P2RY8-CRLF2 Fusion but Not for CRLF2 Over-Expression in Children with Intermediate Risk B-Cell Precursor Acute Lymphoblastic Leukemia

ORIGINAL ARTICLE Poor prognosis for P2RY8-CRLF2 fusion but not for CRLF2 over-expression in children with intermediate risk B-cell precursor acute lymphoblastic leukemia

C Palmi1,9, E Vendramini2,9, D Silvestri3,4, G Longinotti1, D Frison2, G Cario5, C Shochat6,7, M Stanulla5, V Rossi1, AM Di Meglio2, T Villa1, E Giarin2, G Fazio1, A Leszl2, M Schrappe5, G Basso2, A Biondi4, S Izraeli6, V Conter4,8, MG Valsecchi3,9, G Cazzaniga1,9 and G te Kronnie2,9

Pediatric B-cell precursor acute lymphoblastic leukemia (BCP-ALL) has achieved an 80% cure rate as a result of a risk-adapted therapy largely based on minimal residual disease (MRD) monitoring. However, relapse is still the most frequent adverse event, occurring mainly in the patients with intermediate MRD levels (intermediate risk, IR), emphasizing the need for new prognostic markers. We analyzed the prognostic impact of cytokine receptor-like factor 2 (CRLF2) over-expression and P2RY8-CRLF2 fusion in 464 BCP-ALL patients (not affected by Down syndrome and BCR-ABL negative) enrolled in the AIEOP-BFM ALL2000 study in Italy. In 22/464 (4.7%) samples, RQ-PCR showed CRLF2 over-expression (X20 times higher than the overall median). P2RY8-CRLF2 fusion was detected in 22/365 (6%) cases, with 10/22 cases also showing CRLF2 over-expression. P2RY8-CRLF2 fusion was the most relevant prognostic factor independent of CRLF2 over-expression with a threefold increase in risk of relapse. Significantly, the cumulative incidence of relapse of the P2RY8-CRLF2 þ patients in the IR group was high (61.1%±12.9 vs 17.6%±2.6, Po0.0001), similar to high-risk patients in AIEOP-BFM ALL2000 study. These results were confirmed in a cohort of patients treated in Germany. In conclusion, P2RY8-CRLF2 identifies a subset of BCP-ALL patients currently stratified as IR that could be considered for treatment intensification.

Leukemia (2012) 26, 2245–2253; doi:10.1038/leu.2012.101 Keywords: CRLF2; pediatric BCP-ALL; prognosis

INTRODUCTION (CRLF2)8–13 genes. The latter abnormality includes small The cure rate of new diagnosed pediatric B-cell precursor acute deletions within the pseudoautosomal region (PAR1) of the sex lymphoblastic leukemia (BCP-ALL) is higher than 80%. However, chromosomes as well as the translocation of this region to the the probability of survival of patients who relapse after treatment IGH@ locus on chromosome 14. Several studies have provided is only 30%. In the AIEOP-BFM ALL2000 study, risk group evidence for the mechanistic basis of the over-expression of CRLF2 8,10 stratification was largely based on minimal residual disease as a consequence of these chromosomal abnormalities. PAR1 (MRD) monitoring as a measure of early response to therapy. deletion juxtaposes the first non-coding exon of P2RY8 to the Approximately, 30% of patients were at standard risk (SR), 55% at first exon of CRLF2, leading to CRLF2 expression driven by the intermediate risk (IR) and 15% at high risk (HR) of relapse. In spite promoter of P2RY8. The translocation of PAR1 region to IGH@ of the risk-adapted therapy, the majority of relapses occur in the locus, however, brings CRLF2 under the control of IGH@ enhancer large heterogeneous IR group.1 Therefore, the identification of elements. Elevated CRLF2 expression was found to be associated disease features with prognostic values within current risk groups with IKZF1 deletions and activating JAK2 or CRLF2 point 10–12,14 remains a formidable challenge in childhood ALL. mutations. Chromosomal translocations that were identified as key factors CRLF2 rearrangements were shown to be correlated with poor 12,14 in the pathogenesis of ALL and provided essential prognostic outcome in BCP-ALL patients, but it is still under debate hallmarks have been incorporated in patient risk stratification. whether this prognostic value is associated to CRLF2 over- There is a need to identify which of the recently discovered expression or to specific CRLF2 aberrancies. Moreover, CRLF2 genetic alterations have the potential to improve patient aberrations were not proven to bear significant prognostic value 13 stratification2 as well as for the development of targeted in children with ALL treated on UK protocols. therapeutic approaches.3–5 Here, we present data on the incidence and prognostic impact In a subset of BCP-ALL patients without known chromosomal of CRLF2 over-expression, specifically P2RY8-CRLF2 fusion, at aberrations, two genomic abnormalities have been reported diagnosis in 464 Italian BCP-ALL children treated with the protocol that involve Ikaros (IKZF1)6,7 and cytokine receptor-like factor 2 of the Associazione Italiana Ematologia Oncologia Pediatrica

