High incidence of Philadelphia chromosome-like acute lymphoblastic leukemia in older adults with B-ALL.

Treatment of patients with Philadelphia chromosome-positive B-cell acute lymphoblastic leukemia (Ph+ B-ALL) with tyrosine kinase inhibitors (TKIs) such as imatinib and dasatinib has improved survival in younger adults and allowed de-escalation of conventional chemotherapy, particularly in older adults.[1] Similar approaches may also benefit patients with the Philadelphia chromosome-like (Ph-like) B-ALL subtype recently described in children and adolescents/young adults (AYAs)[2, 3]. Ph-like ALL has a “kinase-activated” gene expression profile similar to Ph+ ALL, but distinguishes itself by absence of BCR-ABL1 fusion and presence of activating mutations in kinase- and cytokine receptor-associated signaling pathways.[3, 4] In children and AYAs, approximately 50% of Ph-like ALL activates JAK/STAT and PI3K/mTOR signaling via rearrangement of cytokine receptor like factor 2 (CRLF2-R) with common co-occurrence of JAK2 or JAK1 point mutations. An additional 10–15% of Ph-like ALL harbors JAK2 or EPOR rearrangements, while another 15–20% has translocations or fusions involving ABL1, ABL2, CSF1R, or PDGFRB (ABL-class lesions).[3, 5] As in Ph+ ALL, deletions of IKZF1, CDKN2A/B, and other lymphoid-associated transcription factors are common.[4, 6] While Ph-like ALL comprises 20–25% of B-ALL in AYAs and is associated with high relapse rates and poor survival,[2, 4] the frequency, biology, genetics, and clinical features of Ph-like ALL in older populations remain incompletely defined.[7]

Ph+ and Ph-like ALL are rare in young children, but increase in frequency during adolescence.[3] Ph+ ALL incidence rises further in adulthood, peaking at >60 years. Although younger adults can be treated with pediatric protocol-inspired chemotherapy, such intensive regimens are poorly tolerated in middle-aged and older adults.[8] Robust anti-leukemia activity of JAK inhibition (e.g., ruxolitinib) and SRC/ABL inhibition (e.g., dasatinib) has been demonstrated in preclinical models of JAK2/EPOR-mutant and ABL/PDGFRB-mutant Ph-like ALL, respectively.[3, 5, 9] Clinical testing of kinase inhibitors with chemotherapy is now ongoing for children with specific genetic subtypes of Ph-like ALL (NCT02723994, NCT02883049). Improved delineation of the genetic drivers in adult ALL may similarly facilitate development of molecularly-targeted treatment approaches for appropriate patient subsets that can minimize toxicity while retaining potent anti-leukemic efficacy.

Given these therapeutic implications, we sought to define the incidence and clinical characteristics of adult Ph-like ALL via expression profiling of 93 unselected B-ALL specimens. In particular, demonstration of a sizable older adult population with Ph-like ALL would support development of TKI-based therapies alone or with lower-intensity cytotoxic therapies.[1] Bone marrow or peripheral blood specimens were obtained from consenting patients with newly-diagnosed B-ALL via institutional review board-approved research protocols at the University of Pennsylvania, University of Michigan, and Children’s Hospital of Philadelphia (signaling analyses only) in accordance with the Declaration of Helsinki. Patient age, sex, race/ethnicity, diagnostic white blood cell (WBC) count, cytogenetic and molecular testing data, treatment, and clinical outcomes were tabulated when available. Patients with antecedent chronic myeloid leukemia in lymphoid blast crisis were excluded.

Nucleic acids were extracted from B-ALL specimens according to manufacturer’s instructions (Qiagen). Cell pellets for fluorescence in situ hybridization (FISH) assays were prepared as described.[10] RNA specimens were first profiled via a 15-gene low density microarray (LDA) classifier that identifies a kinase-activated gene expression signature associated with both Ph+ and Ph-like ALL.[11] Ph+ specimens were characterized as BCR-ABL1 p190 or p210 breakpoints. LDA signature-positive specimens without BCR-ABL1 fusion (Ph-like ALL) were categorized as CRLF2-overexpressing or non-overexpressing via LDA quantification of expression levels.[11] CRLF2-overexpressing specimens were subsequently assessed for IGH-CRLF2 and P2RY8-CRLF2 rearrangements and JAK1 and JAK2 mutations via FISH, polymerase chain reaction (PCR), or reverse transcriptase (RT)-PCR with confirmatory Sanger sequencing as described.[12] Non-CRLF2-overexpressing Ph-like specimens were tested for 39 known kinase fusions involving 3′ partner genes ABL1, ABL2, CSF1R, JAK2, NTRK3, PDGFRB, and TYK2 via multiplex RT-PCR with confirmatory sequencing and/or FISH and for EPOR and other rearrangements via candidate next-generation sequencing (Archer FusionPlex).[5, 12]

