Functional Screening Identifies CRLF2 in Precursor B-Cell Acute

Functional screening identiﬁes CRLF2 in precursor B-cell acute lymphoblastic leukemia

Akinori Yodaa, Yuka Yodaa, Sabina Chiarettib, Michal Bar-Natana, Kartik Mania, Scott J. Rodigc, Nathan Westa, Yun Xiaoc, Jennifer R. Browna, Constantine Mitsiadesa, Martin Sattlera, Jeffrey L. Kutokc, Daniel J. DeAngeloa, Martha Wadleigha, Alfonso Piciocchid, Paola Dal Cinc, James E. Bradnera, James D. Grifﬁna, Kenneth C. Andersona, Richard M. Stonea, Jerome Ritza, Robin Foàb, Jon C. Asterc, David A. Franka, and David M. Weinstocka,1

aDepartment of Medical Oncology, Dana-Farber Cancer Institute, Harvard Medical School, Boston, MA 02115; bDivision of Hematology, Department of Cellular Biotechnologies and Hematology, “Sapienza” University of Rome, 00185 Rome, Italy; cDepartment of Pathology, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA 02115; and dGruppo Malattie Ematologiche dell’Adulto Data Center, 00161 Rome, Italy

Edited by Ross Levine, Memorial Sloan-Kettering Cancer Center, and accepted by the Editorial Board November 12, 2009 (received for review October 9, 2009) The prognosis for adults with precursor B-cell acute lymphoblastic which is believed to serve as a scaffold. As a consequence, JAK leukemia (B-ALL) remains poor, in part from a lack of therapeutic gain-of-function mutants do not confer a transformed phenotype targets. We identified the type I cytokine receptor subunit CRLF2 in the absence of a compatible cytokine receptor (16). in a functional screen for B-ALL–derived mRNA transcripts that can We performed a retroviral cDNA library screen to identify substitute for IL3 signaling. We demonstrate that CRLF2 is overex- gain-of-function mutations from primary B-ALL specimens. We pressed in approximately 15% of adult and high-risk pediatric B- made several improvements to older approaches that increase ALL that lack MLL, TCF3, TEL, and BCR/ABL rearrangements, but the efficiency of library construction and cDNA representation. not in B-ALL with these rearrangements or other lymphoid malignancies. CRLF2 overexpression can result from translocation with In addition, we added selection-based clone recovery, which fi the IGH locus or intrachromosomal deletion and is associated with essentially eliminates false-positive ndings. Using this system, poor outcome. CRLF2 overexpressing B-ALLs share a transcriptional we identified CRLF2, a type I cytokine receptor subunit also signature that significantly overlaps with a BCR/ABL signature, and known as thymic stromal lymphopoietin receptor (TSLPR), as a is enriched for genes involved in cytokine receptor and JAK-STAT proto-oncogene in B-ALL. CRLF2 binds its ligand, thymic signaling. In a subset of cases, CRLF2 harbors a Phe232Cys gain-of- stromal lymphopoietin (TSLP), as part of a heterodimeric function mutation that promotes constitutive dimerization and cy- complex with the IL7 receptor subunit (IL7R) (17). TSLP is tokine independent growth. A mutually exclusive subset harbors produced by epithelial cells at sites of inflammation, where it activating mutations in JAK2. In fact, all 22 B-ALLs with mutant activates myeloid dendritic cells and Th2 immune responses (18, JAK2 that we analyzed overexpress CRLF2, distinguishing CRLF2 19). TSLP also promotes early B-cell development (20) and as the key scaffold for mutant JAK2 signaling in B-ALL. Expression stimulates the growth of some human B-ALLs in vitro (21). of WT CRLF2 with mutant JAK2 also promotes cytokine independ- CRLF2 ent growth that, unlike CRLF2 Phe232Cys or ligand-induced signal- We demonstrate that overexpression is common only TEL, MLL, TCF3, ing by WT CRLF2, is accompanied by JAK2 phosphorylation. Finally, in B-ALL cases that lack rearrangements of cells dependent on CRLF2 signaling are sensitive to small molecule and BCR/ABL (5–8) and confers a poor prognosis among chil- inhibitors of either JAKs or protein kinase C family kinases. Togeth- dren and adults. Similar to recent reports (22, 23), we show that er, these findings implicate CRLF2 as an important factor in B-ALL CRLF2 overexpression can result from locus rearrangement, with diagnostic, prognostic, and therapeutic implications. either translocation or intrachromosomal deletion. We identify CRLF2 Phe232Cys and JAK2 Arg683 gain-of-function muta- JAK2 | TSLPR | TSLP tions in mutually exclusive subsets of CRLF2-overexpressing B- ALL that transform growth factor–dependent cells to factor uring the past decade, studies using oligonucleotide arrays independence. Strikingly, 100% of B-ALL cases with mutant fi Dand high-throughput sequencing have identi ed several JAK2 overexpress CRLF2, suggesting that CRLF2 is the essen- genetic and transcriptional aberrations in B-cell acute lympho- tial, or at least the overwhelmingly predominate, scaffold for blastic leukemia (B-ALL) (1), leading to three conceptual ad- mutant JAK2 activity in B-ALL. Finally, we show that the gene vances. First, genes involved in normal B-cell development (e.g., signature associated with CRLF2 overexpression is highly similar PAX5, IKZF1) are frequently mutated in B-ALL (1–3). Second, in both pediatric and adult cases, and significantly overlaps with a B-ALL is highly heterogeneous and can exist as multiple, ge- netically distinct clones within the same individual (1, 4). Third, BCR/ABL signature. Together, these studies establish CRLF2 as B-ALL transcriptional profiles cluster based on characteristic a key factor in B-ALL, and support its use as a prognostic and chromosomal rearrangements, hereafter defined as rearrange- therapeutic target. ments of TEL, MLL, TCF3, and BCR/ABL (5–8). However, one third of B-ALL cases lack characteristic rear- fi rangements (9). Transcriptional pro les from a subset of these Author contributions: A.Y., S.C., K.M., S.J.R., J.B., C.S.M., J.D.G., J.R., J.C.A., D.A.F., and leukemias cluster with profiles from BCR/ABL–expressing B- D.M.W. designed research; A.Y., Y.Y., S.C., M.B.-N., K.M., S.J.R., N.W., Y.X., J.B., C.S.M., and ALL (3, 5), suggesting that the former harbor cryptic alterations P.S.D.C. performed research; S.C., M.B.-N., S.J.R., M.S., J.L.K., D.J.D., M.W., J.E.B., J.D.G., K.C.A., R.M.S., R.F., J.C.A., and D.A.F. contributed new reagents/analytic tools; A.Y., K.M., in tyrosine kinase signaling. Supporting this notion, mutations in S.J.R., N.W., J.B., C.S.M., A.P., P.S.D.C., J.E.B., D.A.F., and D.M.W. analyzed data; and A.Y., JAKs were recently identified in a small percentage of pediatric J.C.A., D.A.F., and D.M.W. wrote the paper. B-ALL and approximately 20% of ALL in children with Down The authors declare no conflict of interest. – syndrome (10 14). This article is a PNAS Direct Submission. R.L. is a guest editor invited by the Editorial Upon ligand binding to a type I cytokine receptor, JAKs phos- Board. phorylate substrates including STATs, which in turn affect the 1To whom correspondence should be addressed. E-mail: [email protected]. transcription of progrowth and antiapoptotic factors (15). JAK This article contains supporting information online at www.pnas.org/cgi/content/full/ enzymatic activity requires interaction with a cytokine receptor, 0911726107/DCSupplemental.

