STAT5 does not drive steroid resistance in T-cell acute lymphoblastic leukemia despite the activation of BCL2 and BCLXL following glucocorticoid treatment

Abstract

Physiological and pathogenic interleukin-7-receptor (IL7R)-induced signaling provokes glucocorticoid resistance in a subset of patients with pediatric T-cell acute lymphoblastic leukemia (T-ALL). Activation of downstream STAT5 has been suggested to cause steroid resistance through upregulation of anti-apoptotic BCL2, one of its downstream target genes. Here we demonstrate that isolated STAT5 signaling in various T-ALL cell models is insufficient to raise cellular steroid resistance despite upregulation of BCL2 and BCL-XL. Upregulation of anti-apoptotic BCL2 and BCLXL in STAT5-activated T-ALL cells requires steroid-induced activation of NR3C1. For the BCLXL locus, this is facilitated by a concerted action of NR3C1 and activated STAT5 molecules at two STAT5 regulatory sites, whereas for the BCL2 locus this is facilitated by binding of NR3C1 at a STAT5 binding motif. In contrast, STAT5 occupancy at glucocorticoid response elements does not affect the expression of NR3C1 target genes. Strong upregulation of BIM, a NR3C1 pro-apoptotic target gene, upon prednisolone treatment can counterbalance NR3C1/STAT5-induced BCL2 and BCL-XL expression downstream of IL7- induced or pathogenic IL7R signaling. This explains why isolated STAT5 activation does not directly impair the steroid response. Our study suggests that STAT5 activation only contributes to steroid resistance in combination with cellular defects or alternative signaling routes that disable the pro-apoptotic and steroid-induced BIM response.

Introduction

Synthetic steroids remain a vital cornerstone drug in the treatment of pediatric T-cell acute lymphoblastic leukemia (T-ALL). Upon binding with steroids, the glucocorticoid receptor (NR3C1) dimerizes and migrates to the nucleus where it functions as a transcription factor and regulates the expression of multiple steroid-response genes, including the pro-apoptotic gene BIM.^1-3 Resistance to steroid treatment predicts for inferior outcome and therefore remains a major clinical problem in T-ALL.^4,5 An important signaling pathway that interferes with steroid sensitivity is the interleukin-7-receptor (IL7R) pathway. Both physiological IL7 signaling or the presence of activating mutations in various IL7R signaling molecules have been linked to steroid resistance in T-ALL patients.^6,7

Recently, the mechanisms underlying IL7R-dependent survival and/or steroid resistance have been studied. IL7R signaling results in downstream activation of the PI3K-AKT and STAT5 pathways, but also activates MAPK-ERK signaling.^6,8-12

For these individual downstream signaling pathways, different mechanisms have been identified which may contribute to steroid resistance (Figure 1). Activation of the PI3K-AKT pathway is reported to drive phosphorylation of NR3C1, which blocks its migration to the nucleus.¹³ Additionally, PI3K-AKT signaling can upregulate various anti-apoptotic proteins including BCLXL and MCL1.⁶ AKT can also inhibit transcription of the important glucocorticoid receptor target gene BIM via an inhibitory phosphorylation of the FOXO3A transcription factor.¹⁴ Epigenetic silencing of the BIM locus, as found in some ALL patient-derived xenograft models, has also been proposed as an important mechanism of resistance to steroids.^15,16 IL7R signaling mutations, found in approximately 35% of pediatric T-ALL patients, or physiological IL7 signaling activate downstream MAPK-ERK signaling.^{6,7,1 2} Activated ERK phosphorylates BIM-L and BIM-EL proteins, which therefore lose their potential to bind and neutralize anti-apoptotic BCL2 protein family members including BCL2, BCL-XL and MCL1, hence resulting in steroid resistance.¹² MEK inhibitors, and to a limited extent the JAK inhibitor ruxolitinib, showed synergy with steroid treatment by restoring functional BIM levels, thereby representing promising targeted compounds to overcome steroid resistance in T-ALL.^6,12

Activation of the glucocorticoid receptor NR3C1 facilitates transcriptional upregulation of glucocorticoid target genes, including BIM and IL7R.¹⁷ Recently, it was shown that by upregulating IL7Ra, steroid treatment could paradoxically induce steroid resistance,¹⁸ since enhanced IL7-induced signaling activates STAT5 and consequentially enhances the expression of the pro-survival gene BCL2¹⁸ and the PIM1 kinase gene.¹⁹ Thus far, upregulation of BCL2 is regarded as the driving mechanism for IL7-induced survival and steroid resistance.7,1 7,18,20,21

Here we explore the significance of STAT5 signaling downstream of physiological or mutant IL7R signaling in relation to steroid resistance in pediatric T-ALL. We studied whether induced expression of the constitutively active N642H STAT5 mutant can activate BCL2 and drive steroid resistance.

