LETTERS

DNA double-strand breaks activate a multi-functionalgenetic program in developing lymphocytesAndrea L. Bredemeyer1*, Beth A. Helmink1*, Cynthia L. Innes2, Boris Calderon1, Lisa M. McGinnis1,Grace K. Mahowald1, Eric J. Gapud1, Laura M. Walker1, Jennifer B. Collins2, Brian K. Weaver1, Laura Mandik-Nayak1,Robert D. Schreiber1, Paul M. Allen1, Michael J. May3, Richard S. Paules2, Craig H. Bassing4,5 & Barry P. Sleckman1

DNA double-strand breaks are generated by genotoxic agents andby cellular endonucleases as intermediates of several importantphysiological processes. The cellular response to genotoxic DNAbreaks includes the activation of transcriptional programs knownprimarily to regulate cell-cycle checkpoints and cell survival1–5.DNA double-strand breaks are generated in all developing lym-phocytes during the assembly of antigen receptor genes, a processthat is essential for normal lymphocyte development. Here weshow that in murine lymphocytes these physiological DNA breaksactivate a broad transcriptional program. This program trans-cends the canonical DNA double-strand break response andincludes many genes that regulate diverse cellular processesimportant for lymphocyte development. Moreover, the expressionof several of these genes is regulated similarly in response to geno-toxic DNA damage. Thus, physiological DNA double-strandbreaks provide cues that can regulate cell-type-specific processesnot directly involved in maintaining the integrity of the genome,and genotoxic DNA breaks could disrupt normal cellular func-tions by corrupting these processes.

The generation of complete antigen receptor genes in developinglymphocytes requires assembly of the second exon through the pro-cess of V(D)J recombination6. This process is initiated by the recom-binase activating gene (Rag) endonuclease, composed of Rag1 andRag2, which introduces DNA double-strand breaks (DSBs) at theborder of two gene segments and their flanking Rag recognitionsequences (recombination signals)7. Rag cleavage occurs in cells atthe G1 phase of the cell cycle, and the resulting pairs of coding andsignal DNA ends are processed and joined by the non-homologousend-joining (NHEJ) pathway of DNA DSB repair, generating a cod-ing joint and a signal joint, respectively8. All developing lymphocytesmust generate and repair several Rag DSBs as they assemble the genesencoding the heterodimeric B and T cell receptors.

The cellular response to genotoxic DSBs is initiated by the ataxiatelangiectasia mutated (Atm) serine threonine kinase1–3. A centralfeature of this response is the activation of transcriptional programs,such as those initiated by p53, which primarily function to promotecell-cycle arrest and apoptosis of cells with extensive, or persistent,DNA breaks9. Whether physiological DSBs generally activate DSBresponses that include these transcriptional programs has beenunclear. In this regard, stabilization of p53 was observed only inNHEJ-deficient, and not wild-type, thymocytes undergoing V(D)Jrecombination, suggesting that Rag DSBs may only activate tran-scriptional programs in a small fraction of developing lymphocytesin which these breaks are not efficiently repaired and thus pose a

hazard to the cell10. Indeed, the unopposed activation of p53 in res-ponse to transient Rag DSBs could lead to the unwanted apoptosis ofdeveloping lymphocytes that would have otherwise undergone suc-cessful antigen receptor gene assembly. Alternatively, all Rag DSBscould generally activate transcriptional pathways, including thosethat promote survival. In this regard, genotoxic DSBs have beenshown to activate NF-kB, and, although the genes that are regulatedby NF-kB in this setting have not been identified, in other settingsNF-kB regulates the expression of genes with anti-apoptotic acti-vity5,11,12. Moreover, NF-kB promotes the expression of genesinvolved in diverse cell-type-specific processes, raising the intriguingpossibility that genotoxic, and perhaps physiological, DSBs activategenetic programs with broader cellular functions13.

Initially we wished to determine whether Rag DSBs are capable ofactivating NF-kB. To this end, we generated several NHEJ-deficient(Artemis2/2 (Artemis is also known as Dclre1c), Ku702/2 (Ku70 is alsoknown as Xrcc6) and Scid (deficient in Prkdc)) v-abl-transformed pre-B cell lines, hereafter referred to as abl pre-B cells. G1 cell-cycle arrestand Rag expression can be induced in these cells using the v-abl kinaseinhibitor, STI571 (refs 14 and 15). As these cells are NHEJ-deficient,Rag induction results in the accumulation of unrepaired DSBs at theimmunoglobulin (Ig) light (L) chain k locus (Supplementary Fig. 1and data not shown). These DSBs lead to activation of NF-kB asevidenced by the increased nuclear translocation of NF-kB (primarilyp50/p65) in STI571-treated NHEJ-deficient (Artemis2/2, Ku702/2

and Scid) abl pre-B cells as compared to STI571-treated Rag22/2

abl pre-B cells (Fig. 1a and Supplementary Fig. 2a and b). This NF-kB activation was partially dependent on Atm as it was significantlydiminished in Artemis2/2 Atm2/2 abl pre-B cells and in Artemis2/2

abl pre-B cells that were treated with an Atm inhibitor (Fig. 1a andSupplementary Fig. 2c). As was observed for genotoxic DSBs, NEMO(NF-kB essential modulator) is required for the optimal activation ofNF-kB in response to Rag DSBs (Supplementary Fig. 2d)12,16,17.