252–257 | PNAS | January 5, 2010 | vol. 107 | no. 1 www.pnas.org/cgi/doi/10.1073/pnas.0911726107 Downloaded by guest on September 27, 2021 Results A CRLF2 overexpression Mutated CRLF2 Is a Gain-of-Function Oncoprotein in Poor-Prognosis Yes (n=15), range No (n=75), range p value fi Median age (years) 32.5, 19.3-67.1 34.3, 15.5-64.8 0.92 B-ALL. We identi ed CRLF2 in a functional screen for leukemia- WBC (x103/µL) 1 26.9, 1.9-422.0 20.2, 1.4-357.0 0.42 derived cDNA that activate tyrosine kinase signaling (Fig. 1A). Male:Female 11:4 41:34 0.25 In this screen, we infect the murine IL3-dependent cell line BaF3 Median overall 25.5 months Not reached with retroviral cDNA libraries constructed from bone marrow survival Median disease- 17.8 months 37.8 months aspirates involved with more than 80% tumor. Clones that sur- free survival vive IL3 withdrawal invariably harbor tumor-derived cDNAs that obviate the requirement for IL3. After infection with a cDNA Disease-free survival Overall survival B 100 library constructed from a B-ALL specimen with 46, XY kar- p=0.0246 p=0.1738 yotype, we identified IL3-independent clones that contained a CRLF2 80 CRLF2 low/no mutated, full-length cDNA transcript of . expression (n=75) We used a combination of quantitative (q) RT-PCR, immu- 60 CRLF2 low/no fi expression (n=75) nohistochemistry (IHC), and gene expression pro ling (GEP) to 40 CRLF2 high assay CRLF2 expression in adult B-ALL samples from Dana- expression (n=15) 20 CRLF2 high Farber Cancer Institute (DFCI; n = 97) and Gruppo Malattie expression (n=15) Ematologiche dell’Adulto (GIMEMA; n = 157) cohorts (Fig. 1B (%) remaining Proportion 0 Months 020 40 60 80 100 020 40 60 80 100 and Fig. S1). Cases with CRLF2 overexpression were clearly distinct, and overexpression correlated completely between as- Fig. 2. Demographic and outcome analysis for patients with B-ALL that lack says (Fig. 1B and Fig. S1). Overall, CRLF2 was overexpressed in characteristic rearrangements, based on CRLF2 expression. (A) A comparison 15 of 120 (12.5%) adult B-ALL cases that lacked characteristic of 90 adult patients with B-ALL who lack characteristic cytogenetic re- = = gene rearrangements, compared with 0 of 134 with these rearrangements, pooled from GIMEMA (n 70) and DFCI specimens (n 20). All − values are at the time of diagnosis. Median follow-up for the full cohort is 39.4 arrangements (P < 10 4). CRLF2 overexpression was not present P = months. P values were derived using the Wilcoxon test for quantitative var- in 69 cases of T-cell ALL assayed by GEP ( 0.001 vs. 15 of 120). iables and a Fisher exact test for qualitative variables. 1Values were available We analyzed 90 adult patients with B-ALL that lacked char- for only 85 of 90 patients. (B) Disease-free survival and overall survival from acteristic rearrangements who had available demographic and the time of diagnosis were estimated using the Kaplan-Meier product-limit outcome information, pooled from the DFCI (n = 20) and GI- method and compared in univariate analysis by the log-rank test. MEMA (n = 70) cohorts (Fig. 2A). CRLF2-overexpressing and nonoverexpressing cohorts had similar median age, sex dis- tribution, and white blood cell counts at presentation. However, CRLF2 Overexpression in Pediatric B-ALL and Other Lymphoid disease-free survival (Fig. 2B) was significantly shorter among Malignancies. To determine the frequency of CRLF2 over- patients with CRLF2 overexpression (median, 17.8 mo vs. 37.8 expression in pediatric B-ALL, we reviewed the Gene Expres- mo; P < 0.03). There was also a trend for shorter overall survival sion Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/ among these patients (median, 25.5 mo vs. not reached; P = 0.17). geo/) for B-ALL samples assayed on the Affymetrix U133 plat- form. Affymetrix HG-U133A and HG-U133Aplus2 arrays con- Thus, CRLF2 overexpression is a marker of poor outcome among CRLF2 patients with B-ALL that lack characteristic rearrangements. tain a single probe set (208303_s_at) that targets the transcript (both complete and partial coding sequence, as well as expression sequence tags). We identified 9 datasets (Table S1 and Table S2) with a total of 1,253 pediatric ALL cases. Among the 3 datasets that included A Generate cDNA from IL-3-dependent patient sample Ba/F3 cells only high-risk patients, 52 (14.8%) of 351 B-ALL cases that lacked characteristic rearrangements had CRLF2 overexpression Clone into Repackage P < −4 retrovirus virus compared with 0 of the remaining 130 B-ALL cases ( 10 ). Infection This may underestimate the frequency of CRLF2 overexpression Sequence in the cohort without characteristic rearrangements, as some Select for integrants Withdraw datasets did not distinguish patients based on karyotype, so pa- IL-3 tients with characteristic rearrangements were presumably in- ALL-73 B Telomeric cluded in the cohort of 351 B-ALL cases. In the six datasets that CRLF2 included both standard-risk and high-risk patients, CRLF2 overexpression was present in only 23 of 559 (4.1%) cases that P < −4 Centromeric lacked characteristic rearrangements ( 10 vs. 52 of 351 IgH from the high-risk–only datasets). In addition, 0 of 51 T-ALL ALL-74 ALL-76 P = Centromeric cases overexpressed CRLF2 ( 0.002 vs. 52 of 351 from the high-risk–only datasets). Telomeric We next asked whether CRLF2 is overexpressed in other

