Immunology and Cell Biology

(2004)

82

, 205–208 doi:10.1046/j.0818-9641.2004.01226.x

© 2004 Australasian Society for Immunology Inc.

Special Feature

The boundaries of the distribution of somatic hypermutation of rearranged immunoglobulin variable genes

R O B E R T V B L A N D E N , A N D R E W F R A N K L I N a n d E D W A R D J S T E E L E

†

Division of Immunology and Genetics, John Curtin School of Medical Research, Australian National University, GPO Box 334, Canberra City ACT 2601, Australia

Summary

Available evidence about the mechanisms and distribution of somatic hypermutation (SHM) ofrearranged immunoglobulin (IgV) genes is reviewed with particular emphasis on the 5

′

boundary. In heavy (H)chain genes, the 5

′

boundary of SHM is the transcription start site; in contrast to

κ

light (L) chain genes, it is locatedin the leader (L) intron. DNA-based models of SHM cannot account for this difference. However, an updatedreverse transcriptase (RT)-based model invoking error-prone RT activity of DNA polymerase

η

copying IgV pre-mRNA templates to produce cDNA of the transcribed strand (TS) of IgV DNA, which then replaces thecorresponding section of the original TS, can explain the difference. This explanation incorporates recent knowledgeof pre-mRNA processing, in particular, binding of the splicing-associated protein termed U2AF to a pyrimidine-richtract in the L intron of pre-mRNA of

κ

L chains that may block RT progression further upstream to the end of thepre-mRNA template (transcription start site). Reasons why this block may not occur in H chains and other aspectsof the updated RT-model are discussed.

Key words

:

5

′

boundary of somatic hypermutation, DNA polymerase

η

, pre-mRNA processing, rearrangedimmunoglobulin genes, reverse transcription.

Introduction

Accumulated knowledge concerning somatic hypermutation(SHM) of rearranged immunoglobulin variable (IgV) genes isnow considerable and is summarized in Table 1. Accidentaldiscovery of an essential role for activation-induced cyti-dine deaminase (AID) in SHM

1

and subsequent elucidationof its mode of action

2,3

has significantly improved ourunderstanding of the initiation of SHM and some of themolecular mechanisms involved. Also, the finding thathumans with the variant form of xeroderma pigmentosum(XPV) caused by a mutation in the gene encoding DNApolymerase eta (pol

η

) have reduced mutation in A-T basepairs in SHM

4

has been critical, indicating a role for pol

η

inA-T mutation.

Many recent reviews have incorporated this new knowl-edge in comprehensive models of SHM. The emerging con-sensus is that AID initiates SHM in the DNA of IgV genes bydeamination of dC to dU,

2,3

which may be removed by uracil-DNA glycosylase activity

5

(primarily UNG2) to create anapyrimidic/apurinic (AP) site. DNA replication then results inmutations opposite dU or AP sites and explains SHM in G-Cbase pairs. Because AID acts more efficiently on single-stranded than on double-stranded DNA, for example, in a

transcription bubble, and favours the non-transcribed strand(NTS) over the TS,

6,7

explanations of the requirement fortranscription and occurrence of strand bias in SHM, bothpuzzling historical observations, are now available. AP endo-nuclease (APE) nicking of the phosphodiester backbone ofDNA, and possible further endonuclease activity, providesites for error-prone synthesis of DNA by pol

η

(and possiblyother low fidelity DNA polymerases). Pol

η

is characterizedby extreme error-proneness opposite template dT, misincor-porating dG instead of dA,

8,9

thus providing an explanationfor A-T base pair mutation in SHM and the prominentsignature of A

→

G transitions in the NTS.

9

Thus, most workers in the field have reached the viewthat the major mechanisms of SHM have been revealed andthat all that remains is to clear up a few minor details.However, the discovery that pol

η

has RT activity by Franklin

et al.

