Investigating the response of paediatric leukaemia‐propagating cells to BCL‐2 inhibitors

Summary Relapse of paediatric acute lymphoblastic leukaemia (ALL) may occur due to persistence of resistant cells with leukaemia‐propagating ability (LPC). In leukaemia, the balance of B‐cell lymphoma‐2 (BCL‐2) family proteins is disrupted, promoting survival of malignant cells and possibly LPC. A direct comparison of BCL‐2 inhibitors, navitoclax and venetoclax, was undertaken on LPC subpopulations from B‐cell precursor (BCP) and T‐cell ALL (T‐ALL) cases in vitro and in vivo. Responses were compared to BCL‐2 levels detected by microarray analyses and Western blotting. In vitro, both drugs were effective against most BCP‐ALL LPC, except CD34−/CD19− cells. In contrast, only navitoclax was effective in T‐ALL and CD34−/CD7− LPC were resistant to both drugs. In vivo, navitoclax was more effective than venetoclax, significantly improving survival of mice engrafted with BCP‐ and T‐ALL samples. Venetoclax was not particularly effective against T‐ALL cases in vivo. The proportions of CD34+/CD19−, CD34−/CD19− BCP‐ALL cells and CD34−/CD7− T‐ALL cells increased significantly following in vivo treatment. Expression of pro‐apoptotic BCL‐2 genes was lower in these subpopulations, which may explain the lack of sensitivity. These data demonstrate that some LPC were resistant to BCL‐2 inhibitors and sustained remission will require their use in combination with other therapeutics.


Introduction
Treatment of childhood acute lymphoblastic leukaemia (ALL) has been very successful in recent years. 1 However, relapse rates remain high in some subgroups. [2][3][4] This may be due to the persistence of resistant clones or the emergence of clones following therapy that were not detected at diagnosis. 5,6 We and others have previously identified ALL cells that have the ability to initiate and maintain the disease in murine models, known as leukaemia-propagating cells (LPC). [7][8][9][10][11][12][13][14][15] A common finding, regardless of subtype, was that expression of antigens commonly associated with B-cell precursor (BCP)-ALL (CD34, CD10, CD19) or T-ALL (CD4, CD7) was not a prerequisite for disease-propagating ability in vivo. Indeed, we demonstrated that the frequency of LPC was often higher in CD34 + /CD19 À and CD34 À subpopulations than in CD34 + /CD19 + cells in BCP-ALL, with similar findings using CD34 and CD7 in T-ALL cases. 12 Moreover, we have shown that some of these LPC populations, such as CD34 + /CD19 À , CD34 + /CD7 À and CD34 À cells, can be resistant to treatment in vivo and in vitro. 8,12,16 Hence, it is important to assess therapeutic agents on all LPC subpopulations that have capacity to cause relapse.
Bi-specific T-cell engagers and chimaeric antigen receptor T cells against CD19 have demonstrated the efficacy of immunotherapeutic approaches to treat ALL. [17][18][19][20] However, such therapies cannot target LPC that lack CD19 and there is evidence of tumour escape, through lack of targeting, lineage switching and masking of the CD19 epitope. 21,22 An alternative target of focus is B-cell lymphoma (BCL)-2, which has been shown to be overexpressed in >66% of BCP-ALL cases compared to normal bone marrow (NBM) donors 23 and is associated with oncogenesis in several cancers. 24,25 Consequently, BCL-2 inhibitors have been investigated in ALL. [26][27][28] ABT-263 (navitoclax) binds to and inhibits BCL-2, BCL-extra large (BCL-xL) and Bcl-2-like protein 2 (BCL-w) and kills cells in a BCL-2-associated X protein (BAX)-and BCL-2-antagonist/killer protein (BAK)dependent manner. 29,30 However, thrombocytopenia caused by BCL-xL inhibition, compromises its use. 31   (venetoclax) specifically inhibits BCL-2 allowing release of pro-apoptotic molecules. It does not bind to BCL-xL and consequently does not affect platelets. 32,33 Both navitoclax and venetoclax have been investigated in xenograft models of paediatric ALL with reports of extended survival of NOD.Cg-Prkdc scid /IL2rc tm1Wjl /SzJ (NSG) mice using navitoclax. 34,35 Responses to venetoclax in BCP-ALL xenografts were more modest, 36 although results improved when used in combination with vincristine and dexamethasone. 37 To date, there is no information on the effects of BCL-2 inhibitors in LPC subpopulations, so the effects on potentially drug-resistant cells are unknown. Consequently, we have undertaken a direct comparison of navitoclax and venetoclax on unsorted cells and LPC from primary BCPand T-ALL cases.

