Although leukemia alters BM niches, whether these changes directly affect hematopoietic cell production has not been carefully studied. To determine if and which hematopoietic cell lineages are affected by leukemia, we transplanted 3 million pre-B-cell precursor ALL cells expressing the BCR-ABL1 oncogene (from here on referred as ALL; BCR-ABL1 reported by YFP expression) into non-irradiated C57Bl6/J mice and analyzed its impact on lymphoid, myeloid, and erythroid cell production over time. As expected, ALL cells expanded rapidly in BM (Figure 1A). Conversely, non-leukemic developing B cells and mature recirculating B cells declined sharply 2 weeks after ALL transplantation (Figure 1B). Monocyte numbers reduced between the first and second weeks by four- to fivefold, but cell numbers recovered to normal levels at 3 weeks (Figure 1C). This contrasted with a moderate twofold decline in neutrophil numbers at 2 weeks that remained stable until 3 weeks (Figure 1D). Changes in immature erythrocytes (Ter119+CD71+ cells) were similar to the reductions seen in B cell progenitors: erythroid cells progressively reduced at 6, 14, and 21 days after ALL transplantation, reaching >10-fold reductions at 3 weeks (Figure 1E). In summary, ALL expansion induces a strong decline in lymphopoiesis and erythropoiesis while their impact on myeloid cell production is modest.

In previous studies we noted that transplanted ALL cells and Artemis-deficient (pre-leukemic) pre-B cells led to IL7 and CXCL12 downregulation in MSCs (Fistonich et al., 2018), which could explain the negative impact of ALL in non-malignant lymphopoiesis. While the mechanism(s) responsible for IL7 and CXCL12 downregulation remained undefined, earlier studies suggested a role for LTβR signaling in BM stromal cells in development of some lymphoid lineages (Wu et al., 2001; Kim et al., 2014). In recent studies, we found that MSCs express LTβR and that LTβR signaling controls IL7 expression in vivo (Zehentmeier et al., 2022). Furthermore, when mRNA levels of LTB were analyzed in pediatric samples of B-ALL (Children’s Oncology Group Study 9906 for High-Risk Pediatric ALL) and associated with clinical outcome at the time of diagnosis, we noted an inverse correlation between LTB transcript abundance and relapse-free survival that reached statistical significance (Figure 2—figure supplement 1A). These observations led us to hypothesize that leukemic cells express LTβR ligands and induce LTβR signaling in MSCs in vivo. In mice, BCR-ABL1 expressing pre-B ALL cells also express higher LTα/LTβ amounts than non-leukemic pre-B, immature, and mature B cells (Figure 2A and B). The presence of ALL cells in the BM environment did not change LTα and LTβ expression on non-leukemic host pre-B cells (Figure 2A). Importantly, when mouse ALL cells were engineered to over-express LTα/LTβ, these ALLs induced stronger IL7 downregulation in BM MSCs and were lethal more quickly than empty vector transduced ALL cells (Figure 2—figure supplement 1B–E). Combined, these studies suggest a pathogenic role for the LTβR pathway in leukemia progression.

To test if LTβR signaling impacts ALL growth and non-malignant hematopoiesis, we transplanted 3 million ALL cells into WT syngeneic recipient mice (C57BL6/J) treated weekly with a soluble LTβR-Ig decoy (a fusion between LTβR ectodomain and the Fc domain of a mouse IgG1 recognizing Hen Egg Lysozyme) or with control Hel-Ig. Transplanted ALL cells reduced lymphopoiesis significantly, which was reverted with LTβR-Ig treatment (Figure 2C and D). In contrast, LTβR signaling blockade did not restore erythropoiesis or myelopoiesis (Figure 2E–G). Importantly, ALL growth was significantly reduced at 2 weeks (Figure 2H), which reflected in a small but significant extension of mouse survival (Figure 2I).

