Prognostic and therapeutic role of targetable lesions in B-lineage acute lymphoblastic leukemia without recurrent fusion genes

To shed light into the molecular bases of B-lineage acute lymphoblastic leukemia lacking known fusion transcripts, i.e. BCR-ABL1, ETV6-RUNX1, E2A-PBX1, and MLL rearrangements (B-NEG ALL) and the differences between children, adolescents/young adults (AYA) and adults, we analyzed 168 B-NEG ALLs by genome-wide technologies. This approach showed that B-NEG cases carry 10.5 mutations and 9.1 copy-number aberrations/sample. The most frequently mutated druggable pathways were those pertaining to RAS/RTK (26.8%) and JAK/STAT (12.5%) signaling. In particular, FLT3 and JAK/STAT mutations were detected mainly in AYA and adults, while KRAS and NRAS mutations were more frequent in children. RAS/RTK mutations negatively affected the outcome of AYA and adults, but not that of children. Furthermore, adult B-NEG ALL carrying JAK/STAT mutations had a shorter survival. In vitro experiments showed that FLT3 inhibitors reduced significantly the proliferation of FLT3-mutated primary B-NEG ALL cells. Likewise, PI3K/mTOR inhibitors reduced the proliferation of primary cells harboring RAS and IL7R mutations. These results refine the genetic landscape of B-NEG ALL and suggest that the different distribution of lesions and their prognostic impact might sustain the diverse outcome between children, adults and partly AYA - whose genomic scenario is similar to adults - and open the way to targeted therapeutic strategies.


INTRODUCTION
B-lineage acute lymphoblastic leukemia (B-ALL) represents 75-85% of all ALLs and is characterized by the accumulation of immature B cells, which results in the suppression of normal hematopoiesis. B-ALL is the most frequent cancer in children, with the highest incidence between 2 and 5 years, while it is relatively rare in adults, with a peak after the age of 50 [1][2][3]. While in childhood the survival rates are in the order of 80%, the outcome in adults is still unsatisfactory, with long-term survival rates of about 40-50% of cases [4][5][6][7][8]. An intermediate group is represented by the adolescents/young adults (AYA), who show an improved outcome when treated with pediatric-

Incidence and outcome of B-NEG ALLs stratified for age cohorts
The meta-analysis of the incidence of known molecular aberrations in a cohort of over 5,000 ALL [3] revealed that within B-ALL, B-NEG cases represent 70.5% of childhood, 72.7% of AYA and 42.7% of adults. Furthermore, within the population included in the current study, we evaluated the outcome of the above mentioned age cohorts. Clinical outcome data were available for 142 patients (48 children, 50 AYA and 44 adults) with a median follow-up of 65 months. Overall survival (OS) was significantly different among children (83%), AYA (55%) and adults (29%), resembling that of ALL in general ( Figure 1) [3].

Genomic overview of B-NEG ALL
Whole exome sequencing (WES) of the discovery panel cases revealed 136 mutated genes (10.5/case), with a similar load across the 3 age cohorts (Figure 2A). Mutated genes were highly heterogeneous across B-NEG ALL samples with only FLT3, PREX2, PAX5 -recurrently mutated, being found in 2 of the 13 samples (Table S1). In spite of this, in silico analyses revealed a significant enrichment of the following gene categories: focal adhesion/ECM interaction, small GTPase mediated signal transduction, ion transport and protein kinase activity, as described in Supplemental Material and Table S2.

Identification and screening of the recurrently mutated genes
Given the low number of recurrently mutated genes detected in the discovery panel, WES was broadened to 68 additional B-NEG cases (screening panel 1) and the bioinformatic analysis focused on the genes mutated in the discovery panel and/or reported as affected by mutations or CNAs in published ALL series [25][26][27][28][29]. This approach yielded a shorter list of genes frequently targeted by mutations, which overall clustered in the RAS/RTK and JAK/STAT cascades, or affected B-lineage differentiation, Figure 1: OS of B-NEG ALL patients, included in the present study, stratified for age cohorts. www.impactjournals.com/oncotarget and also highlighted intriguing differences amongst the age cohorts (Table S3).
To validate the latter findings we sequenced the most frequently mutated genes within the abovementioned pathways -namely FLT3, KRAS, NRAS, IL7R, JAK2, CRLF2, JAK1 and PAX5 -in 87 additional B-NEG ALLs (screening panel 2). The results described below refer to a total population of 168 cases (50 children, 61 AYA and 57 adults).

