High VLA-4 expression is associated with adverse outcome and distinct gene expression changes in childhood B-cell precursor acute lymphoblastic leukemia at first relapse

Abstract

Background Resistance to therapy and subsequent relapse remain major challenges in the clinical management of relapsed childhood acute lymphoblastic leukemia. As the bone marrow environment plays an important role in survival and chemotherapy resistance of leukemia cells by activating different signaling pathways, such as the VLA-4 and PI3K/Akt pathways, we studied the prognostic and biological impact of VLA-4 expression in leukemia cells from children with relapsed B-cell precursor acute lymphoblastic leukemia and its influence on the sensitivity of the leukemia cells to drugs.Design and Methods VLA-4 expression was quantified by real-time polymerase chain reaction in leukemia cells from 56 patients with relapsed acute lymphoblastic leukemia enrolled in the ALL-REZ BFM 2002 trial of the Berlin-Frankfurt-Münster study group. Gene expression changes related to VLA-4 expression were investigated by microarray-based mRNA profiling. The effect of VLA-4 signaling on proliferation and drug resistance was studied in co-cultures of leukemia and stromal cells.Results High expression of VLA-4 at first relapse was associated with adverse prognostic factors, poor molecular response to therapy and significantly worse probabilities of event-free and overall survival. VLA-4 expression was an independent prognostic parameter. Comparing gene expression profiles of leukemia cells with high versus low VLA-4 expression, we identified 27 differentially expressed genes primarily involved in the PI3K/Akt, ephrin and Rho GTPase pathways. Blocking of VLA-4 signaling in combination with cytarabine treatment abolished the growth supportive effect of stromal cells.Conclusions Our results show that high VLA-4 expression is a marker of poor prognosis and a potential therapeutic target in children with relapsed acute lymphoblastic leukemia and confirm that cellular interactions and biological effects related to VLA-4 play a decisive role in the survival of leukemia cells and response to therapy.

Introduction

The survival rate of children at their initial diagnosis of acute lymphoblastic leukemia (ALL) has improved to about 80%.1,2 However, the overall outcome of children with relapsed disease remains poor at less than 40%.3 The main predictors of outcome in relapsed ALL are duration of first remission, immunophenotype of the leukemia cells, site of relapse4 and molecular response to therapy (minimal residual disease, MRD).5 Based on these biological and clinical risk factors, three main risk groups with significantly different event-free survival rates (standard risk: S1, intermediate risk: S2 and high risk: S3/S4) were defined in the ALL relapse trial of the Berlin-Frankfurt-Münster (BFM) study group: ALL-REZ BFM 2002. High-risk parameters are short duration of first remission (very early/early time point of relapse), T-cell ALL and bone marrow involvement. In fact, in more than 80% of childhood ALL the site of first relapse is the bone marrow.4,6 Bone marrow stromal cells have been shown to protect normal and malignant hematopoietic cells from damaging events, including drug-induced cell death.7–12

Normal hematopoiesis is regulated by the adhesive interaction of hematopoietic cells with the bone marrow microenvironment, as well as by growth factors and cytokines expressed by different types of bone marrow cells.12 The bone marrow microenvironment consists of stromal cells and extracellular matrix proteins such as fibronectin, collagen and laminin. Among these extracellular matrix proteins, fibronectin has been shown to support survival and proliferation of hematopoietic stem cells and progenitor cells through their fibronectin receptors such as very late antigen-4 (VLA-4; ITGA4; α₄β₁). α₄ integrins act as both adhesive and signaling receptors by interacting with their ligands, fibronectin and vascular cell adhesion molecule-1 (VCAM-1) expressed by stromal cells.13 In hematopoietic malignancies (e.g. acute myeloid leukemia), VLA-4-mediated signaling has been shown to support the survival of leukemia cells through the phosphatidylinositol-3-kinase (PI3K)/Akt/Bcl2 signaling pathway.14–17 Furthermore, PI3K signaling has a central role in several decisive processes for cancer progression, including metabolism, growth, survival and motility.18,19 Given the importance of the bone marrow environment for leukemia cell development and survival and considering the poor outcome of children with relapsed ALL, we hypothesized that high expression of VLA-4 in leukemia cells could be associated with an adverse outcome in relapsed childhood ALL. We, therefore, investigated the clinical and prognostic impact of VLA-4 expression in bone marrow leukemia cells from 56 children with B-cell precursor (BCP) ALL at diagnosis of first relapse. Subsequently, gene expression changes related to VLA-4 expression were investigated by microarray-based mRNA profiling, and the effect of VLA-4 signaling on leukemia cell survival and drug resistance was studied in co-cultures of leukemia cells and bone marrow stromal cells.

