SRC/ABL inhibition disrupts CRLF2-driven signaling to induce cell death in B-cell acute lymphoblastic leukemia

Abstract

Children with B-cell precursor acute lymphoblastic leukemia (BCP-ALL) overexpressing the CRLF2 gene (hiCRLF2) have poor prognosis. CRLF2 protein overexpression leads to activated JAK/STAT signaling and trials are underway using JAK inhibitors to overcome treatment failure. Pre-clinical studies indicated limited efficacy of single JAK inhibitors, thus additional pathways must be targeted in hiCRLF2 cells. To identify additional activated networks, we used single-cell mass cytometry to examine 15 BCP-ALL primary patient samples. We uncovered a coordinated signaling network downstream of CRLF2 characterized by co-activation of JAK/STAT, PI3K, and CREB pathways. This CRLF2-driven network could be more effectively disrupted by SRC/ABL inhibition than single-agent JAK or PI3K inhibition, and this could be demonstrated even in primary minimal residual disease (MRD) cells. Our study suggests SCR/ABL inhibition as effective in disrupting the cooperative functional networks present in hiCRLF2 BCP-ALL patients, supporting further investigation of this strategy in pre-clinical studies.

Keywords: acute lymphoblastic leukemia; cell signaling; mass cytometry; minimal residual disease; signal transduction inhibitors.

Figure 1. TSLP stimulation induces simultaneous activation of multiple signaling pathways in hiCRLF2 BCP-ALL primary samples

(A) Overview of TSLP-induced signaling in blast cells (gated as shown in Supplementary Figure 1) from BCP-ALL primary samples (column 1 - 6 hiCRLF2 patients; column 7 - 12 loCRLF2 patients). Each row represents the arcsinh ratio of a phosphoprotein in TSLP-treated cells over baseline levels from unstimulated cells (decreased phosphorylation (blue) versus increased phosphorylation (yellow) compared to their basal level). Asterisks indicate statistically significant differences between hiCRLF2 and loCRLF2 phosphoproteins, calculated by using an unpaired two-sided student’s t test (^* p<0.5, ^** p<0.01, ^*** p<0.001). (B) Heatmap of the DREMI scores summarizing the signaling connections present within the TSLP-activated phosphoproteins in the hiCRLF2 patients cohort. The red boxes highlight the strongest connections that are also reported in the diagram on the left.

Figure 2. SRC/ABL inhibition targets coordinated TSLP-induced signaling

(A) Schematic representation of TSLPR heterodimer signaling transduction in childhood hiCRLF2 BCP-ALL and agents used in this study to target activated signaling nodes. The proteins not included in our CyTOF panel and the signaling connections not tested by DREMI are shadowed and grey. (B) Effects on TSLP-activated phosphoproteins (pSTAT5, prpS6, pERK and pCREB) in hiCRLF2 and loCRLF2 BCP-ALL primary cells following short-term single agent kinase inhibition. Each row represents a treatment condition and each phosphoprotein is showed as the arcsinh ratio of treated (TSLP or agent+TSLP) condition over baseline levels. (C) Overview of the DREMI scores across treatment conditions in the hiCRLF2 cohort. The red box highlights the effect of SRC/ABLi in disrupting all the pairwise signaling relationships present within the 4 activated phosphoproteins comprising the TSLP-driven network. Arrow directs to DREVI visualization of the disruption of TSLP dependencies between pCREB → pSTAT5 and pCREB → prpS6. SRCi=dasatinib, JAKi=ruxolitinib, PI3Ki=NVP-BEZ235, A10=A10 anti-CRLF2 mAb.