Samples
BM cells from children (median age, 7 years; range, 2-17) with BCP-(n = 20) and T-ALL (n = 11) at diagnosis or relapse were collected with approval of University Hospitals Bristol NHS Trust (Table I). NBM and cord blood (CB) were obtained from healthy donors. Mononuclear cells were cryopreserved as described. 15,16 Viability on thawing was 69Á6 AE 16Á1% (ALL) and 73Á0 AE 9Á3% (normal) samples.

In vitro drug sensitivity
ALL and CB cells were treated with increasing concentrations of navitoclax or venetoclax for up to 48 h (Stratech Scientific Limited, Newmarket, UK), according to the manufacturer's recommendations. [38][39][40] Apoptosis and viability were assessed by flow cytometry using annexin V-fluorescein isothiocyanate and PI. Half-maximal inhibitory concentration (IC 50 ) was calculated using non-linear regression analyses of dose response.

Microarray analyses
Gene expression analysis of BCP-, T-ALL and NBM samples was performed using Agilent Whole Human Genome Oligo Microarrays (Waldbronn, Germany), as previously described and detailed in supplementary files. 15 Microarray data are available in the ArrayExpress database (www.ebi.ac.uk/arrayex press), accession number E-MTAB-4006.
In vivo studies ALL was established in NSG mice using primary samples as described previously, and detailed in Data S1. 12,15 Briefly, once the level of human cells in murine peripheral blood (PB) was ≥0Á5%, animals were given BCL-2 inhibitors (100 mg/kg/day) for 21 days by oral gavage. Separate cohorts of mice were treated with standard chemotherapy or placebo. PB aspirates were monitored weekly by flow cytometry and animals maintained until they began to exhibit symptoms of disease.

Statistical analyses
Full details are provided in Data S1.

Comparison of in vitro sensitivity and BCL-2 protein expression
To determine whether these responses were associated with BCL-2 expression, flow cytometric analyses were performed. The proportion of BCL-2 + cells was higher in BCP-ALL cases (60Á0 AE 23Á0%) compared to T-ALL (32Á8 AE 31Á2%, P = 0Á1) and normal cells (27Á2 AE 5Á0%, P = 0Á08, Fig 2A). Likewise, BCL-2 median fluorescence intensity (MFI) was significantly higher in BCP-ALL (median 190, range 35-397) than in T-ALL (62, range 23-109, P = 0Á02) and >2-fold higher than normal cells (81, range 75-111, P = 0Á09, Fig 2B). Western blotting confirmed BCL-2 expression was higher in BCP-than T-ALL cases and at least equivalent to that in normal CD19 + B cells (Fig 2C,D). Of the anti-apoptotic proteins, BCL-xL was the most prominent. Myeloid cell leukaemia 1 protein (MCL1 was present in all cases but did not differ from NBM cells, except in pt. 11, where higher levels were observed. BCL-w could not be detected in this cohort. Interestingly, pt. 2, who had the highest BCL-2, had the lowest sensitivity to both drugs. In T-ALL cases, BCL-2 levels were lower than normal T cells in 5/6 cases and comparable in pt. 29 ( Fig 2D). The prosurvival proteins BCL-xL and MCL1 were expressed in most T-ALL cases but not in normal CD7 + cells. Amongst the samples that were more