Design and Methods

Patients and samples

VLA-4 expression was determined retrospectively in bone marrow samples from 56 children and adolescents obtained at diagnosis of first relapse of BCP-ALL with bone marrow involvement. The study was confined to relapses of BCP-ALL because BCP-ALL represent the majority of childhood ALL (85%). Bone marrow aspirates were selected to contain more than 75% leukemia cells based on morphological evaluation of bone marrow smear preparations. All patients (n=56) were enrolled and 51 patients were treated according to the relapse trial ALL-REZ BFM 2002 protocol approved by the Institutional Review Board of the Charité-Universitätsmedizin Berlin. Patients were included, in accordance with the above mentioned criteria, from the start of 2002 until the end of 2003.20 The total cohort of patients registered in the ALL-REZ BFM trial during the same time period was compared to the cohort of patients analyzed for VLA-4 mRNA expression with respect to frequencies of clinical and therapeutic parameters to assess the representativeness of the group, and showed no selection bias (Table 1). Written informed consent was obtained from patients or guardians prior to treatment.

Methods and statistical analysis

Detailed information about the quantification of VLA-4 mRNA by real-time polymerase chain reaction (QRT-PCR) analysis, VLA-4 protein assessment by FACS and immunocytochemistry (ICC), gene expression analysis, cell culture experiments (cell lines, cell culture, western blot analysis, proliferation and adhesion assays) and the statistical analysis are provided in the Online Supplementary Design and Methods.

Results

VLA-4 expression is associated with clinical features and outcome at first relapse

As a prerequisite for our mRNA-based study we analyzed the correlation between VLA-4 mRNA expression and protein level. We, therefore, compared both expression levels in five BCP-ALL cell lines and in 11 patients’ samples, by flow cytometry, ICC and QRT-PCR. The comparative protein-mRNA expression analysis showed that the relative VLA-4 mRNA expression correlated well with protein level (R=0.76), allowing us to study the impact of VLA-4 expression in leukemia cells by QRT-PCR (Online Supplementary Figure S1).

VLA-4 expression was determined by QRT-PCR in 56 samples of bone marrow leukemia cells. Expression levels of VLA-4 ranged from 1.0 to 148.1 in relation to the reference gene ABL1. We compared expression levels of VLA-4 in relevant clinical and biological subgroups of ALL relapse (Table 1 and Online Supplementary Figure S2). VLA-4 expression was significantly higher in leukemia cells from patients who were younger at the time of diagnosis of relapse (P=0.018) and from those with very early/early relapse (P=0.011). Accordingly, leukemia cells from patients stratified into the high-risk groups S3/S4 showed higher VLA-4 expression compared to those in the intermediate risk group S2 (P=0.002, Online Supplementary Figure S2C). Furthermore, we compared the VLA-4 expression levels between four different risk stratification subgroups (S2-, MRD low; S2+, MRD high; S3 and S4). The VLA-4 expression levels differed significantly between these groups (Kruskal Wallis test: P=0.012, Figure 1G). Within the intermediate risk group S2, VLA-4 expression was higher in the group with a high MRD load after induction therapy (Table 1). The median VLA-4 expression in leukemia cells from patients in first relapse who did not respond to therapy (non-responders, median VLA-4 expression=29) and who suffered a subsequent relapse (median=9.0) was significantly higher than that in leukemia cells from the patients remaining in continuous complete remission (median=3.7) (Table 1). In particular, of the four patients with high VLA-4 levels who died during induction therapy or of a therapy-related cause (Table 1), three had a poor response to therapy (Online Supplementary Table S1), and died of sepsis. High VLA-4 expression was also associated with a more immature proB-cell immunophenotype (P=0.025). VLA-4 expression was higher in leukemia cells from four patients with MLL/AF4 fusion genes. There was no correlation between the presence of TEL/AML1 fusion genes and VLA-4 expression.