Genetic lesions of PAX5
Among the mutations of B-lineage development genes, those affecting PAX5 were the most common, occurring in 20 patients (11.9%). Mutations clustered in the paired-box domain with roughly 50% being represented by the hotspot P80R ( Figure S1A). The comparison among age groups highlighted a significant difference between the percentage of mutations detected in children and adults (2% vs 19.3%, p = 0.005, Figure S1B); AYA displayed an intermediate behavior, with mutations observed in 13.1% of cases.
By integrating the mutational with the CN status, we confirmed that PAX5 lesions increase with age progression, being detected in 13.3% of children, 28.6% of AYA and 44.6% of adults: the difference was statistically significant between children and adults, as well as by combining AYA and adults (p = 0.001, p = 0.004, Figure S1C).

Recurrent deletions of B-cell differentiation and cellcycle genes
Beside PAX5 deletions, the screening of CNAs of selected genes involved in B-cell development (i.e. IKZF1, EBF1, BTG1) as well as the cell-cycle related genes (i.e. CDKN2A/B and RB1) revealed that deletions were significantly more frequent in AYA and adult patients than in children. Overall, B-cell development genes were Oncotarget 13889 www.impactjournals.com/oncotarget deleted in 26.6% of pediatric, 64.3% of AYA and 66.1% of adult cases. Similarly, 15.5% of pediatric, 58.9% of AYA and 50% of adult patients harbored deletions targeting cell-cycle genes. In particular, these differences achieved a statistical significance for IKZF1, BTG1 and CDKN2A, as detailed in Figure 3A.
As a result, these selected CNAs tended to accumulate with age, with 46.6% of children, 82.4% of AYA and 87.5% carrying at least one CNA ( Figure 3B).
KRAS and NRAS mutations accumulated mostly in childhood: in fact, collectively RAS mutations were detected in 30% of children and in 14.7% of AYA and 14% of adults (p = 0.052, p = 0.045).
Finally, we found no overlap between FLT3 mutations and NRAS, KRAS G12-13 mutations; nonetheless, 1 sample carrying a FLT3 ITD and KRAS G60D mutations was detected. Moreover, 3 cases harbored both NRAS and KRAS mutations. Notably, concomitant NRAS and KRAS mutations were only observed in AYA and children.

Distribution and features of JAK/STAT pathway mutations
Mutations in genes converging into the JAK/STAT pathway -JAK2, IL7R, CRLF2 and JAK1 -accounted for 12.5% of cases (21/168). Missense mutations targeting arginine 863 of JAK2 and indels in exon 6 of IL7R were the most frequent, detected in 5.3% and 4.8% of cases, respectively. Less than 2% of cases harbored CRLF2 F232C mutations (N = 3) or JAK1 mutations (N = 2) ( Table 2). JAK/STAT mutations were uniquely detected in patients >15 years (p < 0.001): none was in fact found in children, whereas 13% of AYA and 17.5% of adult patients were affected ( Figure 4B). Furthermore, JAK/STAT mutations were associated with male gender (p = 0.01) and higher white blood cell (WBC) counts (90.6 vs 43.5 x10 9 /L in mutated vs WT cases, p = 0.002). GEP analysis of JAK/STAT-mutated samples, revealed that 10 of the 21 mutated samples (47.6%) displayed a BCR-ABL1-like profile, thus indicating that JAK/STAT mutations, though characterizing the BCR-ABL1-like cases, are not exclusive of this cohort ( Figure S4).
Mutations targeting the JAK/STAT genes were mainly mutually exclusive, with only 1 sample carrying 2 mutations: one targeting JAK2 and the other CRLF2, the latter detected at the subclonal level.