To determine whether the activation of NF-kB, and possibly othertranscription factors, by Rag DSBs leads to alterations in gene express-ion, we carried out gene expression profiling on three independentlyderived Rag22/2 and Artemis2/2 abl pre-B cells treated with STI571for 48 h (Supplementary Fig. 3). These analyses showed that theexpression of 364 genes was changed $twofold in all threeArtemis2/2 abl pre-B cells relative to the mean expression level inthe three Rag22/2 cells (Fig. 1b, Supplementary Fig. 3a andSupplementary Tables 1, 2 and 3). Several of the gene expressionchanges were confirmed through quantitative real-time PCR (qRT–PCR) and protein expression analyses (Fig. 1c and Supplementary

*These authors contributed equally to this work.

1Department of Pathology and Immunology, Washington University School of Medicine, St. Louis, Missouri 63110, USA. 2Environmental Stress and Cancer Group, and NIEHSMicroarray Group, National Institute of Environmental Health Sciences, Research Triangle Park, North Carolina 27709, USA. 3School of Veterinary Medicine, and 4Department ofPathology and Laboratory Medicine, Center for Childhood Cancer Research, Children’s Hospital of Philadelphia, University of Pennsylvania School of Medicine, Philadelphia 19104,USA. 5Abramson Family Cancer Research Institute, Philadelphia, Philadelphia 19104, USA.

Vol 456 | 11 December 2008 | doi:10.1038/nature07392

819

©2008 Macmillan Publishers Limited. All rights reserved

Figs 4 and 5). These changes were not due to Artemis deficiency per se,as they were also observed in response to Rag DSBs generated inKu702/2 and Scid (Prkdc-deficient) abl pre-B cells, but not inSTI571-treated Artemis2/2 Rag22/2 abl pre-B cells (SupplementaryFig. 6). Thus, Rag DSBs regulate the expression of a large cohort ofgenes. A significant portion of this cohort is dependent on the activa-tion of Atm by Rag DSBs as evidenced by the analyses ofArtemis2/2 Atm2/2 abl pre-B cells (Fig. 1, Supplementary Figs 3–5and 7, and Supplementary Tables 2 and 3). Notably, however,approximately half of the genes are Atm-independent, demonstratingthat proteins other than Atm can initiate responses to Rag DSBs in G1-phase cells. Furthermore, analyses of Artemis2/2 abl pre-B cellsexpressing a dominant negative amino-truncated form of IkBa(IkBa-DN), which inhibits NF-kB activation, showed that 75 of theAtm-dependent gene expression changes were also NF-kB-dependent(Fig. 1b and c, Supplementary Figs 3 and 4, and Supplementary Table3). As only some of the Rag DSB-dependent genes are dependent onNF-kB, additional transcriptional pathways must be integrated into

the Rag DSB response. Thus, Rag DSBs promote a broad geneticprogram, in part, through the activation of Atm and NF-kB.

A similar genetic program is activated by Rag DSBs in developing Bcells in vivo. This is evidenced by gene expression profiling of purifiedB2201 IgM2 bone marrow B cells, which are primarily pre-B cells,from Artemis2/2 and Rag12/2 mice expressing an IgH transgene(Artemis2/2 IgHtg and Rag12/2 IgHtg, respectively; SupplementaryFig. 8 and Supplementary Table 4). Of the 364 Rag-DSB-responsivegenes identified by the analysis of Artemis2/2 abl pre-B cells, 109showed significant changes in expression in Artemis2/2 IgHtg pre-Bcells, as compared to Rag12/2 IgHtg pre-B cells (Fig. 1b,Supplementary Fig. 8 and Supplementary Tables 1–4). The asyn-chronous nature of Rag DSB induction in developing pre-B cells isprobably responsible, in part, for the reduced magnitude and numberof the gene expression changes in primary Artemis2/2 IgHtg pre-Bcells, as compared to the Artemis2/2 abl pre-B cells. Markedly, thesegene expression changes do lead to significant increases in theexpression of the proteins they encode (Fig. 2a and SupplementaryFig. 9). That these genes are Rag DSB-responsive was further demon-strated by the analysis of Artemis2/2 IgHtg and Rag12/2 IgHtg bonemarrow interleukin (IL)-7 pre-B cell cultures. Withdrawal of IL-7from Artemis2/2 IgHtg, but not from Rag12/2 IgHtg, bone marrowIL-7 cultures led to an increased accumulation of unrepaired IgLkcoding ends and a robust upregulation in the expression of many of