CRLF2/IgH lymphoid malignancies. We selected a cross-section of chronic fusion lymphocytic leukemia (CLL) specimens, based on karyotype, Fig. 1. Identification of CRLF2 in B-ALL. (A) Tumor-derived cDNA is pack- IGHV somatic hypermutation, ZAP-70 and CD38 expression, aged into retroviral particles that are used to infect BaF3 cells. Integrants CLL family history, and clinical features (Table S3). All 30 (gray) survive in puromycin selection. Surviving clones (black) are isolated specimens had no detectable CRLF2 mRNA. A review of gene after IL3 withdrawal and their integrated cDNAs are sequenced and re- expression profiles from GEO dataset GSE6477 (n = 162) failed packaged within retrovirus and confirmed. (B) IHC using anti-CRLF2 antibody to identify significant CRLF2 expression in normal plasma cells, and interphase FISH. ALL-73 has no detectable CRLF2 expression by IHC, monoclonal gammopathy of undetermined significance, smol- whereas both ALL-74 and ALL-76 have visible expression and surface local- dering multiple myeloma (MM), newly diagnosed MM, or re- ization. (Left) FISH demonstrates 2 fusions of the CRLF2 (yellow) and IGH CRLF2 (green) probes in ALL-74, consistent with a reciprocal translocation, but not lapsed MM. Low or undetectable expression was also in ALL-73. (Right) FISH shows one probe centromeric to CRLF2 (yellow) lost in confirmed in a panel of T-ALL (n = 22) and other (n = 14) cell ALL-76, consistent with an intrachromosomal deletion. lines (Table S4). MEDICAL SCIENCES

Yoda et al. PNAS | January 5, 2010 | vol. 107 | no. 1 | 253 Downloaded by guest on September 27, 2021 CRLF2 Locus Rearrangements. Russell et al. (23) recently demon- A C strated that the CRLF2 locus, which is located in the pseudoau- CRLF2 Phe232 CRLF2 Phe232Cys Nonreducing 148 tosomal regions of chromosomes X and Y, can undergo two types ALL-73 genomic ALL-40 T G T 98 T T T DNA of rearrangement: intrachromosomal deletion or translocation 64 with the Ig heavy chain (IGH) locus. Mullighan et al. (22) sub- 50 sequently mapped the deletions and demonstrated that they jux- Reducing 148 tapose CRLF2 3′ of the first noncoding exon of P2RY8. In both cDNA 711T cases, the chromosome X/Y breakpoints are upstream of CRLF2 711T>G 98 CRLF2 64 and place the full coding sequence under alternate tran- 50 scriptional control. CRLF2 and IGH are close to the telomeres of CRLF2 wt - FC FC FC FC wt chromosome X/Y and 14, respectively. Thus, we designed a FISH IL7R -IL-3 IL7R strategy using probes against regions flanking CRLF2 and IGH B D CRLF2 wild-type vector (Fig. S2). Whereas the CRLF2 and IGH loci are clearly separate in Phe232Cys BCR/ABL CRLF2 B 102 UT7 103 BaF3 CRLF2 wild-type cells that lack CRLF2 expression (Fig. 1 ), FISH in 3 of 6 B-ALL 2 JAK2 wild-type 10 10 5/14 (35.7%) specimens with high CRLF2 expression demonstrated joining of CRLF2 F232C CRLF2 and IGH probes, consistent with a reciprocal chromoso- 1 10 JAK2 wildtype 0.1 1 3/14 (21.4%) mal translocation (Fig. 1B). Two of the 3 remaining specimens had -2 CRLF2 wild-type 10 0.1 JAK2 R683 mutant -2 loss of a centromeric chromosome X/Y probe, consistent with an 10-3 10 6/14 (42.9%) intrachromosomal deletion (Fig. 1B). The final specimen had day 0 1 2 3 4 5 6 day 0 1 2 3 4 5 4 3 neither a deletion nor a translocation, suggesting an additional 10 10 ALL-72 ALL-46 ALL-74 3 BaF3 BaF3 Arg683Gly Arg683Thr Arg683Ser 10 2 mechanism for CRLF2 overexpression. 2 10 CRLF2/IGH translocations are akin to lymphoma-associated 10 10 rearrangements involving BCL2 and BCL1 that result from Cell count relative to day 0 10 1 2541A>G 2542G>C 2543A>C aberrant V(D)J recombination (24). In 6 specimens, we PCR 0.1 1 fi IGHJ -2 CRLF2/IL7R+TSLP CRLF2+TSLP ampli ed der(14) translocation junctions between seg- 10 + CRLF2 F232C/IL7R+TSLP + CRLF2/IL7R 0.1 CRLF2 F232C+TSLP CRLF2 CRLF2/JAK2 R683G CRLF2 ments on chromosome 14 and the region centromeric of JAK2 R683G CRLF2 F232C/IL7R CRLF2/JAK2 R683S IL7R+TSLP A JAK2 R683S -2 CRLF2 F232C on chromosome X/Y (Fig. 3 ). Junctions involved sequence CRLF2/JAK2 wt JAK2 wt 10 + IL7R approximately 8 to 16 kb upstream of the CRLF2 translation start site, with multiple cases clustering near putative V(D)J Fig. 4. CRLF2 and JAK2 mutations in B-ALL. (A) The CRLF2 Phe232Cys muta- recombinase recognition signal sequences (Fig. 3B). Thus, tion that results from 711T > G is present in genomic and cDNA, as shown for CRLF2/IGH translocations appear to result from aberrant V(D) ALL-40. (B) UT7 cells and BaF3 cells that stably express CRLF2 and/or JAK2 alleles were grown in the absence of GM-CSF (UT7) or IL3 (BaF3). (Bottom Right) J recombination that can involve cryptic recognition signal se- TSLP 1 ng/mL was added at d 0 as indicated. Growth is the number of viable quence in the pseudoautosomal regions. cells relative to the number of cells initially seeded on d 0. Error bars represent 1SD.(C) Unmodified BaF3 cells (-) and BaF3 cells that stably express WT CRLF2, CRLF2 Phe232Cys Is a Gain-of-Function Mutation. Sequencing of CRLF2 Phe232Cys (FC), and/or IL7R were grown in the presence of IL3 except CRLF2 in 35 B-ALL specimens, including 14 overexpressing for the lane marked (-IL3). Protein lysates were separated by gel electro- cases, identified multiple single nucleotide variants. We assayed phoresis in the presence of reducing or nonreducing conditions and im- the function of the 4 nonsynonymous variants (711T > G, 746G munoblotted with an anti-CRLF2 antibody. (D) The 14 sequenced cases with > A, 789A > G, 984C > T) by retroviral expression in BaF3 cells. CRLF2 overexpression fall into 3 categories based on CRLF2 and JAK2. Se- > A quence traces for 3 different JAK2 alleles amplified from genomic DNA are Of these, only CRLF2 711T G (Phe232Cys; Fig. 4 ) conferred shown. cytokine independence in murine BaF3 and human UT7 mega-