To confirm activated kinase signaling previously reported in childhood Ph-like ALL, 31 adult ALL specimens were incubated in vitro with human cytokines, ruxolitinib, or dasatinib and fixed and permeabilized as described.[13, 14] Phosphorylated CrkL, STAT5, and/or S6 were measured by leukemia cell-specific phosphoflow cytometry to identify basal signaling activation and TKI-induced signaling inhibition and analyzed with Cytobank as described.[13, 14]

Adult Ph-like ALL patient-derived xenograft (PDX) models were established in NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice as described.[14] Engrafted animals (≥5% human CD19+/CD45+ ALL in peripheral blood) were randomized to vehicle, ruxolitinib 90 mg/kg (LC Labs), or dasatinib 10 mg/kg (LC Labs) treatment twice daily via oral gavage for 28 days to assess in vivo inhibition of leukemia proliferation.[3, 9] Human leukemia in spleens from sacrificed animals at end of treatment was quantified by flow cytometry as described.[9, 14]

Means, medians, survival, and multivariate analyses for clinical data were calculated using Prism (GraphPad) or STATA. Unpaired two-sided t-tests were used to compare leukemia cell counts in spleens from vehicle- and TKI-treated PDX mice.

In these studies, we profiled 93 diagnostic B-ALL specimens for the Ph-like gene expression signature and associated genetic alterations (Supplemental Table 1). Four samples failed LDA testing due to insufficient material. For the remaining 89 specimens, median patient age at diagnosis was 46 years (range 18–88), 52.4% of patients were male, and median WBC count was 35.1 × 103 cells/uL blood (range 1–436) (Supplemental Table 2). Chemotherapy, TKI, and hematopoietic stem cell transplant (HSCT) status are delineated in Supplemental Table 1.

LDA analysis identified 51/89 patients (57.3%) with a kinase-active expression signature. Of these, 33 (37.1% of the entire cohort) were BCR-ABL1-rearranged (Ph+), and 18 (20.2%) were BCR-ABL1-negative by FISH or RT-PCR (Ph-like) (Figure 1A). Among the 38 non-Ph+/non-Ph-like specimens, 13 had KMT2A (MLL) rearrangements. Patients with Ph+ ALL and Ph-like ALL had similar sex distribution and levels of hyperleukocytosis (Supplemental Table 2).

Figure 1

Ph-like ALL occurs commonly in adults with B-ALL and is associated with poor outcomes

(A) Incidence of Ph-like and other genetic subtypes in adults with B-ALL with subset analyses of patients ≥40 and <40 years of age. Age was not available for two patients. (B) Kaplan-Meier survival analysis with log-rank comparison test of patients with B-ALL for whom outcomes data were available (n=81) with indicated p-value. When appropriate, survival was censored at last known clinical evaluation.

Fourteen of the 18 (77.8%) Ph-like ALL specimens were CRLF2-R, and seven of these had concomitant JAK2 mutations (Supplemental Table 1). Of the four non-CRLF2-R Ph-like ALL specimens, one had a cytogenetically-cryptic EPOR rearrangement (UP_240).[5] Ph-like kinase alterations were not detected in specimen UP_5049 by RT-PCR or FusionPlex assays, although next-generation sequencing identified FLT3 (fms-related tyrosine kinase 3) internal tandem duplication that may have caused its activated kinase signature (Supplemental Table 1). The remaining two specimens (UM_06, UM_07) had insufficient material for further genetic characterization.

Overall survival was assessed for patients for whom outcomes data were available (n=81). Median survival was 1.6, 2.7, 0.7, and 2.7 years for Ph-like, Ph+, KMT2A-R, and other B-ALL subsets, respectively (Figure 1B). Log-rank testing demonstrated a trend towards differential survival among the ALL subgroups (p=0.0798), although small patient numbers precluded definitive statistical analysis. Thirty-one patients underwent HSCT, including eight patients with Ph-like ALL (Supplemental Tables 1 and 2).

Concordant with prior observations in childhood Ph-like ALL,[13] phosphoflow cytometric analysis of adult CRLF2-R Ph-like ALL specimens demonstrated basal and/or cytokine-inducible pSTAT5 and pS6, which were inhibited by ruxolitinib (Figure 2A and Supplemental Figure 1). Analogously, ABL1-rearranged Ph-like and Ph+ ALL cells had basal pCrkL activation, which was inhibited by dasatinib (Figure 2A). We further observed significant inhibition of leukemia proliferation in adult CRLF2-R/JAK2-mutant Ph-like ALL PDX models treated with ruxolitinib and in ABL1-rearranged models treated with dasatinib (Figure 2B). These results, consistent with earlier pediatric studies,[3, 9, 14] confirm that JAK- and ABL-targeted TKI therapies also have therapeutic potential in relevant subsets of adult Ph-like ALL.