0 12 24 36 48 0 12 24 36 48 0 12 24 36

Artemis–/– Artemis–/– Atm–/–

175 genesAtm-independent

189 genesAtm-dependent

Artemis–/– IκBα-∆NIκBα-∆N

114 genesNF-κB-independent

75 genesNF-κB-dependent

Cd62l Common g-chain

P = 0.02 P = 0.03

Rel

ativ

e ex

pre

ssio

n

Cd69

P = 0.02P = 0.003

P = 0.0002

P = 0.03

Artemis–/–

Artemis–/– Atm–/–

48 STI571 (h)

NFY

NF-κB

or364 genes ≥2-fold

Rag2–/–

Rag2–/–

050

100150200250

0

2

4

6

Untreated +STI571 Untreated +STI571 Untreated +STI571

Untreated +STI571 Untreated +STI571 Untreated +STI571

Pim2 Cd40

0

50

100

150

0100200300400500600700

050

100150200250

0

10

20

30

Bcl3

Rel

ativ

e ex

pre

ssio

n

a

b

c

IκBα-∆N

Artemis–/–

Art–/– Atm–/–

Art–/–

Rag2–/–

Figure 1 | Rag DSBs activate a broad genetic program. a, NF-kB EMSA ofnuclear lysates from Rag22/2, Artemis2/2 and Artemis2/2 Atm2/2 abl pre-Bcells treated with STI571 for the indicated number of hours. NFY EMSA isshown as a control. Results are representative of three experiments.b, Schematic of gene expression changes in response to Rag DSBs bymicroarray analyses. The Atm-independent gene expression changes arereported in Supplementary Table 1. The Atm-dependent gene expressionchanges are reported in Supplementary Table 2 (NF-kB-independent) andSupplementary Table 3 (NF-kB-dependent). c, qRT–PCR analysis ofmessenger RNA isolated from Rag22/2 (white bars), Artemis2/2 (red bars),Artemis2/2 Atm2/2 (Art2/2 Atm2/2, grey bars) and Artemis2/2 IkBa-DN(Art2/2 IkBa-DN; blue bars) abl pre-B cells treated with STI571 for 0 or 48 h.Mean and s.d. from two experiments are shown; P values were calculatedusing a Student’s one-tailed t-test.

CD40 CD69 CD62L CD80

P = 0.002

P = 0.005

Artem

is–/

– IgHtg

Rag1–/

– IgHtg

Artem

is–/

– IgHtg

Rag1–/

– IgHtg

P = 0.0006

Rel

ativ

e ex

pre

ssio

n

P = 0.0002

30

40

50

20

60

02468

1012

0

25

50

75

100

125

0

5

10

15

0

10

20

30

40

50

+IL-7–IL-7

Rel

ativ

e ex

pre

ssio

n

Art

emis

–/–

(%)

P = 0.001 c

Artemis–/–

IgHtg

Rag1–/–

IgHtg

b

Bone marrow B220+IgM–B cellsa

IsoCD62L

IsoCD80

IsoCD69

IsoCD40

PB BM

Pim2

Cd40

Bcl3

Cd69

Figure 2 | Rag DSB-dependent gene expression changes in developing Bcells in vivo. a, Flow cytometric analyses of CD40, CD69, CD80 and CD62Lexpression by B2201 IgM2 bone marrow B cells (primarily pre-B cells) fromRag12/2 IgHtg and Artemis2/2 IgHtg mice (gated as shown inSupplementary Fig. 8). Histograms for specific antibodies (red) and isotypecontrols (Iso, blue) are shown. b, qRT–PCR of gene expression in bonemarrow cultures derived from Rag12/2 IgHtg and Artemis2/2 IgHtg micebefore (1IL-7, white bar) and 48 h after (2IL-7, red bar) the removal of IL-7.Mean and s.d. from two experiments are shown; P values were calculatedusing a Student’s one-tailed t-test. c, Percentage of Artemis2/2 abl pre-B cellsin the peripheral blood (PB) and bone marrow (BM) of Rag-deficient miceafter co-injection of STI571-treated Artemis2/2 and Rag-22/2 abl pre-B cellsat a 1:1 ratio. The mean (line) of each set of mice analysed is indicated and Pvalues were calculated using a two-tailed Wilcoxon matched-pairs test.

LETTERS NATURE | Vol 456 | 11 December 2008

820

©2008 Macmillan Publishers Limited. All rights reserved

the genes that are regulated by Rag DSBs in abl pre-B cells (Fig. 2b,Supplementary Fig. 10 and data not shown)18. Together, these find-ings demonstrate that Rag DSBs activate a broad genetic program indeveloping pre-B cells. Notably, these DNA breaks also promote geneexpression changes in pro-B and pro-T cells, suggesting that tran-scriptional programs are activated by Rag DSBs in all developinglymphocytes (Supplementary Fig. 11).

The genetic program induced by Rag DSBs includes genes that areactive in many cellular processes. Several have functions that wouldbe part of canonical DNA damage responses; for example, the anti-apoptotic proteins Pim2 and Bcl3 could antagonize pro-apoptoticpathways, such as those initiated by p53, promoting the survival ofdeveloping lymphocytes with Rag DSBs19–21. However, this programalso includes genes that have no known role in canonical DNAdamage responses; instead, these genes participate in diverse pro-cesses, many of which are important for lymphocyte developmentand function. These include common c-chain (also known as Il2rg),SWAP70 (SWA-70), Notch 1, CD69, CD40, CD80 (B7.1) and CD62L(L-selectin). Notably, CD62L, SWAP70 and CD69 have known func-tions in lymphocyte homing and migration, suggesting that thesegene expression changes could regulate the localization of developinglymphocytes with Rag DSBs22–26. Indeed, we observed a biased accu-mulation of Artemis2/2 abl pre-B cells with Rag DSBs, as compared toRag22/2 abl pre-B cells, in the bone marrow after co-injection of thesecells into Rag-deficient mice (Fig. 2c and Supplementary Fig. 12). Asimilar bias was not observed in the peripheral blood of these mice(Fig. 2c). Moreover, inhibition of Atm signalling in Artemis2/2 ablpre-B cells led to the diminished localization of these cells in the bonemarrow (Supplementary Fig. 12c). Similar results were obtained fromanalyses of independently derived Artemis2/2 and Rag22/2 abl pre-Bcells (data not shown). Lymphocytes traverse different microenviron-ments as they develop in the thymus and bone marrow; our findingssuggest that one of the functions of the Rag DSB-dependent geneticprogram may be to modulate this migration27–30.

