Characterisation of the genomic landscape of CRLF2‐rearranged acute lymphoblastic leukemia

Abstract Deregulated expression of the type I cytokine receptor, CRLF2, is observed in 5–15% of precursor B‐cell acute lymphoblastic leukaemia (B‐ALL). We aimed to determine the clinical and genetic landscape of those with IGH‐CRLF2 or P2RY8‐CRLF2 (CRLF2‐r) using multiple genomic approaches. Clinical and demographic features of CRLF2‐r patients were characteristic of B‐ALL. Patients with IGH‐CRLF2 were older (14 y vs. 4 y, P < .001), while the incidence of CRLF2‐r among Down syndrome patients was high (50/161, 31%). CRLF2‐r co‐occurred with primary chromosomal rearrangements but the majority (111/161, 69%) had B‐other ALL. Copy number alteration (CNA) profiles were similar to B‐other ALL, although CRLF2‐r patients harbored higher frequencies of IKZF1 (60/138, 43% vs. 77/1351, 24%) and BTG1 deletions (20/138, 15% vs. 3/1351, 1%). There were significant differences in CNA profiles between IGH‐CRLF2 and P2RY8‐CRLF2 patients: IKZF1 (25/35, 71% vs. 36/108, 33%, P < .001), BTG1 (11/35, 31% vs. 10/108, 9%, P =.004), and ADD3 deletions (9/19, 47% vs. 5/38, 13%, P =.008). A novel gene fusion, USP9X‐DDX3X, was discovered in 10/54 (19%) of patients. Pathway analysis of the mutational profile revealed novel involvement for focal adhesion. Although the functional relevance of many of these abnormalities are unknown, they likely activate additional pathways, which may represent novel therapeutic targets.

the pseudoautosomal region (PAR1) of chromosomes X and Y (P2RY8-CRLF2); rare but recurrent CRLF2 mutations. All three CRLF2-r result in overexpression of CRLF2 mRNA and protein; however, alone they are insufficient to cause overt leukaemia. 5,7 Interestingly, studies identifying CRLF2-d patients by mRNA and protein expression have shown that some patients do not harbor one of the three known genomic rearrangements. 8 The cause of this overexpression is currently unknown. The incidence of CRLF2-r is high in patients with Down syndrome ALL (DS-ALL) (>50%) and intrachromosomal amplification of chromosome 21 (iAMP21) (25%), 5,7,9211 with other established cytogenetic abnormalities rarely associated. However, data indicating whether CRLF2-r is a primary or secondary event are scarce. 12 It is well documented that deletions of genes involved in B-cell differentiation and cell cycle control are recurrent in these patients. 5,[13][14][15] Mutations of the Janus kinase family, in particular JAK2, and mutations of IL7R, are also recurrently observed and together result in IL-3 independent growth of mouse BaF3 cells. 11,16 Recently, mutations affecting other kinase genes have also been reported in CRLF2-d ALL. 4,17 Both BCR-ABL1-like and CRLF2-d ALL have been associated with poor outcome and increased risk of relapse. 2,3,10 Although MRD-directed treatment intensification improves survival, 18 outcome for CRLF2-d ALL remains less favorable than for patients with good risk cytogenetics. Hence, novel therapeutic strategies are required to improve survival and quality of life. In this study, we have explored the clinical and genetic landscape of patients with known CRLF2 rearrangements. The aims of this study were: (1) to identify the clinical and genomic differences that may exist between patients with IGH or P2RY8 driven overexpression of CRLF2; (2) to undertake pathway analysis of whole genome (WGS) and whole exome sequencing (WES) data to highlight additional pathways that may co-operate with rearrangements of CRLF2.

| Cytogenetics and fluorescence in situ hybridisation
For details on genetic testing, see Supporting Information Figure 1.
Karyotype data were collected from UK cytogenetics laboratories (Supporting Information Table 1). Cytogenetic analysis and FISH were carried out on the same diagnostic patient samples. The involvement of IGH was determined using the LSI IGH Dual Color Break-Apart Rearrangement Probe (Abbott Molecular, Green Oaks, IL). 6,22 Additional FISH probes used in this study are shown in Supporting Information Table 3. Five control slides of fixed cells from normal individuals were hybridised with all probe combinations 5 (Supporting Information Table 3) to determine the cut-off percentages for false positive results (6 3 3 standard deviations).
A minimum of 100 nuclei were scored by eye for each FISH test by two independent analysts. When combining three or more florophores, capture and scoring was carried out using an automated Olympus BX-61 8bay stage florescence microscope. Images were analysed using the Cyto-Vision 7.1 SPOT counting system (Leica Microsystems, Gateshead, UK).