1Centro Ricerca Tettamanti, Clinica Pediatrica, Universita` di Milano Bicocca, Ospedale San Gerardo, Monza, Italy; 2Laboratory of Oncohematology, Department of Pediatrics, Universita` di Padova, Padova, Italy; 3Centro di Biostatistica per l’Epidemiologia Clinica, Universita` di Milano Bicocca, Monza, Italy; 4Clinica Pediatrica, Universita` di Milano Bicocca, Ospedale San Gerardo, Monza, Italy; 5Department of Pediatrics, University Hospital Schleswig-Holstein, Campus Kiel, Kiel, Germany; 6Sheba Medical Center, Tel Hashomer, Ramat Gan and Tel Aviv University, Tel Aviv, Israel; 7Migal-Galilee Bio-Technology Center, Kiryat-Shmona, Israel. and 8Pediatria, Ospedali Riuniti, Bergamo, Italy. Correspondence: Dr G Cazzaniga, Centro Ricerca Tettamanti, Clinica Pediatrica, Universita` di Milano Bicocca, Ospedale San Gerardo, Via Pergolesi, 33, 20900 Monza, Italy. E-mail: [email protected] 9These authors contributed equally to this work. Received 14 October 2011; revised 21 March 2012; accepted 30 March 2012; accepted article preview online 9 April 2012; advance online publication, 11 May 2012 Poor outcome of P2RY8-CRLF2 in childhood ALL C Palmi et al 2246 (AIEOP) and Berlin-Frankfurt-Munster (BFM) group (‘AIEOP-BFM P2RY8-CRLF2 rearrangement was tested in 365 patients at diagnosis ALL2000 protocol’) and the potential impact of P2RY8-CRLF2 and in 26 paired diagnosis and relapse samples for which leftover RNA was fusion within MRD-based subgroups. available. IGH@-CRLF2 translocation was screened in 8 out of 12 patients We also present data on CRLF2 over-expression and P2RY8- who were positive for CRLF2 over-expression (X20 times higher than CRLF2 fusion at BCP-ALL relapse and speculate on P2RY8-CRLF2 overall median) but negative for P2RY8-CRLF2 fusion at diagnosis. DNA was available from 34 patients positive for CRLF2 over-expression or P2RY8- as a secondary event in disease progression. CRLF2 fusion at diagnosis, and the following were analyzed: CRLF2 mutations (in 25 out of 34 patients), JAK2 mutations (in 32 out of 34 patients) and IKZF1 deletions (in 33 out of 34 patients). CRLF2 and JAK2 PATIENTS AND METHODS mutations were also analyzed in 6 and 32 paired diagnosis and relapse Patients samples, respectively. BCP-ALL patients consecutively enrolled in the AIEOP-BFM ALL2000 study The clinical characteristics of the analyzed patients compared with those in AIEOP Centers from February 2003 to July 2005, not affected by Down not analyzed (AIEOP-BFM ALL2000 study patients diagnosed between syndrome (DS) nor Philadelphia chromosome positive (Ph þ ), were September 2000 and July 2006) are shown in Supplementary Table 1, and included in the study cohort. CRLF2 expression was analyzed in 464 their event-free survival (EFS) curve is shown in Supplementary Figure 1a. patients at diagnosis and in 33 paired diagnosis and relapse samples, while Details of the study cohort are shown in the Supplementary Materials.

a 1000

100

1 expression

0.1 CRLF2 0.01

1 100 200 300 400 Rank (464 pts) CRLF2 expression P2RY8-CRLF2

bc 1000 100

10 100 1 expression expression 10 0.1 CRLF2 CRLF2

1 5 10 15 20 1510 Rank (22 pts) Rank(12 pts)

CRLF2 expression IGH@-CRLF2 P2RY8-CRLF2 CRLF2 expression P2RY8-CRLF2 JAK2 mutations JAK2 mutations CRLF2 mutations IKZF1 deletions CRLF2 mutations IKZF1 deletions

Total P2RY8-CRLF2 IGH@-CRLF2 CRLF2 mut JAK2 mut IKZF1 del

hiCRLF2 22/464(4.7%) 10/20 1/8 2/22 4/22 5/22

P2RY8-CRLF2 pos 22/365 (6.0%) - - 2/14 4/21 4/21

hiCRLF2 & P2RY8-CRLF2 pos 10/365 (2.7%) - - 2/10 3/10 3/10 hiCRLF2 & P2RY8-CRLF2 neg 10/365 (2.7%) - 1/8 0/10 1/10 1/10 loCRLF2 & P2RY8-CRLF2 pos 12/365 (3.3%) --0/4 1/10 2/11 mut, mutations; del, deletions. Figure 1. CRLF2 expression and genomic alterations. (a) CRLF2 expression in 464 BCP-ALL patients. For each case results are reported as fold changes on the median expression value of all patients in the cohort. Expression ranged from 0.006- to 810-fold change. (b) Additional genomic aberrations in hiCRLF2 patients. In bright blue are cases not tested for P2RY8-CRLF2 fusion. (c) Additional genomic aberrations in loCRLF2 P2RY8-CRLF2-positive patients. (d) Details on the combination of CRLF2 expression, P2RY8-CRLF2 fusion and other genomic aberrations.

Leukemia (2012) 2245 – 2253 & 2012 Macmillan Publishers Limited Poor outcome of P2RY8-CRLF2 in childhood ALL C Palmi et al 2247 In addition, 194 SR and 286 IR consecutive patients (non-DS and Ph À ) Protocol stratification enrolled in the AIEOP-BFM ALL2000 study and treated in German Centers Patient risk groups were defined as follows. The HR group included 12 (BFM-G) were analyzed by reverse transcriptase PCR (RT-PCR) for P2RY8- patients with any of the following criteria: t(4;11) or MLL/AF4, prednisone CRLF2 rearrangement, as a validation cohort. poor response (X1000 blasts/ml on day 8 peripheral blood after 7 days of Informed consent to participate in the study was obtained for all prednisone and one dose of intrathecal methotrexate on day 1), inability to patients by parents or legal guardians. Risk group definitions and final achieve clinical remission after induction phase IA and high burden stratification, treatment outlines, details and differences between AIEOP (X10 À 3) of PCR-MRD at day 78. The SR group included patients who and BFM were previously reported15 and briefly summarized below. lacked high-risk criteria and tested negative to PCR-MRD performed by

Table 1. Clinical features of AIEOP study cohort patients positive or negative for CRLF2 over-expression and P2RY8-CRLF2 fusion

Characteristics P-value hiCRLF2 P-value P2RY8-CRLF2

No Yes No Yes

N % N % N % N %

All patients 442 95.30 22 4.70 343 93.97 22 6.03

Gender 0.47 0.19 Female 216 48.87 9 40.91 167 48.69 14 63.64 Male 226 51.13 13 59.09 176 51.31 8 36.36

Age 0.18 0.52 1–5 Yrs 288 65.16 12 54.55 224 65.31 12 54.55 6–9 Yrs 87 19.68 8 36.36 69 20.12 6 27.27 X10 Yrs 67 15.16 2 9.09 50 14.58 4 18.18

WBC( Â 1000/ml) 0.45 0.66 o20 309 69.91 18 81.82 235 68.51 14 63.64 20–100 101 22.85 4 18.18 82 23.91 7 31.82 X100 32 7.24 0 0.00 26 7.58 1 4.55

Translocations t(4;11) — — Pos 5 1.13 0 0.00 5 1.46 0 0.00 Neg 435 98.42 22 100.00 336 97.96 22 100.00 Unknown 2 0.45 0 0.00 2 0.58 0 0.00 t(12;21) 0.01 0.59 Pos 92 20.81 0 0.00 71 20.70 3 13.64 Neg 328 74.21 22 100.00 254 74.05 19 86.36 Unknown 22 4.98 0 0.00 18 5.25 0 0.00