We have established that persistent Rag DSBs in NHEJ-deficientlymphocytes activate a diverse genetic program in vitro and in vivo. Inwild-type lymphocytes, most Rag DSBs are repaired rapidly; thus, forthis program to regulate processes that are generally important fordevelopment, it must also be initiated by these transient DSBs. In thisregard, EMSA analyses did not show a marked increase in NF-kBactivation in wild-type abl pre-B cells undergoing V(D)J recombina-tion (Supplementary Fig. 13). However, this analysis might not besensitive enough to detect the activation of NF-kB in response totransient Rag DSBs. In agreement with this notion, we found thatNF-kB-dependent, Rag DSB-responsive genes are upregulated in anAtm-dependent manner in wild-type developing B cells undergoingV(D)J recombination (Fig. 3a and Supplementary Figs 9 and 14).

To determine directly whether transient Rag DSBs activate tran-scriptional pathways, we developed a sensitive single-cell assay for NF-kB activation. To this end, a retroviral reporter, pMSCV-NRE-GFP,that contains five tandem NF-kB responsive elements (NREs) drivingexpression of the green fluorescent protein (GFP) complementaryDNA on the antisense strand, was generated (Supplementary Fig.15a). pMSCV-NRE-GFP was introduced into Rag22/2 abl pre-B cellsthat contained the pMX-DELCJ retroviral V(D)J recombination sub-strate, hereafter referred to as Rag22/2 NRE cells (Fig. 3b andSupplementary Fig. 15)14. These cells do not express significant levelsof GFP in response to STI571, but they do upregulate GFP in an Atm-and NF-kB-dependent manner in response to DSBs induced by ioni-zing radiation (Fig. 3b and Supplementary Fig. 16).

To determine whether transient Rag DSBs activate NF-kB,Rag22/2 NRE cells were transduced with a retrovirus encodingRag2 (hereafter referred to as Rag22/2 NRE R2 cells), which permitsefficient V(D)J recombination after induction of Rag1 expression(Fig. 3c and data not shown). As STI571-treated cells are not divid-ing, GFP produced in response to transient Rag DSBs will persist,marking any cell that has activated NF-kB (Fig. 3b and

Supplementary Fig. 15c). Notably, after 48 h of STI571 treatment,53% of Rag22/2 NRE R2 cells were GFP1 (Fig. 3b). This NF-kBactivation was Atm-dependent, as an Atm inhibitor abrogated GFPexpression (Fig. 3b). If NF-kB activation was stimulated by transientRag DSBs, then there should be a higher level of completed rearran-gements in the GFP1 cells. Indeed, PCR analyses demonstrated a

GFP

2 9 6

8533

GFP–

GFP+

CD40 CD62LCD69

B22

0

IgM

Iso Iso IsoCD40 CD69 CD62L

WT

Bone marrow B220+IgM–

STI571 STI571 + iAtm

Rag2–/– NRE

Rag2–/– NRE R2

Atm–/–

a

Untreated

CJ

IL-2

DNA

+ – STI571GFP+

+GFP–

+

Rag2–/– NRE Rag2–/– NRE R2c

b

24 h

72 h

GFP CD40 CD69

Rag2–/– NRE R2d

Figure 3 | NF-kB activation in response to transient Rag DSBs. a, Flowcytometric analyses of CD40, CD69 and CD62L expression by B2201:IgM2

bone marrow B cells (gated cells in dot plot) from wild type (WT) andAtm2/2 mice. Histograms for the specific antibodies (red) and isotypecontrol (Iso, blue) are shown. b, Flow cytometric analysis of GFP expressionin Rag22/2 NRE and Rag22/2 NRE R2 cells treated with STI571 for 48 h inthe presence or absence of the Atm inhibitor KU-55933 (iAtm). Thepercentage of GFP1 cells is indicated. Results are representative of threeexperiments. c, Rag22/2 NRE and Rag22/2 NRE R2 abl pre-B cells wereuntreated (2) or treated with STI571 for 48 h (1), and GFP1 and GFP2

Rag22/2 NRE R2 abl pre-B cells were isolated by flow cytometric cell sorting.Serial fourfold dilutions of genomic DNA from all cells were assayed forpMX-DELCJ rearrangement by PCR (see Supplementary Fig. 15b). Resultsare representative of two experiments. CJ, coding joint. The IL-2 gene PCR isshown as a DNA quantity control. d, Flow cytometric analysis of CD40 andCD69 protein expression on GFP1 (blue histograms) and GFP2 (redhistograms) Rag-22/2 NRE R2 cells treated with STI571 for 24 or 72 h.Results are representative of two experiments.

