reasons, this estimate of frequency may be an
underestimate.
84
Figure 4-3. R-loop molecules at the murine core Sµ in stimulated B cells in
the C57BL/6 mouse. Primary splenic B cells were stimulated in culture for 2
days with LPS+IL-4. The genomic DNA was prepared and then treated with
sodium bisulfite as described in the Methods and materials. A single round of
PCR (30 cycles) was done using two regular (native sequence) primers, KY293
and KY294 (shown as black arrows). The PCR product is 1,518 bp, and cloned
into the E.coli. About 1,106 clones were transferred to a filter membrane by
colony lifting. After hybridization with a probe, four positive clones were
identified. The sequence of these is shown here, and they have long stretches
of conversion, indicating long stretches of single-strandedness. The longest one
is 650 nts long. In the diagram at the top, the long ellipse represents the core
Sµ region, which is 1,287 bp long. The region between the core Sµ and the Cµ
is 1,535 bp. The red line in the diagram presents the location of the probe.
Each long line represents an independent molecular clone. The small vertical
bar on each line indicates a C on the sequence converted to a T. The bottom
line displays every C on the top strand of the PCR product.

85

86
The boundaries of the R-loops at murine Sµ in wild type and in the core Sµ
deletion (∆SµTR) mice
Class switch recombination breakpoints are not restricted to the core Sµ, in
contrast to most breakpoints at the acceptor switch regions (Dunnick, Hertz et al.
1993). At Sµ, somewhat more than half of the breakpoints are in the core Sµ
(Fig. 4-1). About one third are upstream of the core Sµ, and about 8 % are
downstream of the core Sµ. This raises the possibility that the R-loops at the Sµ
region may not be restricted to the core repeats of Sµ. To determine the
boundaries of the R-loops at Sµ, converted primers were designed for the
upstream or downstream regions of the core Sµ. For the upstream region of the
core Sµ, the PCR product is 770 bp. Upon cloning and sequencing, molecules
having different lengths of long stretches of single-strandedness were detected
(Fig. 4-4B). The shortest one was 54 nts long, and the longest one was 581 nts.
Abundant R-loops can initiate from upstream of the core Sµ region. This finding
is consistent with the facts that one third of CSR breakpoints observed in the
region. Therefore, locations of R-loops correlate to locations of CSR breakpoints.
The correlation implies that the important role of R-loops in class switch
recombination. For the downstream region of the core Sµ, the PCR product is
87
721 bp. Four molecules having long stretches of conversion were detected (Fig.
4-4D).
RNase H treatment was done to confirm that these molecules are R-loops,
and the treatment was done in the same way as for core Sµ. After the bisulfite
modification assay and PCR, no PCR product was detectable for the upstream
region of the core Sµ, indicating that these are, indeed, R-loops, and, therefore,
cannot be amplified by the converted primer after RNase H treatment (Table 4-1).
For the region downstream of core Sµ, the PCR product was cut out and cloned.
After sequencing 23 molecules, no molecule was found to have long stretches of
conversion. The fact that the long single-stranded DNA regions are eliminated
by RNase H documents that these regions are part of an R-loop.

88
Figure 4-4. Single-strandedness on the nontemplate strand upstream and
downstream of the murine core Sµ in stimulated B cells. (A) Top
(nontemplate) strand sequence at the upstream region of the core Sµ in the core
Sµ deletion mouse (∆SµTR). The genomic DNA was derived from B cells (from
the ∆SµTR mouse) stimulated for 2 days with LPS + IL-4. A single round of PCR
(30 cycles) was done using one regular (native sequence) primer and one
converted primer (shown as the green arrow) whose sequence is complementary
to the top strand where the C's have been converted to U's. FTH119, the
converted primer containing 3 C's converted to 3 T's, is located 400 bp upstream
of the core Sµ. FTH50, the native primer, is located 50 bp downstream of the Iµ
exon. The diagram at the top presents the map of the immunoglobulin µ gene of
the mouse. Eµ, intronic enhancer; Iµ, I µ exon; Cµ, µ constant region exon.
The vertical bar labeled as 2 indicates the boundary of knockout in the core Sµ
deletion mouse (∆SµTR). The vertical bar labeled as 3 indicates the location of
the boundary of the knockout in the larger deletion of Sµ mouse. All symbols
are the same as in Figure 4-2. (B) Top strand sequence at the upstream region
of the core Sµ in the C57BL/6 mouse. The genomic DNA was the same as in
Figure 4-2. (C) Top strand sequence at the downstream region of the core Sµ in
89
the core Sµ deletion mouse (∆SµTR). A single round of PCR (30 cycles) was
done using one regular (native sequence) primer and one converted primer.
FTH111, the converted primer containing 3 C's converted to 3 T's, is located 440
bp downstream of the core Sµ. FTH94, the native primer, is located 400 bp
upstream of the Cµ. (D) Top strand sequence at the downstream region of the
core Sµ in the C57BL/6 mouse.

90
Table 4-1. Single-strandedness was destroyed after in vitro RNase H
treatment. The numbers indicate molecules containing long stretches of
conversion divided by the total number of sequenced molecules. The RNase H
treatment removes the RNA in the R-loops. (The cut-off for long stretches of
conversion is 50 nts long.)

91
The same primer sets were used for the core Sµ deletion mouse (∆SµTR
mouse) to detect any R-loops in the stimulated B cells of this mouse. At these
core Sµ deleted alleles, the class switch recombination efficiency is only about
one third that of the wild type mouse (Fig.4-1). (The efficiency varies depending
on which acceptor switch region is involved.) R-loops were detected upstream
and downstream of where the core Sµ region was formerly located (Fig. 4-4A & C),
and these were indistinguishable from the R-loops found in these regions of the
wild type allele mentioned above. The R-loop nature of these molecules was
confirmed by RNase H treatment (Table 4-1). Therefore, R-loops can initiate
upstream and can extend downstream of the core Sµ region.
The frequency of the R-loops at murine Sµ in the core Sµ deletion (∆SµTR) B
cells
To determine the frequency of R-loops in the ∆SuTR murine B cells, a pair of
native primers was designed to amplify the Iµ-Cµ region from the stimulated B
cells of these mice. One primer is in the Iµ exon, and the other one is in the Cµ
region. The PCR product is 2,251 bp. The composite probe for hybridization,
which is specific for converted top strand molecules, contains five different
oligonucleotides probes: three were positioned at different locations in the
92
region upstream of the former core Sµ, and two were positioned downstream.
After testing 932 colonies (see Methods), 311 were determined to provide
information about the top (nontemplate) strand, whereas the remainder reflected
the bottom (nontemplate) strand. Among these colonies, eleven molecules
having long stretches of conversion were identified (Fig.4-5). Therefore, the
frequency of R-loops at the ∆SµTR allele at day 2 of IL-4+LPS-stimulation is 3.5
% (11 out of 311).

93
Figure 4-5. The R-loop molecules at the murine Iµ-Cµ region in stimulated B
cells from the core Sµ deletion (∆SµTR) mouse. After the bisulfite treatment
of the genomic DNA, a single round of PCR (30 cycles) was done using two
(native sequence) primers, FTH50 and FTH51. The PCR product was 2,251 bp
and was cloned into E.coli. About 932 clones were put onto a filter membrane by
colony lifting. After probe hybridization, eleven positive signal clones having long
stretches of conversion were identified. The diagram at the top shows the map
of the Iµ-Cµ at the ∆SµTR allele. The arrow in the diagram indicates the loxP
sequence at this allele. The dashed line indicates no sequence information in
that portion of the clone. The vertical hollow box marks the position of
separation of the dashed line from the solid line. All symbols are the same as in
Figure 4-2.