10

and the potential provision of priming sites forcDNA synthesis of the TS on an IgV pre-mRNA template

via

the action of AID, UNG and APE on the original TS hasraised the possibility that the omission of RT as a majormechanism of SHM in the recent reviews is premature. Thegeneration of mutations by RT activity of pol

η

is compati-ble with all pertinent published data with the exception ofthe error signature of DNA synthesis by pol

η

using a Vkappa (V

κ

) template.

9

Until the error signature of RTactivity of pol

η

is determined using the same V

κ

templatesequence, this issue will remain unresolved. In this paper wefocus on the features of SHM that can be explained eitherexclusively or best by RT as a major mechanism of SHM,and that cannot be explained by purely DNA-based events.Also, we speculate on answers to questions not addressed bythe prevailing current paradigm, as typified in recent reviews.

8,11

Correspondence: Robert V Blanden, Division of Immunology &Genetics, John Curtin School of Medical Research, AustralianNational University, GPO Box 334, Canberra, ACT 2601, Australia.Email: Robert.Blanden@anu.edu.au

†

Present address: Genomic Interactions Group, Research School of Biological Sciences, Australian National University, GPO Box 475, Canberra, ACT 2601, Australia.

Received 1 December 2003; accepted 1 December 2003.

206

RV Blanden

et al

.

The distribution of mutations in rearranged IgV genes

Figure 1 shows a typical rearranged IgV gene representing Hand

κ

chains (not lambda) for which considerable publisheddata sets exist illustrating the distribution of mutationsresulting from SHM;

12

the figure embodies the first threeunanswered questions posed in Table 1. The diagram does notextend past intronic enhancer/matrix attachment region (Ei/MAR), because SHM does not affect regions 3

′

of this point.

12

Presumably this 3

′

boundary in H and

κ

chains is caused bypremature termination of transcription at or before Ei/MAR, anotion consistent with observations that germinal centre (GC)B cells undergoing SHM lack surface Ig (sIg).

13

This lossof the original sIg would be essential prior to the display ofmutated sIg and selection by Ag in the GC. The cause ofpremature termination of transcription is unknown, but onepossible mechanism is

via

overexpression of a B-cell-specifictranscription factor termed BrIgHt, which binds to DNAsequence motifs

14

that are present in Ei/MAR and adjacentareas of the J-C intron.

The distribution of SHM, shown in Figure 1, is typified byan abrupt beginning at the 5

′

boundary, which is the transcrip-tion start site for H chains

15

and a site in the leader (L) intronfor

κ

chains.

16

After maxima approximately over the rear-ranged variable (diversity) joining region (V(D)J), there is agradual decline to the 3

′

boundary within the J-C intron nearEi/MAR. The RT model still provides the best explanation forthe gradual decline as explicitly stated by Steele, Rothenfluhand Both over a decade ago.

12

In this model, a family of error-filled cDNA (produced by the RT activity of pol

η

), initiatedat various evenly spaced priming sites over the V(D)J and J-C intron of the pre-mRNA template, replace the original TSDNA in the populations of B cells assayed for SHM. Synthe-sis of cDNA on the pre-mRNA template should continue tothe end of the template unless blocked by RNA structure or -at the 5

′

boundary. The different initiation points, provided bypriming sites described above, explain the gradual decline inSHM down to the 3

′

boundary. An alternative explanationbased upon AID activity being focused over the V(D)J anddeclining as Ei/MAR is approached has no current scientificbasis.

Figure 1

Frequency of somatic hypermutation (SHM) in relation to the structure of a rearranged IgV gene, represented by the 5

′

to 3

′

non-transcribed (coding) DNA strand with leader (L) and V(D)J exons in rectangular boxes. The curves represent typical data for V

H

andV

κ

with sharp increases at the 5

′

boundary of SHM, maxima roughly over the rearranged V(D)J and a slow decline to the 3

′

boundary nearintronic enhancer/matrix attachment region (Ei/MAR).

Table 1

What we know about SHM

1. Activation-induced cytidine deaminase (AID) is essential.

1

2. AID deaminates cytosine

→

uracil.

2,3

3. AID operates most efficiently, but not exclusively, on single-stranded DNA, e.g. in a transcription bubble.

25,26

4. In

Escherichia coli

, AID acts preferentially on the non-transcribedstrand (NTS) over the transcribed strand (TS).