Table 1.Clinical and biological characteristics of the studied BCP-ALL patients and VLA-4 mRNA expression in correlation with these parameters.

Figure 1.VLA-4 expression is associated with outcome of BCP-ALL at first relapse. (A) Kaplan-Meier analysis of event-free survival (EFS) is shown for the total VLA-4 BCP-ALL patient cohort (n=51). (B) Kaplan-Meier analysis of overall survival (OS) is shown for the total VLA-4 BCP-ALL patient cohort (n=51). (C) Kaplan-Meier analysis of EFS is shown for BCP-ALL patients (n=51) with VLA-4 expression levels in leukemia cells lower and higher than the median (8.3). (D) Kaplan-Meier analysis of OS is shown for BCP-ALL patients (n=51) with VLA-4 expression levels in leukemia cells lower and higher than the median. (E) Kaplan-Meier analysis of EFS for three groups of BCP-ALL patients (n=51) with VLA-4 low (lower quartile, < 25%; LEV), intermediate (intermediate quartiles, >25%–<75%; IEV) and high expression level (upper quartile, >75%; HEV). (F) Kaplan-Meier analysis of OS for three groups of BCP-ALL patients (n=51) with VLA-4 low (LEV), intermediate (IEV) and high (HEV) expression level. (G) Comparison of VLA-4 expression levels between four different risk stratification groups of the ALL-REZ-BMF 2002 trial (Kruskal Wallis test: P=0.012). S2-: S2 group with MRD < 10³ at day 36; S2+: S2 group with MRD ≥ 10⁻³ at day 36; S3 and S4, high-risk groups.

Figure 2.VLA-4 expression is associated with distinctive gene expression changes. (A) Expression levels of 27 genes identified by comparison of relapsed BCP-ALL patients with low (lower quartile, < 25%; LEV, green), intermediate (intermediate quartiles, >25%–<75%; IEV blue), and high (upper quartile, >75%; HEV, red) expression of VLA-4. Expression levels are shown as differences with respect to the mean over all patients (n=43). Patients were ordered according to VLA-4 expression levels. Genes are clustered according to down-and up-regulation in HEV- versus LEV-groups. (B) Box plots of down-regulated genes in HEV- versus LEV-groups obtained by chip analysis in the three groups (LEV: green; IEV: blue; HEV: red). Signal intensities (SI) of expression of five representative genes: EFNB2, FOXO1, TCF7L2, IRF4, and ARHGAP29 are shown. (C) Box plots of up-regulated genes in the HEV- versus LEV-groups, obtained by chip analysis in the three groups (LEV: green; IEV: blue; HEV: red). Signal intensities (SI) of expression of two representative genes are shown: ICAM2 and F13A1. Statistical analysis was done by ANOVA (*P<0.05; **P<0.001; ***P<0.0001).

To study the impact of VLA-4 expression on treatment outcome, we divided the patients into groups with high and low VLA-4 expression, defined as those with an expression level above or below the median VLA-4 expression (median=8.3), respectively (Table 1). The probabilities of event-free survival (23% ± 6) and overall survival (35%±7) of the total study cohort are shown in Figure 1A-B. Kaplan-Meier estimates revealed that patients with VLA-4 levels higher than median in leukemia cells had a significantly lower probability of event-free survival (8%±5 versus 40%±10; P<0.001; Figure 1C) and overall survival (15%±7 versus 55%±10; P<0.001; Figure 1D) than those with VLA-4 levels lower than the median at a follow-up of 8 years. To exclude biases related to genetic alterations in leukemia cells, separate Kaplan-Meier analyses were conducted excluding MLL/AF4 (n=3, high VLA-4 expression) and TEL/AML1 patients (n=6). As expected, the exclusion of either group did not diminish the prognostic significance of VLA-4 expression (MLL/AF4: event-free survival: 08%±05 versus 43.5%±.10; overall survival: 20 ±08 versus 55%±10; Online Supplementary Figure S3A-B; TEL/AML1: event-free survival: 08% ± 06 versus 33%±.10; overall survival: 16%±08 versus 46%±11; Online Supplementary Figure S3C-D).