Outcome of B-NEG ALL stratified for age cohorts
To assess whether the mutations identified had a role on patients' outcome, we evaluated the impact on OS and DFS of RAS/RTK and JAK/STAT pathway mutations. Provided the relatively low incidence of the single gene mutations, clonal mutations (detected in   Subclonal mutations: detected in <20% of tumor population. Oncotarget 13893 www.impactjournals.com/oncotarget >20% of tumor population) were grouped into the JAK/ STAT (IL7R, JAK2, JAK1, CRLF2 mutations and CRLF2rearrangements) and RAS/RTK (FLT3, KRAS, NRAS) pathways. Among the latter, only the most common hotspot mutations -namely, ITD and D835 FLT3 mutations and those targeting the G12-13 aminoacidic residues of KRAS and NRAS -were considered [37,38]. When analyzing the impact of mutations targeting one pathway, samples harboring mutations of the other pathway were filtered out from the WT cohort.
Firstly, the entire B-NEG cohort -comprising children, AYA and adults -was analyzed and we observed that the RAS/RTK pathway mutations did not impact on long-term survival. At variance, the OS of JAK/STATmutated cases was significantly shorter than that of WT cases (29.4% vs 62.8% at 4 years; 95% CI 14.1-61.4 and 53.3-73.9, p = 0.003) ( Figure 5A).
Samples carrying mutations in both pathways (N = 4) experienced the worse OS when compared to the samples harboring only JAK/STAT or RAS/RTK (p = 0.006); in fact, all the events occurred within 30 months ( Figure 5B).
Given the diverse distribution of mutations and the heterogeneous treatments across age cohorts, we repeated these analyses by grouping AYA and adults, whose mutational profile and therapeutic regimens were comparable. This approach led to the results described below.
Contrariwise, in the pediatric cohort the survival of RAS/RTK-mutated cases resembled that observed in WT cases.

Impact of JAK/STAT mutations on B-NEG ALL patient's outcome
The evaluation of the impact of JAK/STAT mutations on B-NEG ALL patient's outcome was confined to AYA and adult cases, since these mutations were absent in children. This analysis documented the negative impact of JAK/STAT mutations on the OS of adult patients only, with JAK/STAT-mutated cases having a 10% (95% CI 1.6-64.2) survival probability at 4 years, as opposed to 31.3% (95% CI 16.6-59.2) for WT patients ( Figure 7A). JAK/ STAT-mutated cases also had a DFS at 4 years of 16.7% (95% CI 2.8-99.7) compared to 46.9% (95% CI 25.6-86) for WT patients (p = 0.009, Figure 7B).

In vitro sensitivity assays show the efficacy of target specific inhibition
Finally, in order to examine the potential of targeting the RAS/RTK and JAK/STAT pathways in B-NEG ALLs, we performed in vitro sensitivity experiments by treating primary cells from cases harboring known mutations with gene or pathway specific inhibitors.
When the sensitivity of primary cells carrying JAK/ STAT pathway mutations was tested, ruxolitinib (0.1 μM) proved effective in reducing to 59% the proliferation rate of JAK2-mutated cells from 1 case and rapamycin (0.1 μM) reduced to 35.4% the proliferation of IL7Rmutated primary cells from another patient, in line with a downstream PI3K/mTOR cascade activation.
Beside their anti-proliferative activity, the abovementioned drugs also exerted a pro-apoptotic effect, particularly evident for FLT3 inhibitors.
To assess whether the response to the selected inhibitors was specifically associated to the mutational status of B-ALL primary cells, we compared it with B-ALL cases that proved negative for FLT3, KRAS and NRAS mutations. The reduction of proliferation induced by crenolanib, quizartinib and ponatinib (0.1 μM) was significantly higher in FLT3-mutated than in FLT3-WT cases (N = 4, p = 0.002, p < 0.0001, p = 0.0004, respectively) ( Figure 8C). Similarly, the proliferation rate of RAS-mutated cases was considerably reduced in response to rapamycin and selumetinib (0.1 μM) when compared to RAS-WT B-ALL samples (N = 3, p = 0.024, p = 0.037) ( Figure 8D). At variance, BEZ235, which simultaneously targets mTOR and PI3K, was also partly active in reducing the proliferation (50.5±21.4%, 0.1 μM) of RAS-WT primary cells.