karyoblastic leukemia cells (Fig. 4B). The 711T > G mutation was present in 3 of 14 overexpressing cases (21.4%) and was a Chr. 14 Chr. X/Y A somatic mutation. Sequencing of cDNA from all three cases IGHJ6 IGHJ5 IGHJ1 CRLF2 demonstrated expression of the mutant allele (Fig. 4A). The other three nonsynonymous variants were polymorphisms, as Clone IgH Deleon N nucleodes Distance from ATG they were present in germline specimens and were recovered ALL2 J6 15bp TATTAAGGCTAATA 8659bp from healthy donor peripheral blood lymphocytes (PBLs). ALL44 J6 14bp TGGGGAGACAA 11726bp ALL39 J6 65bp TCCCAAAGTGGGGGACCCTACG 15875bp The CRLF2 Phe232 residue is near the junction of the extrac- ALL71 J6 4bp CCCCCCCCAAAAGTCC 16052bp ellular and transmembrane domains. Mutations that introduce ALL73 J6 16bp CCCCCACAAATTAA 15914bp ALL77 J5 2bp CTGAGCGCGCCACTGCACTCCA 16186bp cysteine residues in this region of other receptor tyrosine kinases, GCCTGGCGAGGGCGACCCC such as the erythropoietin receptor, can activate signal trans- duction through intermolecular disulfide-bonded dimers (25, 26). 16186bp 5' of CRLF2 ATG B ALL39 fi 311bp To con rm that CRLF2 Phe232Cys promotes constitutive dime- MUTZ5 133bp rization, we performed immunoblots in BaF3 cells expressing WT CAGAGTGagactctgtctcAAAAAAAAG CRLF2 or CRLF2 Phe232Cys under both reducing and non- ALL77 7243 reducing conditions (Fig. 4C). Under nonreducing conditions, the 134bp ALL71 molecular weight of the CRLF2 Phe232Cys band, but not the 272bp ALL73 WT band, was doubled, consistent with constitutive dimerization through the cysteine residues. Fig. 3. CRLF2/IGH junctions. (A) Junctions were amplified on der(14) using a common IgHJ primer and spaced primers 5′ (centromeric) of CRLF2 coding JAK2 Mutations Are Highly Associated with CRLF2 Overexpression. sequence. N nucleotides are nontemplated insertions. (B) Breakpoints for 4 of the 6 cases (black arrows) clustered near a single putative V(D)J re- The absence of gain-of-function CRLF2 mutations in most cases combinase recognition signal sequence (RSS, blue). The expected sites of with CRLF2 overexpression raised the possibility that other cleavage by the V(D)J recombinase are in red. Cases MUTZ-5 and 7243 were factors within the same signaling cascade may harbor mutations. reported by Russell et al. (23). JAK mutations were recently reported in children with B-ALL

254 | www.pnas.org/cgi/doi/10.1073/pnas.0911726107 Yoda et al. Downloaded by guest on September 27, 2021 fi (11, 13, 14). Thus, we sequenced for previously identi ed mu- A P-JAK2 tations in JAK1 and JAK2 and identified JAK2 Arg683Gly (n = JAK2 4), Arg683Ser (n = 1), and Arg683Thr (n = 1) substitutions (Fig. P-STAT5 D 4 ) in 6 of 14 adult B-ALL cases that overexpress WT CRLF2. STAT5 Of note, expression of CRLF2 Phe232Cys and JAK2 mutant D P-AKT alleles was mutually exclusive (Fig. 4 ), suggesting that they AKT function within the same pathway. Mutant JAK proteins can transform growth factor–dependent P-ERK cells only when expressed in combination with a type I cytokine ERK receptor (11). The specific receptor that the JAK associates with, along with the particular JAK mutation, can affect the trans- TSLP stimulated X IL3 stimulated X formed phenotype (27). Thus, we asked whether overexpression Growth in TSLP X X X X of CRLF2 is essential for B-ALL associated with mutant JAK2. IL7R X X X X CRLF2 X X X X X X X X X We linked gene expression (GEO GSE11877) and JAK mutation CRLF2 F232C X X status from a cohort of 207 patients with high-risk pediatric B- JAK2 X A JAK2 R683G X X X ALL (Fig. S3 ) (3, 14). CRLF2 expression among the 207 pa- JAK2 R683S X X X tients was clearly bimodal, with overexpression in 29 cases A (14.0%; Fig. S3 ). Twenty of the 207 patients also had mutations B 20 3.5 in JAK1 (n = 3), JAK2 (n = 16), or JAK3 (n = 1) (14). Strikingly, 18 3 all 16 patients (100%) with JAK2 exon 16 mutations overex- 16 PIM1 BCL-xL pressed CRLF2, as did 2 of the 3 patients with JAK1 mutations. 14 2.5 *^ 12 2 10 * CRLF2 and JAK2 Contribute to Cytokine-Independent Growth. The * 8 * 1.5 finding that all cases of B-ALL with JAK2 mutations overex- * pressed CRLF2 raised the possibility that JAK2 associates 6 1 4 directly with CRLF2, either in the presence or the absence of 0.5 2 IL7R. The stoichiometry of a WT CRLF2/IL7R complex is 0 0 believed to be 1:1 (17). RT-PCR and gene expression profiling demonstrated that IL7R expression did not differ between the +TSLP +TSLP Starved CRLF2 overexpressing and nonoverexpressing cases (Fig. S3B). Starved Relative expression to starved Stimulated This argues that mutated JAK2 signals with CRLF2 in the absence Stimulated JAK2 R683S JAK2 R683S JAK2 R683G JAK2 R683G