To assess whether a gradual division by VLA-4 expression would further strengthen discrimination of prognostic groups, we split the patients’ samples into three groups according to whether they had low expression of VLA-4 (LEV; lower quartile, <25%), intermediate expression (IEV; interquartile range, >25%-<75%) or high expression (HEV; upper quartile, >75%). Patients with LEV had a significantly better probability of event-free survival (Figure 2E) and overall survival (Figure 2F) than those with IEV and HEV at a follow-up of 8 years (event-free survival: P<0.01; overall survival: P=0.001), indicating a clear correlation between VLA-4 expression levels and outcome in these groups. In addition, the median event-free survival was significantly (P=0.002) shorter in the HEV-group (0.37 years) than in the IEV (1.5 years) and LEV groups (5.3 years). Thus, through analysis of VLA-4 expression we identified prognostic groups with high/low event-free and overall survival probabilities and early/late subsequent events. The multivariate Cox regression analysis (Table 2), including VLA-4 expression, time point, age, immunophenotype and MLL/AF4 fusion gene, revealed an independent prognostic significance of VLA-4 expression (hazard ratio, HR=3.43, P=0.001) next to time point of relapse (HR= 3.81, P=0.006). For patients stratified into the intermediate risk group S2, MRD was additionally included in the multivariate Cox regression model but only VLA-4 expression was found to be of independent prognostic significance (P=0.014) since this group comprises only patients with time point early/late and MRD was already used for risk assessment (Table 2).

VLA-4 expression is associated with distinctive gene expression changes

To identify pathways associated with VLA-4 expression at ALL relapse, we studied gene expression profiling data available for a subset of 43 patients with relapsed ALL from the total cohort of 56 patients. The event-free and overall survival probabilities of these patients are shown in Online Supplementary Figure S4A-B. Gene expression analysis confirmed our VLA-4 level results obtained with QRT-PCR in the defined groups (IEV, LEV, HEV; P<0.001, Figure 2A). We identified 414 genes differentially expressed in correlation with VLA-4 expression between the three groups with a P-value of 0.01 and a minimal relative fold change of 1.3 between the HEV and the LEV groups (Online Supplementary Table S3: 142 down-regulated genes; Online Supplementary Table S4: 272 up-regulated genes). According to the fold change and P-value, we selected 27 genes related to known pathways (Figure 2A). The differential expression, fold change and statistical analysis are summarized in Table 3. When comparing the most extreme groups, HEV versus LEV, 12 of the genes were down-regulated by up to 6-fold (e.g. EFNB2, FOXO1, TCF7L2, IRF4, ARHGAP29; Figure 2B) and 16 were up-regulated up to 2.8-fold (e.g. ICAM2, F13A1, PIK3CA; Figure 2C). Most of these genes are involved in PI3K/Akt, Pten, Ephrin (PIK3CA, ILK, FOXO1, ICAM2, RAC2, IRS-2, ITGAL, PAK2, F13A1, EFNB2), Wnt/beta-catenin (TC7FL2, HDAC1), and Rho GTPase signaling pathways (Rac2, ARHGAP29, ARHGEF6).

Blocking of VLA-4 signaling overcomes the supportive effect of stromal cells

Having shown that VLA-4 expression is associated with prognosis and distinctive gene expression changes, we next evaluated the effect of VLA-4 signaling on BCP-ALL cell adhesion, survival and drug resistance using the BCP-ALL cell line REH which expresses high amounts of VLA-4 (Online Supplementary Figure S1A,B) as a model system. In co-culture with stromal cells (L87/4) the adhesion of REH cells was reduced after treatment with VLA-4 blocking antibodies (Figure 3A). This effect was not observed in the absence of stromal cells. In addition, the stromal cell-induced proliferation of REH cells was less pronounced after treatment with VLA-4 blocking antibodies (Figure 3B). To assess whether blocking of VLA-4 signaling overcomes the protection of leukemia cells from cytarabine-induced cytotoxicity by stromal cells,21 we treated REH cells in mono-culture and in co-culture with cytarabine (Ara-C) in the presence or absence of anti-VLA-4 antibodies. Blocking of VLA-4 in leukemia cells significantly abolished the cytoprotective effect of stromal cells (Figure 3B). Consistent with these findings, expression of the anti-apoptotic protein BCL-2, was also decreased in leukemia cells (Figure 3C), revealing that VLA-4 signaling has the ability to regulate the sensitivity of leukemia cells to chemotherapy.