Differential expression of BCL-2 molecules
As navitoclax and venetoclax both affect other BCL-2 family members, gene expression levels of all anti-apoptotic and pro-apoptotic molecules were determined in a subset of cases and data compared to haemopoietic stem cells (HSC). There was no clear trend in upregulated or downregulated genes in ALL samples (Table II, Figures S2-S4). Some LPC had higher levels of anti-apoptotic genes (BCL-2 alpha, BCL-xL and BFL1/A1) than HSC. However, higher levels of proapoptotic genes including BMF (P ≤ 0Á01), BAD and BIM (P ≤ 0Á05), BAX (P = 0Á05) and BCL2L13 were observed  (Table SIII).
Navitoclax is superior to venetoclax in vivo NSG mice with established leukaemia, from low, intermediate and high risk cases, were treated with navitoclax, venetoclax and, in some cases, standard chemotherapeutic agents (dexamethasone, vincristine, L-asparaginase AE daunorubicin). In BCP cases, navitoclax treatment decreased leukaemia burden or maintained it at the same level for 7-29 days from treatment start (P < 0Á0001, Fig 4A). Survival was extended up to 54 days. Using venetoclax, disease burden remained significantly below the levels in placebo groups in pts. 9 & 14 (P < 0Á0001) during treatment and for a further seven days in pt. 14. However, in pt. 4 (Ph + , low risk) disease burden significantly decreased to 2Á0 AE 2Á0% (P < 0Á0001) and progression was significantly delayed by three weeks. Survival was improved with both drugs, particularly in pt. 4 (54 days, P = 0Á01) with more modest improvements in pts. 9 & 14 (four days, P ≤ 0Á02). These responses were similar to those observed using standard induction therapy for 28 days.
In mice engrafted with T-ALL samples, navitoclax reduced disease burden and significantly delayed progression in all cases compared to controls (P ≤ 0Á001, Fig 4B). Survival in navitoclax-treated animals was extended by up to 40 days but disease burden was more effectively reduced using standard chemotherapeutics. Venetoclax had no effect on this T-ALL cohort.
Analyses of BM from terminated animals revealed an increase in the proportion of CD34 + /CD19 À cells following treatment with navitoclax (8-fold, P ≤ 0Á02) and venetoclax (13-fold, P ≤ 0Á03) in pt. 4 compared to the proportions in the diagnostic sample (Table III). In pts. 9 & 14, the proportions of CD34 À /CD19 À cells increased significantly following treatment with BCL-2 inhibitors (P ≤ 0Á03) and following standard therapy, in pt 14 (P = 0Á002). In T-ALL cases, the CD34 À / CD7 À subpopulation, which was just detectable at diagnosis, had increased significantly in pt. 22, following treatment with both inhibitors (P ≤ 0Á01). Likewise, in pt. 25, the majority of cells recovered from the BM were CD34 À /CD7 À , representing a >15-fold increase (P < 0Á004) in the proportion of these cells compared to the primary sample inoculated, while the CD34 + / CD7 + and CD34 + /CD7 À cells were significantly depleted (P < 0Á02). In pt. 31, in contrast, the CD7 À subpopulations were most affected by treatment.