| Low depth paired-end and whole exome sequencing
Sufficient diagnostic DNA was available from 11 CRLF2-r patients (representitive of cohort and all patients have SNP analysis completed), which was prepared for library construction (300-500 bp), flow cell preparation and cluster formation using the Illumina no-PCR library protocol (Illumina Inc, San Diego, CA), and 50 bp reads were performed using the Illumina Genome Analyzer IIx instrument following the manufactures guidelines.
Structural variants were selected as previously reported. 23 Sufficient diagnostic and germ-line DNA from the same 11 patients was prepared for Illumina paired-end sequencing with subsequent exome enrichment using the Agilent SureSelect Human All Exon 50 MB kit (Agilent Technologies LTD, Berkshire, UK). Guidelines for the Illumina Genome Analyzer IIx instrument were followed for flow-cell preparation, cluster generation and paired-end sequencing of 75 bp reads. 23,24 Whole † These authors contributed equally to this study.
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. V C 2017 Wiley Periodicals, Inc. genome and exome sequencing data are available using EGA accession numbers EGAD00001002007 and EGAD00001002008, respectively.

| Structural variant detection and validation
Sequencing reads were mapped to the human genome (hg19) with a minimum and mean fold coverage of 5.61 and 8.46, respectively. The algorithm BRASS 25 was used to identify groups of 2 discordantly mapped (distance or orientation) paired-end reads. Rearrangements observed in 5 or more paired-end reads were validated (excluding those where both reads were in the same intron or outside the coding regions) by conventional PCR and Sanger sequencing using diagnostic and remission DNA. If a rearrangement failed to validate using two sets of primers, repeated twice over a temperature gradient, it was regarded as a fail. Breakpoints within immunoglobulin loci were not validated.
MLPA and SNP data were also used to validate CNA. A total of 476 SV were identified from WGS of 11 patients with CRLF2-r ALL (average 43 SV/patient, range 22-90), including 416 intrachromosomal and 57 interchromosomal rearrangements.

| Mutation and in/del detection and validation
A total of 458 SNVs and insertion/deletions (in/dels) were identified from the same 11 patients. A minimum sequencing coverage of 30-fold was required for each sample. Default setting of BWA 24 and CaVEMan was used to align the reads and detect somatically acquired single nucleotide substitutions as previously reported 23 . The algorithm PIN-DEL 26 was used to detect in/dels. 23 Conventional PCR was used to validate the substitutions, with subsequent 454 pyrosequencing for confirmation. 23 Conventional PCR validated the in/dels, with subsequent sequencing using the ABI terminator Cycle Sequencing Kit (Applied Biosystems) to confirm each in/del. 23

| Targeted JAK mutation screening
Primers were used to amplify exon 14 of the JAK2 and JAK1 genes only 27 and Sanger sequencing confirmed the presence of the mutation.
As far as we are aware CRLF2 mutations have not been reported in patients with IGH-CRLF2 or P2RY8-CRLF2, therefore screening for these mutations was not carried out.

| Statistical analysis
The distribution of categorical variables was examined using Fisher's exact test. Due to the investigative nature of this analysis, we did not apply stringent multiple comparisons adjustment (all tests were conducted at the 5% significance level). All analyses were performed using Intercooled Stata 14.1 for Windows (Stata Corporation, College Station, TX).

| A novel fusion between USP9X and DDX3X
An interstitial deletion of chromosome sub-band Xp11.4, fusing USP9X to DDX3X (Supporting Information Figure 3A), was observed in 2/26 cases by SNP6.0 arrays and 7/24 cases by FISH, including one DS patient, giving an incidence of 19% (Supporting Information Figure 3B).
No cases were identified among the publically available SNP6.0 data. 32 FISH revealed that the fusion was present in both minor and major clones (8-91% of nuclei) in both male (n 5 4) and female patients (n 5 5). All male patients had either an additional copy of normal chromosome X or derived X involved in the CRLF2-r. It was identified in patients with both IGH and P2RY8 involvement (4 vs. 5, respectively).
Real-time PCR confirmed that the fusion was in-frame and was expressed at the mRNA level (Supporting Information Figure 3C).
Sanger sequencing verified fusion of USP9X exon 31 to exon 2 of DDX3X (Supporting Information Figure 3D). The DNA breakpoint sequence from three patients consistently fell within intron 31 of   (Table 1), with six patients harboring either a mutation or CNA in two or more kinase genes.