Methods

Cell preparation and cytotoxicity assays

An extended description of the generation of SUPT-1 and P12-ICHIKAWA cell lines can be found in the Online Supplementary Methods. For quantitative real-time reverse-transcription polymerase chain reaction (RTQ-PCR), immunoblot analysis and cytotoxicity assays, SUPT-1 and CCRF-CEM cells were plated in RPMI-1640 medium as described in the Online Supplementary Methods. For activation of IL7R signaling, the medium of the CCRF-CEM cells was supplemented with 10 ng/mL IL7 (R&D systems). For RTQ-PCR and immunoblot analysis experiments, cells were plated at a concentration of 1x10⁶ cells/mL overnight. For cytotoxicity assays, cells were plated at a concentration of 0.2x10⁶ cells/mL, and cell viability was determined after 4 days by a methylthiazolyldiphenyl-tetrazolium bromide (MTT, Sigma Aldrich) assay.

Immunoblot analysis and immunoprecipitation

Cell pellets of treated cell suspensions were lysed using kinase lysis buffer, and protein eluates were loaded on BioRad Mini-Protean^® TGX^TM any-kd precast gels.¹² Proteins were transferred to 0.2 mm nitrocellulose membranes using the Trans-Blot^® Turbo^TM Transfer System (BioRad). Primary antibodies used for immunoblot analysis are listed in the Online Supplementary Methods. Loading controls were equivalent in experiments in which multiple membranes were used. Immunoprecipitation was performed as previously described,¹² and as described in more detail in the Online Supplementary Methods.

Figure 1.Schematic overview of IL7-receptor- and steroid-induced signaling. In the presence of IL-7, IL7Ra (left) and IL7Rγ (right) subunits heterodimerize, allowing transphosphorylation of JAK1 and JAK3 kinases and phosphorylation of IL7Ra as a docking site for STAT5. As a result, downstream PI3K-AKT, MAPK-ERK and STAT5 signaling pathways are activated. Activated STAT5 migrates to the nucleus as a homodimer to regulate the transcription of canonical STAT5 target genes, including anti-apoptotic BCL2 and BCLXL. Activated PI3K-AKT signaling upregulates anti-apoptotic BCLXL and MCL1 and can regulate the nuclear translocation of the activated glucocorticoid receptor (NR3C1). Activated MAPK-ERK signaling phosphorylates and therefore inactivates proapoptotic BIM. Upon exposure to steroids, NR3C1 migrates to the nucleus as a homodimer and induces the expression of NR3C1 target genes (including pro-apoptotic BIM) at glucocorticoid response element sites.

Quantitative real-time reverse-transcription polymerase chain reactions

Isolation of RNA, cDNA synthesis and RTQ-PCR were performed as previously described.^22,23 The expression of NR3C1 or STAT5 target genes was calculated using the delta CT (dCT) method as percentage of GADPH expression. Fold expression change was calculated relative to the non-doxycycline-induced (-dox), non-prednisolone treated (-pred) condition. The primers used are described in the Online Supplementary Methods.

Processing and visualization of chromatin immunoprecipitation sequencing data

An extended description of chromatin immunoprecipitation (ChIP)-sequencing can be found in the Online Supplementary Methods. DNA eluates were sequenced using the Illumina NextSeq500 platform of the Utrecht Sequence facility. Raw reads were aligned to the CRCh38 human genome, using the Burrows-Wheeler Aligner tool²⁴ with default settings. Narrowpeak calling with default settings from MACS2²⁵ was used for peak calling. BamCoverage from deeptools²⁶ was used to normalize input signal for figures, using the reads per genomic content method. Peaks were visualized using Integrative Genomics Viewer.²⁷ Bedtools²⁸ was used to select unique and overlapping peaks for STAT5/NR3C1. These are visualized using the plotheatmap function from deeptools. MEME-ChIP was used to detect motifs within regions of 50 bp up or down of the summit of NR3C1 or STAT5 peak summits (101 bp window), using default settings.²⁹

Data availability

Affymetrix U133 Plus2 microarray data for the 117 patients as previously published³⁰ were normalized using the robust multi-array average, computed by the affy package.³¹ Microarray data are available at http://www.ncbi.nlm.nih.gov/geo/ under accession number GSE26713. Data analysis is described in the Online Supplementary Methods. A selection of STAT5 target genes ChIP-sequencing files are available at Gene Expression Omnibus under GEO series accession number: GSE171976, h ttps://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE171 976.