94

95
The secondary structure formed within the R-loop due to the loxP site
When we analyzed the data from the ∆SµTR locus (Fig. 4-5), we found a
consistent, but slightly variable lack of conversion in the DNA at the center of the
R-loops within the boundaries of where the knock out deletion occurred (deletion
boundaries designated by the number 2 in Fig. 4-5) in ten of eleven molecules.
We were initially quite perplexed by this because we felt that the R-loops should
extend right through this 93 bp region of prokaryotic DNA. Some of the sites that
failed to react with bisulfite were CpG sites, and DNA methylation of the C in these
could make them unreactive with bisulifte. Though this could explain some of
the instances when these C's were unreactive, it could not explain all aspects of
the observed pattern of reactivity and unreactivity. We then realized that the
palindromic loxP site is located in the middle of this region. The palindromic
nature of the loxP site permits it, when single-stranded, to form a 13 bp stem and
an 8 nt terminal loop (Fig. 4-6). The C's in the loop always reacted with bisulfite,
but three C's in the stem only rarely did. Two of these 3 are complicated by
possible CpG methylation, but the remaining one was not a CpG site and is best
explained by base stacking in the stem of the stem-loop. A DNA methylation
assay confirmed that the non-reactivity of that C in the stem of the stem-loop
96
could not be accounted for by DNA methylation (Fig. 4-7). Moreover, two other
C's were substantially more unreactive than their methylation status could
account for. This lack of reactivity in the 3 C's of the stem is quite consistent with
protection of the stem region from bisulfite reactivity in the native bisulfite
modification assay for assessing single-strandedness. These results provide an
independent indication that this region in the middle of the non-template DNA
strand of the R-loop is single-stranded. This strand can form secondary
structures that interfere with bisulfite reactivity by permitting stacking of the C's,
but this is not usually observed, except, for example, in palindromic regions like
the one described. The lack of such interruptions in a majority of R-loops that
extend through this region on the wild type allele is most consistent with a lack of
any other secondary structure (such as G-quartet formation).

97
Figure 4-6. The stem-loop structure of the loxP site in the middle of the
R-loop at the Iµ-Cµ region in the ∆SµTR mouse. The top diagram presents
the putative structure at the Iµ-Cµ region in the ∆SµTR mouse. A stem-loop (at the
lox P site) can form in the middle of the R-loop within the nontemplate DNA strand.
The bold arrow represents the RNA annealed to the template DNA strand. The
bottom diagram is the magnified stem-loop, shown as detailed sequence. The
region contains both C's of CpG's and non-CpG's. The first row of numbers
indicates the converted frequency of the C's from ten molecules of Figure 4-5.
The higher this fraction, the more single-strandedness at the specified C. The
numbers in italics indicate the DNA methylation analysis of each C. The higher
this number, the more DNA methylation at this C (DNA palindromes can
sometimes interfere with DNA methylation analysis). The detailed DNA
methylation pattern of individual molecular clones (alleles) is shown in Fig. 4-7.
When the bisulfite modification fraction is higher, it indicates that these C's are
more vulnerable to bisulfite modification, because the region is single-stranded
(this assumes that DNA methylation does not block the bisulfite reactivity). Note
that the unconverted C's of the CpG's can be protected by DNA methylation or by
base stacking, but the unconverted C's of the non-CpGs can be protected only by
98
stacking. For example, the third C is 100% converted by bisulfite in the bisulfite
modification analysis (10/10) because it is single-stranded, and there is no DNA
methylation (0/30) that would interfere with the bisulfite modification analysis.
The fourth C, a non-CpG, is only 20% converted, because it is in the stem part of
the stem-loop, and therefore protected from bisulfite by stacking. The sixth and
seventh C's are always converted by bisulfite, and they are in the loop part of the
stem-loop (and DNA methylation does not interfere with either one of them).

Figure 4-7. DNA methylation analysis of the lox P sequence in the middle
of the Iµ-Cµ region of the ∆SµTR allele. In the diagram, each circle presents
the C in the exogenous (lox P) sequence in the middle of the Iµ-Cµ region in the
∆SµTR mouse. The detailed sequence is shown in Fig. 4-6. A cross in the
circle indicates that the C is a CpG. A filled circle indicates that the C is
unconverted in the DNA methylation assay, which means that it is methylated.

99
Figure 4-7. DNA methylation analysis of the lox P sequence in the middle
of the Iµ-Cµ region of the ∆SµTR allele. In the diagram, each circle presents
the C in the exogenous (lox P) sequence in the middle of the Iµ-Cµ region in the
∆SµTR mouse. The detailed sequence is shown in Fig. 4-6. A cross in the
circle indicates that the C is a CpG. A filled circle indicates that the C is
unconverted in the DNA methylation assay, which means that it is methylated.

100
R-loops are not detected at alleles from mice that have a wider deletion of core Sµ
and surrounding sequences
A larger region between Iµ-Cµ was previously removed in a different knock
out allele in mice (Khamlichi, Glaudet et al. 2004). In the B cells of these mice,
the class switch recombination efficiency is only about 2% of that seen in the wild
type mouse (Khamlichi, Glaudet et al. 2004) (Fig.4-1). (The efficiency varies
depending on the acceptor switch region.) The converted primer, FTH52,
containing 6 C's converted to T's, was designed to test for the presence of
R-loops in the B cells of this mouse. This is a strongly enriching primer because
of the large number of C's converted to T's. The PCR product is 505 bp. After
sequencing 14 molecules, no molecules containing long stretches of conversion
were found, despite use of the very strongly enriching primer (Fig. 4-8).
Therefore, no R-loops are detected at alleles that have a more extensive deletion
surrounding Sµ.
Although no R-loops are detected by the enrichment method, the colony-lift
hybridization was done to further search for any R-loops at this more
widely-deleted allele. To avoid missing any R-loops at the upstream region,
where the loci of CSR breakpoints are located in B cells of mice with this allele,
101
the PCR product was amplified using one more upstream native primer, which is
in the Iµ exon, paired with the same downstream native primer, as was used in the
∆SuTR mice. The PCR product was 1,018 bp. The composite probe for
hybridization, which is specific for converted top (nontemplate) strand molecules,
contains three different oligonucleotides probes. After testing 1,029 colonies
(see Methods), 450 were determined to provide information about the top strand,
whereas the remainder reflected the bottom (template) strand. Among these
colonies, no molecules containing long stretches of conversion were detected.
Therefore, the frequency of R-loops at this allele is below one per 450.

102
Figure 4-8. Test for single-strandedness on the nontemplate strand at
murine Iµ-Cµ in the stimulated B cells from mice having the larger deletion
around Sµ. The genomic DNA was from B cells stimulated for 2 days with
LPS+IL-4 from the mice with a larger deletion around Sµ. A single round of PCR
(30 cycles) was done using one regular (native sequence) primer and one
converted primer. FTH52, the converted primer containing 6 C's converted to 6
T's, is located 200 bp downstream of the Iµ exon. FTH51, the native primer, is
70 bp downstream of the start of the Cµ region. The PCR product is 505 bp,
including the 175 bp heterologous DNA from the construct for making the deletion
allele. Among 14 different clones, no molecule had long stretches of conversion.
All symbols are the same as in Figure 4-2