6,27

5. UNG2 removes the uracils produced by AID

5

(probably

∼

95%done by UNG2 but others

can

do it too). Thus AID-UNG2 are G-C mutators.

6. AID-GFP is

not

found in nucleus

20

(but AID must operate in thenucleus to achieve SHM).

7. DNA polymerase (pol)

η

is involved in SHM as shown by patientswith xeroderma pigmentosum variant (XPV) who have reduced A-T mutation signature in SHM.

4

Thus, pol

η

is an A-T mutator inSHM.

8. Pols

η

,

ι

and

κ

(Y family DNA pols) are reverse transcriptases(RT).

10

9. Pols

η

and

ι

bind to each other

21

and

η

binds to proliferating cellnuclear antigen (PCNA) which causes

η

and

ι

to co-localize inDNA replication foci.

22

10. Pol

η

is enhanced as a DNA pol and RT by PCNA, replicationprotein A (RPA) and replication factor C (RFC) (proteinsassociated with DNA replication and repair).

10,23

11. UNG2 co-localizes with PCNA and RPA in replication foci.

28

12. Intronic enhancer/matrix attachment region (Ei/MAR) isessential

29

but no other locus-specific factor has been found.13. The 5

′

boundary of SHM is the transcription start site for V

H15

andthe leader intron for V

κ

;

16

the 3

′

boundary is Ei/MAR.

15

14. The SHM mechanisms produce A-T strand bias.

24,30

Unanswered questions

1. How is mutation terminated at

∼

Ei/MAR?2. Why are the 5

′

boundaries sharp and the 3

′

boundaries gradual?3. What locates the 5

′

and 3

′

boundaries? In particular, why are theV

H

and V

κ

boundaries different?4. Why is Ei/MAR essential?5. How do AID, UNG2 and pol

η

and rearranged IgV come together?

6. Why are there occasional mutations upstream of transcriptionstart sites? How do these occur?

7. How is strand bias produced?

Boundaries of SHM

207

What about the difference in 5

′

boundary between H(transcription start) and

κ

(L intron)? Again, purely DNA-based models of SHM have no explanation, but the RT modelcan explain this difference based upon detailed knowledge ofthe processing of pre-mRNA.

In order to produce the mature mRNA that is exportedfrom the nucleus, pre-mRNA must be processed

via

capping(at the 5

′

terminus), splicing to remove introns and poly-adenylation at the 3

′

terminus. Even though these events arechemically distinct and occur at different sites, recent workindicates that one can influence the efficiency of another.

17

Inthe present context, capping and splicing are relevant; theyoccur co-transcriptionally, with capping being initiated afterthe first 20–30 nucleotides of pre-mRNA have been synthe-sized. In turn, the cap structure enhances splicing out of theclosest intron (the L intron in the case of rearranged IgVgenes). The 5

′

and 3

′

exon–intron junctions are marked bysequence motifs, the 5

′

motif being the consensus sequenceAG/GURAGU (where/is the splice site and R = purine). Thefirst chemical step in splicing occurs when the 5

′

exon–intronjunction is attacked by a nucleophile embodied in a conservedadenosine (providing the 2

′

OH required for the attack) withinthe branch point motif CURA

2

′

OH

Y (Y = pyrimidine) followedby a pyrimidine-rich tract.

This interaction requires the participation of 5 smallnuclear RNA (snRNA), U1, U2, U4, U5 and U6 complexedwith proteins (snRNP) that bind to the pre-mRNA sequencemotifs. Thus, U1 snRNA binds to the 5

′

splice site to initiatespliceosome assembly. The 3

′

splice site and pyrimidine-richtract are bound by a dimeric factor termed U2AF, whichpromotes binding of U2 snRNP to the branch point.

There are two key differences between H and

κ

withrespect to transcription and L intron structure which, coupledto the scenario outlined above for pre-mRNA processing, canexplain the different locations of the 5

′

boundary of SHM inH and

κ

loci. First, it has been reported that in the rearranged

κ

locus, RNA polymerase II can pause at a position 45–89base pairs downstream of the transcription start site.