Prednisone response 1.00 1.00 Good 413 93.44 21 95.45 317 92.42 21 95.45 Poor 27 6.11 1 4.55 25 7.29 1 4.55 Unknown 2 0.45 0 0.00 1 0.29 0 0.00

MRD 0.09 0.69 SR 134 30.32 3 13.64 108 31.49 6 27.27 IR 205 46.38 13 59.09 154 44.90 12 54.55 HR 6 1.36 1 4.55 3 0.87 0 0.00 Unknown 97 21.95 5 22.73 78 22.74 4 18.18

Final protocol strata 0.28 0.94 SR 126 28.51 3 13.64 100 29.15 6 27.27 IR 283 64.03 17 77.27 214 62.39 15 68.18 HR 33 7.47 2 9.09 29 8.45 1 4.55 DNA index 1.00 0.006 1.16–1.6 92 20.81 5 22.73 80 23.32 0 0.00 Other 319 72.17 17 77.27 240 69.97 21 95.45 Unknown 31 7.01 0 0.00 23 6.71 1 4.55

hiCRLF2 — o0.001 No — — — — 333 97.08 12 54.55 Yes — — — — 10 2.92 10 45.45 Unknown — — — — 0 0.00 0 0.00

P2RY8-CRLF2 o0.001 — — — — — No 333 75.34 10 45.45 — — — — Yes 12 2.71 10 45.45 — — — — Unknown 97 21.95 2 9.09 — — — — Abbreviations: HR, high risk; IR, intermediate risk; MRD, minimal residual disease; SR, standard risk; WBC, white blood cell count.

& 2012 Macmillan Publishers Limited Leukemia (2012) 2245 – 2253 Poor outcome of P2RY8-CRLF2 in childhood ALL C Palmi et al 2248 using two sensitive markers (X1 Â 10 À 4) at both day 33 and day 78. The IR 30 s, annealing at 63 1C for 30 s, extension at 72 1C for 1 min and final group included the remaining patients, and those not evaluated by PCR- extension at 72 1C for 10 min (PCR reagents from Roche). MRD. Details of PCR-MRD are reported in the Supplementary Materials. RT-PCR to detect the P2RY8-CRLF2 fusion in the BFM-G cohort was performed as previously described.12 IGH@-CRLF2 translocation was inquired by fluorescence in situ hybridiza- Quantitative expression of CRLF2 tion (FISH) on interphase nuclei using the Vysis LSI IGH Dual Color Break- CRLF2 transcript levels on diagnostic and relapse samples were analyzed Apart Rearrangement Probe (Abbott Molecular, Abbott Park, IL, USA) and using TaqMan Gene Expression Assay Hs00913509_s1 (Applied Biosystems, the CEP (chromosomes enumeration probes) DNA FISH Probe to detect Foster City, CA, USA); the housekeeping GUS gene transcript was tested as chromosome X (Abbott Molecular). The nuclei were counterstained with an internal control by using Universal Probe Library System (Roche DAPI (40,6-diamino-2-phenylindol). Results were recorded using a fluores- Diagnostics, Basel, Switzerland), following the manufacturer’s instruc- cence Leica DMRB microscope fitted with a Â 100/1.30 oil objective, CCD tions.16 Optimal primers and probe for GUS amplification were selected camera and digital imaging software from Metasystem (ISIS, FISH imaging using the Roche ProbeFinder software (https://www.roche-appliedscience. system). com/sis/rtpcr/upl). Each cDNA sample (20 ng RNA equivalent) was tested in To identify CRLF2 mutations, intronic primers of human CRLF2 sequence duplicate (Ct range between replicates 1.5). The amplification reaction (accession number NM_022148.2) were used to amplify exons 1–6 of o 11 was performed on the 7900HT FAST Real Time PCR System instrument the gene with PCR. Fragments were analyzed by denaturing (Applied Biosystems) with the following protocol: initial step at 95 1C for high-performance liquid chromatography (WAVE, Transgenomic, Omaha, 10 min, then 50 cycles at 95 1C for 15 s and at 60 1C for 1 min. NE, USA). Fragments with abnormal chromatography patterns were Relative gene expression (indicated as fold change) was quantified by sequenced. the 2 À DDCt method.17 The DDCts were calculated by subtracting the median of the DCt of all the (n ¼ 464) tested patients at diagnosis to the DCt of each sample. Other genetic aberrations High resolution melting (HRM) analysis was performed to identify JAK2 mutations in exon 16 using High Resolution Melting Master (Roche CRLF2 aberrations Diagnostics), following the manufacturer’s instructions. In detail 20 ng of The presence of the fusion transcript P2RY8-CRLF2 was investigated by DNA was amplified in a final volume of 20 ml with 10 mlofResolution RT-PCR. The patients analyzed at diagnosis were representative of the Melting master mix (Roche Diagnostics), 2.8 ml of magnesium chloride, entire study cohort (Supplementary Table 2 and Supplementary Figure 1b). 3.5 mM (Roche Diagnostics) and 4 pmol of each primer (forward: 50-TTTGGG In order to verify that the RT reaction worked properly, the house-keeping GGCTTGAACATACT-30 and reverse: 50-TTCAAGGAAAATTAACAACATGC-30). gene Abelson tyrosine kinase (ABL) expression was tested in parallel for each The reaction was performed on the LightCycler 480 instrument (Roche cDNA sample. RT-PCR for P2RY8-CRLF2 was performed using primers Diagnostics) with the following protocol: initial denaturation at 95 1C for designed in the first exon of P2RY8 (50-GGACAGATGGAACTGGAAGG-30) 10 min, 45 cycles of amplification at 95 1C for 10 s, 53 1C for 10 s and 72 1C and the third exon of CRLF2 (50-GTCCCATTCCTGATGGAGAA-30), giving rise for 15 s, then for HRM analysis samples were denatured at 95 1C for 1 min, to a PCR fragment of B511 bp. Amplification conditions were: initial cooled at 40 1C for 1 min and heated from 60 to 95 1C with a ramping rate denaturation at 94 1C for 2 min; 40 PCR cycles: denaturation at 94 1C for of 0.02 1C/s. DNA from a pool of peripheral blood buffy coat of healthy

a 1.0 b 1.0 N. pts N. rel. 5 yrs Cum. Incidence 0.9 loCRLF2 442 71 15.2%(1.7) 0.8 0.8 hiCRLF2 22 8 37.1%(10.5) 0.7 0.6 0.6 0.5 EFS 0.4 0.4 0.3