Table 2.Univariate and multivariate analysis for survival after relapse.

Discussion

In this study, we assessed the clinical and prognostic relevance of VLA-4 expression in bone marrow-derived leukemia cells of pediatric patients with first relapse of BCP-ALL. Patients with higher VLA-4 levels in their leukemia cells had significantly worse event-free and overall survival probabilities than those with lower expression. Accordingly, high levels of VLA-4 expression were associated with clinical and biological high-risk parameters, i.e. early time point and high-risk stratification. In multivariate Cox regression analysis VLA-4 expression was a significant independent prognostic determinant of event-free survival. ALL patients at first relapse with VLA-4 levels higher than the median rarely remained in continuous complete remission (8%) compared to those with VLA-4 levels lower than median (40%; Table 1). However, there was no significant difference in subsequent relapse rates between these groups (54% versus 48%), perhaps because of the presence of different minor subclones at first relapse that could not be analyzed for VLA-4 expression. Addressing whether VLA-4 expression levels vary between first to second relapse, we analyzed ten matched leukemia cell samples. Five patients showed similar VLA-4 expression in second relapse, three had higher and two had lower VLA-4 expression levels (Online Supplementary Table S2). Although in these two cases, VLA-4 expression was lower at second relapse, the relative VLA-4 expression was still high (HEV/IEV group). Nevertheless, the median VLA-4 expression in leukemia cells from patients who suffered a subsequent relapse or who did not respond to therapy was higher than the median in patients in continuous complete remission, suggesting an involvement of VLA-4 in the persistence of leukemia cells during relapse therapy (Table 1). The detailed analysis of outcome revealed that the significant difference between the VLA-4 groups was largely accounted for by the disparity in response to therapy, next to subsequent relapse. In particular, also the patients experiencing death during induction therapy (2/2) or in continuous complete remission (1/2) showed a poor response to therapy (Online Supplementary Table S1), indicating that response to therapy is the main factor responsible for the divergence of these groups. In line with this, patients with higher MRD load showed higher VLA-4 levels than those with low MRD burden.

Table 3.Differential expression, fold change and pathway analysis of 27 identified genes.

In a few previous studies of adult and pediatric ALL, VLA-4 expression was associated with different patho-physiological and clinical features, such as peripheral blood cell count,14,22–25 but significant correlations between VLA-4 expression levels and outcome could not be clearly established,14,22–24 probably due to the consideration of heterogeneous, non-uniformly treated groups and small numbers of patients. In acute myeloid leukemia, more systematic studies have been published, with contradictory results, some finding that high VLA-4 expression was associated with adverse outcome16,26 while others found that it was associated with improved outcome.27,28 In our study of uniformly treated patients at first relapse, we observed that high VLA-4 expression was associated with an adverse outcome as well as with high-risk features, i.e. age,29 immunophenotype and time of relapse.4

To identify molecular determinants underlying the prognostic relevance of VLA-4 expression at ALL relapse, we studied gene expression profiles of leukemia cells from 43 ALL patients in our study cohort. Different VLA-4 expression levels were characterized by distinctive gene expression changes. Most of the genes differentially expressed are involved in the PI3K/Akt, ephrin, Wnt and Rho GTPase signaling pathways. FOXO1, a member of the family of Forkhead transcription factors, is primarily regulated by the PI3K/Akt pathway and its inactivation/activation plays a crucial role in B-cell proliferation and tumor suppression by inducing apoptosis.30–33 Down-regulation of FOXO1 and TCF7L2 has been shown to correlate with metastasis and poor prognosis in renal cell carcinoma,34 as observed in the HEV-group. Furthermore, the low expression level of EFNB2 in the HEV-group is consistent with a recently published study showing that epigenetic silencing of EPH/EPHRIN family genes contributes to ALL pathogenesis.35 Rho GTPases (Rac2, Cdc42),36,37 ARHGEF638 and PAK2 (p21-activated kinases) transduce signals leading to migration, adhesion, and survival of many blood cells and are often over-expressed in multiple cancer types.39 Rac genes have been shown to regulate leukemogenesis,40,41 and B-cell development.37 In line with these findings Rac2, PAK2 and ARHGEF6 were up-regulated in the HEV-group with worse event-free and overall survival probabilities. It has been recently shown that VLA-4 mediates survival in chronic lymphocytic leukemia through two different pathways: by binding to Vcam-1 and subsequent PI3K/Akt pathway activation or by interacting with MMP9/CD44 leading to STAT3 phosphorylation and MCL-1 expression.42 We did not observe any differential expression of CD44 and MCL-1 in our data set (data not shown), suggesting that VLA-4 signaling probably contributes via the PI3K/Akt pathway to leukemia cell survival in relapsed BCP-ALL.