CRLF2 CRLF2 F232C of IL7R. To test this possibility, we coexpressed WT and mutant CRLF2 CRLF2 F232C /IL7R CRLF2 /IL7R CRLF2 versions of CRLF2 and JAK2 in the presence or absence of IL7R B in BaF3 cells (Fig. 4 ). Although WT CRLF2 or mutant JAK2 C alone were insufﬁcient to confer IL3 independence in BaF3 cells, 1 the combination readily transformed the cells to IL3-independent growth in the absence of IL7R. Of note, the combination of WT 0.8 CRLF2 with WT JAK2 provided a growth advantage versus JAK2 CRLF2 alone (Fig. 4B). 0.6 +IL3 We next asked whether CRLF2 Phe232Cys renders cells hypersensitive to TSLP. Unlike in cells that express CRLF2/ 0.4 CRLF2 IL7R, the addition of TSLP had no effect on the growth of BaF3 F232C cells expressing CRLF2 Phe232Cys (Fig. 4B). However, TSLP 0.2 promoted the growth of cells that express CRLF2 Phe232Cys 0 B -3 -2 and IL7R (Fig. 4 ), suggesting that CRLF2 Phe232Cys retains 00.10.31 3 10 10 0.1 1 10 Relative survival to vehicle the ability to signal within a heterodimer. In contrast, TSLP did JAK inhibitor-1 (µM) k252a (µM) not promote the growth of cells that express CRLF2/mutant CRLF2/JAK2 R683S BCR/ABL JAK2 (Fig. S3C). CRLF2/JAK2 R683G +IL3 Both CRLF2 Phe232Cys and CRLF2/mutant JAK2 promoted CRLF2 F232C the phosphorylation of the JAK targets STAT5 and ERK (Fig. Fig. 5. Signal activation and dependence on JAK signaling. (A) BaF3 cells 5A). Phosphorylation of STAT5 by CRLF2 Phe232Cys was less expressing CRLF2, IL7R, and/or JAK2 proteins were starved of cytokine for 6 h, robust than by CRLF2/mutant JAK2 but comparable to CRLF2/ starved then stimulated with cytokine (TSLP 20 ng/mL or IL3 500 pg/mL) or IL7R/TSLP. CRLF2 Phe232Cys also promoted the up-regulation grown in the presence of TSLP 1 ng/mL. Protein lysates were subjected to of transcriptional targets downstream of JAKs (BCL-xL and immunoblotting with antibodies against total and phosphorylated (P-) JAK2, PIM1) to a comparable or greater extent than WT CRLF2/mutant STAT5, AKT, and ERK. (B) Quantitative RT-PCR of transcriptional targets in JAK2 (Fig. 5B). BaF3 cells that stably express CRLF2, IL7R, and/or JAK2 proteins grown in the absence of cytokine, starved of cytokine for 6 h, or stimulated with TSLP 20 ng/ Whereas cells expressing WT CRLF2/mutant JAK2 had con- mL after 6 h cytokine starvation. *P < 0.05 versus cells that express CRLF2/IL7R stitutively phosphorylated JAK2, cells expressing CRLF2 grown in the presence of TSLP. ^P < 0.05 versus CRLF2/JAK2R683G and CRLF2/ Phe232Cys and CRLF2/IL7R cells treated with TSLP had no JAK2R683S. (C) BaF3 cells that stably express BCR/ABL, CRLF2, and/or JAK2 detectable phospho-JAK2 (Fig. 5A). Yet, cells expressing either were grown in the absence of cytokines except where +IL3 indicates 500 pg/mL the CRLF2 Phe232Cys or CRLF2/mutant JAK2 were highly IL3. Data represents the ratio of viable cells exposed to JAK inhibitor-1 (Cal- sensitive to a small-molecule JAK inhibitor (Fig. 5C), suggesting biochem) or k252a at varying concentrations for 72 h, compared with the same that CRLF2 signaling involves a JAK other than JAK2. Cells ex- cell line exposed to vehicle. Error bars indicate 1 SD. pressing CRLF2 Phe232Cys had little or no AKT phosphorylation, in contrast with cells expressing CRLF2/mutant JAK2 (Fig. 5A). However, cells expressing CRLF2/IL7R grown in the presence of starvation clearly promoted AKT phosphorylation (Fig. 5A). TSLP also had minimal or no AKT phosphorylation despite the Thus, the absence of AKT phosphorylation in CRLF2 F232C- fact that stimulation of these cells with TSLP after a brief cytokine expressing cells may reﬂect a limitation in the sensitivity of im- MEDICAL SCIENCES

Yoda et al. PNAS | January 5, 2010 | vol. 107 | no. 1 | 255 Downloaded by guest on September 27, 2021 munoblotting. As expected, TSLP had no effect on STAT5, ERK, acute myelogenous leukemia, the presence of CRLF2 over- AKT, or JAK2 phosphorylation in the absence of IL7R (Fig. 5A). expression could guide the decision whether to pursue allogeneic stem cell transplantation or conventional chemotherapy for CRLF2 Transcriptional Signature in Adult and Pediatric B-ALL. To patients with B-ALL (35, 36). Of note, screening for CRLF2 characterize the dysregulated genes associated with CRLF2 over- identifies a larger cohort of high-risk cases than JAK2 sequenc- expression, we performed a supervised analysis of gene expression ing, as only some cases that overexpress CRLF2 harbor JAK2 profiles from the 22 B-ALL cases (8 CRLF2 overexpressing, 14 mutations but all cases with JAK2 mutations overexpress CRLF2. nonoverexpressing) that were validated by qRT-PCR (Fig. S1B). CRLF2/IGH translocations appeared to involve cryptic rec- CRLF2 overexpression defined a 130–probe set (105-gene) ognition signal sequences for the V(D)J recombinase centro- “CRLF2 adult signature” (Fig. S4A and Table S5). Up-regulated meric of CRLF2 (Fig. S3). Mullighan et al. recently reported that genes in the CRLF2 adult