N. pts N. events 5 yrs EFS Cum. Incidence loCRLF2 0.2 442 81 82.6%(1.8) 0.2 hiCRLF2 22 8 62.9%(10.5) 0.1 0.0 0.0 012345012345 YEARS FROM DIAGNOSIS YEARS FROM DIAGNOSIS p-value=0.05 p-value=0.02

c 1.0 d 1.0 N. pts N. rel. 5 yrs Cum. Incidence 0.9 P2RY8-CRLF2 - 343 52 14.5%(1.9) 0.8 0.8 P2RY8-CRLF2 + 22 9 42.8%(10.9) 0.7 0.6 0.6 0.5 EFS 0.4 0.4

N. pts N. events 5 yrs EFS 0.3 Cum. Incidence 0.2 P2RY8-CRLF2 - 343 59 83.5%(2) 0.2 P2RY8-CRLF2 + 22 9 57.2%(10.9) 0.1 0.0 0.0 012345012345 YEARS FROM DIAGNOSIS YEARS FROM DIAGNOSIS p-value=0.004 p-value=0.001 Figure 2. Association of CRLF2 over-expression and P2RY8-CRLF2 fusion to treatment outcome. (a) EFS and (b) CIR of study cohort patients according to CRLF2 expression: hiCRLF2 and loCRLF2.(c) EFS and (d) CIR of investigated patients according to the presence or absence of P2RY8-CRLF2 fusion gene.

Leukemia (2012) 2245 – 2253 & 2012 Macmillan Publishers Limited Poor outcome of P2RY8-CRLF2 in childhood ALL C Palmi et al 2249 donors was used as a wild-type reference, whereas DNA from the MUTZ5 competing risks of other events and comparison performed with the Gray human cell line (DSMZ), carrying R683G Jak2 mutation, was used as a test. positive control. All the fragments with abnormal melting profile were The Cox regression model was applied to evaluate the prognostic value sequenced. In particular, 200 ng of DNA samples positive for HRM analysis of P2RY8-CRLF2 fusion and hiCRLF2 expression on the cause-specific hazard were amplified using 2.5 U High Fidelity Polymerase Optimase (Transge- of relapse after adjusting by main prognostic features: sex, age, white nomic, Glasgow, UK) and 20 pmol JAK2 oligos described above at the cell count and MRD. following PCR conditions: 95 1C for 5 min, 35 cycles at 95 1C for 30 s, 54 1C Follow-up was updated in February 2010. for 40 s, 72 1C for 45 s and a final extension at 72 1C for 7 min. The PCR products were cloned by the Zero Blunt PCR Cloning Kit (Invitrogen, Carlsbad, CA, USA), following the manufacturer’s instructions, and then sequenced using the ABI-3130 Genetic Analyzer instrument RESULTS (Applied Biosystems). CRLF2 aberrations at diagnosis IKZF1 deletions were investigated by Multiplex Ligation-dependent CRLF2 expression, evaluated by real-time quantitative (RQ)-PCR on Probe Amplification (MLPA) technique using the Salsa MLPA kit P335-A3 464 pediatric BCP-ALL patients at diagnosis, ranged from 0.006- ALL-IKZF1 kit (MRC-Holland, Amsterdam, the Netherlands), according to 810-fold change compared to the overall median value to the manufacturer’s instructions. Samples of pediatric ALL patients in complete remission were used as wild-type controls. The fragments (Figure 1a). Outcome was evaluated within subgroups defined were separated with the ABI-3130 Genetic Analyzer instrument (Applied by different categories of CRLF2 expression. In particular, we Biosystems) and the data analyzed using Coffalyser software (http://old. evaluated potential cut-points starting from five times the overall mlpa.com/coffalyser). median value and moving by steps of 5. The lowest threshold for CRLF2 expression showing a significant difference in CIR between two groups was 20 times the median that was then adopted as Statistical analysis cut-point (Supplementary Figure 2). Higher values of the threshold did not show an increase in discriminating power of the EFS and survival were calculated from the date of diagnosis to the date of event, which for EFS was resistance, relapse, death or second malignant CIR for the two groups separated by the threshold. neoplasm, whichever occurred first, and for survival was death from any Twenty-two patients (4.7%) presented an expression higher cause. EFS and survival curves at 5 years were estimated according to than 20 times (hiCRLF2). Kaplan–Meier, and compared according to log-rank test. Cumulative The P2RY8-CRLF2 fusion was detected in 22 patients of the 365 incidence of relapse (CIR) at 5 years was estimated by adjusting for cases analyzed (6.0%). In all, 10 of them were also hiCRLF2 (10/20

ab1.0 1.0

0.8 0.8

0.6 0.6

0.4 0.4 Survival Survival

N. pts N. deaths 5 yrs Survival N. pts N. deaths 5 yrs Survival 0.2 loCRLF2 442 34 92.1%(1.3) 0.2 P2RY8-CRLF2 - 343 22 93.4%(1.4) hiCRLF2 22 3 85.9%(7.6) P2RY8-CRLF2 + 22 4 80.9%(8.6)

0.0 0.0 012345 012345 YEARS FROM DIAGNOSIS YEARS FROM DIAGNOSIS p-value=0.35 p-value=0.04

c IR patients 1.0

0.8

0.6

0.4 Survival

N. pts N. deaths 5 yrs Survival 0.2 P2RY8-CRLF2 - 214 14 93.1%(1.8) P2RY8-CRLF2 + 15 4 72.7%(11.7)

0.0 012345 YEARS FROM DIAGNOSIS p-value=0.005

Figure 3. Association of CRLF2 over-expression and P2RY8-CRLF2 fusion to overall survival (OS) in the whole cohort and in MRD IR subgroup. (a) OS of study cohort patients according to CRLF2 expression: loCRLF2 and hiCRLF2.(b) OS of investigated patients according to the absence or presence of P2RY8-CRLF2 fusion gene. (c) OS of IR patients according to the absence or presence of P2RY8-CRLF2 fusion gene.