DISCUSSION
We describe the first in-depth genomic analysis of a broad cohort of B-ALLs lacking known fusion transcripts, including pediatric, AYA and adult cases. This "genetically orphan" subgroup deserves special attention, since it constitutes 70.5% of childhood, 72.7% of AYA and 42.7% of adult ALL cases [3], represents a suitable setting to identify additional lesions sustaining the heterogeneous clinical scenario between children and adults, and reflects the clinical outcome observed in ALL, where the worse prognosis is observed in adults.
By WES of the discovery panel we found that the load of somatic mutations in B-NEG ALL resembles what observed in other leukemias, being 10.5 per sample [26,28,39]. Since few recurrent mutations were detected, the study was broadened to two screening panels, for a total of 168 B-NEG ALL samples. This analysis, focused on the genes commonly altered in ALL, revealed an enrichment of mutations targeting genes involved in B-cell development and protein kinases signaling [25][26][27][28][29].
Among the B-cell development genes, PAX5 mutations were the most common event -occurring in 19% of cases -and increased with age. When PAX5 deletions were also considered, we observed the same pattern, with 44.6% of adults and only 13.3% of children carrying PAX5 genetic lesions. Remarkably, similar differences were found when the analysis of CNAs was extended to other B-cell development genes and cell-cycle genes commonly affected by deletions in B-ALL that tend to accumulate with age. Taken together, these results indicate that the genetic background of children is considerably different from patients older than 15 years old.
When focusing on protein kinases cascades, we found that 26.8% of B-NEG ALL cases carry mutations in the RAS/RTK pathway. The deregulation of this pathway may be due to alterations of upstream signaling molecules, like the receptor tyrosine kinase FLT3, or of essential players of the pathway like NRAS and KRAS, as occurred in our series [37].
Within B-ALL, the role of FLT3, KRAS and NRAS mutations has been investigated mostly in the pediatric setting and has been linked to hyperdiploid, MLL-rearranged ALL, hypodiploid, as well as high-risk pediatric cases, as elucidated by NGS [26,[30][31][32][33][34][35][36]. Our findings encourage a thoughtful reconsideration of their role since we documented that FLT3, KRAS and NRAS mutations are also a common event in B-NEG ALL with a different distribution across age cohorts. In particular, FLT3 mutations were specific of AYA and adults, while NRAS and KRAS alterations were prominent in children: in fact, in the latter cohort they collectively accounted for 30% of cases, consistent with the previously reported association with younger age at presentation [36]. Importantly, RAS/RTK mutations were not confined to hyperdiploid cases, as previously reported: in fact, in our cohort mutations were detected in both hyperdiploid and non-hyperdiploid cases.
The clinical impact of RAS/RTK pathway mutations has been explored with contradictory results and seems to be mainly subgroup-dependent. In fact, while several studies, including the largest analysis performed in pediatric cases [36], failed to document any significant effect of RAS mutations on survival, a recent paper focusing on MLL-rearranged infant ALL documented an independent prognostic factor of NRAS/KRAS [33,35,36]. Similarly, while the impact of FLT3 mutations has been widely explored in acute myeloid leukemia [40], it is unclear whether FLT3 mutations affect the outcome of ALL patients [26,30,31]. In our series, given the low number of mutations affecting each single gene we grouped them according to their pathway. When we evaluated the outcome of patients carrying RAS/RTK mutations, we observed a peculiar pattern: AYA and adult mutated cases -all enrolled in clinical trials for adults -had a significantly reduced OS and DFS when compared to WT patients, while the outcome of pediatric cases was not influenced by these mutations, in line with previous reports, most likely because of the effect of more aggressive regimens.
Overall, these results prompted us to investigate the efficacy of specific inhibitors, in light of their availability. Among the FLT3 inhibitors, the newly introduced quizartinib and crenolanib are promising against ITD and D835 mutations [41]. In our in vitro assays, both compounds, as well as the third-generation pan-TKI ponatinib were effective in reducing the proliferation of FLT3-mutated primary cells.
Since RAS proteins have proven challenging to inhibit, an alternative strategy is to interfere with the RAS signaling at different levels of the pathway [42,43]. Encouraging results have been obtained with the inhibitors of PI3K/mTOR and MEK1/2 [26,44]. Consistently, in the current study, samples carrying NRAS or KRAS mutations proved highly sensitive to rapamycin and to the dual PI3K/ mTOR inhibitor BEZ235; the proliferation rate was also inhibited by the MEK1/2 inhibitor selumetinib, though at a lower extent.
These findings qualify FLT3, PI3K/mTOR and MEK1/2 inhibitors as alternative therapeutic agents for B-NEG ALL patients, especially in adults/elderly who frequently suffer of comorbidities and often cannot undergo intensive chemotherapeutic approaches.
The other set of mutations that captured our attention were those leading to the JAK/STAT signaling activation, being druggable and frequently altered in other poor prognosis ALL subsets [23-25, 34, 45-47]. Overall, we detected JAK/STAT mutations in 12.5% of B-NEG cases and, importantly, only AYA and adults were affected, with an incidence of 14.7% and 21%, respectively. This skewed distribution of lesions might be ascribed to the characteristics of our study population: in fact, unlike previously published papers describing high-risk and BCR-ABL1-like pediatric ALL cases [23-25, 34, 45-47], our cohort was not chosen for unfavorable clinical features nor for a peculiar GEP, and included all age cohorts. Therefore, our data indicate that unselected pediatric B-ALLs are rarely affected by these mutations, whereas their incidence increases with age. We provide direct evidence that considering only the mutational status of JAK/STAT members, regardless of GEP and concomitant rearrangements, it is possible to single out adult B-NEG ALLs experiencing a worse OS and DFS. As a matter of fact, the cohort of JAK/STAT-mutated only partly overlap with the BCR-ABL1-like profile, thus indicating that the mutational characterization may integrate the identification of BCR-ABL1-like cases, presently not completely elucidated nor straightforward.
The identification of JAK/STAT lesions bears also important therapeutic implications given the possibility of interfering with JAK members or downstream pathways as shown by previous experimental models [42,43,48] and our pilot experiments on primary cells.
In conclusion, our study, that was carried out in B-ALL cases without major molecular rearrangements www.impactjournals.com/oncotarget of all ages, implements the current knowledge on ALL. Firstly, we hereby show that the genetic background is significantly different between children, AYA and adults, with the number of deleterious lesions incrementing with age progression; this phenomenon might, at least in part, account for the different outcome between age cohorts. Second we show that RAS/RTK pathway mutations negatively impact on OS and DFS while they do not affect children's outcome. Along the same line, we show that JAK/STAT mutations confer a worse prognosis within the adult population. Finally, we show that by targeting these lesions we can reduce the in vitro leukemic cell proliferation, thus indicating that specific inhibitors might be integrated in the therapeutic backbone of ALL.

