Expression feature of BACH2 is associated with risk stratification and early treatment responses
To determine the clinical relevance of BACH2 in p-ALL, we firstly analyzed the expression values of BACH2 based on one published microarray data from 284 children with ALL at newly diagnosis (ND). Compared with normal BM CD19+CD10+ cells from 4 healthy donors, leukemic cells from p-ALL samples showed reduced BACH2 levels (Fig. 1A). Interestingly, we found lower BACH2 levels in patients with T-cell ALL (T-ALL) who have poorer outcomes than those with B-cell ALL (B-ALL) (28) (Fig. 1B). Among B-ALL samples, patients with unfavorable BCR-ABL1 fusion gene contained remarkably lower BACH2 levels than patients without BCR-ABL1 (Fig. 1C). In contrast, BACH2 expression in patients who have favorable genetic subtype (ETV6-RUNX1+) or low-to-intermediate risk subtype (TCF3-PBX1+) of B-ALL were higher compared with patients who did not (Supplementary Fig. 1A and Fig. 1D-E), suggesting that differential expression of BACH2 could facilitate risk classification of p-ALL.
Given that minimal residual disease (MRD) tracking plays a crucial role in early outcome prediction for p-ALL, we next analyzed the correlation between BACH2 expression and MRD response. As shown in Fig. 1F, patients with lower BACH2 levels at ND were more inclined to occur positive MRD (MRD+) at day 19 (d19) from diagnosis (Fig. 1F), and this correlation becomes more significant at d46 (Fig. 1G). Intriguingly, MRD+ at d19 turned into MRD negative (MRD−) at d46 in patients with higher BACH2 levels in B-ALL compared with those with lower BACH2 levels, and the same is true in T-ALL (Supplementary Fig. 1B-C). Strikingly, BACH2 levels in different subtypes at ND also coincided with the MRD monitoring at d19 or d46. For example, the highest BACH2 levels were observed in patients with TCF3-PBX1 (Fig. 1E), while the smallest proportion of MRD+ patients were found at either d19 or d46 (Supplementary Fig. 1D). These data suggest that aberrant expression of BACH2 at ND is very likely a predictor for early treatment response.
In addition to ND samples, further analysis from another microarray dataset revealed that early MRD response is also predictive for the degree of BACH2 expression at relapse (RE): the higher percentages of MRD at d36, the lower levels of BACH2 at RE (Fig. 1H), and a much stronger inverse correlation between %MRD and BACH2 expression was observed in T-ALL (Supplementary Fig. 1E), suggesting a reciprocal dependency of BACH2 and MRD on their respective role in outcome prediction.
BACH2 is a sensitive predictor of clinical outcome
To validate above microarray analysis, we examined the mRNA and protein levels of BACH2 in an independent cohort of p-ALL samples at ND (n = 12). Indeed, BACH2 levels were significantly lower in p-ALL compared with patients with immune thrombocytopenic purpura (ITP), a non-tumorous hematologic disorder of megakaryocyte without disturbing lymphocytes (Fig. 2A). Of note, in addition to a patient with T-ALL, there is one patient with B/T-cell mixed-phenotype acute leukemia (B/T MPAL), a high-risk subtype of ALL with a uniformly poor outcome (29). BACH2 levels were much lower in T-ALL and B/T MPAL cases compared to B-ALL cases (Fig. 2B), and the lowest expression of BACH2 was observed in a B/T MPAL patient (Fig. 2C). Interestingly, amongst B-ALL cases, there is one special case (Pt #12) that showed the lowest levels of BACH2 compared with the others (Fig. 2D). When reviewing the clinical information for this patient, we discovered that she had a very high tumor burden in peripheral blood at ND (90% of blasts), and passed away soon after induction therapy (Supplementary Table 1). This, together with a close relationship between BACH2 and MRD as indicated above, further supported the possibility that BACH2 may serve as a sensitive predictor for risk classification and clinical outcome, although additional evidence from more clinical samples are needed.
Similarly, immunoblots of BACH2 showed that patients with a favorable subtype (ETV6-RUNX1+) exhibited higher expression of BACH2 compared with patients with unfavorable subtypes (BCR-ABL1+, B/T MPAL and T-ALL) (Fig. 2E), in agreement with both the microarray analysis and the mRNA findings.