Discussion
Over the last decade there has been considerable interest in BCL-2 inhibitors for the treatment of cancers. However, their efficacy has not been assessed on LPC subpopulations, which have variable repopulating capacity, [10][11][12][13] and some are refractory to treatment with conventional (dexamethasone, vincristine) and novel therapeutics, such as parthenolide and Hsp90 inhibitors. 8,12,16 This is the first report directly comparing the sensitivity of paediatric BCP-and T-ALL samples and their respective LPC subpopulations to navitoclax and venetoclax.
BCL-2, BCL-xL and other BCL-2 family proteins were differentially expressed in ALL pts., resulting in a range of responses to BCL-2 inhibitors. In BCP-ALL, 8/11 samples were sensitive to both drugs in vitro and three cases responded to navitoclax only. In contrast, all six T-ALL cases responded to navitoclax but only one case responded to venetoclax.
Most BCP LPC subpopulations responded to both drugs, but more so navitoclax, with the exception of CD34 À /CD19 À cells. T-ALL LPC were also responsive to navitoclax, although CD34 À /CD7 À LPC were the least sensitive. In contrast, venetoclax was not effective against T-ALL LPC, with the exception of one pt. where the drug killed most LPC,  9,11,14,15,19,20) were sorted into four subpopulations using antibodies against CD34 and CD19. Each subpopulation was treated with navitoclax or venetoclax for up to 48 h and cell survival was measured by flow cytometry using annexin V and propidium iodide (PI). (B) Cells from six T-cell (T-)ALL cases (21,25,(27)(28)(29)(30) were sorted into four subpopulations using antibodies against CD34 and CD7 and treated with BCL-2 inhibitors, as above. Unsorted cells from five T-ALL cases were resistant to venetoclax (pts. 21,25,27,28,30) and 1one was sensitive (pt. 29, depicted separately). Viability is shown as mean AE SD across samples, expressed as a percentage of untreated controls. [Colour figure can be viewed at wileyonlinelibrary.com] with limited effects on the CD34 À /CD7 À subpopulation. Our varied findings in vitro are in agreement with reports on the activity of venetoclax in ALL cell lines and patient derived xenograft (PDX) samples, 36,41 acute myeloid leukaemia cell lines, 36,42 and of navitoclax in ALL cell lines. 34 The lack of response of T-ALL cases to venetoclax also concurs with the findings by of Changaile et al., who reported that only the early T-cell precursor (ETP) subtype was sensitive to venetoclax, indicating this phenotype may be BCL-2-dependent. 43 The reason why a sample responds to navitoclax and not venetoclax may be due to varied expression of BCL-2 targets. All BCP cases investigated had high levels of BCL-2, BCL-xL and MCL1, in accordance with previous reports, 26,35 and responded to navitoclax. Likewise, T-ALL cases that responded to navitoclax had higher levels of BCL-xL and/or MCL1. BCL-2 expression was higher in BCP-than in T-ALL samples, which may explain why more BCP-ALL samples responded to venetoclax. Indeed, venetoclax sensitivity in BCP-ALL has been linked to high expression of BCL-2 and lower expression of MCL1. 41 In the present study, venetoclax resistance in T-ALL may be associated with low levels of BCL-2. 36,37 The single T-ALL case that responded to both drugs was the only one with high BCL-2, although it was not ETP subtype.
Navitoclax has been reported to have broad efficacy in xenograft models of paediatric ALL, with survival extended by 25-30 days, 34,35 with more modest effects using venetoclax. 36 In our direct comparison, the majority of samples responded to navitoclax in vivo, with extended survival observed in both subtypes. While venetoclax improved survival in BCP engrafted mice, it was ineffective on those Data shows the immunophenotype of primary patient samples at diagnosis and the immunophenotype of cells recovered from the bone marrow of NSG mice engrafted with these patient samples then treated with navitoclax, venetoclax or standard chemotherapeutics (Std Chem: dexamethasone, vincristine, L-asparaginase AE daunorubicin). Bold text indicates significant difference from inoculated primary sample (*, P < 0Á05; **, P ≤ 0Á01; ***, P ≤ 0Á001). ALL, acute lymphoblastic leukaemia; BCP, B-cell precursor; T-ALL, T-cell ALL.  Red asterisks correspond to comparisons between navitoclax-and placebo-treated mice and blue asterisks between venetoclax-and placebo-treated mice. *, P ≤ 0Á05; **, P ≤ 0Á01; ***, P ≤ 0Á001; ****, P ≤ 0Á0001. engrafted with T-ALL. In accordance with reports to date, disease burden increased on withdrawal of treatment. [34][35][36][37] The lack of response of T-ALL cases to venetoclax concurs with previous reports, 36 although Seyfried et al. reported delayed progression in one of three PDX samples. 41 The superior effects of navitoclax in vivo support the findings of Khaw et al., 36 and concur with our findings in vitro. However, in two BCP-ALL cases (intermediate and high risk), promising responses to both drugs in vitro were not confirmed in vivo, where effects were transient at best. Since the LPC populations only had moderate responses to navitoclax in vitro, with T-ALL LPC largely unaffected by venetoclax, and sustained remission in treated mice was lacking, direct in vivo investigations using these inhibitors on LPC were not pursued. However, the proportions of LPC recovered from murine BM were compared to those in the inoculated patient samples. Interestingly, in intermediate and high risk BCP cases, the proportions of CD34 À / CD19 À cells increased significantly following treatment with BCL-2 inhibitors and standard therapeutics, while CD34 + / CD19 À cells increased in the low risk case. This concurs with our findings in vitro and provides robust functional evidence of resistance in the CD34 À /CD19 À subpopulation. Likewise, a significant increase in CD34 À /CD7 À cells was observed in T-ALL cases, analogous to the poor in vitro response observed in this subpopulation, with a corresponding decrease in the proportion of CD7 + cells. We have previously shown that some LPC subpopulations (CD34 + / CD19 À in BCP-ALL and CD34 + /CD7 À and CD34 À in T-ALL), that have self-renewal capacity over several generations in NSG mice, are resistant to therapy. 8,12,16 The results in this study add to the evidence of refractoriness in these subpopulations, which would also be resistant to chimaeric antigen receptor (CAR) T cells directed at CD19 or CD7. Therefore, alternative therapeutic options will be required to eradicate these subpopulations.
Collective results indicate that BH3 mimetics may not be effective as single agents for paediatric leukaemias, especially as the LPC subpopulations expressed BCL-2 genes at various levels. An alternative approach may be to combine BH-3 mimetics with inhibitors of survival pathways (e.g. tyrosine kinase inhibitors) to increase apoptosis with potentially fewer side effects than using conventional cytotoxic therapies.

Supporting Information
Additional supporting information may be found online in the Supporting Information section at the end of the article.
Data S1. Supplemental methods. Table SIII. Median fold change in expression of genes in acute lymphoblastic leukaemia (ALL) LPC compared to normal HSC. Fig S1. Gating strategy for sorting leukaemia subpopulations. Fig S2. Expression of anti-apoptotic and pro-apoptotic BCL-2 genes in BCP acute lymphoblastic leukaemia (ALL) and normal cells. Fig S3. Microarray analysis of anti-apoptotic and proapoptotic BCL-2 genes in T-cell acute lymphoblastic leukaemia (T-ALL) and normal cells. Fig S4. Functional grouping of differentially expressed genes.