103
DISCUSSIONS
In vivo R-loop formation has been described at prokaryotic replication origins
(Masukata and Tomizawa 1990), yeast mitotic recombination hotspots (Huertas
and Aguilera 2003), avian G-rich sequences (under specific circumstances) (Li
and Manley 2005), and immunoglobulin class switch regions (Shinkura, Tian et al.
2003; Yu, Chedin et al. 2003; Zarrin, Tian et al. 2005; Huang, Yu et al. 2006).
Despite these key findings in various systems, the in vivo sequence determinants
of R-loop formation have yet to be explored. The studies here provide an initial
set of functionally relevant sequences that help delimit what regions do and do not
form R-loops in vivo.
Sites of R-loop initiation, of class switch recombination, and of AID deamination
R-loops are found at the ∆SµTR allele at a frequency that is similar to that at
Sµ in the wild type mouse. No R-loops are detected in the Iµ-Cµ deletion mouse.
Therefore, the sequences upstream of the core Sµ repeats can be important sites
for initiating R-loops. The findings correlate well with the findings of CSR
efficiency (Luby, Schrader et al. 2001) (Khamlichi, Glaudet et al. 2004), which
implies that the R-loops are the targets in the class switch recombination. For
the wild type and the ∆SµTR loci, respectively, we found a frequency of one
104
R-loop per 25 and per 28 alleles using native primers (no enrichment) in B cells
stimulated with LPS+IL4. The agreement of these numbers is encouraging,
given that the native primers used were different. (The primers for the wild type
allele must be adjacent to the core Sµ repeats to detect a PCR product.)
The results of this study indicate a substantial correlation between the
location of R-loops, the location of recombination breakpoints, and the location of
AID sites of C to U conversion (Figs. 4-9 & 4-10). Though we believe that many
R-loops initiate within the core Sµ repeats, unexpectedly, we find that many
R-loops can begin upstream of the core Sµ repeats (Fig. 4-3). In fact, we think
that the majority (but not all) of R-loops initiate in the region upstream of the core
Sµ, because there is not much difference in the R-loop frequency between wild
type and ∆SµTR alleles.

105
Figure 4-9. Plots of R-loop location, class switch recombination breakpoints,
AGCT sites, and G density for murine Sµ through the initial portion of the
first constant exon. (A) The relative position of the 313 bp Eµ, 454 bp Iµ exon,
the 1,234 bp Iµ-Sµ intervening region, the 1,517 bp Sµ, the 1,370 bp Sµ-Cµ
intervening region, and the first 1,517 bp of the Cµ region (exon/intron boundaries
not specified) are shown. R-loop upstream and downstream boundaries based
on Figure 4-4 are shown as lines above the diagram. The locations of class
switch recombination breakpoints in different mice are shown as arrows in the
diagram. (B) The AGCT, WRCW, 4G cluster, and 3G cluster plot of the
sequence in (A). (C) The G-density plot of the sequence in (A).

106

107
Figure 4-10. Plots of R-loop location, mutation frequency, AGCT sites, and G
density for the region upstream of the core Sµ. (A) A magnified region
upstream of the core Sµ of Fig.4-9. Each line represents different R-loops
initiating at different places upstream of the core Sµ. (B) The mutational
frequency of the corresponding region from the msh2-/-ung-/- mice published by
others (Xue, Rada et al. 2006). (Note that only the top (nontemplate) strand
mutations are shown (provided by Dr. M. Neuberger), rather than the composite
top (nontemplate) and bottom (template) strand mutation data that is published
(Xue, Rada et al. 2006).) The R-loop data in the current paper permitted
prediction of where the nontemplate strand mutation frequency would start to
increase without being provided with the detailed data by Xue et al.) (C) The
AGCT, WRCW, 4G cluster, and 3G cluster plot of the (A) sequence. (D) The
G-density plot of the (A). The dashed box highlights the region where major
R-loops start and the mutation frequency rises.

108

109
What might be the mechanism for initiation of R-loops at these upstream
locations? R-loops that begin upstream of Sµ do so at clusters of G's. We do
not know the precise combination of clustered G's that are important for the
initiation of R-loop formation, but we know that relatively random DNA sequence
does not cause R-loop formation in the mammalian genome. For example, even
the most strongly enriching converted primers (7 C to T conversions out of 22 nt)
did not permit detection of R-loops in the Cγ3 region under conditions where
R-loops could be readily detected at Sγ3. Moreover, even when two enriching
(converted) primers were used at the Cγ3 region to achieve even greater
sensitivity, no single-strandedness was detected on either the template or
nontemplate strand. Hence, R-loop formation does not occur in the constant
regions, despite being in the same transcription unit. Therefore, specific patterns
and lengths of G clustering increase the probability of R-loop formation, and
deviations from those patterns make R-loops increasingly less likely.
The majority of R-loops that initiate upstream of Sµ begin at a specific site
with the sequence, GGGGCTGGGG, which is within a 50 bp zone that is 50%
G-rich on the nontemplate strand (Fig. 4-4, 4-5, 4-10). Interestingly, this is also
close to where AID-mediated C to U deaminations begin to show a distinct
110
increase on the top strand (within the zone indicated by a dashed rectangle in Fig.
4-10). In a recent paper from the Neuberger laboratory, it was unclear what
could account for the increase in AID mutation frequency at this region because
the AGCT (and other preferred WRC sites) density did not yet increase until
several hundred base pairs downstream (Xue, Rada et al. 2006). The peak of G
density and clusters of GGGG or GGG in this region where most of the upstream
R-loops initiate appear to account for the sharp rise in AID sites of action here
(Figs. 4-9 & 4-10) (Xue, Rada et al. 2006; Xue, Rada et al. 2006).
Among R-loops detected in wild type or core Sµ deleted (∆SµTR) B cells,
nine R-loops begin even further upstream than those demarcated within the
dashed rectangle in Figure 4-10. Seven of these initiate within a region with
clusters of 3 G's, and the remaining two also initiate near several clusters of 3 or 4
G's. The latter two initiate upstream of the boundary where the Iµ-Cµ deletion
boundary (designated by the number 3) is located. Rare R-loops initiating here
may explain the low but distinct occurrence of class switch recombination events
that have breakpoints within the residual region between Iµ and Cµ in these B
cells (Khamlichi, Glaudet et al. 2004).