Figure 3. Signaling heterogeneity exists within hiCRLF2 cells

(A) Left: contour plot of pSTAT5 activation after TSLP stimulation in a representative hiCRLF2 case. The red rectangle highlights the pSTAT5 negative cell subset while the black rectangle denotes the pSTAT5 positive cell subset. Right: prpS6 levels in pSTAT5 non-responsive cells. Each column represents a hiCRLF2 patient and each row the level of prpS6 in the different treatment conditions showed as the arcsinh ratio of the treated condition over baseline levels. (B) Graphic representation (dark purple) of the only pairwise connection (pCREB → prpS6) present in the pSTAT5 non-responsive cells assessed by DREMI analysis. The absent connections in this cell subset are shadowed. (C) DREVI visualization of conditional density function for pCREB → prpS6 present in the pSTAT5 not responsive cells. Treatment effects on this interaction are showed in the respective condition plots and in each plot are reported the sigmoidal response functions and the DREMI score. SRCi=dasatinib, JAKi=ruxolitinib, PI3Ki=NVP-BEZ235, A10=A10 anti-CRLF2 mAb.

Figure 4. Optimal disruption of the TSLP-driven signaling network results in decreased cellular survival

(A) Percentage of live cells in single or combined treatments in Ba/F3 hiCRLF2-IL7Rɑ expressing cells, human REH cells (loCRLF2), human MHH-CALL4 cells (hiCRLF2/JAK2 mutated) and in patient-derived primary cells (Pt#2 hiCRLF2/JAK2 mutated). Cells were cultured in presence of TSLP (10 ng/mL) and treated with agents. Combined treatments were compared to the untreated condition and single agent treated cells via one-way ANOVA using Tukey’s post-test for multiple comparisons with ɑ=0.05. The statistics in the figure is reported only for the most effective treatment (SRC/ABLi) compared to either the untreated condition or the combination treatments. ^* p<0.5, ^** p<0.01, ^*** p<0.001, ^**** p<0.0001. (B) Histogram overlay of pSTAT5, prpS6 and pCREB levels after treatment of MHH-CALL4 cells in vitro. The histograms are colored based on the median levels. (C) Heatmap representation of the DREMI score in MHH-CALL4 treated cells showing the pairwise dependencies present in the untreated condition and strongly affected by SRC/ABLi treatment as highlighted by the red box. SRCi=dasatinib, JAKi=ruxolitinib, PI3Ki=NVP-BEZ235, A10=A10 anti-CRLF2 mAb.

Figure 5. Primary chemotherapy-resistant cells display the CRLF2-driven pro-survival network

(A) Leukemia response to induction therapy in 3 hiCRLF2 BCP-ALL primary samples (rows) at 3 different time-points (columns). According to the MRD protocol, at diagnosis and at Day 15, bone marrow samples were analyzed, while, at Day 8, analysis was performed on peripheral blood cells. (B) Histogram overlays of prpS6 and pCREB in 3 hiCRLF2 BCP-ALL primary samples at diagnosis, Day 8 and Day 15 post induction initiation. Histograms are colored based on the raw median of prpS6 and pCREB mean at diagnosis, Day 8 and Day 15. (C) Graphic representation of the only pairwise connections (pCREB → prpS6), present at diagnosis (shadowed circles), and enriched at Day 8 and Day 15 (dark purple circles) in Pt#13. The DREMI scores at each time-point and for each connection are reported. (D) Heatmaps of pSTAT5, prpS6, pERK and pCREB levels (columns of the heatmaps) in 5 different conditions (rows) at diagnosis, Day 8 and Day 15 in one hiCRLF2 patient (Pt #13). The heatmap in the SRC/ABLi condition at Day 8 (showed separately) was obtained by a different experiment and the level of the phosphoproteins were normalized based on the basal state of that experiment. Heatmaps are colored with a scale from blue to yellow based on phosphoproteins levels calculated as arcsinh ratio of their levels in the treated conditions (TSLP, JAKi, A10 mAb or SRCi) over baseline levels. (E) Heatmap representation of the DREMI score of MRD cells of Pt#13 at Day 8 in the following conditions: basal, TSLP alone, SRCi+TSLP. The red boxes highlight the connections present in the MRD cells and their disruption after treatment with the SRCi. SRCi=dasatinib, JAKi=ruxolitinib, PI3Ki=NVP-BEZ235, A10=A10 anti-CRLF2 mAb.