& 2012 Macmillan Publishers Limited Leukemia (2012) 2245 – 2253 Poor outcome of P2RY8-CRLF2 in childhood ALL C Palmi et al 2250 tested for the deletion, 50.0%), whereas the remaining 12 cases a 1.0 presented low levels of CRLF2 expression (less than 20 times the median, range: 0.006–18.5, ‘loCRLF2’; Figure 1a). Within the hiCRLF2 group, only one out of eight (negative for P2RY8-CRLF2) was 0.8 positive for the IGH@ translocation (Figure 1b). We observed JAK2 mutations in 3/10 hiCRLF2 P2RY8-CRLF2 0.6 positive patients, in 1/12 hiCRLF2 P2RY8-CRLF2 negative patients and 1/11 loCRLF2 P2RY8-CRLF2 positive patients (two R683G, EFS one R683S and two insertion mutations: L681-I682insGEEL and 0.4 T705delinsRWE). IKZF1 deletions were detected in 5/22 hiCRLF2 N. pts N. events 5 yrs EFS patients, 3 associated with P2RY8-CRLF2 fusion, and in 2/11 cases hiCRLF2 POS 10 6 37.5%(16.1) 0.2 positive for P2RY8-CRLF2 but loCRLF2. The CRLF2 F232C mutation hiCRLF2 NEG 10 1 90%(9.5) was detected in two hiCRLF2 patients, positive for both P2RY8- loCRLF2 POS 12 3 75%(12.5) loCRLF2 NEG 333 58 83.3%(2.1) CRLF2 and IKZF1 deletions (Figures 1b, c and d). 0.0 Clinical characteristics of patients positive for CRLF2 aberrations at 012345 diagnosis are described in Table 1 and in Supplementary Results. YEARS FROM DIAGNOSIS All patients with P2RY8-CRLF2 fusion had DNA index o1.16, while 5/22 hiCRLF2 patients were classiﬁed as high hyperdiploid b 1.0 N. pts N. rel. 5 yrs Cum. Incidence (Table 1). 0.9 hiCRLF2 POS 10 6 62.5%(16.1) All hiCRLF2 patients were negative for the recurring chro 0.8 hiCRLF2 NEG 10 1 10%(9.5) mosomal translocations t(4;11) and t(12;21). In particular, by loCRLF2 POS 12 3 25%(12.5) regression analysis t(12;21) and CRLF2 expression show a 0.7 loCRLF2 NEG 333 51 14.6%(2) signiﬁcant negative correlation. 0.6 0.5 Prognostic impact of CRLF2 over-expression and P2RY8-CRLF2 0.4 fusion at diagnosis

Cum. Incidence 0.3 HiCRLF2 patients have a significantly inferior EFS (62.9%±10.5 vs 82.6%±1.8, P ¼ 0.05) and increased CIR (37.1%±10.5 vs 15.2%± 0.2 1.7, P ¼ 0.02) compared with loCRLF2 patients. All the eight events 0.1 in hiCRLF2 patients were relapses (Figures 2a and b). The 5-year 0.0 survival estimates were 85.9%±7.6 and 92.1%±1.3, respectively 012345 (P ¼ 0.35; Figure 3a). YEARS FROM DIAGNOSIS In all, 9 out of the 22 patients carrying P2RY8-CRLF2 fusion experienced relapse. Kaplan–Meier analysis revealed an inferior Figure 4. Association of combined CRLF2 over-expression EFS (5-year EFS of 57.2%±10.9 vs 83.5%±2.0, P ¼ 0.004) and and P2RY8-CRLF2 fusion to treatment outcome. (a) EFS and (b) CIR increased CIR (42.8%±10.9 vs 14.5% ±1.9, P 0.001) for patients according to hiCRLF2 and loCRLF2 expression and to the presence ¼ (POS) or absence (NEG) of P2RY8-CRLF2 fusion gene. carrying the fusion gene (Figures 2c and d). The 5-year survival estimates were significantly different, 80.9%±8.6 and 93.4%±1.4, respectively (P ¼ 0.04; Figure 3b). Table 2. Results of the multivariate analyses. Cox model on hazard The worst outcome (EFS 37.5%±16.1 and CIR 62.5%±16.1) of relapse in 365 patients was observed for patients with the combination of hiCRLF2 and Characteristics P-value Hazard ratio 95% CI P2RY8-CRLF2 fusion, while hiCRLF2 patients who are negative for P2RY8-CRLF2 fusion seem to have an outcome comparable with hiCRLF2 patients lacking CRLF2 over-expression and fusion. Twelve patients No 1.00 were loCRLF2 and positive for P2RY8-CRLF2 fusion, three of whom Yes 0.93 1.05 0.37–2.97 relapsed (Figures 4a and b). Cox model analysis was performed to assess the relative P2RY8-CRLF2 prognostic value of P2RY8-CRLF2 fusion and hiCRLF2 expression No 1 after adjusting for main prognostic features (Table 2). Results show Yes 0.01 3.30 1.29–8.48 that P2RY8-CRLF2 fusion retains prognostic significance with a Gender threefold increase in the risk of relapse for positive versus negative Female 1 patients (hazard ratio 3.30; 95% CI 1.29–8.48; P ¼ 0.01), while Male 0.80 0.94 0.56–1.58 hiCRLF2 is not an independent prognostic factor (hazard ratio 1.05; 95% CI 0.37–2.97; P ¼ 0.93). Age 1–9 Yrs 1 10–17 Yrs 0.35 1.37 0.71–2.65 Outcome and risk group We further analyzed the prognostic value of P2RY8-CRLF2 fusion WBC ( Â 1000/ml) separately within SR and IR patient groups. P2RY8-CRLF2 fusion o100 1 was similarly distributed within the two non-HR subgroups, being X100 0.006 2.8 1.35–5.8 found in 6 out 106 SR patients (5.7%) and 15 out of 229 IR patients (6.7%; Table 1). However, none of the 6 P2RY8-CRLF2-positive MRD SR patients relapsed (Figure 5a), while 9 out of 15 cases in SR 1 IR 0.001 4.74 2.12–10.62 the IR group relapsed with a significant difference in the CIR o HR o0.001 38.54 7.67–193.58 (61.1%±12.9 vs 17.6%±2.6, Po0.0001; Figure 5b), and in survival Unknown 0.44 1.51 0.53–4.34 (72.7%±11.7 vs 93.1%±1.8, P ¼ 0.005; Figure 3c). EFS curves are provided as Supplementary Figure 3. Four of the nine relapses had Abbreviations: HR, high risk; IR, intermediate risk; MRD, minimal residual BM localization, three were extramedullary and two combined disease; SR, standard risk; WBC, white blood cell count.