To gain further insight into the mechanisms used by ALL cells for reducing non-malignant hematopoiesis, we analyzed the MSC transcriptome in homeostasis, during ALL expansion, and in mice with ALL but treated with LTβR-Ig. To identify gene expression differences between the three groups, we performed principal component analyses (PCA) on the transcriptome datasets from three to four independent replicates. The first two principal components (PC1 and PC2) represent the main axes of variation within these datasets and explained 46% and 17% of variation, respectively. Samples from control and ALL groups separated by PC1, and within ALL cohorts, samples from LTβR-Ig versus Hel-Ig treated ALL also segregated from each other, thus indicating major transcriptional changes induced by ALL growth in vivo, of which a significant fraction was sensitive to LTβR blocking (Figure 3A). Unsupervised clustering of the top 1000 most variable genes also independently segregated the three groups (Figure 3B). Comparisons between control and ALL treated with Hel-Ig samples revealed 322 differentially expressed genes (DEGs; Padj <0.05, |log2FC|>1), of which 74 were downregulated and 248 were upregulated in MSCs of control mice (Supplementary file 1). Comparisons between the ALL groups (Hel-Ig versus LTβR-Ig) revealed 226 DEGs of which 149 were upregulated and 77 were downregulated in MSCs of mice with ALL and treated with Hel-Ig (Supplementary file 2). Gene set enrichment analyses revealed a strong inflammatory gene signature induced by ALL with a strong statistical significance in interferon α- and γ-induced genes, complement, and cytokines IL2, IL6, and TNFα signaling (Figure 3C). Of note, LTβR blocking further increased the interferon stimulated gene signature, while it reduced the expression of genes associated with TNFα signaling (Figure 3C), consistent with the fact that LTβR is a TNF superfamily member that activates canonical and non-canonical nuclear factor kappa-binding transcription factors (NFκB) (Norris and Ware, 2007). Importantly, of the several hematopoietic cytokines expressed by MSCs, KITL, IL7, IGF1, and CSF1 were significantly downregulated by ALL cells (Figure 3D). These transcriptional changes in MSCs were similar to those described in mice with acute myeloid leukemia (Baryawno et al., 2019). However, of these hematopoietic cytokines, only IL7 downregulation was blocked by LTβR-Ig treatment (Figure 3D). Furthermore, blocking other NFκB-inducing cytokines, such as TNFα and IL1β, did not prevent IL7 downregulation nor did it rescue non-malignant lymphopoiesis or myelopoiesis (Figure 3—figure supplement 1A–C) and did not impact ALL expansion in vivo (Figure 3—figure supplement 1D). Even though ALL cells promoted an interferon-induced gene expression signature in MSCs (Figure 3C), blocking IFNα or IFNγ signaling did not rescue IL7 downregulation and non-malignant hematopoiesis, nor did it reduce ALL growth in vivo (Figure 3—figure supplement 1E–I). Combined, these results show a major impact of ALL expansion in the MSC transcriptome, with a large fraction of DEGs being sensitive to LTβR blocking.

To test if LTβR signaling in MSCs impacts ALL growth, non-malignant hematopoiesis, and mouse survival, we transplanted ALL cells into mice conditionally deficient in Ltbr in MSCs (Ltbr^fl/fl; Lepr^Cre/+ mice, from here on referred as LTβR∆) that also report Il7 transcription via GFP expression (Il7^GFP/+). We ruled out a role for LTβR signaling in MSCs in promoting ALL homing to the BM by transplanting 3×10⁶ ALL cells into control or LTβR∆ mice (Figure 4A), in agreement with prior studies showing that LTβR signaling in MSCs does not control CXCL12 expression under homeostatic conditions (Zehentmeier et al., 2022). Transplanted ALL cells induced IL7 downregulation in control mice (WT, Lepr^+/+; Ltbr^fl/fl) but not in LTβR∆ mice (Figure 4B), as expected (Figure 3D). These changes in IL7 production corresponded with reduced lymphopoiesis in WT mice whereas lymphopoiesis was largely unaffected in LTβR∆ mice (Figure 4C–E). In contrast, ALL-induced reductions in myeloid and erythroid lineages were largely independent of LTβR signaling in MSCs (Figure 4G–J). The inability to induce LTβR signaling in MSCs also impacted ALL growth in vivo (Figure 4F and K) such that it extended mouse survival by approximately 1 week (Figure 4L). To further test if ALL cells directly induce LTβR signaling in MSCs, we generated ALL cells genetically deficient in Ltb (Figure 4M). Indeed, Ltb-deficient ALL cells were unable to induce IL7 downregulation in MSCs and to block non-malignant lymphopoiesis (Figure 4N and O). Furthermore, Ltb-deficient ALL cells proliferated significantly less than Ltb-sufficient ALL cells (Figure 4P), which extended mouse survival significantly (Figure 4Q). Finally, to account for reduced Ltb-deficient ALL growth in vivo, we measured changes in Il7 expression in mice transplanted with 3×10⁶ Ltb^+/+ ALLs, 3×10⁶ Ltb^-/- ALLs, and 9×10⁶ Ltb^-/- ALLs (3×Ltb^-/-). Importantly, IL7 expression was unchanged even in mice transplanted with threefold higher number of Ltb-deficient ALLs (Figure 4R and S). Combined, these studies show that the ALL-induced IL7 downregulation that we reported in previous studies (Fistonich et al., 2018; Zehentmeier and Pereira, 2019) is mediated by direct delivery of lymphotoxin ligands to LTβR expressed on BM MSCs.