111
The core repeats are not essential for R-loop formation (this study) or for
CSR (Luby, Schrader et al. 2001). Initiation of R-loops upstream of the core
repeats may be evolutionarily advantageous because it permits a greater fraction
of the zone of WRC sites in the core Sµ repeats to be included in a zone of
single-strandedness. Of course, the GGGGT portions of the Sµ repeat can also
initiate R-loops, accounting for the R-loops initiating within the core Sµ.
In the Sµ core deletion mice (designated by the 2 in Fig. 4-5), a lower level of
residual CSR breakpoints occur upstream or downstream of where the deletion
boundary is located (Min, Rothlein et al. 2005) (Min, Schrader et al. 2003), and
the ones occurring upstream would be particularly difficult to explain if R-loops
initiated only in the core Sµ region. Because we now know that the R-loops can
begin upstream, the breakpoints upstream of core Sµ in those mice can now be
more easily understood. The CSR breakpoints downstream of the core Sµ are
not difficult to explain, given that R-loops extend past the repetitive zone (Fig.
4-4C & D).
Heterogeneity of R-loop termination points
Like Sγ3 and Sγ2b, the Sµ R-loops are heterogeneous in their termination
points. Many R-loops terminate within the core Sµ repeats. We have not been
112
able to quantify the abundance of R-loops within the core region relative to the
region upstream or downstream because PCR through the core Sµ is only
successful when the primers are close to the borders of the core Sµ repeats. But
using native/converted primer pairs, we find abundant numbers of R-loops in the
core Sµ region (Fig.4-2). The termination points of R-loops that extend
downstream of the core Sµ repeats reside within the region where the G density is
declining but still higher in density than random DNA. This is similar to what was
found at the Sγ3 and Sγ2b repeats (Huang, Yu et al. 2006), and it is consistent
with the emerging view that clusters of G's (and, indirectly, overall G density) are
important for R-loop formation and stability.
Collapsed R-loops and the location of long consecutive regions of AID action
In the ung-/-msh2-/- null mice. the pattern of AID C to U deamination sites
appears to occur in long discontinuous tracts of up to several hundred base pairs
(Xue, Rada et al. 2006). This is far too long to be attributable to any processivity
of AID along one strand, which is limited to much shorter lengths (Pham,
Bransteitter et al. 2003). The separation of DNA strands during normal
transcription occurs for a length of approximately 9 bp (Westover, Bushnell et al.
2004), which is also far too short to account for the very long tracts of AID action.
113
Finally, separation of strands due to transient negative supercoiling associated
with transcription would result in shorter tracts of single-strandedness, and this
would occur on both strands (Liu and Wang 1987).
The several hundred base pair tracts of AID conversion in ung-/-msh2-/- B
cells on the top (nontemplate) strand (upstream, downstream and within core Sµ)
could be attributed to R-loops, based on the findings here. But what could be the
basis for tracts of conversion on the bottom (template) strand in these
ung-/-msh2-/- B cells (Xue, Rada et al. 2006; Xue, Rada et al. 2006)? We
believe that exposure of single-stranded regions on the template strand of the
R-loops can arise in either of two ways. The RNA of the R-loop is vulnerable to
the endogenous endonucleolytic action of RNase H1 or H2 within the cell. First,
partial action by RNase H may give rise to the gapped regions of
single-strandedness on the top strand that we have observed here for Sµ and
previously for Sγ3 and Sγ2b (Yu, Chedin et al. 2003) (Fig. 4-11A). Second,
complete action by RNase H may give rise to misalignment due to collapse of a
repeat on the nontemplate strand that does not match the corresponding repeat
on the template strand (Fig. 4-11B) (Yu, Chedin et al. 2003; Yu and Lieber 2003).
We have previously described evidence for both of these mechanisms of RNase
114
H-mediated exposure of regions of single-strandedness on the template strand.
Using purified plasmid R-loops digested in vitro, we have shown evidence for
collapsed R-loops (Yu, Chedin et al. 2003). Incomplete digestion of
chromosomal R-loops with exogenous RNase H gives results that are quite
consistent with partially-digested R-loops (Yu, Chedin et al. 2003; Huang, Yu et al.
2006). Hence, both of these RNase H-digestion products of R-loops may exist
(Yu and Lieber 2003) (Fig. 4-11A & B). Either mechanism would leave long
tracts of single-strandedness on the template strand of the type suggested by the
pattern of AID action (Xue, Rada et al. 2006). In the ung-/-msh2-/- B cells, it is
noteworthy that the discontinuous tracts of AID conversion on the nontemplate
strand are, on average, about twice as long as the ones on the template strand,
despite the fact that the density of C's on the template strand is about 1.5-fold
higher (Xue, Rada et al. 2006). The two mechanisms for template strand
single-strandedness suggested here would be consistent with the favoring of
longer stretches on the nontemplate strand seen in the ung-/-msh2-/- B cells.
For R-loops in the repetitive zone, collapsed R-loops and the partially
digested R-loop are both reasonable explanations. For R-loops that begin
upstream or extend downstream of the core Sµ repeats, the DNA is not markedly
115
repetitive. Hence, tracts of AID action are more directly explained by the partial
RNase H model (Fig. 4-11A).
Secondary structure within the G-rich DNA strand of the R-loop
Another group has proposed that G-quartets might form along the G-rich
DNA strand of the R-loops (Duquette, Handa et al. 2004). There is no in vivo
data for G-quartets at these sites to our knowledge. One might have expected
that G-quartets would cause base pairing of other sites along the G-rich DNA
strand of the R-loop, and this would cause single nucleotide interruptions in the
continuous sites of bisulfite conversion on the G-rich strand. We do not see
evidence of this here or previously (Yu, Chedin et al. 2003). One could wonder if
the bisulfite method would be sensitive to such fine-structural features along the
G-rich DNA strand. Our ability to detect a small stem-loop structure on the
G-rich DNA strand indicates that small regions of secondary structure within a
strand are readily discernable (Fig. 4-5 and Fig. 4-6).

116
Figure 4-11. Two mechanisms by which endogenous RNase H activity
could generate single-stranded regions on the bottom (C-rich) DNA strand
in or around the switch regions. (A) Partial endonucleolytic action by
endogenous RNase H removes part of the RNA, but not other parts. This may
expose regions of single-strandedness on the C-rich template strand. The loops
on the bottom strand may account for regions of AID action on this strand (Xue,
Rada et al. 2006). The regions of single-strandedness on the top strand would
be a direct result of the R-loop itself. (B) Complete endonucleolytic action by
endogenous RNase H removes the RNA, perhaps at more than one location
along the RNA. The repeats on the top (nontemplate) strand anneal with the
repeats on the bottom (template) strand, but not necessarily in the correct order.
This would generate heterologous loops on the bottom strand, in addition to loops
on the top strand (Yu, Chedin et al. 2003; Yu and Lieber 2003). The loops on the
top and bottom strands may account for regions of AID action (Xue, Rada et al.
2006).
117