Results

Activation of STAT5 target genes does not predict steroid resistance in T-cell acute lymphoblastic leukemia

To study the significance of STAT5 activation in relation to steroid resistance in pediatric T-ALL, we investigated the expression of STAT5 target genes (e.g., BCL2, BCL2L1 (BCLXL), PIM1, CISH, OSM1 and SOCS2) in our historic microarray dataset of 117 samples from treatment-naïve pediatric T-ALL patients.³⁰ The leukemia subtype and the IL7R signaling pathway mutational status of these patients, as previously determined,^6,30 are displayed in Figure 2A. Cluster analysis based on the expression of STAT5 target genes separated T-ALL patients into two major clusters, with cluster B representing patients with high expression of STAT5 target genes. In relation to leukemic subtype, cluster B was significantly enriched for TLX3-rearranged leukemias (P<0.0001), whereas TLX1-/NKX2.1- rearranged patients or TAL1/2- and LMO1/2-rearranged T-ALL patients were strongly associated with cluster A (P=0.0126 and P<0.0001, respectively) (Figure 2B). Moreover, activating IL7R signaling mutations in IL7R, JAK1, JAK3 and/or STAT5B genes were also strongly associated with STAT5 transcriptional activity (P<0.0001) (Figure 2C), in line with previous observations that these mutations are also associated with TLX3-rearranged patients.^6,30,32 Enrichment of TLX3-rearranged leukemia and IL7R signaling mutations in samples that display high expression of these STAT5 target genes was verified by performing similar analysis on RNA-sequencing data available from the St. Jude database (Online Supplementary Figure S1).³² These results highlight the contribution of IL7R, JAK1, JAK3 and/or STAT5B mutations to the activation of STAT5 in T-ALL. In relation to ex-vivo prednisolone cytotoxic response levels, as determined for 84 out of these 117 T-ALL patients’ samples, we observed that the prednisolone concentrations lethal to 50% of the T-ALL cells (LC₅₀) were comparable for patients with high or low STAT5 transcriptional activity (Figure 2D). Moreover, no difference in event-free survival was observed between the two groups (log-rank test P=0.24) (Figure 2E). This suggests that the predicted STAT5 activity does not alone predict for sensitivity to steroid treatment at diagnosis, either by not having an immediate impact on steroid sensitivity or due to the presence of alternative resistance mechanisms.

STAT5 activation does not drive steroid resistance in IL7R-mutant cell lines

To further explore the relation between the STAT5-regulated transcriptional program and steroid resistance, we generated SUPT-1 or P12-Ichikawa derivative lines that can be induced to express wild-type IL7Rα (IL7R^WT) or cysteine-mutant IL7Rα molecules (IL7R^{PILT240-244RFCPH} or IL7R^{PIL240-242QSPSC}) following exposure to doxycycline. We previously demonstrated that induced expression of cysteine-mutant IL7Rα molecules, in contrast to wild-type IL7Rα, provoked steroid resistance in otherwise steroidsensitive T-ALL SUPT-1 cells.¹² Overexpression of mutant IL7Rα resulted in STAT5 phosphorylation and thus STAT5 activation, while no effect was seen upon wild-type IL7Rα overexpression (Figure 3A). Examining the activation of STAT5 target genes, both mutant IL7Rα molecules strongly induced BCL2 and BCL-XL protein expression, but only in the presence of prednisolone. To further validate this concept, we studied the expression of BCL2 and BCLXL and various other STAT5 target genes such as PIM1 and CISH in SUPT-1 cells expressing IL7R^WT or a cysteine-mutant IL7R variant. Expression of these STAT5 target genes was also studied in the presence of prednisolone with or without targeted inhibitors of JAK1/2 (ruxolitinib), MEK (selumetinib) and/or AKT (MK2206) (Figure 3B).^6,7,12 Again, steroid treatment of IL7R-mutant overexpressing cells led to a nearly two-fold increase in BCL2 and BCLXL expression (Figure 3C). As expected, the JAK inhibitor ruxolitinib effectively blocked downstream STAT5B, MAPK-ERK and AKT pathways (Online Supplementary Figure S2), and also blocked steroid-dependent upregulation of BCL2, BCLXL, CISH and PIM1 (Figure 3C). Treatment with the MEK inhibitor selumetinib, the AKT inhibitor MK2206, or their combination could not block transcriptional upregulation of these genes. Similar results were obtained in the context of IL7-induced signaling in the T-ALL cell line CCRFCEM, indicating that this reflects a general mechanism in T-ALL cells (Online Supplementary Figure S3A, B).

Figure 2.STAT5 transcriptional activity in pediatric patients with T-cell acute lymphoblastic leukemia does not predict steroid resistance. (A) Unsupervised clustering of STAT5 target gene expression of 117 treatment-naïve pediatric patients with T-cell acute lymphoblastic leukemia (T-ALL) using Affymetrix U133 Plus2 microarrays. Genes used for clustering were already described STAT5 target genes BCL2, BCL2L1 (BCLXL), PIM1, CISH, OSM1 and SOCS2. For most of these genes, multiple probes were used. Euclidean distances were used to determine clusters. The leukemic subtype and IL7R-pathway mutational status for these patients had been previously determined. Cluster B represents patients with the highest STAT5 transcriptional activity. The four major T-ALL subtypes (characterized by specific oncogenic rearrangements, clustered as described by Homminga et al.³⁰ are represented in green, yellow, pink and blue. Patients who harbor a mutation in the IL7R, JAK1, JAK3 or STAT5 gene are indicated in red (versus no mutations in these genes in black). (B) Representation of the four major leukemic T-ALL subtypes in both STAT5 transcriptional clusters. Statistical differences were calculated by the χ² test; *P<0.05, ** P<0.01, *** P<0.001, **** P<0.0001. (C) IL7R/JAK1/JAK3/STAT5B mutation status of primary patients’ blasts related to STAT5 transcriptional clusters. Statistical differences were calculated by the χ² test; *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001. (D) Statistical analysis of in vitro prednisolone sensitivity of patients in low versus high STAT5 transcriptional clusters (Mann-Whitney-test). (E) Kaplan-Meier analysis of event-free survival between patients in low versus high STAT5 transcriptional clusters (long-rank test). LC50: concentrations lethal to 50% of cells.