NATURE | Vol 456 | 11 December 2008 LETTERS

821

©2008 Macmillan Publishers Limited. All rights reserved

fivefold higher level of pMX-DELCJ rearrangement (coding joint, CJ,formation) in the GFP1 cells as compared to GFP2 cells (Fig. 3c).Moreover, the GFP1 cells showed increased expression of genes regu-lated by Rag DSBs, including Pim2, Bcl3, Cd40 and Cd69 (Fig. 3d andSupplementary Fig. 17). CD69 expression in the GFP1 populationwas transient, similar to that observed in response to persistent RagDSBs (Fig. 3d and Supplementary Figs 4a and 5b). Together, thesedata demonstrate that transient Rag DSBs generated during normalV(D)J recombination activate DNA DSB responses, including Atmand NF-kB, which lead to the initiation of a diverse genetic programin all developing lymphocytes.

As physiological Rag DSBs activate a broad genetic program, wereasoned that genotoxic DSBs may also activate components of thisprogram in developing lymphocytes. Indeed, treatment of STI571-treated Rag22/2 abl pre-B cells with either ionizing radiation oretoposide—which both generate DNA DSBs—led to the Atm-dependent induction of many genes that are also induced by RagDSBs (Fig. 4a, b and Supplementary Fig. 18). Furthermore, treatmentof bone marrow from Rag12/2 IgHtg mice with either ionizing radi-ation or etoposide led to a B-cell-specific (B2201), Atm-dependent

induction of CD69 expression (Fig. 4c). These findings demonstratethat, in developing B cells, genotoxic DSBs can induce lymphocyte-specific gene expression changes that are normally induced by RagDSBs.

We have shown that physiological DNA DSBs generated duringantigen receptor gene assembly activate a broad genetic program indeveloping lymphocytes. Notably, this program is initiated by tran-sient Rag DSBs that are rapidly repaired during normal V(D)J recom-bination in all developing lymphocytes. Thus, we propose that RagDSBs provide cellular cues that regulate processes—such as migra-tion and homing—that are integral components of normal lympho-cyte development. As Atm is required to activate a portion of this RagDSB-dependent genetic program, we expect that the defects in theexpression of these genes contribute to the lymphopenia observed inataxia-telangiectasia.

Our findings establish that DNA DSBs generated during physio-logical processes can regulate the cell-type-specific expression ofgenes that participate in functions not directly involved in maintain-ing genomic stability or suppressing malignant transformation. Wespeculate that DNA DSBs generated in other physiological settings,such as immunoglobulin class-switch recombination, meiosis andDNA replication, may have similar effects. Furthermore, genotoxicDSBs generated by agents such as ionizing radiation and chemother-apeutic drugs could corrupt normal cellular processes through theaberrant activation of these cell-type-specific genetic programs.Thus, the effect of DNA DSB responses on cell physiology and func-tion is probably more pervasive than previously appreciated.

METHODS SUMMARY

Standard methodologies used for cell culture, retroviral vector generation, west-

ern blotting, Southern blotting, EMSA, gene expression profiling, PCR and flow

cytometry are described in detail in the Methods. All v-abl-transformed pre-B

cell lines were generated from mice harbouring the EmBcl-2 transgene as prev-

iously described14. Flow cytometric analyses were performed on a FACSCaliber

(BD Biosciences) with the normalized geometric mean fluorescence intensity

calculated as the difference between the geometric mean fluorescence for the

specific antibody and the isotype control. Flow cytometric cell sorting (FACS)

was performed using a FACSVantage (BD Biosciences). To generate pMSCV-

NRE-GFP, the NF-kB responsive elements and TATA box from pNF-kB-Luc

(Stratagene) were amplified using the following oligonucleotides: pNRE-F,

CCAAACTCATCAATGTATCTTATCATG and pNRE-R, TACCAACAGTA-

CCGGAATGC. This PCR product was cloned upstream of a cDNA encoding

enhanced GFP (Clontech) on the bottom strand of pMSCV containing an IRES

upstream of the Thy1.2 cDNA. Gene expression profiling was performed using

the Affymetrix 430 v2.0 mouse genome microarray. To assay abl pre-B cell

localization, cells were treated for 48 h with STI571, labelled with 0.1 mM

CFSE (Invitrogen) or 5 mM SNARF (Invitrogen) at 37 uC for 15 min in PBS at

2–5 3 106 ml21, washed three times, and resuspended in DMEM at a concen-

tration of 80 3 106 ml21. Labelled cells were then mixed in a 1:1 ratio, and a total

of 40 3 106 cells were injected into the tail vein of each Rag-deficient mouse.

Bone marrow and peripheral blood was collected 30–60 min after injection and

analysed by FACS for presence of CFSE- and SNARF-labelled abl pre-B cells.

Full Methods and any associated references are available in the online version ofthe paper at www.nature.com/nature.

Received 16 April; accepted 4 September 2008.Published online 12 October 2008.

1. Shiloh, Y. ATM and related protein kinases: safeguarding genome integrity. NatureRev. Cancer 3, 155–168 (2003).

2. Rouse, J. & Jackson, S. P. Interfaces between the detection, signaling, and repair ofDNA damage. Science 297, 547–551 (2002).