To test if increased lymphopoiesis due to excess IL7 is directly responsible for reduced ALL growth in vivo, we treated mice transplanted with B-ALL cells with recombinant IL7 complexed with a neutralizing anti-IL7 (aIL7, clone M25) monoclonal antibody (Figure 4—figure supplement 1A), which increases the half-life of recombinant IL-7 in vivo (Boyman et al., 2008). Indeed, mice treated with IL7/aIL7 had significantly higher numbers of developing B cell subsets in the BM (Figure 4—figure supplement 1B), particularly of IL7-dependent proB and pre-B cells. Conversely, ALL numbers in BM were significantly reduced, which reflected in significant reductions in peripheral blood and spleen (Figure 4—figure supplement 1C). Combined, these data demonstrate that the lymphotoxin-mediated attenuation of IL7 production reduces lymphopoiesis, which results in accelerated ALL growth.

Sleckman and colleagues used an elegant mouse model consisting of Artemis-deficient mice crossed with immunoglobulin heavy chain transgenic mice (Vh147) and with EµBCL2 transgenic mice to study the impact of double stranded DNA break (DSB) repair pathway in B cell development. In these studies, they reported NFκB activation in pre-BCR-expressing pre-B cells that could not repair DSBs due to deficiency in Artemis (Bredemeyer et al., 2008; Bednarski et al., 2016). In previous studies using this model, we showed that Artemis-deficient pre-B cells could also induce IL7-downregulation in MSCs (Fistonich et al., 2018; Zehentmeier and Pereira, 2019), suggesting that the LTβR pathway may also be engaged in pre-leukemic states. Consistent with this possibility, when comparing the transcriptome of Artemis-deficient (Dclre1c^−/−) and Rag1-deficient (Rag1^−/−) pre-B cells, we noted that Artemis-deficient pre-B cells also expressed significantly higher amounts of LTα and LTβ transcripts (Bredemeyer et al., 2008; Bednarski et al., 2016). In agreement with these observations, we detected higher amounts of LTα and LTβ protein on the cell surface of Dclre1c^−/− pre-B cells than on Rag1^−/− pre-B cells (Figure 5—figure supplement 1A). Transplantation of Dclre1c^−/− BM cells into lethally irradiated Il7^GFP/+ LTβR∆ mice or control littermate revealed LTβR-dependent Il7 downregulation (Figure 5—figure supplement 1B), which resulted in a trend towards increased numbers of Artemis-deficient B cells (Figure 5—figure supplement 1C). In contrast, Artemis deficiency did not impact myeloid-erythroid production (Figure 5—figure supplement 1D and E). Artemis deficiency renders cells unable to repair DSBs which causes a supraphysiological activation of the DNA damage response pathway (Bednarski and Sleckman, 2012). To test if the DNA damage response controls LTα and LTβ expression, we treated ALL cells with Etoposide, a chemotherapeutic agent that prevents DSB repair. ALLs upregulated LTα on the cell surface in an Etoposide dose-response manner (Figure 5—figure supplement 1F and G). The DNA damage response pathway signals activation of NFκB. Interestingly, treatment with IMD0354 (a small molecule inhibitor of IKKβ-mediated IκB phosphorylation) significantly reduced lymphotoxin α1β2 expression in Etoposide-treated ALLs (Figure 5—figure supplement 1H). Combined, these studies demonstrate that LTα and LTβ expression can be activated by DNA damage response pathway.