Unexpectedly, whereas expression of both mutant IL7R isoforms strongly raised steroid resistance, this effect seemed independent of upregulation of pro-survival proteins BCL2 and BCL-XL via STAT5; while the JAK inhibitor ruxolitinib inhibits upregulation of BCL2 and BCL-XL and reverts steroid resistance in this model, both AKT-inhibitor treatment and combined MEK- and AKT- inhibitor treatment also sensitized SUP-T1 cells to prednisolone treatment (Figure 3D) regardless of the STAT5-driven and steroid-induced BCL2 and BCL-XL induction (Figure 3C). Similar findings were observed for T-ALL CCRF-CEM cells, in which these inhibitors reverted IL7-induced steroid resistance independently of the expression levels of BCL2 and BCL-XL via STAT5 (Online Supplementary Figure S3B,C). Therefore, we conclude that STAT5 activation and consequent upregulation of BCL2 and BCL-XL do not have a direct negative impact on steroid sensitivity as studied in SUPT-1 cell line models (Figure 2D).

STAT5N642H-induced SUPT1 cells remain steroid sensitive despite enhanced BCL2 and BCLXL levels

To further exclude a direct role of active STAT5 signaling in relation to steroid resistance, we generated doxycycline-inducible SUPT-1 and P12-Ichikawa derivative lines that express wild-type STAT5B (STAT5^WT) or the constitutively active mutant isoform STAT5^N642H (Figure 4A). This activating and transforming STAT5B mutation, found in 4/117 patients of our patient cohort (i.e., 3.4%), solely activates the STAT5-signaling pathway and therefore serves as a STAT5-focused model.^6,33-37 Expression of STAT5B^N642H, but not STAT5^WT, led to activated STAT5B signaling without affecting the MAPK-ERK or PI3K-AKT signaling pathways (Figure 4A, Online Supplementary Figure S4A).^6,33-37 In line with these results, expression of STAT5B^WT was ineffective at activating the expression of downstream target genes, while STAT5B^N642H induced the expression of canonical STAT5 target genes (Figure 4B, Online Supplementary Figure S4B). Moreover, the expression of STAT5 target genes was further boosted in STAT5^N642H cells upon steroid treatment (Figure 4B). In addition to enhanced expression of BCLXL and other classical STAT5 target genes such as PIM1, CISH and OSM1, BCL2 expression was greatly dependent on the steroid treatment. Despite upregulation of BCL2 and BCL-XL anti-apoptotic molecules in the presence of active STAT5 signaling (STAT5^N642H) and steroid treatment, the cytotoxic steroid response did not change for SUPT-1 or P12-Ichikawa cells (Figure 4C, D). This again demonstrates that STAT5 signaling alone does not provoke steroid resistance despite the upregulation of anti-apoptotic BCL2 and BCL-XL.

NR3C1-STAT5B co-binding enhances the expression of canonical STAT5 target genes

The enhanced STAT5 transcriptional activity during steroid treatment has previously been attributed to steroid-induced transcriptional upregulation of IL7RA by NR3C1, which can further enhance STAT5 activation and subsequently the upregulation of BCL2 in the presence of IL7.¹⁸ As the induction of BCL2 and IL7RA also occurs in parallel in IL7-exposed and inhibitor-treated CCRF-CEM cells (Figure 3B, Online Supplementary Figure S5A), an alternative hypothesis could be that NR3C1 and STAT5B act in a single transcriptional complex and co-regulate an identical set of target genes. To explore this possibility, we performed NR3C1 co-immunoprecipitation experiments using SUPT-1 STAT5B^WT and STAT5B^N642H cells and identified that NR3C1 and STAT5B could indeed physically interact (Figure 5A). This interaction seemed to be independent of the phosphorylation status of STAT5B, as both (unphosphorylated) wild-type STAT5 and (phosphorylated) mutant STAT5B bound NR3C1 to equal extents following exposure to steroids. Moreover, NR3C1 also bound to STAT5 in the non-prednisolone treated condition (Online Supplementary Figure S5B). To explore whether (mutant) STAT5 and NR3C1 could bind to the same transcriptional regulatory regions, we performed ChIP-sequencing for NR3C1 and STAT5. While wild-type STAT5 was expressed in a non-phosphorylated, transcriptionally inactive form, we observed that it could still bind to many regulatory sites that were also bound by the phosphorylated and constitutively active STAT5B^N642H isoform (Figure 5B). While binding of STAT5B^WT or STAT5B^N642H to regulatory sites was independent of steroid exposure, NR3C1 only bound to DNA upon prednisolone treatment, which was expected based on its nuclear translocation upon interaction with steroids.¹ We identified genomic regions that were predominantly bound by either STAT5B or NR3C1, denoted as STAT5 or NR3C1 unique binding sites. Interestingly, many genomic regions were bound by both STAT5B and NR3C1 upon steroid exposure, suggesting that STAT5B and NR3C1 can co-regulate a common set of target genes. Motif analysis for NR3C1 and NR3C1/STAT5 binding sites revealed significant enrichment of classical glucocorticoid response element motifs³⁸ and STAT5 motifs, but not at binding sites that were uniquely bound by STAT5 (Figure 5C, Online Supplementary Table S1).