3. Zhou, B. B. & Elledge, S. J. The DNA damage response: putting checkpoints inperspective. Nature 408, 433–439 (2000).

4. Innes, C. L. et al. ATM requirement in gene expression responses to ionizingradiation in human lymphoblasts and fibroblasts. Mol. Cancer Res. 4, 197–207(2006).

5. Rashi-Elkeles, S. et al. Parallel induction of ATM-dependent pro- andantiapoptotic signals in response to ionizing radiation in murine lymphoid tissue.Oncogene 25, 1584–1592 (2006).

6. Tonegawa, S. Somatic generation of antibody diversity. Nature 302, 575–581(1983).

IR Etp

CD69

Untreated IR Etp

CD69

B22

0

B22

0

CD69

P = 0.002P = 0.0003

P = 0.001P = 0.005

0.0

0.5

1.0

1.5

2.0

0

300

600

900

0

10

20

30

40

0

10

20

30

40No IRIRIR + iAtm

Cd40 Cd69 Il-12αCommong-chain

Rag2–/– abl pre-B cells + STI571

Rel

ativ

e ex

pre

ssio

n

Rag2–/– abl pre-B cells + STI571

Rag1–/– IgHtg bone marrow

No IR/EtpIR/EtpIR/Etp + iAtm

iAtm

Vehicle

a

c

b

Figure 4 | Genotoxic DSBs promote changes in expression of lymphocyte-specific genes. a, qRT–PCR analysis of gene expression in Rag22/2 abl pre-B cells 2 h after receiving 0 Gy (white bars) or 4 Gy of ionizing radiation (IR)in the presence (grey bars) or absence (red bars) of the Atm inhibitor KU-55933 (iAtm). All cells were treated with STI571 for 24 h before ionizingradiation. Results are the mean and s.d. from two experiments; P values werecalculated using a Student’s one-tailed t-test. b, Flow cytometric analysis ofCD69 expression on STI571-treated Rag22/2 abl pre-B cells treated withionizing radiation as described in a or 2 h after being treated with 5 mMetoposide (Etp). Results are representative of three experiments. c, Flowcytometric analysis of B220 and CD69 expression on bone marrow cells fromRag12/2 IgHtg mice 2 h after irradiation (0.5 Gy) or etoposide treatment(5 mM) in the presence or absence of the Atm inhibitor KU-55933. Resultsare representative of two experiments.

LETTERS NATURE | Vol 456 | 11 December 2008

822

©2008 Macmillan Publishers Limited. All rights reserved

7. Fugmann, S. D. et al. The RAG proteins and V(D)J recombination: complexes,ends, and transposition. Annu. Rev. Immunol. 18, 495–527 (2000).

8. Rooney, S., Chaudhuri, J. & Alt, F. W. The role of the non-homologous end-joiningpathway in lymphocyte development. Immunol. Rev. 200, 115–131 (2004).

9. Meek, D. W. The p53 response to DNA damage. DNA Repair (Amst.) 3,1049–1056 (2004).

10. Guidos, C. J. et al. V(D)J recombination activates a p53-dependent DNA damagecheckpoint in scid lymphocyte precursors. Genes Dev. 10, 2038–2054 (1996).

11. Perkins, N. D. Integrating cell-signalling pathways with NF-kB and IKK function.Nature Rev. Mol. Cell Biol. 8, 49–62 (2007).

12. Huang, T. T., Wuerzberger-Davis, S. M., Wu, Z. H. & Miyamoto, S. Sequentialmodification of NEMO/IKKc by SUMO-1 and ubiquitin mediates NF-kB activationby genotoxic stress. Cell 115, 565–576 (2003).

13. Gerondakis, S. et al. Unravelling the complexities of the NF-kB signalling pathwayusing mouse knockout and transgenic models. Oncogene 25, 6781–6799 (2006).

14. Bredemeyer, A. L. et al. ATM stabilizes DNA double-strand-break complexesduring V(D)J recombination. Nature 442, 466–470 (2006).

15. Muljo, S. A. & Schlissel, M. S. A small molecule Abl kinase inhibitor inducesdifferentiation of Abelson virus-transformed pre-B cell lines. Nature Immunol. 4,31–37 (2003).

16. May, M. J. et al. Selective inhibition of NF-kB activation by a peptide that blocksthe interaction of NEMO with the IkB kinase complex. Science 289, 1550–1554(2000).

17. Wu, Z. H., Shi, Y., Tibbetts, R. S. & Miyamoto, S. Molecular linkage between thekinase ATM and NF-kB signaling in response to genotoxic stimuli. Science 311,1141–1146 (2006).

18. Grawunder, U. et al. Expression of DNA-dependent protein kinase holoenzymeupon induction of lymphocyte differentiation and V(D)J recombination. Eur. J.Biochem. 241, 931–940 (1996).

19. Kashatus, D., Cogswell, P. & Baldwin, A. S. Expression of the Bcl-3 proto-oncogenesuppresses p53 activation. Genes Dev. 20, 225–235 (2006).

20. Amaravadi, R. & Thompson, C. B. The survival kinases Akt and Pim as potentialpharmacological targets. J. Clin. Invest. 115, 2618–2624 (2005).

21. Hayden, M. S. & Ghosh, S. Signaling to NF-kB. Genes Dev. 18, 2195–2224 (2004).22. Rosen, S. D. Ligands for L-selectin: homing, inflammation, and beyond. Annu. Rev.