Sequencing analyses
Tumor and germline genomic DNA (gDNA, 3 µg) from the discovery panel patients and gDNA (1 µg) from the screening panel 1 tumor samples were enriched in protein coding sequences by the SureSelect Human All Exon 50Mb and the SureSelectXT2 Human All Exon kits, respectively (Agilent Technologies, Santa Clara, CA). Captured targets were subjected to massively parallel sequencing (HiSeq 2000 analyzer, Illumina, San Diego, CA) with the paired-end 2 × 100 bp read option (Tables  S7 and S8).
Mapping to the hg19/NCBI GRCh37 and identification of tumor-specific variants of the discovery panel was performed as described [39,49], while the analysis of screening panel 1 required more stringent criteria (Supplemental Material).

CNA and GEP analyses
CNA analysis of the discovery panel was performed with the Genome-Wide Human SNP Array 6.0 (Affymetrix, Santa Clara, CA; GSE67405) and processed as previously described [39]. Recurrent deletions and CRLF2 rearrangements were screened by the Salsa MLPA P335-A3 ALL-IKZF1 kit (MRC-Holland, Amsterdam, the Netherlands) and analyzed as reported [50,51].
GEP of B-NEG ALLs carrying JAK/STAT mutations was performed by using the HGU133 Plus 2.0 gene chips (Affymetrix) according to manufacturer's instructions and GEP analyses were carried out by dChip software (http:// www.hsph.harvard.edu/cli/complab/dchip/). In order to recognize BCR-ABL1-like cases, gene expression levels of a set of genes specific of the BCR-ABL1-like profile, selected in the context of a previously described dataset [52], were evaluated in JAK/STAT-mutated cases.

Statistical analyses
Patients' characteristics were compared by chisquared or Fisher's exact test for categorical variables. Differences in distributions were assessed by Wilcoxon test for continuous data. OS and DFS curves were estimated by Kaplan-Meier product-limit method and compared using the log-rank test. Multivariate analysis was performed with Cox proportional hazards regression model to adjust the effect of RAS and JAK/STAT pathway mutations for WBC and platelets counts, hemoglobin levels, age and gender, on OS and DFS. All tests were 2-sided, accepting p < 0.05 as index of statistical significance. All analyses relied on SAS v9.4 software.