Figure 3.Steroid treatment and inhibition of MAPK-ERK and PI3K-AKT signaling enhance the expression of STAT5 target genes in mutant IL7R cell lines. (A) Ligand-independent STAT5B pathway activation in wild-type and cysteine-mutant IL7R overexpressing SUPT-1 cell lines. Protein band intensity for BCL2 and BCL-XL are represented relative to the condition without doxycycline induction and without prednisolone treatment (-dox-pred). Loading controls were equivalent between different membranes used. (B) Graphic representation of Online Supplementary Figure S1: selective inhibition of IL7R^MUT activated pathways by ruxolitinib (1 mM, JAK1/2 inhibitor), selumetinib (1 mM, MEK inhibitor) and MK2206 (0.5 mM, AKT inhibitor). (C) Expression of STAT5B target genes in wild-type and mutant IL7R overexpressing SUPT-1 cells. Cells were treated with targeted inhibitors for 30 min before doxycycline induction. Steroid-exposed cells were treated with prednisolone (250 mg/mL) for 16 h. Representative data from two independent experiments with identical conditions. (D) Steroid sensitivity of wild-type and mutant IL7R overexpressing SUPT-1 cells in the absence or presence of targeted inhibitors. Steroid sensitivity was determined by a 4-day MTT read-out. Representative data from a biological duplicate.

We then specifically studied co-occupancy of STAT5 and NR3C1 at the regulatory sites of selected STAT5 target genes (Figure 5D, Online Supplementary Figure S5C). For BCLXL and PIM1, we identified STAT5 regulatory sites that were bound by STAT5B^N642H but not by wild-type STAT5B, indicating that binding of STAT5 at these regulatory sites is dependent on STAT5 phosphorylation. Interestingly, we found that NR3C1 could also bind to these sites upon steroid exposure in mutant STAT5 cells. For the BCL2 locus, two NR3C1 peaks were identified and bound by NR3C1 only in STAT5^N642H cells, with one site harboring a STAT5 binding motif. Interestingly, no clear STAT5 binding was observed at the BCL2 locus, suggesting that STAT5 might be unable to independently regulate BCL2 expression in the absence of steroid treatment or nuclear NR3C1. This is also reflected by the lack of BCL2 expression in STAT5^N642H-over-expressing cells in the absence of steroid treatment (Figure 4B, Online Supplementary Figure S3B). We did not identify glucocorticoid response elements at any of the NR3C1-bound sites in BCLXL, PIM or BCL2, suggesting that NR3C1 may be recruited to these binding sites by binding to STAT5B rather than physical binding to DNA. Combined, these data revealed that NR3C1 can interact with STAT5B and bind in or near STAT5-bound genomic sites, even in the absence of a conserved glucocorticoid response element binding motif. Binding of STAT5 to various ‘canonical STAT5 target genes’ seems dependent on its active (phosphorylated) form, which enables recruitment of NR3C1 to these sites following steroid exposure to further boost the expression of these target genes.

Figure 4.Overexpression of STAT5B^N642H does not provoke steroid resistance despite enhanced expression of anti-apoptotic molecules upon steroid treatment. (A) Activation of STAT5B signaling and expression of anti-apoptotic Bcl2 family proteins in STAT5B wild-type and mutant overexpressing SUPT-1 cells. Protein band intensity for BCL2 and BCL-XL represented relative to the condition without doxycycline induction and without prednisolone treatment (-dox-pred). Loading controls were equivalent between different membranes used. (B) Expression of STAT5 target genes in STAT5B^WT and STAT5B^N642H overexpressing SUPT-1 cells in the absence and presence of prednisolone treatment (16 h, 250 mg/mL). Data from biological triplicates with standard deviation indicated. (C) Steroid sensitivity of wild-type and mutant STAT5B overexpressing SUPT-1 and P12 ICHIKAWA cell lines. Steroid sensitivity was determined by a 4-day MTT read-out. Data from biological triplicates with standard deviation indicated.