Immunol. 22, 129–156 (2004).

23. Feng, C. et al. A potential role for CD69 in thymocyte emigration. Int. Immunol. 14,535–544 (2002).

24. Matloubian, M. et al. Lymphocyte egress from thymus and peripheral lymphoidorgans is dependent on S1P receptor 1. Nature 427, 355–360 (2004).

25. Shiow, L. R. et al. CD69 acts downstream of interferon-a/b to inhibit S1P1 andlymphocyte egress from lymphoid organs. Nature 440, 540–544 (2006).

26. Pearce, G. et al. Signaling protein SWAP-70 is required for efficient B cell homingto lymphoid organs. Nature Immunol. 7, 827–834 (2006).

27. Ladi, E., Yin, X., Chtanova, T. & Robey, E. A. Thymic microenvironments for T celldifferentiation and selection. Nature Immunol. 7, 338–343 (2006).

28. Johnson, K. et al. Regulation of immunoglobulin light-chain recombination by thetranscription factor IRF-4 and the attenuation of interleukin-7 signaling. Immunity28, 335–345 (2008).

29. Tokoyoda, K. et al. Cellular niches controlling B lymphocyte behavior within bonemarrow during development. Immunity 20, 707–718 (2004).

30. Nagasawa, T. Microenvironmental niches in the bone marrow required for B-celldevelopment. Nature Rev. Immunol. 6, 107–116 (2006).

Supplementary Information is linked to the online version of the paper atwww.nature.com/nature.

Acknowledgements We thank J. Bednarski, B. Van Honten and M. Diaz for criticalreview of the manuscript, F. W. Alt for providing us with the Artemis2/2 mice andD. Ballard for providing us with the IkBa-DN construct. This research is supportedby the National Institutes of Health (NIH, grant AI47829) and the WashingtonUniversity Department of Pathology and Immunology (to B.P.S.); the Departmentof Pathology and Center for Childhood Cancer Research of the Children’s Hospitalof Philadelphia, and the Abramson Family Cancer Research Institute (to C.H.B.);and the intramural research program of the NIH, National Institute ofEnvironmental Health Sciences (to R.S.P.). B.P.S. is a recipient of a Research ScholarAward from the American Cancer Society. C.H.B. is a Pew Scholar in the BiomedicalSciences. A.L.B. is supported by a post-doctoral training grant from the NIH.

Author Information The microarray gene expression data have been deposited inNCBI’s Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/) underaccession number GSE9024. Reprints and permissions information is available atwww.nature.com/reprints. Correspondence and requests for materials should beaddressed to B.P.S. (Sleckman@immunology.wustl.edu).

NATURE | Vol 456 | 11 December 2008 LETTERS

823

©2008 Macmillan Publishers Limited. All rights reserved

METHODSMice. The Atm2/2, Scid, Ku702/2, Artemis2/2, VH147 IgH transgenic and

EmBcl-2 transgenic mice have been described previously14,31–33. Mice were bred

and maintained under specific pathogen-free conditions at the Washington

University School of Medicine and were handled in accordance to the guidelines

set forth by the Division of Comparative Medicine of Washington University.

Cell culture. Three independently derived Rag22/2 (R.1, R.2 and R.3), three

independently derived Artemis2/2 (A.1, A.2 and A.3), two independently

derived Artemis2/2 Atm2/2 (AA.1 and AA.2) and individual Scid and Ku-

702/2 v-abl-transformed pre-B cell lines containing the EmBcl-2 transgene weregenerated. The three Artemis2/2 IkBa-DN cells (A.3DN1, A.3DN2 and A.3DN3)

were generated by standard transfection of the A.3 cells with a cDNA encoding

IkBa-DN34. Expression of the IkBa-DN protein in these cells was confirmed by

western blotting. STI571 treatments were carried out using 3 mM STI571 as

previously described14. The KU-55933 Atm inhibitor (Sigma) was used at

15 mM, the BAY-11-7085 inhibitor (Calbiochem) was used at 20mM, and etopo-

side was used at 5mM. Irradiation was carried out with a Cs137 source, at doses of

either 0.5 or 4 Gy.

Southern blotting. Southern blot analysis was performed as described prev-

iously14,35.

Western blotting and EMSA. Western blotting was carried out as described

previously, using an antibody to Pim2 (Santa Cruz, 1D12)36. The secondary

reagent was horseradish peroxidase (HRP)-conjugated goat anti-mouse IgG

(Zymed). EMSAs were run as described previously and analysed using a Li-

Cor Odyssey Infrared scanner37. Supershifts were performed using anti-p50

and anti-p65 antibodies (Santa Cruz, sc-114 X and sc-372 X, respectively).

Gene expression profiling. All abl pre-B cells were collected 48 h after treatment

with STI571. RNA was isolated using Qiagen RNeasy technology following themanufacturer’s instructions. Gene expression analysis was conducted using

Affymetrix Mouse Genome 2.0 GeneChip arrays (Mouse 430 v2, Affymetrix).