signature included CD10, PKC iota, and both P2RY8-CRLF2 (22) and IKZF1 (encodes the lymphoid the STAT5-induced negative regulator SOCS2, whereas genes differentiation factor IKAROS) (2) rearrangements can also showing decreased expression included the class I and class II involve aberrant V(D)J recombination. Translocations resulting HLAs. Higher SOCS2 expression among CRLF2 overexpressing from aberrant V(D)J recombination are common in more ma- cases was confirmed by qRT-PCR (123.1 fold higher than donor ture B-cell neoplasms, leading to the suggestion that V(D)J re- PBLs vs. 53.5 fold for CRLF2 nonoverexpressing cases; P < 0.05). combination during receptor revision is more error-prone than V Interestingly, several agents with activity against PKC family kina- (D)J recombination earlier in B-cell ontogeny (37). The dis- ses had selective toxicity in BaF3-CRLF2 Phe232Cys cells, com- covery of CRLF2 rearrangements, as well as less frequent IGH pared with BaF3 cells expressing WT CRLF2 (Fig. 5C and Fig. S5). translocations in B-ALL (38–40), challenges this possibility. We applied a similar supervised approach to identify differ- The association of transcriptional signatures between CRLF2 entially expressed genes based on CRLF2 expression in the pe- overexpressing B-ALL and BCR/ABL–expressing B-ALL suggests diatric B-ALL GEO datasets GSE11877 and GSE12995, and that activation of JAK/STAT signaling by CRLF2 Phe232Cys, compared these with the CRLF2 adult signature of 105 genes CRLF2 with mutant JAK2, or CRLF2 with another partner can using Gene Set Enrichment Analysis (GSEA). Both pediatric mimic BCR/ABL signaling. In agreement, IKZF1 mutations are signatures showed significant enrichment of the adult set (P < common in BCR/ABL–expressing B-ALL and lymphoid blast 0.001; false discovery rate (FDR) of 0; Fig. S4B). crisis of chronic myelogenous leukemia (2, 3), and were present in To identify pathways up-regulated in CRLF2-overexpressing 14 of the 16 cases of B-ALL with JAK2 mutations reported by B-ALL, we applied GSEA to the CRLF2 differential expression Mullighan et al. (14). Importantly, the use of imatinib has drasti- signatures (from both GSE11877 and GSE12995) using gene sets cally improved outcomes for patients with BCR/ABL–expressing from the Kyoto Encyclopedia of Genes and Genomes (KEGG) B-ALL (41). Similar improvements may be possible by targeting database (28). Four pathways, all involved in cytokine receptor CRLF2/JAK signaling, either with small-molecule kinase in- signaling, were identified as significant (P < 0.01 and FDR < hibitors or monoclonal antibodies against CRLF2. 0.10) in both datasets: “cytokine–cytokine receptor interaction,” The finding that all 22 B-ALLs with JAK2 mutations (6 from our “JAK-STAT signaling,”“neuroactive ligand receptor inter- cohort, 16 from GEO GSE11877) overexpressed CRLF2 suggests action,” and “ECM receptor interaction.” The enrichment of the that CRLF2 is the essential, or at least overwhelmingly predom- JAK-STAT signaling pathway (Fig. S4C) further suggests that inate, scaffold for mutant JAK2 signaling in B-ALL. Interestingly, CRLF2 overexpression in B-ALL supports aberrant JAK-STAT the JAK2 V617F mutation associated with myeloid proliferations activation. In agreement with the BaF3 data (Fig. 5B), BCL-xL has not been observed in B-ALL. This raises the question whether and PIM1 were among the up-regulated genes enriched in the JAK2 V617F is capable of signaling in combination with CRLF2. JAK-STAT signaling pathway. Alternatively, JAK2 V617F could confer a different substrate Finally, we hypothesized that CRLF2 overexpression promotes selectivity from the JAK mutants observed in B-ALL (14). a similar gene expression pattern to BCR/ABL. To address this, a Unlike most proto-oncogenes, which are activated by a single BCR/ABL signature was obtained from the Oncomine Concepts genetic event that results in either overexpression or somatic Map (http://www.oncomine.org), which identified the top up- mutation, CRLF2 appears to require 2 genetic events. First, regulated and down-regulated genes (29). Using GSEA, the pe- CRLF2 is overexpressed through intrachromosomal deletion, diatric CRLF2 overexpression signatures from both datasets translocation to IGH, or possibly a third mechanism (22, 23). showed striking enrichment of the BCR/ABL-associated genes (P Second, somatic mutation activates either CRLF2, JAK1, JAK2, < 0.001, FDR = 0; Fig. S4D). Together, these findings establish or possibly other factors in JAK/STAT signaling (14, 22). Of common expression patterns between BCR/ABL–positive and note, the combined overexpression of WT CRLF2 and JAK2 in CRLF2-overexpressing B-ALL in both children and adults. BaF3 cells conferred a growth advantage in the absence of cytokines (Fig. 4B), suggesting that the first of the 2 events is likely Discussion to be the CRLF2 locus rearrangement, which provides some Using a selection-based cDNA library screen, we identified the advantage to a preleukemic clone. In addition, overexpression of cytokine receptor subunit CRLF2 as a factor in poor-risk B-ALL. WT CRLF2 in mouse fetal liver cells supports the proliferation The screen, which uses tumor-derived mRNA to