Figure 2

Ph-like ALL in adults demonstrates signaling hyperactivation and sensitivity to tyrosine kinase inhibition

(A) Thirty-one primary adult B-ALL specimens with the designated genetic alterations (n=9 CRLF2-R Ph-like, n= 2 ABL1-R Ph-like, n=10 Ph+, n=KMT2A-R, n=5 other B-ALL) were incubated in vitro with 25 ng/mL thymic stromal lymphopoietin (TSLP; ligand for CRLF2 receptor), 100 nM dasatinib, or 1 uM ruxolitinib for 30 minutes (TSLP) or 60 minutes (inhibitors) at 37 °C, then fixed with paraformaldehyde and permeabilized with methanol.[13] Stimulated or inhibited intracellular phosphoproteins within human CD19+ CD45+ (and CD10+ and TSLPR+ if applicable) ALL cells were analyzed by phosphoflow cytometry. Data are depicted as %change of basal median fluorescent intensity (MFI) with colorimetric normalization to basal MFI levels (grey) for each phosphoprotein in each specimen and summarized as a heatmap. Red and blue indicate increased and decreased phosphorylation, respectively. Gating strategy and representative histograms for individual samples are displayed in Supplemental Figure 1. (B) Patient-derived xenograft (PDX) models of adult Ph-like ALL established in 6–8 week old male and female NSG mice were randomized to treatment with vehicle, ruxolitinib 90 mg/kg twice daily, or dasatinib 10 mg/kg twice daily (n=5 mice/treatment cohort) via oral gavage for 28 days after initial demonstration of ≥5% human CD19+CD45+ ALL in murine peripheral blood.[9, 14] Animals were monitored daily for physical signs of disease and weekly by venous bleeding for flow cytometry (FC) quantification of leukemia in peripheral blood during treatment. Human ALL cells in murine spleens from sacrificed animals were also quantified by FC as above after four weeks of treatment at planned study endpoint. Ruxolitinib or dasatinib markedly reduced total human ALL cell number (y-axis) versus vehicle-treated controls in Ph-like ALL PDX models harboring CRLF2 or ABL1 rearrangement, respectively (p<0.0001 via two-tailed t-test). All animal studies were conducted on CHOP Institutional Animal Care and Usage Committee-approved research protocols in compliance with NIH and ARRIVE guidelines. Animals were monitored daily for physical signs of illness and sacrificed at planned study endpoint or sooner if ill-appearing. Note that clinical data associated with specimens from young adults at the Children’s Hospital of Philadelphia (CHOP) used for functional studies were not included in the main analyses in Figure 1 or Supplemental Tables. UP = University of Pennsylvania, UM = University of Michigan.

A potential bias of our study is possible preferential banking of ALL specimens from patients with higher WBC counts, which may have enriched detection of Ph+, Ph-like, and other high-risk subtypes of ALL. Multivariate analyses of pediatric/AYA B-ALL patients suggest that presenting WBC ≥100,000 cells/uL is highly suggestive of Ph-like ALL and portends poor prognosis, although leukocytosis alone remains an imperfect clinical predictor.[2]

Over-representation of GATA3 ALL susceptibility variants and IKZF1 deletions has been reported in pediatric/AYA Ph-like ALL with greatest incidence in Hispanic/Latino and Native American patients.[15] While we did not specifically investigate these polymorphisms or other copy number alterations, 8.3% of our cohort was Hispanic/Latino, which may have also influenced the Ph-like frequency identified in our population.

In summary, we identified a 20.2% incidence of Ph-like ALL in adults and, importantly, demonstrate that >50% of adult B-ALL is associated with activated kinase signaling via BCR-ABL1 or BCR-ABL1-like rearrangements. While frequent kinase fusions also occur in pediatric/AYA Ph-like ALL, our study and a recent European study report CRLF2 rearrangement as the predominant genetic lesion in adult Ph-like ALL.[7] We thus propose that routine ALL diagnostic testing algorithms include flow cytometric assessment of surface CRLF2 (TSLPR) overexpression with subsequent genetic confirmation of specific CRLF2 rearrangements.[13]

Our study used LDA testing as a “gold standard” to define a Ph-like gene expression signature in adult ALL.[11, 12] This validated test derived from a gene expression array classification defining Ph-like ALL rapidly identifies both Ph+ and Ph-like ALL and enables prioritization of targeted downstream molecular characterization of Ph-like specimens. While our study analyzed a relatively small patient cohort, we nonetheless demonstrate a substantial population of adults ≥40 years old with Ph-like ALL and poor outcomes who may benefit from TKI addition to appropriately dose-intensive chemotherapy. Clinical translation of these findings will potentially transform therapeutic approaches and improve clinical outcomes for older adults with Ph-like ALL.