One microgram of total RNA was amplified as directed in the Affymetrix One-

Cycle cDNA Synthesis protocol. Fifteen micrograms of amplified biotin-com-

plementary-RNAs were fragmented and hybridized to each array for 16 h at

45 uC in a rotating hybridization oven using the Affymetrix Eukaryotic Target

Hybridization Controls and protocol. Array slides were stained with streptavidin

and phycoerythrin using a double-antibody staining procedure, and then

washed using the EukGE-WS2v5 protocol of the Affymetrix Fluidics Station

FS450 for antibody amplification. Arrays were scanned in an Affymetrix

Scanner 3000 and data was obtained using the GeneChip Operating Software

(Version 1.2.0.037). Data pre-processing, normalization and error modelling

was performed with the Rosetta Resolver system (Version 6.0)38.

To identify differentially expressed genes between the Rag22/2 and the

Artemis2/2 abl pre-B cells, an error-weighted analysis of variance (ANOVA)

using the Benjamini–Hochberg false discovery rate was performed using

Rosetta Resolver. At P # 0.05, this analysis yielded 1,578 probe sets. Fold changes

were generated in Resolver, based on the ratio of each individual cell line relativeto the average of the three Rag22/2 cells with 100-fold being the maximum

change that Resolver can report. Of the 1,578 probes found using the ANOVA

analysis, only those with a fold-change $2.0 (not including anti-correlated

genes) in all ratios for a given genotype were considered for further analysis. A

Venn diagram of the resulting ANOVA and fold-change data sets for each geno-

type comparison to Rag22/2 was then used to determine which expression

changes were dependent on Atm and NF-kB.

For gene expression profiling of primary developing B cells, B2201 cells were

isolated from Rag12/2 IgHtg and Art2/2 IgHtg bone marrow using Mouse Pan B

(B220) Dynabeads from Invitrogen. RNA was isolated using the Qiagen RNeasy

Mini Kit. Microarrays were treated as described above for the abl pre-B cells. For

the probe sets that were identified as being differentially regulated in response to

Rag DSBs in the abl pre-B cells, we compared the average expression in the twoRag12/2 IgHtg samples to each of the two Art2/2 IgHtg samples. The probe sets

that were not anti-correlated and changed $1.5-fold in the average of these

comparisons are included in Supplementary Table 4.

Flow cytometry. Flow cytometric analyses were performed using fluorescein

isothiocyanate (FITC)-conjugated anti-CD45R/B220, allophycocyanin (APC)-conjugated anti-IgM, phycoerythrin (PE)-conjugated anti-CD40, PE-conju-

gated anti-CD62L, PE-conjugated anti-CD69, and the appropriate isotype con-

trols (BD Biosciences).

Analysis of pMX-DELCJ rearrangement. PCR was carried out as described

previously on fourfold dilutions of genomic DNA14. IL-2 PCR primers were

IMR42 and IMR43, and the product was probed with IMR042-2, all of whichhave been described previously1. Primer sequences for pMX-DELCJ are as fol-

lows: pA, CACAGGATCCCACGAAGTCTTGAGACCT; pB, ATCTGGATCC-

GTGCCGCCTTTGCAGGTGTATC; pD, AGACGGCAATATGGTGGA.

qRT–PCR gene expression analysis. qRT–PCR was performed using a

Stratagene Mx3000P real-time PCR machine and Platinum SYBR Green qPCR

SuperMix-UDG (Invitrogen). PCRs were carried out using the following pro-gram: 95 uC, 30 s; 55 uC, 60 s; 72 uC, 30 s; 40 cycles. Relative expression is calcu-

lated as the difference between expression of b-actin (C1) and expression of the

gene of interest (C2), using the equation relative expression 5 22 (C1 2 C2). C is

the PCR cycle in which the product detection curve becomes linear. Primers are

listed in Supplementary Table 5.

31. Rooney, S. et al. Leaky Scid phenotype associated with defective V(D)J coding endprocessing in Artemis-deficient mice. Mol. Cell 10, 1379–1390 (2002).

32. Gu, Y. et al. Growth retardation and leaky SCID phenotype of Ku70-deficientmice. Immunity 7, 653–665 (1997).

33. Mandik-Nayak, L., Racz, J., Sleckman, B. P. & Allen, P. M. Autoreactive marginalzone B cells are spontaneously activated but lymph node B cells require T cellhelp. J. Exp. Med. 203, 1985–1998 (2006).

34. Brockman, J. A. et al. Coupling of a signal response domain in IkBa to multiplepathways for NF-kB activation. Mol. Cell. Biol. 15, 2809–2818 (1995).

35. Ramsden, D. A. & Gellert, M. Formation and resolution of double-strand breakintermediates in V(D)J rearrangement. Genes Dev. 9, 2409–2420 (1995).

36. Khor, B. et al. Proteasome activator PA200 is required for normalspermatogenesis. Mol. Cell. Biol. 26, 2999–3007 (2006).

37. Weaver, B. K., Bohn, E., Judd, B. A., Gil, M. P. & Schreiber, R. D. ABIN-3: a molecularbasis for species divergence in interleukin-10-induced anti-inflammatory actions.Mol. Cell. Biol. 27, 4603–4616 (2007).

38. Weng, L. et al. Rosetta error model for gene expression analysis. Bioinformatics 22,1111–1121 (2006).

doi:10.1038/nature07392

©2008 Macmillan Publishers Limited. All rights reserved