Figure 5.NR3C1 co-regulates the expression of transcriptional STAT5 target genes by co-binding at STAT5-regulated genes. (A) Immunoblot analysis of NR3C1-immunoprecipitation in STAT5B^WT and STAT5B^N642H SUPT-1 cells. Steroid-exposed cells were treated with prednisolone (250 mg/mL) for 16 h. Loading controls were equivalent between different membranes used. (B) Reads per genomic content-normalized centered heatmap of unique and overlapping NR3C1 and STAT5 chromatin-immunoprecipitation (ChIP) sequencing binding peaks of doxycycline-induced STAT5B^WT and STAT5B^N642H SUPT-1 cells with (+dox+pred) or without (+dox-pred) prednisolone treatment (16 h, 250 mg/mL). Blue represents more reads, indicating confirmed binding of protein at the genome. (C) MEME-ChIP motif analysis of NR3C1 and STAT5 motifs significantly enriched in NR3C1/STAT5 unique or overlapping peak sets. (D) ChIP sequencing identified binding of NR3C1 and STAT5 transcription factors at the STAT5 target genes BCL2, BCLXL and PIM1.

In a reciprocal manner, ChIP-sequencing analysis identified STAT5 binding near NR3C1-binding sites that lack conserved STAT5 binding motifs. Examination of various NR3C1 target genes including BIM, KLF13, FKBP5 and GILZ and the newly proposed NR3C1 target genes BMF and MCL1 revealed various NR3C1 binding sites that harbor conserved glucocorticoid response element sequences in BIM, KLF13 and FKBP5 (Figure 6A, B, Online Supplementary Figure S6A, B). Remarkably, binding of STAT5^WT and STAT5^N624H was observed at glucocorticoid response element sites of KLF13 and FKBP5 that were not flanked by conserved STAT5 binding sequences. This suggests that STAT5 molecules can be recruited to these sites by (direct) interaction with NR3C1, independently of their activation (phosphorylation) state. The significance of STAT5 binding at these sites remains unknown, as the expression of NR3C1 target genes, including BIM, following steroid exposure remained unaffected upon co-incubation with inhibitors of JAK1/2, MEK, or AKT or of both MEK and AKT (Figure 6C).

NR3C1-induced BIM binds to enhanced BCL2 and BCL-XL protein in STAT5BN642H cells

We then explored whether the pro-apoptotic steroid response (upregulation of pro-apoptotic BIM via NR3C1) could effectively counter-balance the strong upregulation of anti-apoptotic BCL2 and BCL-XL molecules downstream of activated STAT5 during steroid treatment. To do this, we performed BIM-immunoprecipitation experiments in non-induced or doxycycline-induced STAT5B^N642H SUPT-1 cells that were exposed to prednisolone treatment. As shown in Figure 7, steroid treatment led to increased expression of BIM in its active and unphosphorylated form (total lysate, lanes 2 and 4). In line with previous results,¹² active BIM strongly bound to BCL2, BCL-XL and MCL1 (immunoprecipitation, lane 2). Steroid treatment again enhanced the expression of BCL2 and BCL-XL in doxycycline-induced STAT5B^N642H cells (total lysate, lane 4). However, immunoprecipitated BIM could effectively bind the upregulated BCL2 and BCL-XL, despite their increased expression (immunoprecipitation, lane 4). Since STAT 5^N642H overexpression in SUPT-1 cells does not confer steroid resistance (Figure 4C, D), BIM seems to counter the enhanced STAT5/NR3C1 driven anti-apoptotic induction of BCL2 and BCL-XL and therefore preserves a sensitive steroid response. Thus, our study highlights that steroid sensitivity is not solely defined by the upregulation of anti-apoptotic proteins, but is regulated by a tight balance between anti- and pro-apoptotic molecules.

Discussion

Resistance to synthetic steroids remains a problem in the treatment of pediatric ALL. Aberrant activation of the IL7R signaling cascade is frequently observed in T-ALL and is due either to production of IL7 by stromal cells in the leukemia microenvironment⁷ or to activating mutations in the IL7R signaling pathway.^6,12 Activation of the IL7R pathway has been strongly related to steroid resistance via various pro-survival mechanisms downstream of the PI3K-AKT, JAK-STAT and MAPK-ERK signaling pathways. Understanding these mechanisms of resistance is of utmost importance, as it may reveal new therapeutic strategies to revert steroid resistance using targeted agents.

Figure 6.Intact NR3C1 transcriptional activity outbalances steroid-dependent enhanced expression of anti-apoptotic molecules. (A) Chromatin immunoprecipitation (ChIP) sequencing identified binding of NR3C1 and STAT5 transcription factors at glucocorticoid response elements of the NR3C1 target genes BIM, FKBP5 and KLF13. (B) Gene expression of NR3C1 target genes (MCL1, KLF13, FKBP5, BIM, BMF and GILZ) in mutant STAT5 overexpressing SUPT-1 cells in the absence or presence of overnight steroid treatment (250 mg/mL). Data from biological triplicate with standard deviations indicated. (C) Gene expression of NR3C1 target genes (BIM and GILZ) in wild-type and mutant IL7R overexpressing SUPT-1 cells in the absence and presence of targeted inhibitors. Cells were treated with targeted inhibitors 30 min before doxycycline-induction. Steroid-exposed cells were treated with prednisolone (250 mg/mL) for 16 h. Representative data from biological duplicates.