118
CHAPTER 5: Concluding Remarks
In mammals, SHM and CSR are two genetic modification events permitting
high specificity and different effector functions for antibodies in the body. AID is
the important enzyme and indispensable in SHM and CSR. AID causes point
mutations in VDJ and VJ segments and double-strand breaks in switch regions in
SHM and CSR, respectively. Since AID is a single-stranded DNA deaminase, it
is highly mutagenic. However, AID rarely deaminates non-immunoglobulin
genes. To date, it’s still unclear what the targeting mechanism of the AID action
is. How could AID target the immunoglobulin gene but not non-immunoglobulin
genes in the genome? How could AID target the VDJ segment and switch
regions but spare constant regions?
In our studies, we investigated the targeting mechanism of the AID in CSR.
In chapter two, we found 5’-WRCr-3’ is the AID sequence preference motif based
on biochemical studies with purified mouse GST-AID overexpressed in insect
cells. The AID hotspot motif corresponds to the hotspot motif in somatic
hypermutation, 5’-RGYW-3’, which indicates that the SHM hotspots are derived
from AID activity (Yu, Huang et al. 2004). In chapter three, the downstream
boundaries of R-loops at murine Sγ2b and Sγ3 are correlated with the G density of
119
the region. In addition, the frequency of the R-loop at murine Sγ3 is one out of
570 alleles at day two of stimulation (Huang, Yu et al. 2006). In chapter four, to
further understand the AID targeting mechanism, we analyzed R-loops at the
murine Sµ region and the frequency of them is one out of 25 alleles at day 2 of
stimulation using either LPS only or LPS plus IL-4. The frequency of R-loops is
related to the CSR efficiency. Furthermore, the R-loop initiation site is correlated
with the AID mutational pattern, which indicates that the R-loop is the target for
AID to act in CSR (Huang, Yu et al. 2007).
The difference between the donor switch Sµ and downstream acceptor switch
regions
CSR occurs between the donor switch region, Sµ, and one of the
downstream switch regions upon different cytokine stimulations. Therefore, the
Sµ plays a critical role in CSR. We found that the R-loop frequency is much
lower in the acceptor switch region, Sγ3, (1/570) than the donor switch region
(1/25). It’s not surprising that the donor switch region behaves differently than
the acceptor switch region, since the donor switch region has many features,
which are discussed below, different from acceptor switch regions. Therefore,
the detailed switching mechanism may be different in Sµ and other switch regions.
120
First, the identified CSR breakpoints in lymphomas and myelomas are exclusively
in the switch repeat regions in acceptor switch regions (rather than upstream or
downstream of the repetitive core of the acceptor switch regions) (Dunnick, Hertz
et al. 1993). In contrast, while most CSR breakpoints are in the core Sµ region,
one third of CSR breakpoints are in region upstream of the core Sµ, and 7 percent
of them are downstream of the core Sµ region. Second, the Sµ region is
constitutively transcribed in B cells, but other switch regions are only transcribed
upon cytokine stimulation. The main promoter is closer to Sµ than any of the
acceptor switch regions. Hence, the germline transcription of Sµ is more
abundant than those of the acceptor switch regions. Third, the switch repeats in
the acceptor switch regions are more critical than those in the donor switch region.
The impact of switch repeat deletion in KO mice is more severe if a downstream
acceptor switch region is deleted than if Sµ is deleted. The switching of B cells
to Ig1 is abrogated in the Sγ1 deletion mice (Shinkura, Tian et al. 2003).
However, the CSR efficiency only drops twofold or threefold in the core Sµ
deletion mice (Selsing 2006). Furthermore, the splicing of the germline
transcripts is required for the switching to Ig1 as shown in the splicing site of Sγ1
knockout mice (Lorenz, Jung et al. 1995) (Hein, Lorenz et al. 1998).
121
Nevertheless, the splicing site of Sµ is dispensable for the CSR to any isotype
(Kuzin, Ugine et al. 2000).
Evolutionary point of view of switch regions
Among three genetic modification mechanisms in the adaptive immune
system- VDJ recombination, SHM and CSR, CSR evolved the most recently
(Stavnezer and Amemiya 2004). Cartilaginous fishes have VDJ and SHM, but
not CSR. Although AID is present in the shark, amphibians are the first creature
known to have CSR. Through evolution, some features of switch regions
change, like switch repeat unit sequence, the length of the switch region, and in
particular, the GC content of the switch region. The only unchanged feature is
that switch regions consist of the switch repeat, which contains a high density of
WGCW motif (the palindromic SHM hotspot motif). In Xenopus switch
sequences, they are highly AT rich (65 %), and the repeat unit is 150 bp
(Mussmann, Courtet et al. 1997). In bird switch sequences, they are enriched in
GC content (55-60 %), and the G:C ratio is 1.2 and 0.6 in chicken and duck Sµ,
respectively (Stavnezer and Amemiya 2004). In human and mouse switch
sequences, the GC content is similar as in birds. Furthermore, they have the GC
asymmetry, and the G: C ratio is 3. The average GC content of the genome is
122
40.9 % for Xenopus, 45.0% for chicken, and 40.3 % for human and mouse (Kitao,
Arakawa et al. 2000). The role of the GC asymmetry in CSR has been
investigated. Upon transcription, the switch region can produce an RNA-DNA
hybrid because of the GC asymmetry. We and subsequently others further
found that the advantage of the GC asymmetry in mouse and human switch
sequences is its capability of the R-loop formation in vivo, which can provide the
single-strandedness for AID to act on. The R-loop only forms upon the
transcription producing the G-rich RNA, but not the other direction. Since the
R-loop provides the single-strandedness for AID to act, it’s possible that the
R-loop formation enhances the CSR efficiency. In the Sγ1 inversion mouse,
which can’t produce the R-loop at the Sγ1 region, the CSR efficiency drops
fourfold relative to the wild type mouse (Shinkura, Tian et al. 2003).
Xenopus switch sequence (4kb) can replace the murine Sγ1 (10 kb) and
sustains one fourth of the CSR efficiency of the wild type (Zarrin, Alt et al. 2004).
Considering the length of the replacement, the Xenopus switch sequence has
about 60% of CSR efficiency as the wild type. Since the Xenopus switch
sequence is AT rich, it’s unlikely that they can form the R-loop. No R-loop was
found in this knockin chimera (unpublished). How could the Xenopus switch
123
sequence support CSR in the mouse? The Xenopus switch sequence has a
high density (5.99%) of WGCW motif (the palindromic WRC motif), compared
with 3.84 % in the murine Sγ1. Therefore, the Xenopus switch sequence has
the ability to compensate the loss of R-loop formation in the murine genome by
having a WGCW density that is 1.6-fold higher than mouse Sγ1.
Mutation vs. R-loop formation
We proposed that mammalian switch regions provide a higher density of
WGCW sites (palindormic WRC motif) and higher G content (the R-loop formation)
to provide more AID target sites and single-strandedness to enhance the CSR
efficiency. In our studies, we provided results indicating that the R-loop is the
target for AID to act in CSR.
First, we found that the capability of R-loop formation is correlated with the in
vivo CSR ability by the comparison of two Sµ deletion mice, ∆SµTR and the larger
Sµ deletion mice, with the wild type mice. In the ∆SµTR mice, the CSR
efficiency is about one third of that of wild type and has a similar R-loop formation
ability. It can be explained that the R-loop formation is saturated in this mouse.
However, without enough WGCW sites (the core Sµ contains abundant WGCW
sites), fewer DSBs created at the Sµ region and results in a lower CSR efficiency.
124
In the larger Sµ deletion mice, the CSR is severely impaired and no R-loop is
detected.
Second, the strongest evidence supporting the R-loop as the target for AID to
act is the mutation profile at Sµ region in the Ung-/-Msh2-/- mice (Xue, Rada et al.
2006). The AID mutation profile is more correlated to the R-loop profile than the
WGCW profile. If the high content of WGCW site is the target of AID, then AID
would simply deaminate its preferred hotspots, and the mutation profile would
corresponds to the WGCW profile, which is the hotspot motif for AID. However,
this is not the case based on two observations. The AID mutations increase at
the start points of R-loops, but not at the initiation points of increasing WGCW
sites. The R-loop initiation sites can explain why 35% of CSR breakpoints are
found in the upstream region of the core Sµ, since in this region, there are fewer
WGCW sites compared to the core Sµ region. Moreover, the Cµ region contains
similarly low numbers of WGCW sites as the region upstream of the core Sµ and
the mutation frequency is low. No R-loops are found extending to the Cµ region.
Unsolved questions in the targeting mechanism of AID in CSR and SHM
There are several hypotheses for the targeting mechanism in CSR; however,
none has direct evidences yet. Our studies suggest that the R-loop is the
125
mechanism for targeting AID at CSR regions. The other possible mechanisms
are other transcription-related DNA structures (described in Chapter 3),
chromosomal structure, and other protein factors guiding AID to the target genes.
They will be discussed in detailed in the following paragraphs.
Chromosomal structure and modifications: One possible mechanism of AID
targeting mechanisms is the status of chromosomal modifications of switch
regions and variable regions for CSR and SHM, respectively. Several labs
checked different chromosomal modifications in B cell activation, including
histones H3 and H4 acetylation and methylation. One laboratory reported that
histones H3 of switch regions in primary murine B cells are hyperacetylated,
which is likely correlated to germline transcription (Li, Luo et al. 2004). Upon
activation of CSR in B cells, the acetylation of histones H4 is slightly increased at
switch regions. The histones H3, H4 acetylation is abolished in AID knockout
mice (Wang, Whang et al. 2006). In addition, phosphorylation of histones H2B is
higher at switch regions and variable regions of Igλ in activated cells than in naïve
B cells. It is also dependent on AID (Odegard, Kim et al. 2005). There are two
possible reasons that the chromosomal modifications are related to AID
expression. First, AID may affect other transcription factors, which result in the
126
alteration of the chromosomal modification status. Second and more likely,
changes of chromatin structure result from DSBs following AID action. It is
proposed that a DSB can trigger the change of chromosomal modifications
surrounding the DSB (Fernandez-Capetillo and Nussenzweig 2004). Therefore,
these chromosomal modifications may be involved in later steps of CSR, which is
DSB formation, rather than the AID targeting.
AID cofactors: The other possible targeting mechanism is that there are other
protein factors binding to AID and recruit it to the target genes. From the
site-directed mutagenesis assay, the C-terminal 16 amino acid residues of AID
are required for CSR activity, but not SHM activity (Revy, Muto et al. 2000; Barreto,
Reina-San-Martin et al. 2003; Ta, Nagaoka et al. 2003; Zhu, Nonoyama et al.
2003). This region contains the NES signal. In contrast to that, the N terminal
domain of AID is required for SHM activity, but not CSR activity (Shinkura, Ito et al.
2004). This region contains the NLS signal. The reason for the importance of
these two regions for different pathways is still uncertain, and there are several
hypotheses for it. First, one hypothesis is the AID dimerization ability. However,
none of the regions tested affect the dimerization ability of AID (Wang, Shinkura et
al. 2006). Second, the other hypotheis is AID localization. Since these
127
domains overlapping with NES and NLS signals, these domains may be crucial
for shuttling of AID from the nucleus to the cytoplasm and affect AID function.
However, the N-terminal domain mutant still has CSR ability in the presence of
leptomycin B (LMB), an inhibitor of exportin-1-dependent nuclear export
(Shinkura, Ito et al. 2004). This result suggests that the localization of AID is not
correlated with the CSR efficiency (Muramatsu, Nagaoka et al. 2007). The other
possible hypothesis is that different domains of AID have different cofactors in the
cell, and recruit AID to V (variable) and S (switch) region in SHM and CSR,
respectively. (Ta, Nagaoka et al. 2003) (Shinkura, Ito et al. 2004). This
hypothesis is favored by the field, however, so far, none of these SHM- and
CSR-specific factors has been found.
The current model of the class switch mechanism
From genetic studies, many proteins are involved in CSR. According to their
known biological functions, model for CSR is proposed. However, further
investigation is still needed to confirm these speculated detailed steps.