Leukemia (2012) 2245 – 2253 & 2012 Macmillan Publishers Limited Poor outcome of P2RY8-CRLF2 in childhood ALL C Palmi et al 2251 (compared with only 6/39 extramedullary relapses in the P2RY8- de novo P2RY8-CRLF2 fusion was detected. CRLF2 mutations CRLF2-negative IR patients). remained stable at relapse, while JAK2 mutations were lost in 1/34 In order to validate those findings, we analyzed the prognostic patients at relapse (2.9%), acquired in 2/34 cases (5.9%) and in value of P2RY8-CRLF2 fusion for SR and IR patients enrolled in 1/34 case the mutation was different between diagnosis and the AIEOP-BFM ALL2000 study and treated in German Centers relapse (respectively, R683G and L681-I682insRD; Figure 6b). (BFM-G).12 Detailed description of CRLF2-expression level and genomic In the BFM-G cohort, 2 out of 8 SR patients carrying the fusion aberrations of diagnosis/relapse pairs is shown in Supplementary experienced relapse, whereas 6 relapses out of 11 IR patients were Table 3. observed (Figures 5c and d). In the same BFM-G cohort, IR patients positive for P2RY8-CRLF2 fusion showed a significant higher CIR than patients without the fusion (54.5%±15 vs 10.7%±1.9, DISCUSSION Po0.0001). Two of the six relapses had BM localization, three were In childhood ALL, risk-adjusted therapy has brought about extramedullary and one combined (compared with 10/31 extra- impressive improvements of survival rates during the last decades. medullary relapses in the P2RY8-CRLF2-negative IR patients). In this context, clinicians and researchers have two major BFM-G SR patients with or without the fusion have a CIR of objectives: to find new prognostic markers to identify upfront 25.0%±15.3 vs 5.6%±1.7, P ¼ 0.05. Thus, apparently even in the patients with high probability to relapse,1,5 including relapses BFM-G cohort, the SR patients who are positive for the fusion have during long-term follow-up, and to recognize good prognostic a lower incidence of relapse than IR patients. features that can be used to guide treatment-intensity reductions, aiming to spare children from acute and long-term side effects.15 Although the advent of MRD monitoring has revolutionized CRLF2 expression and genomic aberrations at relapse the stratification approach of the risk-based treatments,15,18 new Thirty-three paired diagnostic and relapse specimens were genomic discoveries are expected to be very useful, for analyzed for CRLF2 expression level. Samples at relapse showed recognizing previously undetectable subgroups of childhood a median value of CRLF2 expression two times higher than the ALL patients with specific risk levels, thus complementing MRD. respective samples at diagnosis (2.60 vs 1.25; Figure 6a). Here, for the first time, we report the incidence and prognostic The relapse sample with the highest CRLF2 expression level was relevance of CRLF2 over-expression and P2RY8-CRLF2 rearrange- the only sample positive for both P2RY8-CRLF2 rearrangement and ment in a study cohort of MRD-stratified childhood BCP-ALL, JAK2 mutation (L681-I682insLR). Two P2RY8-CRLF2-positive treated in Italy with the AIEOP-BFM ALL2000 protocol, excluding patients at diagnosis lost the rearrangement at relapse and no the confounding effect of DS and Ph þ subgroup of patients,

abSR AIEOP patients IR AIEOP patients 1.0 1.0 N. pts N. rel. 5 yrs Cum. Incidence N. pts N. rel. 5 yrs Cum. Incidence P2RY8-CRLF2 - 100 7 93.9%(2.4) 0.9 P2RY8-CRLF2 - 214 39 17.6%(2.6) P2RY8-CRLF2 + 60 P2RY8-CRLF2 + 0.8 100%(0) 0.8 15 9 61.1%(12.9) 0.7 0.6 0.6 0.5 0.4 0.4 0.3 Cum. Incidence Cum. Incidence 0.2 0.2 0.1 0.0 0.0 012345012345 YEARS FROM DIAGNOSIS YEARS FROM DIAGNOSIS p-value<0.0001

c SR BFM-G patients d IR BFM-G patients 1.0 1.0 N. pts N. rel. 5yrs Cum. Incidence N. pts N. rel. 5 yrs Cum. Incidence 0.9 0.9 P2RY8-CRLF2 - 186 12 5.6%(1.7) P2RY8-CRLF2 - 275 31 10.7%(1.9) P2RY8-CRLF2 + 0.8 8 2 25%(15.3) 0.8 P2RY8-CRLF2 + 11 6 54.5%(15) 0.7 0.7 0.6 0.6 0.5 0.5 0.4 0.4 0.3 0.3 Cum. Incidence Cum. Incidence 0.2 0.2 0.1 0.1 0.0 0.0 012345012345 YEARS FROM DIAGNOSIS YEARS FROM DIAGNOSIS p-value<0.0001 Figure 5. Association of P2RY8-CRLF2 fusion to treatment outcome in MRD subgroups. CIR from diagnosis of (a) SR and (b) IR AIEOP patients according to the presence or absence of P2RY8-CRLF2 fusion gene. CIR from diagnosis of (c) SR and (d) IR BFM-G patients according to the presence or absence of the P2RY8-CRLF2 fusion gene.

& 2012 Macmillan Publishers Limited Leukemia (2012) 2245 – 2253 Poor outcome of P2RY8-CRLF2 in childhood ALL C Palmi et al 2252 a 1000.0 P2RY8-CRLF2 fusion but with low expression of CRLF2, we might speculate that the deletion is present in subclones of the leukemic blast cell population. The latter would be consistent with the marked subclonal heterogeneity in ALL,19,20 but further 100.0 experiments will be necessary to show this. The lack of absolute correlation between CRLF2 fusion and 20.0 over-expression prompted us to explore the prognostic impact of 10.0 both P2RY8-CRLF2 fusion and CRLF2 over-expression, alone and in combination. Both hiCRLF2 and P2RY8-CRLF2 fusion were associated with a