18

Thishas not been observed for H loci. If it occurred duringtranscriptional events required for SHM, it could enableassembly of the cap structure before transcription progressedthrough the L intron, thus accelerating the splicing process.

Second, the L intron of

κ

genes is much longer than that ofH genes. For example, the L intron of V

κ

Ox, the gene used todefine the 5

′

border of SHM by the Milstein laboratory,

16

is171 bases long. In contrast, the L intron of V

H

9-related genesis 81 bases,

19

and that of V

H

186.2-related genes is 82 bases.

15

These H locus introns each contain a single pyrimidine-richtract of 18 and 20 bases, respectively, preceded by theCURA

2

′

OH

Y branch point consensus, albeit incomplete ineach case (CUXA

2

′

OH Y in VH9 and XURA2′OH Y in VH186.2,where X is a non-consensus base). The long intron of VκOxcontains 3 pyrimidine-rich tracts of 21, 20 and 24 bases (from5′ to 3′ direction). The first two have incomplete CURA2′OH

Y branchpoint motifs (XXRA2′OH Y in each case), but thethird has a complete motif and presumably is preferred in thesplicing process. However, it is the first tract that is relevantto SHM because it defines the 5′ boundary, as explicitlystated by Rada et al. in 1994.16

The RT model provides an explanation for this 5′ boundaryof SHM as follows. Priming sites for RT activity producing

cDNA of the TS DNA require progress of pre-mRNA synthe-sis from the transcription start site towards Ei/MAR. Thefurther transcription has progressed, the more likely primingsites for RT will be created by the action of AID, UNG andAPE on the TS DNA. Because very little mutation occurs innormal VκOx genes upstream of the first pyrimidine-richtract in the L intron, AID action and priming of RT upstreamof this point must be rare. Furthermore, when priming occursdownstream and RT is progressing on the pre-mRNA templatetowards the 5′ terminus (transcription start site and capstructure), it would be prevented from progressing furtherupstream by binding of U2AF to the pyrimidine-rich tracts inthe L intron. Presumably it is at the first (5′) tract that thisoften occurs in the case of VκOx. Furthermore, since the Hlocus SHM 5′ boundary is the transcription start site, it seemsthat U2AF binding has not impeded RT progression to the5′ terminus of the H locus pre-mRNA. This is plausibly due tomore rapid progress of transcription to potential RT primingsites in the case of the H locus because the L intron is 90bases shorter and there is no documented pausing of RNApolymerase II in H loci. Additionally, this more rapidprogress would reduce the likelihood of completion of the capstructure that contributes to accelerated splicing (and acceler-ated binding of U2AF).

Remaining unanswered questions

There are no definitive answers to the remaining questions(4, 5, 6, 7) posed in Table 1. We speculate that Ei is requiredto achieve a level of transcription that enables AID activity onsingle-stranded DNA and provides pre-mRNA as a templatefor the RT activity of pol η. MAR presumably enables thelocation of the rearranged IgV gene at the nuclear matrix,which may be necessary to bring AID and the transcribingIgV DNA together. AID is a dangerous potential mutator ofall transcribing genes and there must be strong selectionpressure for mechanisms that limit its activity to particulartargets where mutation and selection are useful (such asrearranged IgV genes). The failure to find AID free in thenucleus20 may reflect this selection pressure and its strictlocation at the nuclear membrane; thus, the target IgV gene isbrought to AID by MAR binding to the nuclear matrix at thenuclear membrane. The other important factors in SHM,UNG2 and pol η, are found in DNA replication foci.21,22 Itseems plausible that collision between such foci and a tran-scription complex on rearranged IgV loci will assemble allthe major factors required for SHM, including proliferatingcell nuclear antigen (PCNA), RPA and RFC which enhancethe efficiency of pol η.10,22,23

The boundaries of SHM, particularly the 5′ boundary, arenot absolute. In both H and κ loci, occasional mutations areseen upstream of the transcription start site, the majority ofthem in A-T base pairs. We speculate that these mutations areproduced by pol η RT activity utilizing rare, aberrant pre-mRNA species in which transcription is initiated upstream ofthe usual site by cryptic regulatory elements.