128
Figure 5-1. The current model of the class switch recombination. There
are four steps in the recombination: targeting, cleavage, synapsis, and rejoining.
Many proteins are proposed involving in different steps. Among them, AID is the
only required factor in the CSR.

129
Bibliography
Abner, C. W., A. Y. Lau, et al. (2001). "Base excision and DNA binding activities of
human alkyladenine DNA glycosylase are sensitive to the base paired with
a lesion." J Biol Chem 276(16): 13379-87.

Barreto, V., B. Reina-San-Martin, et al. (2003). "C-terminal deletion of AID
uncouples class switch recombination from somatic hypermutation and
gene conversion." Mol Cell 12(2): 501-8.

Barreto, V. M., Q. Pan-Hammarstrom, et al. (2005). "AID from bony fish catalyzes
class switch recombination." J Exp Med 202(6): 733-8.

Basu, U., J. Chaudhuri, et al. (2005). "The AID antibody diversification enzyme is
regulated by protein kinase A phosphorylation." Nature 438(7067): 508-11.

Bergsagel, P. L., M. Chesi, et al. (1996). "Promiscuous translocations into
immunoglobulin heavy chain switch regions in multiple myeloma." Proc
Natl Acad Sci U S A 93(24): 13931-6.

Bosma, G. C., J. Kim, et al. (2002). "DNA-dependent protein kinase activity is not
required for immunoglobulin class switching." J Exp Med 196(11): 1483-95.

Bottaro, A., R. Lansford, et al. (1994). "S region transcription per se promotes
basal IgE class switch recombination but additional factors regulate the
efficiency of the process." Embo J 13(3): 665-74.

130
Bransteitter, R., P. Pham, et al. (2004). "Biochemical analysis of hypermutational
targeting by wild type and mutant activation-induced cytidine deaminase."
J Biol Chem 279(49): 51612-21.

Bransteitter, R., P. Pham, et al. (2003). "Activation-induced cytidine deaminase
deaminates deoxycytidine on single-stranded DNA but requires the action
of RNase." Proc Natl Acad Sci U S A 100(7): 4102-7.

Catalan, N., F. Selz, et al. (2003). "The block in immunoglobulin class switch
recombination caused by activation-induced cytidine deaminase deficiency
occurs prior to the generation of DNA double strand breaks in switch mu
region." J Immunol 171(5): 2504-9.

Chaudhuri, J. and F. W. Alt (2004). "Class-switch recombination: interplay of
transcription, DNA deamination and DNA repair." Nat Rev Immunol 4(7):
541-52.

Chaudhuri, J., C. Khuong, et al. (2004). "Replication protein A interacts with AID to
promote deamination of somatic hypermutation targets." Nature 430(7003):
992-8.

Chaudhuri, J., M. Tian, et al. (2003). "Transcription-targeted DNA deamination by
the AID antibody diversification enzyme." Nature 422(6933): 726-30.

Coffman, R. L., D. A. Lebman, et al. (1993). "Mechanism and regulation of
immunoglobulin isotype switching." Adv Immunol 54: 229-70.

Conticello, S. G., M. A. Langlois, et al. (2007). "DNA deamination in immunity: AID
in the context of its APOBEC relatives." Adv Immunol 94: 37-73.
131
Daniels, G. A. and M. R. Lieber (1995). "RNA:DNA complex formation upon
transcription of immunoglobulin switch regions: implications for the
mechanism and regulation of class switch recombination." Nucleic Acids
Res 23(24): 5006-11.

Dempsey, L. A., H. Sun, et al. (1999). "G4 DNA binding by LR1 and its subunits,
nucleolin and hnRNP D, A role for G-G pairing in immunoglobulin switch
recombination." J Biol Chem 274(2): 1066-71.

Di Noia, J. and M. S. Neuberger (2002). "Altering the pathway of immunoglobulin
hypermutation by inhibiting uracil-DNA glycosylase." Nature 419(6902):
43-8.

Dickerson, S. K., E. Market, et al. (2003). "AID mediates hypermutation by
deaminating single stranded DNA." J Exp Med 197(10): 1291-6.

Doi, T., K. Kinoshita, et al. (2003). "De novo protein synthesis is required for the
activation-induced cytidine deaminase function in class-switch
recombination." Proc Natl Acad Sci U S A 100(5): 2634-8.

Drolet, M., S. Broccoli, et al. (2003). "The problem of hypernegative supercoiling
and R-loop formation in transcription." Front Biosci 8: d210-21.

Dunnick, W., G. Z. Hertz, et al. (1993). "DNA sequences at immunoglobulin switch
region recombination sites." Nucleic Acids Res 21(3): 365-72.

Dunnick, W. A., G. Z. Hertz, et al. (1993). "DNA sequence at immunoglobulin
switch region recombination sites." Nucl. Acid Res. 21(3): 365-372.

132
Duquette, M. L., P. Handa, et al. (2004). "Intracellular transcription of G-rich DNAs
induces formation of G-loops, novel structures containing G4 DNA." Genes
Dev 18(13): 1618-29.

Durandy, A., N. Taubenheim, et al. (2007). "Pathophysiology of B-cell intrinsic
immunoglobulin class switch recombination deficiencies." Adv Immunol 94:
275-306.

Fernandez-Capetillo, O. and A. Nussenzweig (2004). "Linking histone
deacetylation with the repair of DNA breaks." Proc Natl Acad Sci U S A
101(6): 1427-8.

Gopal, V., L. G. Brieba, et al. (1999). "Characterization of structural features
important for T7 RNAP elongation complex stability reveals competing
complex conformations and a role for the non-template strand in RNA
displacement." J Mol Biol 290(2): 411-31.

Hein, K., M. G. Lorenz, et al. (1998). "Processing of switch transcripts is required
for targeting of antibody class switch recombination." J Exp Med 188(12):
2369-74.

Huang, F. T., K. Yu, et al. (2007). "Sequence Dependence of Chromosomal
R-Loops at the Immunoglobulin Heavy-Chain S{micro} Class Switch
Region." Mol Cell Biol 27(16): 5921-32.

Huang, F. T., K. Yu, et al. (2006). "Downstream boundary of chromosomal R-loops
at murine switch regions: implications for the mechanism of class switch
recombination." Proc Natl Acad Sci U S A 103(13): 5030-5.

133
Huertas, P. and A. Aguilera (2003). "Cotranscriptionally formed DNA:RNA hybrids
mediate transcription elongation impairment and transcription-associated
recombination." Mol Cell 12(3): 711-21.