assay for of precursor B cells and STAT5 phosphorylation (23). transcripts that can substitute for IL3 signaling, is broadly ap- Similar to CRLF2 F232C, mutations that promote the dimeriza- plicable to other tumor types, as many gain-of-function alter- tion of other cytokine receptors (e.g., RET, erythropoietin receptor) ations in epithelial and mesenchymal malignancies (e.g., mutant occur in both hematologic and nonhematologic tumors (25, 26). EGFR, KIT, ERBB2, ALK, and EWS) confer IL3 independence However, these other receptors canonically signal in response to li- in BaF3 cells (29–34). The alterations identified by the screen are gand as homodimers. In contrast, CRLF2 responds to TSLP as a functionally relevant and therefore attractive therapeutic targets. heterodimer with IL7R and may signal independently of TSLP as a Our data suggests that routine screening of B-ALL for CRLF2 homodimer when harboring the F232C mutation or as a monomer expression, either by IHC or flow cytometry, offers prognostic through mutant JAK2 (Fig. S6). value. CRLF2 overexpression was associated with high-risk Many aspects of CRLF2 signaling remain unclear. First, what disease in children and a dismal prognosis in adults. Marked kinase or kinases associate with CRLF2, either in its WT or enrichment of the adult CRLF2 overexpression signature among mutant forms? In a previous study, TSLP signaling was inhibited pediatric B-ALL suggests that CRLF2 overexpression drives a by a dominant-negative TEC kinase, but not by kinase-deficient similar disease in children and adults. Like FLT3 mutations in JAK1 or JAK2. Thus, activation of STAT5 and other targets

256 | www.pnas.org/cgi/doi/10.1073/pnas.0911726107 Yoda et al. Downloaded by guest on September 27, 2021 induced by TSLP or mutation of CRLF2 may not involve JAKs. Cell Lines. BaF3 cells were maintained in RPMI medium with 10% FCS and Conversely, the sensitivity of cells expressing CRLF2 Phe232Cys 500 pg/mL IL3. UT7 cells were maintained in RPMI medium with 10% FCS and to an enzymatic JAK inhibitor (Fig. 5C) argues that JAKs (or a 500 pg/mL GM-CSF (Bayer). kinase sensitive to this inhibitor) are involved. A final possibility is that JAK activation in this setting does not involve canonical Survival and Inhibitor Assays. Survival of BaF3 and UT7 cells transduced with JAK autophosphorylation. CRLF2 and/or JAK2 constructs was determined by growing all cell lines in the presence of growth factors until d 0. On d 0, cells were seeded at 104 to 105/mL The second question is whether mutant JAK2 signaling in media in the presence or absence of growth factors (IL3 500 pg/mL or TSLP 1 involves CRLF2 alone or a dimeric complex that includes ng/mL) or JAK inhibitor-1 (Calbiochem) and the number of cells was then CRLF2. The low levels of IL7R expression in all cases of B-ALL counted daily. All assays were performed in triplicate or more. (Fig. S3B) suggest that overexpressed CRLF2 signals in the absence of IL7R, and thus independently of TSLP. Finally, what is Immunoblotting. AntibodieswereusedagainstCRLF2(R&DSystems),STAT5(Santa the role of TSLP in signaling by overexpressed CRLF2 and is it Cruz Biotechnology), and phospho-STAT5, JAK2, phospho-JAK2, AKT, phospho- dependent on which mutation is present? Cells that expressed AKT, ERK, and phospho-ERK (Cell Signaling). For nonreducing immunoblot, we CRLF2 Phe232Cys and IL7R proliferated more rapidly in re- used protein dissolved in SDS sample buffer without 2-mercaptoethanol. sponse to TSLP (Fig. 4B), but those that expressed CRLF2/ mutant JAK2 did not (Fig. S3C). Experiments are under way to FISH. FISH was performed as described (6). For each sample, 100 cells were determine whether this difference applies to B-ALL in vivo. scored. For all cases with rearrangements, the frequency of cells with the Levels of TSLP within the bone marrow microenvironment have rearrangement was ≥80% of the fraction of leukemia involvement within not been described, but could be adequate to stimulate B-ALL that bone marrow specimen, as measured by histologic examination. proliferation. Importantly, TSLP liberated from keratinocytes in a Additional methods are outlined in SI Methods. murine model of atopic dermatitis stimulated polyclonal B-cell ACKNOWLEDGMENTS. The authors thank Dr. Bob Distel and Yanan Kuang lymphoproliferation (42). Although no association has been from the Dana-Farber Cancer Institute Translational Research Laboratory for identified between B-ALL and atopy, it will be important to re- assistance with HPLC; Marc Dibona for assistance with FISH; Drs. Shai Izraeli and examine this issue with a focus specifically on CRLF2- A. Thomas Look for helpful discussion; and Dr. Roberto Bellucci, Gorka Murga, overexpressing B-ALL. Ilene Galinsky, RNP, Dr. Ann Mullally, and Dr. Takaomi Sanda for assistance with obtaining primary samples and cell lines. This work was supported by a Burroughs Wellcome Fund Career Award in the Biomedical Sciences (D.M.W.), Materials and Methods the Compagnia di San Paolo, Turin, the Dunkin Donuts Rising Star Program, the The studies were approved by the DFCI institutional review board. Kristen Amico Fund, and the Claudia Adams Barr Program in Cancer Research.