expression at relapse 12,13 1.0 higher relapse rate. As already reported late relapses were prevalent, occurring mostly after the end of the 2-year treatment. Compared to the distribution of relapses in the AIEOP-BFM CRLF2 ALL2000 protocol,19,20 the tendency toward an increased 0.1 percentage of extramedullary relapse localization (either isolated 0.1 1.0 10.0 20.0 100.0 1000.0 or combined) was observed. CRLF2 expression at diagnosis Interestingly, even if the worst outcome was observed for patients with the combination of hiCRLF2 and P2RY8-CRLF2 fusion, b DIAGNOSES vs RELAPSES the latter turned out to be a poor prognostic factor regardless of CRLF2 over-expression by the Cox model, with a threefold increase REL in the risk of relapse for positive versus negative patients even DX NEG POS after adjusting for main prognostic features. Also the data of the OS analysis after 5 years observation time corroborated the finding that P2RY8-CRLF2 is a stronger prognostic marker than NEG 21 - CRLF2 over-expression. Moreover, P2RY8-CRLF2 fusion also proved to be a significant POS 23 factor of poor outcome when patients with the favorable factors t(12;21) and DNA index 1.16–1.6 were excluded from the analysis P2RY8-CRLF2 (N=26) (Supplementary Figure 4). In order to understand how to transfer the prognostic impact of this CRLF2 alteration into clinical practice, P2RY8-CRLF2 fusion was REL NEG POS analyzed within the non-HR subgroups according to protocol DX stratification. In particular, it is relevant that the EFS and CIR of the P2RY8-CRLF2 fusion-positive patients in the IR group was similar to NEG 4- that of HR patients in the AIEOP-BFM ALL2000 study and that their OS was significantly reduced.15 Importantly, this was not the case for SR patients. POS -2 The impact of P2RY8-CRLF2 fusion in the Italian cohort was validated in a similar number of patients enrolled in the AIEOP- CRLF2 MUTATIONS (N=6) BFM ALL2000 protocol in Germany, further supporting the evidence that the detection of P2RY8-CRLF2 in MRD-based IR REL patients would potentially qualify them for a high risk-based DX NEG POS treatment. Overall, among IR patients in AIEOP and BFM cohorts, 21 out of 26 P2RY8-CRLF2 fusion-positive patients were treated in the IR prednisone arm, which is available to all patients in the new NEG 30 2 current AIEOP-BFM ALL2009 protocol. Three out of the twenty six patients were highly MRD-positive at TP1 (d33) and still positive at (ref.15) POS 11* TP2 (d78 ; a new criterion for HR in the current protocol) and 2/3 relapsed. Indeed, 13/23 relapsing patients would still not have JAK2 MUTATIONS (N=34) been stratified by these means as HR patients in the current *different AIEOP-BFM ALL2009 protocol. In agreement with emerging evidence of mutual collaboration Figure 6. CRLF2 expression and genomic alterations at relapse. (a) among several genomic lesions leading to constitutive activation Log–log plot of the CRLF2 expression value for 33 paired diagnosis 10–13 and relapsed specimens. CRLF2 expression at relapse was, on of the JAK2/STAT5 pathway, we found additional genomic median, double the paired diagnostic samples (median fold abnormalities (CRLF2, JAK2 mutations and IKZF1 deletions) in change ¼ 2.60 vs 1.25). (b) Table of genomic aberrations investi- hiCRLF2- and/or P2RY8-CRLF2-positive patients. However, the gated in paired diagnosis and relapsed specimens. limited numbers for each of the combinations of these aberrations prevent any prognostic indication from being drawn. which had been included in other studies.8,10–13 A high CRLF2 Here, for the first time, we report the analysis on CRLF2 expression (defined as an expression level 20 times higher than alterations at relapse of childhood BCP-ALL. A total of 33 of the 79 the median expression of all patients analyzed) was detected in relapses occurring during the observation time of the study were about 5% of patients, while the incidence of the P2RY8-CRLF2 evaluated for CRLF2-expression levels. Samples at relapse showed fusion was estimated as 6%, which is in agreement with the a median value of CRLF2 expression that was slightly higher than incidence found in other cohorts with similar Caucasian the respective samples at diagnosis (2.60 vs 1.25). Interestingly, no composition.12,13 Importantly, for the first time, we observed de novo P2RY8-CRLF2 fusion genes were detected in the 26 relapse that the P2RY8-CRLF2 fusion was not only found among patients cases for which the diagnostic correlate had been tested. On the with hiCRLF2 expression but also among patients with loCRLF2 contrary, two positive patients at diagnosis (with low CRLF2 expression. In the case of patients who are positive for the expression levels) lost the P2RY8-CRLF2 fusion at relapse. This is an