Ichikawa, H. T., M. P. Sowden, et al. (2006). "Structural phylogenetic analysis of
activation-induced deaminase function." J Immunol 177(1): 355-61.

Ito, S., H. Nagaoka, et al. (2004). "Activation-induced cytidine deaminase shuttles
between nucleus and cytoplasm like apolipoprotein B mRNA editing
catalytic polypeptide 1." Proc Natl Acad Sci U S A 101(7): 1975-80.

Khamlichi, A. A., F. Glaudet, et al. (2004). "Immunoglobulin class-switch
recombination in mice devoid of any S mu tandem repeat." Blood 103(10):
3828-36.

Khamlichi, A. A., F. Glaudet, et al. (2004). "Immunoglobulin class-switch
recombination in mice devoid of any Smu tandem repeat." Blood 103:
3828-3836.

Kitao, H., H. Arakawa, et al. (2000). "Class switch recombination of the chicken
IgH chain genes: implications for the primordial switch region repeats." Int
Immunol 12(7): 959-68.

Kuzin, II, G. D. Ugine, et al. (2000). "Normal isotype switching in B cells lacking
the I mu exon splice donor site: evidence for multiple I mu-like germline
transcripts." J Immunol 164(3): 1451-7.

134
Lee, C. G., S. Kondo, et al. (1998). "Frequent but biased class switch
recombination in the S mu flanking regions." Curr Biol 8(4): 227-30.

Leung, H. and N. Maizels (1994). "Regulation and targeting of recombination in
extrachromosomal substrates carrying immunoglobulin switch region
sequences." Mol Cell Biol 14(2): 1450-8.

Li, X. and J. L. Manley (2005). "Inactivation of the SR protein splicing factor
ASF/SF2 results in genomic instability." Cell 122(3): 365-78.

Li, Z., Z. Luo, et al. (2004). "Differential regulation of histone acetylation and
generation of mutations in switch regions is associated with Ig class
switching." Proc Natl Acad Sci U S A 101(43): 15428-33.

Liu, L. F. and J. C. Wang (1987). "Supercoiling of the DNA template during
transcription." Proc Natl Acad Sci U S A 84(20): 7024-7.

Lorenz, M., S. Jung, et al. (1995). "Switch transcripts in immunoglobulin class
switching." Science 267(5205): 1825-8.

Luby, T. M., C. E. Schrader, et al. (2001). "The mu switch region tandem repeats
are important but not required for antibody class switch recombination." J.
Exp. Med. 193: 159-168.

Luby, T. M., C. E. Schrader, et al. (2001). "The mu switch region tandem repeats
are important, but not required, for antibody class switch recombination." J
Exp Med 193(2): 159-68.

135
MacDuff, D. A., M. S. Neuberger, et al. (2006). "MDM2 can interact with the
C-terminus of AID but it is inessential for antibody diversification in DT40 B
cells." Mol Immunol 43(8): 1099-108.

Manis, J. P., D. Dudley, et al. (2002). "IgH class switch recombination to IgG1 in
DNA-PKcs-deficient B cells." Immunity 16(4): 607-17.

Masukata, H. and J. Tomizawa (1990). "A mechanism of formation of a persistent
hybrid between elongating RNA and template DNA." Cell 62(2): 331-8.

Mehta, A., M. T. Kinter, et al. (2000). "Molecular cloning of apobec-1
complementation factor, a novel RNA-binding protein involved in the
editing of apolipoprotein B mRNA." Mol Cell Biol 20(5): 1846-54.

Michael, N., T. E. Martin, et al. (2002). "Effects of sequence and structure on the
hypermutability of immunoglobulin genes." Immunity 16(1): 123-34.

Milstein, C., M. S. Neuberger, et al. (1998). "Both DNA strands of antibody genes
are hypermutation targets." Proc Natl Acad Sci U S A 95(15): 8791-4.

Min, I. M., L. R. Rothlein, et al. (2005). "Shifts in targeting of class switch
recombination sites in mice that lack mu switch region tandem repeats or
Msh2." J Exp Med 201(12): 1885-90.

Min, I. M., C. E. Schrader, et al. (2003). "The Smu tandem repeat region is critical
for Ig isotype switching in the absence of Msh2." Immunity 19(4): 515-24.

136
Muramatsu, M., K. Kinoshita, et al. (2000). "Class switch recombination and
hypermutation require activation-induced cytidine deaminase (AID), a
potential RNA editing enzyme." Cell 102(5): 553-63.

Muramatsu, M., H. Nagaoka, et al. (2007). "Discovery of activation-induced
cytidine deaminase, the engraver of antibody memory." Adv Immunol 94:
1-36.

Muramatsu, M., V. S. Sankaranand, et al. (1999). "Specific expression of
activation-induced cytidine deaminase (AID), a novel member of the
RNA-editing deaminase family in germinal center B cells." J Biol Chem
274(26): 18470-6.

Mussmann, R., M. Courtet, et al. (1997). "Microsites for immunoglobulin switch
recombination breakpoints from Xenopus to mammals." Eur J Immunol
27(10): 2610-9.

Nambu, Y., M. Sugai, et al. (2003). "Transcription-coupled events associating with
immunoglobulin switch region chromatin." Science 302(5653): 2137-40.

Odegard, V. H., S. T. Kim, et al. (2005). "Histone modifications associated with
somatic hypermutation." Immunity 23(1): 101-10.

Okazaki, I. M., H. Hiai, et al. (2003). "Constitutive expression of AID leads to
tumorigenesis." J Exp Med 197(9): 1173-81.

Okazaki, I. M., K. Kinoshita, et al. (2002). "The AID enzyme induces class switch
recombination in fibroblasts." Nature 416(6878): 340-5.

137
Petersen-Mahrt, S. K., R. S. Harris, et al. (2002). "AID mutates E. coli suggesting
a DNA deamination mechanism for antibody diversification." Nature
418(6893): 99-103.

Petersen, S., R. Casellas, et al. (2001). "AID is required to initiate
Nbs1/gamma-H2AX focus formation and mutations at sites of class
switching." Nature 414(6864): 660-5.

Pham, P., R. Bransteitter, et al. (2003). "Processive AID-catalysed cytosine
deamination on single-stranded DNA simulates somatic hypermutation."
Nature 424(6944): 103-7.

Ramiro, A. R., P. Stavropoulos, et al. (2003). "Transcription enhances
AID-mediated cytidine deamination by exposing single-stranded DNA on
the nontemplate strand." Nat Immunol 4(5): 452-6.

Reaban, M. E. and J. A. Griffin (1990). "Induction of RNA-stabilized DNA
conformers by transcription of an immunoglobulin switch region." Nature
348(6299): 342-4.

Revy, P., T. Muto, et al. (2000). "Activation-induced cytidine deaminase (AID)
deficiency causes the autosomal recessive form of the Hyper-IgM
syndrome (HIGM2)." Cell 102(5): 565-75.

Rogozin, I. B., Y. I. Pavlov, et al. (2001). "Somatic mutation hotspots correlate with
DNA polymerase eta error spectrum." Nat Immunol 2(6): 530-6.

138
Ronai, D., M. D. Iglesias-Ussel, et al. (2007). "Detection of chromatin-associated
single-stranded DNA in regions targeted for somatic hypermutation." J Exp
Med 204(1): 181-90.

Rush, J. S., S. D. Fugmann, et al. (2004). "Staggered AID-dependent DNA double
strand breaks are the predominant DNA lesions targeted to S mu in Ig
class switch recombination." Int Immunol 16(4): 549-57.

SantaLucia, J., Jr. (1998). "A unified view of polymer, dumbbell, and
oligonucleotide DNA nearest-neighbor thermodynamics." Proc Natl Acad
Sci U S A 95(4): 1460-5.

Selsing, E. (2006). "Ig class switching: targeting the recombinational mechanism."
Curr Opin Immunol 18(3): 249-54.