Prior studies have shown that Gαi-protein coupled receptor signaling in B-lineage cells promotes lymphotoxin α1β2 expression (Ansel et al., 2000). Engagement of LTβR expressed on secondary lymphoid organ stromal cells increases the production of B cell chemokines, which further increases lymphotoxin α1β2 expression in B cells, thus establishing a feedforward loop (Ansel et al., 2000; Cyster et al., 2000). To test if CXCR4 signaling in ALLs promotes lymphotoxin α1β2 expression, we treated ALLs in vitro with a range of CXCL12 concentrations and measured surface LTα. Indeed, ALLs upregulated LTα after exposure to CXCL12 (Figure 5A). Furthermore, LTα expression was further increased in ALLs treated with CXCL12 and Etoposide (Figure 5B). Although prior studies have also shown that IL7R signaling can promote lymphotoxin α1β2 expression in lymphoid tissue inducer cells (Yoshida et al., 2002), in ALL cells IL7R signaling was not sufficient for upregulating surface LTα even at high IL7 concentrations, even though it promoted IL7Ra internalization (Figure 5—figure supplement 2A and B). To test if CXCR4 signaling is required for lymphotoxin α1β2 expression in ALLs in vivo, we transplanted 3×10⁶ ALL cells into WT mice for 1 week and treated them with AMD3100, or saline, 12 hr prior to sacrifice. Indeed, lymphotoxin α1β2 expression was significantly reduced in ALLs and in non-leukemic pre-B cells of AMD3100-treated mice (Figure 5C). Conversely, AMD3100 increased CXCR4 surface expression in ALL and non-leukemic pre-B cells (Figure 5D), as expected (Broxmeyer et al., 2005; Beck et al., 2014). AMD3100 has a short half-life in vivo thus making it unsuitable for long-term treatment in vivo (Beck et al., 2014; Hendrix et al., 2000). To test if CXCR4 antagonism also prevents ALL-induced IL7 downregulation in BM MSCs, we transferred 3×10⁶ ALL cells into Il7^GFP/+ mice and treated them with an orally bioavailable CXCR4 antagonist (Dale et al., 2020) or with vehicle by daily gavage (Figure 5E). While control-treated mice showed ALL-induced Il7-GFP downregulation in MSCs, mice treated with CXCR4 antagonist maintained IL7 expression within the normal range of mice without ALL (Figure 5F). Similarly, developing B cells were significantly reduced in control-treated mice, but their numbers were normal in CXCR4 antagonist treated mice (Figure 5G). In contrast, ALL numbers were significantly decreased in the BM and periphery of mice treated with CXCR4 antagonist (Figure 5H), which correlated with extended mouse survival (Figure 5I).

Several studies have demonstrated that CXCR4 promotes ALL homing and retention in the BM (Juarez et al., 2007; Sipkins et al., 2005). Paradoxically, our studies revealed that ALLs induce a small, but significant, downregulation of CXCL12 expression in vivo (Figure 3D). However, when comparing CXCR4 protein levels between ALLs and non-leukemic B cell progenitors developing in the same microenvironment, we noted that ALLs express significantly higher amounts of CXCR4 in vivo (Figure 5J) and migrate towards a CXCL12 gradient in vitro significantly more than non-leukemic developing B cell subsets (Figure 5K). These observations suggest that ALLs are retained more efficiently in the BM than non-leukemic B cell progenitors.

As previously mentioned, AML also induces the downregulation of multiple hematopoietic cytokines expressed by MSCs, including IL7 (Baryawno et al., 2019). Furthermore, we also found an inverse correlation between LTB transcript abundance and patient outcome in a cohort of AML patients from the Cancer Genome Atlas (Figure 6—figure supplement 1A). To study the impact of AML growth in non-malignant hematopoiesis, we used a doxycycline (DOX)-inducible mouse model of the MLL-AF9 oncogenic fusion driving AML (Chen et al., 2019). In a mixed BM chimera of DOX-induced MLL-AF9 transgenic and competitor WT cells expressing CD45.1, MLL-AF9 expression increased the frequency of AML cells and of myeloid cell subsets expressing NGFR, a reporter for MLL-AF9 (Figure 6—figure supplement 1B and C). In contrast, both erythroid and lymphoid subsets were significantly reduced 2 and 4 weeks after MLL-AF9 induction (Figure 6—figure supplement 1D and E). Next, we tested if the LTβR pathway is also engaged by AMLs. LTβ expression is higher on MLL-AF9-positive than on MLL-AF9-negative myeloid cells and neutrophils (Figure 6A). MLL-AF9-expressing cells induced IL7 downregulation in MSCs, which could be blocked with LTβR-Ig but not with Hel-Ig treatment (Figure 6B). The effect of LTβR-Ig treatment also restored B lymphopoiesis partially (Figure 6C), but not erythropoiesis (Figure 6D), which correlated with reduced AML growth in vivo and extended mouse survival (Figure 6E and F). Similar findings were obtained with AML cells deficient in Ltb (Figure 6G). Overall, these studies collectively demonstrate that acute lymphoid and myeloid leukemias turn off lymphopoiesis by expressing LTβR ligands and downregulating IL7 production.