| DISCUSSION
In this study investigating the genomic landscape of CRLF2-r ALL, we have confirmed the high incidence of CRLF2-r in DS-ALL, demonstrated its co-existence with other primary chromosomal rearrangements and enrichment of specific chromosomal gains and deletions of IKZF1, BTG1, ADD3, SERP2, TSC22D1, SLX4IP, and PBX3. There were significant differences in CNA profiles between P2RY8-CRLF2 and IGH-CRLF2 patients with an increased incidence of IKZF1, BTG1, and ADD3 deletions and a higher age at diagnosis being observed in the latter.
These disparities correlate with reported differences in outcome, where for example older age and IKZF1 status drive the inferior prognosis observed in adult IGH-CRLF2. 14 A fifth of CRLF2-r patients harbor a primary chromosomal abnormality with all but one being P2RY8-CRLF2, suggesting a co-operating role for CRLF2 deregulation. Interestingly, the sole patient with IGH-CRLF2 and a primary abnormality had the BCR-ABL1 fusion in a separate clone. 22 Collectively, these data do not imply that IGH and P2RY8 are driving distinct subgroups but rather that CRLF2-d may play a dual role as both a primary and co-operating driver in ALL, with the latter being more prominent among P2RY8-CRLF2 patients. This model is akin the BCR-ABL1 fusion, which although predominatly a primary driver, has also been reported as a secondary abnormality 22,33 . Independent pathway analysis of the aber- ing, and many groups are now assessing the efficacy of inhibitors to these pathways. The first inhibitor to be tested in CRLF2-r ALL was the JAK1/2 inhibitor, Ruxolitinb. 47 While this study showed a response in vivo, greater effect was observed in JAK-activated, non-CRLF2-r ALL.
Evidence of resistance to type I inhibitors in model systems of B-ALL is already driving the development of type II JAK inhibitors. 48 Targeting of mulitple pathways was an approach taken by Suryani et al, (2015) in assessing the JAK inhibitor, AZD1480, alone and in combination with the MEK inhibitor, Selumetinib. 49 While this study showed a strong anti-leukaemic effect in vitro, only modest effects were seen in vivo, an important consideration for future preclinical testing, which highlights the need to identify other common targets to provide further options for more patients. In addition to JAK-STAT, we have identified several deregulated pathways in CRLF2-r ALL, for example, focal adhesion signaling, one which has not previously been implicated in this subtype of ALL and was enriched in patients with both IGH-and P2RY8-CRLF2 ALL.
Mutations within genes controlling cell adhesion would seem to be  50 Although the functional consequence of these mutations in CRLF2-r ALL remains to be determined, when laminin binding integrin signaling is altered, it activates downstream signaling kinases, such as FAK (focal adhesion kinase) and JUN (MAPK10)-FOS, which impacts on cell proliferation and migration in various cancers. 51,52 In a recent study, the presence of the Ikaros isoform 6 (IK6), IKZF1 haploinsuficiency or mutations in BCR-ABL1-positive disease have been shown to increase expression of adhesion molecules and increased transcription of FAK, rendering these cells insensitive to tyrosine kinase inhibition. 53 The application of inhibitors to FAK in patients with BCR-ABL1-positive ALL resulted in the same abrogation of adhesion and self renewal programmes. 54 Interestingly, all 11 patients with sequencing data either showed mutation/ CNA of IKZF1 or a mutation within a focal adhesion gene. Our data along with emerging literature in leukaemia highlight a role for the focal adhesion pathway in CRLF2-r ALL.
In summary, we show clear clinical and genomic differences between patients with IGH-and P2RY8-CRLF2. We describe these rearrangements in the presence of other established cytogenetic abnormalities, suggesting a secondary role for CRLF2-r in some patients, akin to BCR-ABL1 positive disease. However, due to low patient numbers, we were unable to ascertain any impact of P2RY8-CRLF2 on the prognosis of these patients. This co-occurrence needs to be further assessed in a larger patient cohort. The treatment of patients across four independent trails precluded meaningful survival analysis between patients IGHand P2RY8-CRLF2. It is clear from our data that CRLF2-r ALL is heterogeneous, requiring a combination of genetic abnormalities in functionally relevant genes to co-operate with deregulated expression of CRLF2. Although the functional relevance of some of the deletions and mutations presented in this descriptive study are currently unknown, there are pointers to activation of additional targetable pathways.
There is a clear requirement for studies addressing the biological effect of these aberrations, which together with the identification of mutations in cell adhesion genes and a high incidence of IKZF1 deletions provide interesting targets for pre-clinical testing. Inhibitors to the focal adhesion pathway, as one example, may provide an insight into a new realistic therapeutic approach to improve outcome in CRLF2-r ALL.

ACKNOWLEDGMENTS
The authors would like to thank member laboratories of the United