Sen, D. and W. Gilbert (1988). "Formation of parallel four-stranded complexes by
guanine-rich motifs in DNA and its implications for meiosis." Nature
334(6180): 364-6.

Shapiro, G. S., K. Aviszus, et al. (2002). "Evolution of Ig DNA sequence to target
specific base positions within codons for somatic hypermutation." J
Immunol 168(5): 2302-6.

Shinkura, R., S. Ito, et al. (2004). "Separate domains of AID are required for
somatic hypermutation and class-switch recombination." Nat Immunol 5(7):
707-12.

139
Shinkura, R., M. Tian, et al. (2003). "The influence of transcriptional orientation on
endogenous switch region function." Nat Immunol 4(5): 435-41.

Sohail, A., J. Klapacz, et al. (2003). "Human activation-induced cytidine
deaminase causes transcription-dependent, strand-biased C to U
deaminations." Nucleic Acids Res 31(12): 2990-4.

Stavnezer, J. and C. T. Amemiya (2004). "Evolution of isotype switching." Semin
Immunol 16(4): 257-75.

Stavnezer, J., G. Radcliffe, et al. (1988). "Immunoglobulin heavy-chain switching
may be directed by prior induction of transcripts from constant-region
genes." Proc Natl Acad Sci U S A 85(20): 7704-8.

Ta, V. T., H. Nagaoka, et al. (2003). "AID mutant analyses indicate requirement for
class-switch-specific cofactors." Nat Immunol 4(9): 843-8.

Tashiro, J., K. Kinoshita, et al. (2001). "Palindromic but not G-rich sequences are
targets of class switch recombination." Int Immunol 13(4): 495-505.

Tian, M. and F. W. Alt (2000). "Transcription-induced cleavage of immunoglobulin
switch regions by nucleotide excision repair nucleases in vitro." J Biol
Chem 275(31): 24163-72.

Wakae, K., B. G. Magor, et al. (2006). "Evolution of class switch recombination
function in fish activation-induced cytidine deaminase, AID." Int Immunol
18(1): 41-7.

140
Wang, J., R. Shinkura, et al. (2006). "Identification of a specific domain required
for dimerization of activation-induced cytidine deaminase." J Biol Chem
281(28): 19115-23.

Wang, L., N. Whang, et al. (2006). "AID-dependent histone acetylation is detected
in immunoglobulin S regions." J Exp Med 203(1): 215-26.

Westover, K. D., D. A. Bushnell, et al. (2004). "Structural basis of transcription:
nucleotide selection by rotation in the RNA polymerase II active center."
Cell 119(4): 481-9.

Wu, X., P. Geraldes, et al. (2005). "The double-edged sword of activation-induced
cytidine deaminase." J Immunol 174(2): 934-41.

Xu, L., B. Gorham, et al. (1993). "Replacement of germ-line epsilon promoter by
gene targeting alters control of immunoglobulin heavy chain class
switching." Proc Natl Acad Sci U S A 90(8): 3705-9.

Xue, K., C. Rada, et al. (2006). "The in vivo pattern of AID targeting to
immunoglobulin switch regions deduced from mutation spectra in msh2-/-
ung-/- mice." J Exp Med 203(9): 2085-94.

Xue, K., C. Rada, et al. (2006). "The in vivo pattern of AID targeting to
immunoglobulin switch regions deduced from mutation spectra in
msh2-/-ung-/- mice." J. Exp. Med. 203: 2085-2095.

Yan, C. T., C. Boboila, et al. (2007). "IgH class switching and translocations use a
robust non-classical end-joining pathway." Nature.

141
Yoshikawa, K., I. M. Okazaki, et al. (2002). "AID enzyme-induced hypermutation
in an actively transcribed gene in fibroblasts." Science 296(5575): 2033-6.

Yu, K., F. Chedin, et al. (2003). "R-loops at immunoglobulin class switch regions in
the chromosomes of stimulated B cells." Nat Immunol 4(5): 442-51.

Yu, K., F. T. Huang, et al. (2004). "DNA substrate length and surrounding
sequence affect the activation-induced deaminase activity at cytidine." J
Biol Chem 279(8): 6496-500.

Yu, K. and M. R. Lieber (2003). "Nucleic acid structures and enzymes in the
immunoglobulin class switch recombination mechanism." DNA Repair
(Amst) 2(11): 1163-74.

Yu, K., D. Roy, et al. (2005). "Fine-structure analysis of activation-induced
deaminase accessbility to class switch region R-loops." Mol. Cell. Biol. 25:
1730-1736.

Zarrin, A. A., F. W. Alt, et al. (2004). "An evolutionarily conserved target motif for
immunoglobulin class-switch recombination." Nat Immunol 5(12): 1275-81.

Zarrin, A. A., M. Tian, et al. (2005). "Influence of switch region length on
immunoglobulin class switch recombination." Proc Natl Acad Sci U S A
102(7): 2466-70.

Zhu, Y., S. Nonoyama, et al. (2003). "Type two hyper-IgM syndrome caused by
mutation in activation-induced cytidine deaminase." J Med Dent Sci 50(1):
41-6.

Abstract (if available)

Abstract Class switch recombination (CSR) is the mechanism responsible for changing constant domains of antibodies to produce different effector functions of B cells. Somatic hypermutation (SHM) mutates the V(D)J segment to enhance the antigen-binding affinity of antibodies. Activation induced deaminase (AID) is required and responsible for targeting in both SHM and CSR. With DNA repair systems, AID causes mutations in the V(D)J segment and double strand breaks in switch regions for SHM and CSR, respectively. However, the targeting and cleavage mechanism of AID is still not clear yet.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

The mechanism of R-loop formation in mammalian immunoglobulin class switch recombination

PDF

The mechanism of mammalian immunoglobulin class switch recombination

PDF

How phosphorylation affects the biochemical regulation and targeting of activation induced cytidine deaminase (AID)

PDF

Generation of monoclonal antibodies via phage display and in vitro affinity maturation using activation induced deoxycytidine deaminase and DNA polymerase eta

PDF

Biochemistry and reconstitution of V(D)J recombination in a purified system

PDF

Molecular mechanisms of recurrent chromosomal translocations in human leukemias and lymphomas

PDF

Lineage boundaries and cell migration in the patterning of the mammalian skull

PDF

Mechanistic basis for chromosomal translocations at the E2A gene

PDF

AID scanning & catalysis and the generation of high-affinity antibodies

PDF

Studies on the role of Artemis in non-homologous DNA end-joining to understand the mechanism and discover therapies

PDF

Molecular genetic analysis of human genetic lesions

PDF

Mechanisms of nucleases in non-homologous DNA end joining

PDF

IPS and CNS cell models of peroxisomal disorders

Asset Metadata

Creator Huang, Feng-Ting (author)

Core Title The mechanism of mammalian immunoglobulin class switch recombination: R-loop structures and activation-induced deaminase site preferences

School Keck School of Medicine

Degree Doctor of Philosophy

Degree Program Biochemistry and Molecular Biology

Degree Conferral Date 2007-12

Publication Date 11/13/2007

Defense Date 09/28/2007

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

Tag activation-induced deaminase,class switch recombination,immunoglobulin,OAI-PMH Harvest,R-loops

Language English

Advisor Lieber, Michael R. (committee chair), Hacia, Joseph G. (committee member), Haworth, Ian S. (committee member), Hsieh, Chih-Lin (committee member), Lieber, Michael R. (committee member), Stallcup, Michael R. (committee member)

Creator Email fengtinh@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m916

Unique identifier UC1210163

Identifier etd-Huang-20071113 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-590814 (legacy record id),usctheses-m916 (legacy record id)

Legacy Identifier etd-Huang-20071113.pdf

Dmrecord 590814

Document Type Dissertation

Rights Huang, Feng-Ting

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 cisadmin@lib.usc.edu