Leukemia (2012) 2245 – 2253 & 2012 Macmillan Publishers Limited Poor outcome of P2RY8-CRLF2 in childhood ALL C Palmi et al 2253 indirect clue that the deletion, even though it is a strong marker of 4 Schultz KR, Bowman WP, Aledo A, Slayton WB, Sather H, Devidas M et al. poor prognosis as described above, is not a primary lesion and can Improved early event-free survival with imatinib in Philadelphia chromosome- be acquired during the progression of the pre-leukemic clone and positive acute lymphoblastic leukemia: a children’s oncology group study. thus can contribute to the overt disease. Moreover, like P2RY8- J Clin Oncol 2009; 27: 5175–5181. CRLF2, JAK2 mutations appear to be secondary lesions; both do 5 Izraeli S. Application of genomics for risk stratification of childhood acute not remain stable at relapse in all cases. lymphoblastic leukaemia: from bench to bedside? Br J Haematol 2010; 151: 119–131. In conclusion, we have shown that childhood BCP-ALL patients 6 Den Boer ML, van Slegtenhorst M, De Menezes RX, Cheok MH, Buijs-Gladdines JG, with the P2RY8-CRLF2 fusion have a very poor outcome when Peters ST et al. A subtype of childhood acute lymphoblastic leukaemia with poor treated by the AIEOP-BFM approach that could justify a shift from treatment outcome: a genome-wide classification study. Lancet Oncol 2009; 10: the MRD-based intermediate to the high-risk group. It remains to 125–134. be demonstrated that a high risk-based therapy could be effective 7 Mullighan CG, Su X, Zhang J, Radtke I, Phillips LA, Miller CB et al. Deletion of IKZF1 in preventing relapse for these patients. In the long run, and prognosis in acute lymphoblastic leukemia. N Engl J Med 2009; 360: 470–480. alternative therapies that interfere with the activation of the 8 Russell LJ, Capasso M, Vater I, Akasaka T, Bernard OA, Calasanz MJ et al. JAK2/STAT5 signaling pathway may hold a promise for treatment Deregulated expression of cytokine receptor gene, CRLF2, is involved in lymphoid of these patients. transformation in B-cell precursor acute lymphoblastic leukemia. Blood 2009; 114: 2688–2698. 9 Yoda A, Yoda Y, Chiaretti S, Bar-Natan M, Mani K, Rodig SJ et al. Functional screening identifies CRLF2 in precursor B-cell acute lymphoblastic leukemia. CONFLICT OF INTEREST Proc Natl Acad Sci USA 2010; 107: 252–257. The authors declare no conflict of interest. 10 Mullighan CG, Collins-Underwood JR, Phillips LA, Loudin MG, Liu W, Zhang J et al. Rearrangement of CRLF2 in B-progenitor- and Down syndrome-associated acute lymphoblastic leukemia. Nat Genet 2009; 41: 1243–1246. ACKNOWLEDGEMENTS 11 Hertzberg L, Vendramini E, Ganmore I, Cazzaniga G, Schmitz M, Chalker J et al. This study was supported by grants from: Fondazione Citta` della Speranza (Padova), Down syndrome acute lymphoblastic leukemia, a highly heterogeneous disease Fondazione Tettamanti (Monza), Associazione Italiana Ricerca sul Cancro (AIRC; to in which aberrant expression of CRLF2 is associated with mutated JAK2: a report GteK, GB, GiC, AB and MGV), CARIPARO Project of Excellence (to GteK, EV and DF), from the International BFM Study Group. Blood 2010; 115: 1006–1017. MIUR (to AB and GB), Fondazione Cariplo (to AB and GiC), Deutsche Krebshilfe 12 Cario G, Zimmermann M, Romey R, Gesk S, Vater I, Harbott J et al. Presence of the (to GuC, MSt and MSc), Madeleine-Schickedanz-Kinderkrebsstiftung (to GuC, MSt and P2RY8-CRLF2 rearrangement is associated with a poor prognosis in non-high-risk MSc), Israel Science Foundation Legacy Program and Children with Leukaemia UK precursor B-cell acute lymphoblastic leukemia in children treated according to (to SI). We thank Simona Songia, Lilia Corral, Valentina Carrino, Eugenia Mella the ALL-BFM 2000 protocol. Blood 2010; 115: 5393–5397. (Monza), Elena Seganfreddo and Katia Polato (Padova) for AIEOP MRD monitoring; 13 Ensor HM, Schwab C, Russell LJ, Richards SM, Morrison H, Masic D et al. Demo- Silvia Bungaro and Marta Galbiati (Monza) for cytogenetics and FISH; all medical graphic, clinical and outcome features of children with acute lymphoblastic doctors of the AIEOP and BFM centers. leukemia and CRLF2 deregulation: results from the MRC ALL97 clinical trial. Blood 2011; 117: 2129–2136. 14 Harvey RC, Mullighan CG, Chen IM, Wharton W, Mikhail FM, Carroll AJ et al. AUTHOR CONTRIBUTIONS Rearrangement of CRLF2 is associated with mutation of JAK kinases, alteration of IKZF1, Hispanic/Latino ethnicity, and a poor outcome in pediatric B-progenitor CP, GL, EV, DF and CS performed the molecular analyses on CRLF2 aberrations acute lymphoblastic leukemia. Blood 2010; 115: 5312–5321. and additional genetic mutations; CP, EV and SI collaborated in writing the 15 Conter V, Bartram CR, Valsecchi MG, Schrauder A, Panzer-Grumayer R, Moricke A manuscript; DS and MGV collected the Trial data and performed all the et al. Molecular response to treatment redefines all prognostic factors in statistical analyses; AMDM and AL performed cytogenetic analyses; VR, GF and children and adolescents with B-cell precursor acute lymphoblastic leukemia: TV represent the team that performed Trial molecular and MRD analyses; results in 3184 patients of the AIEOP-BFM ALL 2000 Study. Blood 2010; 115: GB and AB supervised the research; EG performed DNA/RNA extraction and cell 3206–3214. banking; VC is responsible of the AIEOP ALL2000 study; MSc, MSt and GuC 16 Shochat C, Tal N, Bandapalli OR, Palmi C, Ganmore I, te Kronnie G et al. Gain-of-function mutations in interleukin-7 receptor-alpha (IL7R) in childhood provided the data of the BFM-G patients; GiC and GtK designed the study, acute lymphoblastic leukemias. J Exp Med 2011; 208: 901–908. supervised the research, analyzed data and wrote the manuscript. 17 Livak KJ, Schmittgen TD. Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) method. Methods 2001; 25: 402–408. REFERENCES 18 van Dongen JJ, Seriu T, Panzer-Grumayer ER, Biondi A, Pongers-Willemse MJ, 1 Pui CH, Carroll WL, Meshinchi S, Arceci RJ. Biology, risk stratification, and therapy Corral L et al. Prognostic value of minimal residual disease in acute lymphoblastic of pediatric acute leukemias: an update. J Clin Oncol 2011; 29: 551–565. leukaemia in childhood. Lancet 1998; 352: 1731–1738. 2 Waanders E, van der Velden VH, van der Schoot CE, van Leeuwen FN, 19 Anderson K, Lutz C, van Delft FW, Bateman CM, Guo Y, Colman SM et al. Genetic van Reijmersdal SV, de Haas V et al. Integrated use of minimal residual disease variegation of clonal architecture and propagating cells in leukaemia. Nature classification and IKZF1 alteration status accurately predicts 79% of relapses 2011; 469: 356–361. in pediatric acute lymphoblastic leukemia. Leukemia 2011; 25: 254–258. 20 Rothman R, Trakhtenbrot L, Bielorai B, Izraeli S, Ishoev G, Amariglio N et al. 3 Ottmann OG, Pfeifer H. Management of Philadelphia chromosome-positive acute Co-existence of multiple subclones in TEL-AML1 at diagnosis of acute lymphoblastic leukemia (Ph þ ALL). Hematol Am Soc Hematol Educ Program 2009; lymphoblastic leukaemia in association with submicroscopic deletion of AML1. 2009: 371–381. Br J Haematol 2005; 129: 491–498.

Supplementary Information accompanies the paper on the Leukemia website (http://www.nature.com/leu)

& 2012 Macmillan Publishers Limited Leukemia (2012) 2245 – 2253