Finally, there is the question of strand bias. That is, whichDNA strand is predominantly targeted during A-T mutation,the subject of another paper published herein.24 It is plausiblethat the preference of AID for the NTS followed by error-prone DNA-dependent short patch synthesis of that strand by

208 RV Blanden et al.

pol η, with misincorporation of dG instead of dA oppositetemplate dT, produces the typically predominant A→G signa-ture and strand bias.9 In the case of the RT model, thepredominant error of pol η RT activity producing cDNA ofthe TS DNA from a pre-mRNA template would need to bemisincorporation of dC opposite template rA causing T→Ctransitions on the TS DNA that are converted to A→Gtransitions on the NTS DNA during mismatch repair.10 Thisoutcome could occur via RNA editing as described in detail inthe accompanying papers.10,24 The clear difference betweenthese alternatives means that a definitive answer should comefrom experiments that determine the error signature of pol ηRT activity and in which DNA strand A-T changes originate.10

References

1 Muramatsu M, Kinoshita K, Fagarasan S, Yamada S, Shinkai Y,Honjo T. Class switch recombination and hypermutation requireactivation-induced cytidine deaminase (AID), a potential RNAediting enzyme. Cell 2000; 102: 553–63.

2 Petersen-Mahrt SK, Harris RS, Neuberger MS. AID mutatesE. coli suggesting a DNA deamination mechanism for antibodydiversification. Nature 2002; 418: 99–103.

3 Di Noia J, Neuberger MS. Altering the pathway of immuno-globulin hypermutation by inhibiting uracil-DNA glycosylase.Nature 2002; 419: 43–8.

4 Zeng X, Winter DB, Kasmer C, Kraemer KH, Lehmann AR,Gearhart PJ. DNA polymerase eta is an A-T mutator in somatichypermutation of immunoglobulin variable genes. Nat. Immunol.2001; 2: 537–41.

5 Rada C, Williams GT, Nilsen H, Barnes DE, Lindahl T,Neuberger MS. Immunoglobulin isotype switching is inhibitedand somatic hypermutation perturbed in UNG-deficient mice.Curr. Biol. 2002; 12: 1748–55.

6 Ramiro AR, Stavropoulos P, Jankovic M, Nussenzweig MC.Transcription enhances AID-mediated cytidine deamination byexposing single-stranded DNA on the nontemplate strand. Nat.Immunol. 2003; 4: 452–6.

7 Pham P, Bransteitter R, Petruska J, Goodman MF. ProcessiveAID-catalysed cytosine deamination on single-stranded DNAsimulates somatic hypermutation. Nature 2003; 424: 103–7.

8 Kunkel TA, Pavlov YI, Bebenek K. Functions of human DNApolymerases eta, kappa and iota suggested by their properties,including fidelity with undamaged DNA templates. DNA Repair(Amst.) 2003; 2: 135–49.

9 Pavlov YI, Rogozin IB, Galkin AP et al. Correlation of somatichypermutation specificity and A-T base pair substitution errorsby DNA polymerase eta during copying of a mouse immuno-globulin kappa light chain transgene. Proc. Natl Acad. Sci. USA2002; 99: 9954–9.

10 Franklin A, Milburn PJ, Blanden RV, Steele EJ. Human DNApolymerase-eta, an A-T mutator in somatic hypermutation ofrearranged immunoglobulin genes, is a reverse transcriptase.Immunol. Cell Biol. 2004; 82: 219–25.

11 Goodman MF. Error-prone repair DNA polymerases in prokary-otes and eukaryotes. Annu. Rev. Biochem. 2002; 71: 17–50.

12 Steele EJ, Rothenfluh HS, Both GW. Defining the nucleic acidsubstrate for somatic hypermutation. Immunol. Cell Biol. 1992;70: 129–44.