C57BL/6NCR (strain code 556, CD45.2+) and B6-Ly5.1/Cr (stain code 564, CD45.1+) were purchased from Charles River Laboratories. Lepr-cre mice were purchased from The Jackson Laboratories. Il7^GFP/+ mice were from our internal colony. Ltb^fl/fl (Tumanov et al., 2002) and Ltbr^fl/fl (Wang et al., 2010) mice were bred at Yale Animal Resources Center. DOX-inducible MLL-AF9 (Hprt^MLL-AF9, Rosa26^rtTA/rtTA) transgenic mice were bred at Yale University. Male and female adult mice (8–12 weeks) were used for all experiments. All mice were maintained under specific pathogen-free conditions at the Yale Animal Resources Center and were used according to the protocol approved by the Yale University Institutional Animal Care and Use Committee (2022-11377).

BCR-ABL-expressing B-ALL cells are developmentally arrested at the pre-B cell stage (kindly provided by Hilde Schjerven, UCSF). B-ALL cells were injected into recipient mice by tail vein, and then analyzed at different time points. For cytokine/cytokine receptor blocking, 150 μg of LTβR-Ig/HEL-Ig (Biogen), anti-TNF antibody (Bio-X-Cell #BE0058), anti-mouse IL1b (Bio-X-Cell #BE0246), anti-mouse IFNAR1 (Bio-X-Cell #BE0241), or anti-mouse IFNγ (Bio-X-Cell #BE BE0055) were injected intravenously via retro-orbital sinus immediately prior to ALL cell injection (tail vein). Then antibody treatment was administered every 5 days with same amount.

BM MSCs were isolated as previously described (Cordeiro Gomes et al., 2016; Zehentmeier et al., 2022; Lim et al., 2023). Briefly, long bones were flushed with HBSS supplemented with 2% of heat-inactivated fetal bovine serum, penicillin/streptomycin, L-glutamine, HEPES, and 200 U/mL Collagenase IV (Worthington Biochemical Corporation) and digested for 30 min at 37°C. Cells were filtered through a 100 μm nylon mesh and washed with HBSS/2% FBS. All centrifugation steps were done at 1200 rpm for 5 min and all stains were done on ice. LEPR stains were done for 1 hr and all other stains for 20 min on ice. BM MSCs were identified as CD45^- Ter119^- CD31^- CD144^- LEPR⁺ cells. For analysis of hematopoietic populations, long bones were flushed with DMEM supplemented with 2% fetal calf serum, penicillin/streptomycin, L-glutamine, and HEPES. Red blood cells were lysed with ammonium chloride buffer.

Hematopoietic cell populations were identified as follows: proB: CD19⁺ CD93⁺ IgM^- cKit⁺; pre-B: CD19⁺ CD93⁺ IgM^- cKit^-; immature B: CD19⁺ IgM⁺ CD93⁺; mature B: CD19⁺ IgM⁺ CD93^-; immature neutrophils: CD115^- Gr1⁺ CD11b^hi CXCR4^hi; mature neutrophils: CD115^- Gr1^hi CD11b^lo; immature monocytes: CD115⁺ Gr1⁺ CXCR4^hi; mature monocytes: CD115⁺ Gr1⁺ CXCR4^lo; GMP: Lineage^- cKit⁺ SCA-1^- CD34⁺ CD16/32^hi; immature and mature erythrocytes: Ter119⁺ CD71^- (mature) or CD71⁺ (mature). The lineage cocktail was as follows: CD19, B220, CD3e, CD4, Gr1, NK1.1, Ter119, CD11b, and CD11c.

Measurements of LTα and LTβ expression were performed using anti-mouse LTα and LTβ antibodies (a gift from Biogen). A list of antibodies and conditions used is provided in Supplementary file 3.

YFP tagged BCR-ABL plasmid was kindly provided by Dr Hilde Schjerven (UCSF). Pre-B cells were sorted from Ltb^+/-or Ltb^-/- mouse BM by gating on CD19+CD93+IgM-cKit-. Sorted pre-B cells were transduced with BCR-ABL YFP retroviruses. After infection, pre-B cells were cultured in DMEM supplemented with 20% FBS, penicillin/streptomycin, L-glutamine, HEPES, 0.05 mM 2-mercaptoethanol, and 100 ng/mL recombinant murine IL7 (peprotech, 217-17). After 2 days of culture, cells were cultured in the same media without IL7 and expanded for transplantation. Recipient mice received 6 Gy gamma irradiation before transplantation.