In the last decade, various steroid resistance mechanisms that are linked to the IL7R pathway have been uncovered. Interestingly, many of these mechanisms involve the inactivation or activation of pro-apoptotic (e.g., BIM, BMF) and anti-apoptotic Bcl-2 family members (e.g., BCL2, BCLXL, MCL1), respectively. In healthy lymphocytes, steroid-induced upregulation of BIM is sufficient to neutralize anti-apoptotic Bcl-2 family members, to antagonize their function and to effectuate cellular apoptosis during early T-cell selection processes.^39,40 Overexpression of BCL2 in healthy thymocytes has been demonstrated to outbalance pro-apoptotic BIM, resulting in the abnormal survival and resistance of cells to steroid-induced cell death.^41-44 Similarly, upregulation of BCL2 downstream of STAT5 has been proposed to drive IL7-induced steroid resistance in T-ALL patients’ samples.^{7,1 8} Here we demonstrate that upregulation of BCL2 and BCLXL by STAT5 is not sufficient to induce steroid resistance in various T-ALL cellular models. Previously, we demonstrated that aberrant activation of the MEK-ERK signaling pathway is one of the major drivers of steroid resistance in IL7R-, JAK1- and RAS-mutant cells, since activated ERK phosphorylates and inactivates proapoptotic BIM.¹² MEK inhibitors (and to a lesser extent the JAK inhibitor ruxolitinib) can re-sensitize IL7-induced or IL7R signaling mutant T-ALL patients’ cells to steroid treatment. Other studies also revealed the importance of the pro-apoptotic BIM, since epigenetic silencing of BIM also results in steroid resistance in T-ALL.^15,16

Our current study suggests that the anti-apoptotic response of activated STAT5 signaling by itself is insufficient to drive steroid resistance. In fact, combined MEK and AKT inhibition in mutant-IL7R T-ALL cell models resensitized these cells to steroid treatment, despite elevated BCL2 and BCLXL expression levels by activated STAT5 in steroidtreated conditions. The NR3C1-induced expression of BIM seems sufficient to counteract the increased expression of BCL2 and BCL-XL that is caused by steroid/STAT5, through direct binding. Activation of STAT5 may contribute to steroid resistance when the steroid-induced pro-apoptotic BIM response is disabled by (epi-)genetic changes or by MAPK-ERK and PI3K-AKT signaling events.^12-16 This is also supported by our observation that active STAT5 signaling, as measured by the expression of various downstream target genes in primary T-ALL patients’ cells, is not associated with steroid resistance in T-ALL patients.

Interestingly, we did not observe clear STAT5 binding at the BCL2 locus, despite this harboring a STAT5-binding motif. We demonstrate that this site is bound by NR3C1 upon steroid treatment, only in STAT5-activated cells, resulting in high BCL2 expression. Our results are in line with those of a previous study,²⁰ suggesting that STAT5 itself is unable to regulate the expression of BCL2 despite the presence of a STAT5-binding site. Combined with our observation that BCL2 is induced in steroid-treated STAT5^N642H cells, we suggest that nuclear NR3C1 is vital for BCL2 expression in STAT5-activated cells. Our study reveals that expression of BCL2 and the enhanced expression other STAT5-regulated genes following exposure to steroids reflect a common mechanism in different STAT5-activated T-ALL models. Whereas regulatory sites of many genes can be bound by both inactive (non-phosphorylated) STAT5^WT and constitutively active (phosphorylated) STAT5^N642H, various canonical STAT5 target genes were only bound by active STAT5. Remarkably, NR3C1 is also recruited to these sites upon steroid treatment, which points to important direct co-regulation of NR3C1 in the induction of STAT5 target genes. Although we only analyzed a limited set of genes in more detail, we observed that NR3C1 can be recruited to these sites in the absence of glucocorticoid response element sequences, confirming that NR3C1 can act as a transcriptional co-factor. In combination with our immunoprecipitation results, we suggest that this co-occupancy is caused by direct binding between NR3C1 and STAT5, rather than genomic binding of NR3C1 and STAT5 at overlapping sites. In line with this, we also observed that STAT5 can be recruited to NR3C1-bound target genes irrespective of its activation state. Various of these sites did not contain STAT5-binding motifs, again implying that STAT5 binds directly to NR3C1 rather than binding directly to specific DNA sequences. In contrast to the expression of STAT5 target genes, binding of STAT5 at NR3C1-bound target genes did not further enhance the relative expression of NR3C1 downstream target genes, and the expression of these genes also remained unaffected by ruxolitinib treatment. Matched transcript ome studies are required to determine whether the interaction between NR3C1 and STAT5 can regulate the expression of specific genes, or whether it causes more general aberrant transcriptional activity. Additional research is also required to study the exact mechanism by which NR3C1 and STAT5 interact at these genomic regions, and in what ways this physical interaction depends on Tyr694 phosphorylation or a non-dimer form of STAT5.