13 MacLennan I. Immunology. The centre of hypermutation.Nature 1991; 354: 352–3.

14 Herrscher RF, Kaplan MH, Lelsz DL, Das C, Scheuermann R,Tucker PW. The immunoglobulin heavy-chain matrix-associatingregions are bound by Bright: a B cell-specific trans-activator thatdescribes a new DNA-binding protein family. Genes Dev. 1995;9: 3067–82.

15 Both GW, Taylor L, Pollard JW, Steele EJ. Distribution ofmutations around rearranged heavy-chain antibody variable-region genes. Mol. Cell. Biol. 1990; 10: 5187–96.

16 Rada C, Gonzalez-Fernandez A, Jarvis JM, Milstein C. The 5′boundary of somatic hypermutation in a V kappa gene is in theleader intron. Eur. J. Immunol. 1994; 24: 1453–7.

17 Proudfoot NJ, Furger A, Dye MJ. Integrating mRNA processingwith transcription. Cell 2002; 108: 501–12.

18 Raschke EE, Albert T, Eick D. Transcriptional regulation of theIg kappa gene by promoter-proximal pausing of RNA poly-merase II. J. Immunol. 1999; 163: 4375–82.

19 Zylstra P, Franklin A, Hassan KA, Powell KL, Steele EJ,Blanden RV. Molecular evolution of V (H) 9 germline genesisolated from DBA, BALB, 129 and C57BL mouse strains andsublines. Immunogenetics 2003; 55: 182–8.

20 Rada C, Jarvis JM, Milstein C. AID-GFP chimeric proteinincreases hypermutation of Ig genes with no evidence of nuclearlocalization. Proc. Natl Acad. Sci. USA 2002; 99: 7003–8.

21 Kannouche P, Fernandez de Henestrosa AR, Coull B et al.Localization of DNA polymerases eta and iota to the replicationmachinery is tightly co-ordinated in human cells. EMBO J. 2003;22: 1223–33.

22 Kannouche P, Broughton BC, Volker M, Hanaoka F,Mullenders LH, Lehmann AR. Domain structure, localization,and function of DNA polymerase eta, defective in xerodermapigmentosum variant cells. Genes Dev. 2001; 15: 158–72.

23 Haracska L, Johnson RE, Unk I et al. Physical and functionalinteractions of human DNA polymerase eta with PCNA. Mol.Cell. Biol. 2001; 21: 7199–206.

24 Steele EJ, Franklin A, Blanden RV. Genesis of the strand biasedsignature in somatic hypermutation of rearranged immuno-globulin variable genes. Immunol. Cell Biol. 2004; 82: 209–218.

25 Dickerson SK, Market E, Besmer E, Papavasiliou FN. AIDmediates hypermutation by deaminating single stranded DNA.J. Exp. Med. 2003; 197: 1291–6.

26 Chaudhuri J, Tian M, Khuong C, Chua K, Pinaud E, Alt FW.Transcription-targeted DNA deamination by the AID antibodydiversification enzyme. Nature 2003; 422: 726–30.

27 Sohail A, Klapacz J, Samaranayake M, Ullah A, Bhagwat AS.Human activation-induced cytidine deaminase causes transcription-dependent, strand-biased C to U deaminations. Nucl. Acids Res.2003; 31: 2990–4.

28 Otterlei M, Warbrick E, Nagelhus TA et al. Post-replicative baseexcision repair in replication foci. EMBO J. 1999; 18: 3834–44.

29 Betz AG, Milstein C, Gonzalez-Fernandez A, Pannell R,Larson T, Neuberger MS. Elements regulating somatic hyper-mutation of an immunoglobulin kappa gene critical role for theintron enhancer/matrix attachment region. Cell 1994; 77:239–48.

30 Manser T. Regulation, timing, and mechanism of antibody Vgene somatic hypermutation: lessons from the arsonate system.In: Steele EJ (ed.). Somatic Hypermutation in V-Regions. BocaRaton: CRC Press 1991; 41–54.