Figure 3.Blocking of VLA-4 signaling overcomes the supportive effect of stromal cells. (A) For cell adhesion assay, the BCP-ALL cells (REH) were fluorescence-labeled and seeded into 24-well plates coated with poly-lysine or the stromal cell line L87/4 in the absence or presence of the blocking anti-VLA-4 monoclonal antibody. The percentage of adherent cells measured by the fluorescence plate reader is shown relative to control samples without blocking antibodies. (B) REH cells were cultured either alone or in co-culture with L87/4 stromal cells with or without blocking anti VLA-4 monoclonal antibody and cytarabine (ARA-C, 1 …M). After 48 h, proliferation was measured by the MTS tetrazolium assay and cell numbers per microliter were calculated from appropriate cell dilution series. Columns, mean of at least three independent experiments; bars, SD (**P<0.01; ***P<0.001). (C) Western-blot analysis of the anti-apoptotic protein BCL-2. REH cells were grown in co-culture with L87/4 stromal cells and were treated with or without blocking anti VLA-4 monoclonal antibody and/or ARA-C for 48 h. β-actin served as the control for equal protein loading. The density of dots was measured. The ratios of BCL-2 density to β-actin density are shown.

Many studies have revealed that VLA-4 regulates adhesion, migration and survival of leukocytes and hematopoietic stem cells.12 Furthermore, blocking VLA-4 signaling increases chemotherapy-induced apoptosis in acute myeloid leukemia16 and decreases the ability of leukemia cells to engraft in bone marrow in a xenograft mouse model.43 Consistent with these findings, our co-culture experiments showed that inhibition of VLA-4 signaling in BCP-ALL overcomes the cytoprotective effect of stromal cells. Thus, targeting receptors/pathways involved in the VLA-4/PI3K pathway may be a promising therapeutic option in relapsed ALL, like the application of clinically approved antibodies against VLA-444,45 or PI3K inhibitors.19 In conclusion, our study reveals that high VLA-4 expression is an indicator of poor prognosis and a potential therapeutic target in relapsed childhood BCP-ALL. Furthermore, our results show that cellular and biological effects related to VLA-4 signaling play a decisive role in survival and/or response to therapy of childhood BCP-leukemia cells.

Acknowledgments

The authors are especially grateful to M. Pfau, W. Keune and J. Proba for their excellent technical assistance, and grateful to Dr. T. Taube and Dr. S. Wellmann for critical discussion. Parts of this work were supported by grants from the Berliner Krebsgesellschaft e.V., the Deutsche Krebshilfe e.V., and KINDerLEBEN e.V., Deutsche Kinderkrebsstiftung, the Federal Ministry for Education and Research in the National Genome Research Network (NGFN2, 01GS0870) Germany and the Deutsche José Carreras Leukämie foundation (SP10/01) for use of Partek®Genomics Suite Software.

Footnotes

• The online version of this article has a Supplementary Appendix.
• Authorship and Disclosures The information provided by the authors about contributions from persons listed as authors and in acknowledgments is available with the full text of this paper at www.haematologica.org.
• Financial and other disclosures provided by the authors using the ICMJE (www.icmje.org) Uniform Format for Disclosure of Competing Interests are also available at www.haematologica.org.

• Received May 19, 2011.
• Revision received July 14, 2011.
• Accepted July 14, 2011.