MLL-AF9 mice were maintained and genotyped as previously described (Chen et al., 2019). Two million BM cells recovered from MLL-AF9 transgenic mice were mixed with 1 million BM cells from WT congenic CD45.1 mice; BM mixtures were transplanted into lethally irradiated CD45.1 recipients. Two weeks after transplantation, mice were fed with 1 g/L Dox in the drinking water sweetened with 10 g/L sucrose. For Ltb-deficient MLL-AF9 experiment, Ltb^-/- mice were crossed with MLL-AF9 transgenic mice to generate Ltb^+/-, MLL-AF9 and Ltb^-/-, MLL-AF9 mice.

CXCR4 receptor antagonist X4P-X4-185-P1 (X4 Pharmaceuticals) was dissolved in 50 mM citrate buffer, pH 4.0. Mice were treated at 100 mg/kg daily by oral gavage 2 days after ALL adoptive transfer. Control mice were treated by oral gavage of citrate buffer alone. Some experiments were performed with the CXCR4 antagonist AMD3100 (Sigma). Briefly, mice were transplanted with 3×10⁶ ALL cells i.v. (tail vein) and ALLs were allowed to expand in vivo for 7 days. Mice were treated every 6 hr with 80 µg AMD3100 dissolved in saline (200 µL/mouse) i.v. for a period of 18 hr prior to sacrifice.

Recombinant murine IL7 of 1.5 µg (peprotech, 217-17) was pre-incubated with 15 µg IL7 antibody (clone M25, Bio-X-Cell) for 20 min at room temperature. Mice were treated i.v. with IL7/aIL7 complex on day 6 and seven after adoptive transfer of BCR-ABL-expressing B-ALL cells. Mice were analyzed 48 hr after the last IL7/aIL7 treatment.

BM stromal cells were isolated as previously described (Cordeiro Gomes et al., 2016; Zehentmeier et al., 2022; Lim et al., 2023). Briefly, hematopoietic cells were depleted by staining with biotin-conjugated CD45 and Ter119 antibodies followed by magnetic depletion with Dynabeads Biotin Binder (Invitrogen #11047). Remaining cells were stained with antibodies against CD31, CD144, and PDGFRα, and MSCs were sorted as CD45^- Ter119^- CD31^- CD144^- PDGFRα⁺ cells using a BD FACS Aria II. Cells were sorted directly into 350 μL RLT plus buffer (Qiagen) and RNA extracted using the RNeasy Plus Micro Kit (Qiagen #74034). RNA-sequencing was performed by the Yale Center for Genome Analysis using the Illumina HiSeq2500 system, with paired-end 75 bp read length. The sequencing reads were aligned onto Mus musculus GRCm38/mm10 reference genome, using the HISAT2 software. The mapped reads were converted into the count matrix with default parameters using the StringTie2 software, followed by the variance stabilizing transformation offered by DESeq2. DEGs were identified using the same software, DESeq2, based on a negative binomial generalized linear models and visualized in hierarchically clustered heatmaps using the pheatmap R package.

We used Student’s t test to determine if differences between experimental grous with normal distribution were statistically significant. Shapiro–Wilk normality test was performed to assess distribution normality (Supplementary file 4). We employed a non-parametric Mann–Whitney test for experimental groups with abnormal distribution. Differences in mouse survival between experimental groups were analyzed by the Kaplan–Meyer method.

The B-ALL gene expression microarray and patient outcome data were obtained from the National Cancer Institute TARGET Data Matrix (http://targetnci.nih.gov/dataMatrix/TARGET_DataMatrix.html) of the Children’s Oncology Group (COG) Clinical Trial P9906 with the GEO database accession number GSE11877 (Kang et al., 2010). The patients were segregated into two groups according whether they had above or below the median expression level of a gene (i.e., the average of multiple probesets for a gene) or above the top 25% or below the bottom 25% expression level of a gene. OS or RFS probabilities were estimated using the Kaplan–Meier method, and log-rank test (two-sided) was used to compare survival differences between the two patient groups. R package ‘survival’ version 2.35-8 was used for the survival analysis.