Figure 7.Steroid-induced BIM in STAT5-activated cells binds the enhanced expression of BCL2 and BCL-XL. Immunoblot analysis of BIM-immunoprecipitation in STAT5B^N642H SUPT-1 cells (left: total lysate, right: BIM-immunoprecipitation). Steroid-exposed cells were treated with prednisolone (250 mg/mL) for 16 h. The protein abundancy of BCL2, BCL-XL and BIM was quantified, and the expression in cells without doxycycline induction that were treated with prednisolone was taken as the reference. Loading controls were equivalent between different membranes used.

Co-binding of STAT5 with NR3C1 has previously been reported to occur during glucocorticoid-induced signaling in T cells, mammary and hepatocyte epithelial cells.^45-48 Apart from acting as a transcription factor, STAT5 is known to regulate chromatin accessibility of the TCRγ-locus or immunoglobulin heavy-chain,^49,50 and induces epigenetic changes at EZH2- and SUZ12-binding sites in STAT5B^N642H mutated cells.³⁷ Therefore, STAT5-dependent chromatin remodeling might render (certain) gene sites accessible for NR3C1. This would give an alternative explanation why certain STAT5-target gene sites are only bound by NR3C1 in SUPT-1 cells that overexpress the constitutively active mutant STAT5 molecule.

Our results warrant more detailed research into the mechanisms of STAT5 and NR3C1 cooperation in STAT5-induced and NR3C1/steroid-induced signaling. The complexity of STAT5-regulated transcription is exemplified by the interaction of STAT5 with the transcription factor TLX1 in T-ALL cases that harbor a NUP214-ABL1 fusion. Co-binding of these transcription factors drives the expression of BCL2 and MYC in these leukemias.⁵¹ STAT5 therefore seems to act as a versatile transcription factor that can interact with various other transcription factors to promote the transcription of STAT5-regulated genes. In fact, TLX1 and STAT5 are predominantly found at the same enhancer sites in NUP214-ABL1-positive leukemias, and BET protein inhibitors diminish the expression of BCL2 and MYC. Moreover, deacetylase inhibitors can also inhibit STAT5-mediated transcription by relocating BET proteins.⁵² These findings and our data highlight the plasticity of STAT5 as a transcription factor, and its ability to induce local or broad epigenetic changes in leukemia.

Figure 8.Model of STAT5/NR3C1-regulated transcription and steroid responsiveness. (A) Sole STAT5 activation leads to the transcription of STAT5-target genes (such as BCLXL, but also PIM1, OSM1 and CISH). (B) In the presence of steroid treatment, NR3C1 functions as a transcriptional factor to induce the expression of NR3C1-target genes (such as BIM) at glucocorticoid response elements. In the presence of active STAT5 signaling, NR3C1 also functions as a co-factor at STAT5 transcriptional binding sites (such as BCLXL). Moreover, NR3C1 facilitates the expression of BCL2 by binding at the BCL2 locus in the presence of active STAT5 signaling. Our work suggests that NR3C1-induced BIM expression is sufficient to neutralize enhanced (STAT5/NR3C1-regulated) BCL-XL and BCL2 expression, therefore securing a steroid-sensitive phenotype. (C) In the presence of BIM-inactivating events by active PI3K-AKT and/or MAPK-ERK signaling (in the context of active IL7R-signaling), BIM abundancy or function is diminished. As a result, BIM is unable to neutralize enhanced BCL-XL and BCL2 expression, thus resulting in steroid resistance.

In conclusion, we identified that NR3C1 can directly coregulate the expression of STAT5 target genes including BCL2 and BCLXL, without influencing the steroid response. We demonstrated that NR3C1/STAT5-regulated expression of BCL2 and BCLXL can be counterbalanced by the upregulation of the pro-apoptotic and steroid response gene BIM (Figure 8). Therefore, in the absence of other BIM-inactivating mechanisms by aberrant IL7 signaling, STAT5 activation itself seems insufficient to provoke steroid resistance in TALL. Multi-omic screening of primary patient material, before and after induction therapy, is required to extrapolate our findings to patients.

Footnotes

• Received November 23, 2021
• Accepted May 30, 2022

Correspondence

°Current afliation: Acerta Pharma, Oss, the Netherlands

J.PP. Meijerink j.meijerink@prinsesmaximacentrum.nl

Disclosures

No conflicts of interest to disclose.

Contributions

JvdZ designed the study, performed research, and wrote the manuscript. JBG, RH, WKS, EV, LG, MN, and VP performed research. RP and VC provided critical input and wrote the manuscript. JM designed and supervised the study and wrote the manuscript.

Data-sharing statement

The data that support the findings of this study are openly available at http://www.ncbi.nlm.nih.gov/geo/ under accession number GSE26713 (microarray data) and at Gene Expression Omnibus under GEO series accession number: GSE171976, h ttps://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE17197 6 (ChIPseq files).