1. Mullighan CG, et al. (2007) Genome-wide analysis of genetic alterations in acute 23. Russell LJ, et al. (2009) Deregulated expression of cytokine receptor gene, CRLF2, is lymphoblastic leukaemia. Nature 446:758–764. involved in lymphoid transformation in B-cell precursor acute lymphoblastic 2. Mullighan CG, et al. (2008) BCR-ABL1 lymphoblastic leukaemia is characterized by the leukemia. Blood 114:2688–2698. deletion of Ikaros. Nature 453:110–114. 24. Küppers R, Dalla-Favera R (2001) Mechanisms of chromosomal translocations in B cell 3. Mullighan CG, et al.; Children’s Oncology Group (2009) Deletion of IKZF1 and lymphomas. Oncogene 20:5580–5594. prognosis in acute lymphoblastic leukemia. N Engl J Med 360:470–480. 25. Lu X, Gross AW, Lodish HF (2006) Active conformation of the erythropoietin receptor: 4. Mullighan CG, et al. (2008) Genomic analysis of the clonal origins of relapsed acute random and cysteine-scanning mutagenesis of the extracellular juxtamembrane and lymphoblastic leukemia. Science 322:1377–1380. transmembrane domains. J Biol Chem 281:7002–7011. 5. Chiaretti S, et al. (2005) Gene expression profiles of B-lineage adult acute lymphocytic 26. Kjaer S, Kurokawa K, Perrinjaquet M, Abrescia C, Ibáñez CF (2006) Self-association of leukemia reveal genetic patterns that identify lineage derivation and distinct the transmembrane domain of RET underlies oncogenic activation by MEN2A mechanisms of transformation. Clin Cancer Res 11:7209–7219. mutations. Oncogene 25:7086–7095. 6. Fine BM, et al. (2004) Gene expression patterns associated with recurrent chromosomal 27. Mullighan CG (2008) JAK2—a new player in acute lymphoblastic leukaemia. Lancet translocations in acute lymphoblastic leukemia. Blood 103:1043–1049. 372:1448–1450. 7. Juric D, et al. (2007) Differential gene expression patterns and interaction networks in BCR- 28. Kanehisa M (2002) The KEGG database. Novartis Found Symp 247:91–101. ABL-positive and -negative adult acute lymphoblastic leukemias. J Clin Oncol 25:1341–1349. 29. Ross ME, et al. (2003) Classification of pediatric acute lymphoblastic leukemia by gene 8. Moos PJ, et al. (2002) Identification of gene expression profiles that segregate expression profiling. Blood 102:2951–2959. patients with childhood leukemia. Clin Cancer Res 8:3118–3130. 30. Shimamura T, et al. (2006) Non-small-cell lung cancer and Ba/F3 transformed cells 9. Faderl S, Kantarjian HM, Talpaz M, Estrov Z (1998) Clinical significance of cytogenetic harboring the ERBB2 G776insV_G/C mutation are sensitive to the dual-specificepidermal abnormalities in adult acute lymphoblastic leukemia. Blood 91:3995–4019. growth factor receptor and ERBB2 inhibitor HKI-272. Cancer Res 66:6487–6491. 10. Malinge S, et al. (2007) Novel activating JAK2 mutation in a patient with Down 31. George RE, et al. (2008) Activating mutations in ALK provide a therapeutic target in syndrome and B-cell precursor acute lymphoblastic leukemia. Blood 109:2202–2204. neuroblastoma. Nature 455:975–978. 11. Bercovich D, et al. (2008) Mutations of JAK2 in acute lymphoblastic leukaemias 32. Slupianek A, et al. (2002) Fusion tyrosine kinases induce drug resistance by associated with Down’s syndrome. Lancet 372:1484–1492. stimulation of homology-dependent recombination repair, prolongation of G(2)/M 12. Gaikwad A, et al. (2009) Prevalence and clinical correlates of JAK2 mutations in Down phase, and protection from apoptosis. Mol Cell Biol 22:4189–4201. syndrome acute lymphoblastic leukaemia. Br J Haematol 144:930–932. 33. Malinge S, et al. (2008) Activating mutations in human acute megakaryoblastic 13. Kearney L, et al. (2009) Specific JAK2 mutation (JAK2R683) and multiple gene leukemia. Blood 112:4220–4226. deletions in Down syndrome acute lymphoblastic leukemia. Blood 113:646–648. 34. Jiang J, et al. (2005) Epidermal growth factor-independent transformation of Ba/F3 14. Mullighan CG, et al. (2009) JAK mutations in high-risk childhood acute lymphoblastic cells with cancer-derived epidermal growth factor receptor mutants induces gefitinib- leukemia. Proc Natl Acad Sci USA 106:9414–9418. sensitive cell cycle progression. Cancer Res 65:8968–8974. 15. Lin TS, Mahajan S, Frank DA (2000) STAT signaling in the pathogenesis and treatment 35. Gaidzik V, Döhner K (2008) Prognostic implications of gene mutations in acute of leukemias. Oncogene 19:2496–2504. myeloid leukemia with normal cytogenetics. Semin Oncol 35:346–355. 16. Lu X, et al. (2005) Expression of a homodimeric type I cytokine receptor is required for 36. Zuckerman T, Rowe JM (2009) Hematopoietic stem cell transplantation for adults JAK2V617F-mediated transformation. Proc Natl Acad Sci USA 102:18962–18967. with acute lymphoblastic leukemia. Curr Opin Hematol 16:453–459. 17. Pandey A, et al. (2000) Cloning of a receptor subunit required for signaling by thymic 37. Wang JH, et al. (2009) Mechanisms promoting translocations in editing and switching stromal lymphopoietin. Nat Immunol 1:59–64. peripheral B cells. Nature 460:231–236. 18. Liu YJ (2009) TSLP in epithelial cell and dendritic cell cross talk. Adv Immunol 101:1–25. 38. Akasaka T, et al. (2007) Five members of the CEBP transcription factor family are 19. Reche PA, et al. (2001) Human thymic stromal lymphopoietin preferentially stimulates targeted by recurrent IGH translocations in B-cell precursor acute lymphoblastic myeloid cells. J Immunol 167:336–343. leukemia (BCP-ALL). Blood 109:3451–3461. 20. Levin SD, et al. (1999) Thymic stromal lymphopoietin: a cytokine that promotes the 39. Russell LJ, et al. (2009) A novel translocation, t(14;19)(q32;p13), involving IGH@ and development of IgM+ B cells in vitro and signals via a novel mechanism. J Immunol the cytokine receptor for erythropoietin. Leukemia 23:614–617. 162:677–683. 40. Russell LJ, et al. (2008) t(6;14)(p22;q32): a new recurrent IGH@ translocation involving 21. Brown VI, et al. (2007) Thymic stromal-derived lymphopoietin induces proliferation of ID4 in B-cell precursor acute lymphoblastic leukemia (BCP-ALL). Blood 111:387–391. pre-B leukemia and antagonizes mTOR inhibitors, suggesting a role for interleukin- 41. Ottmann OG, Pfeifer H (2009) First-line treatment of Philadelphia chromosome- 7Ralpha signaling. Cancer Res 67:9963–9970. positive acute lymphoblastic leukaemia in adults. Curr Opin Oncol 21 Suppl 1S43–S46. 22. Mullighan CG, et al. (2009) Rearrangement of CRLF2 in B-progenitor- and Down 42. Demehri S, et al. (2008) Notch-deficient skin induces a lethal systemic B-lymphoproliferative syndrome-associated acute lymphoblastic leukemia. Nat Genet 41:1243–1246. disorder by secreting TSLP, a sentinel for epidermal integrity. PLoS Biol 6:e123. MEDICAL SCIENCES

Yoda et al. PNAS | January 5, 2010 | vol. 107 | no. 1 | 257 Downloaded by guest on September 27, 2021