Silencing BACH2 increases leukemic cell proliferation and accelerates cell cycle progression
To better delineate the biological roles of BACH2 in leukemic cells, we silenced BACH2 (BACH2KD) in a human pre-B ALL cell line using a lentiviral shRNA-mediated knockdown system. The knockdown efficiency of BACH2 in leukemic cells was evaluated by immunoblots which showed a better knockdown efficiency of BACH2KD-2 than BACH2KD-1 (Fig. 3A); BACH2KD-2 was selected to perform the subsequent experiments. Correspondingly, we generated stable BACH2-overexpressing (BACH2OE) cells (Fig. 3A).
Compared with control cells (BACH2Con), silencing BACH2 significantly increased cell growth, whereas BACH2OE cells showed lower growth rate than the control (Fig. 3B), indicating a potential anti-tumor role of BACH2 in the pathogenesis of leukemia. To elucidate the mechanism involving enhanced cell growth in BACH2KD cells, cell proliferation was analyzed by staining cells with PKH26 dye to track cell division. PKH26 fluorescent labelling was declined rapidly after BACH2 silencing while slowly in BACH2OE cells, indicating that downregulation of BACH2 promotes cell proliferation (Fig. 3C). Further analysis of cell cycle distribution showed approximately 15% more of BACH2KD cells in S-G2/M phases compared with control cells, while the diminished proliferation in BACH2OE cells was associated with decreased S-G2/M population and increased apoptotic Sub-G1 population (Fig. 3D). Intracellular pulse staining for BrdU incorporation further confirmed higher amounts of BACH2KD cells while lower amounts of BACH2OE in S phase (Fig. 3E). These data indicated that silencing BACH2 leads to increased cell proliferation and accelerated cell cycle progression, thus contributing to a dominant growth.
To demonstrate in vivo relevance, we intravenously transplanted manipulated Nalm-6 cells into mice (Supplementary Fig. 2). BACH2KD xenografts developed larger spleens (SP) as compared to the BACH2OE and control xenografts (Fig. 3F). Further analysis or these xenografts displayed increased human CD19+ (hCD19+) cells in the SP and BM upon BACH2 silencing, indicating higher leukemia burden in BACH2KD xenografts; by contrast, lower leukemia burden was observed in BACH2OE xenografts (Fig. 3G).
Decreased BACH2 expression confers chemo-resistant properties to p-ALL
We next questioned the implication of reduced BACH2 levels in p-ALL treatment. Based on a published microarray data from 173 p-ALL cases at ND, patients with lower BACH2 levels predisposed to prednisolone resistant (Fig. 4A), and more obviously correlation was found in B-ALL group (Fig. 4B). In addition to clinical samples, leukemic cell lines with decreased BACH2 expression were also likely to occur cytarabine (Ara-C) resistance (Fig. 4C).
To confirm these findings, we tested whether BACH2 blockade contributes to chemoresistance in leukemic cells. Flow cytometry (FCM) analysis revealed a survival advantage of the BACH2KD cells compared to control cells, whereas higher proportion of apoptotic cells were found in the BACH2OE cells (Fig. 4D), indicating that silencing BACH2 contributes to enhanced leukemic cell survival. After introducing Ara-C into leukemic cells, BACH2 deletion displayed lower drug sensitivity by preventing cell apoptosis, which was reversed by BACH2 overexpression (Fig. 4D). These data demonstrated that BACH2 downregulation confers Ara-C resistance properties to leukemic cells by likely increasing the threshold for drug-induced apoptosis.
BACH2 silencing promotes cell adhesion and chemoresistance by altering stromal microenvironment
Bone marrow stromal cells (BMSCs) are regarded as a safeguard to protect BM-resident leukemic cells from chemotherapy-induced apoptosis by producing multiple growth factors and cytokines, leading to stroma-mediated chemoresistance (30–33). Thus, BM microenvironmental remodeling has become a key parameter and prognostic factor in leukemia (34). Since BACH2KD xenografts showed increased leukemia burden to the BM (Fig. 3H), we then used a coculture model of leukemic cells and BMSCs to investigate the effect of BACH2 on cell adhesion and complex leukemia-stroma network, and how such effect modifies the cytotoxicity of anticancer drugs within the surrounding stroma.
Compared with control cells, silencing BACH2 in leukemic cells resulted in a significant increase in cell adhesion to the HS-5 BMSCs, while decreased cell adhesion was observed in BACH2OE cells (Fig. 5A and Supplementary Fig. 3A). Further analysis using coculture media based on a multiplexed flow cytometric system revealed substantial changes in the secretion of many growth factors and cytokines that play pivotal roles in maintenance of normal BM microenvironment (35–37) (Fig. 5B). Coculturing HS-5 with BACH2KD cells resulted in significant upregulation of GM-CSF, IL-6 and IL-8 compared with control, whereas IL-6, IL-8 and MIP-1α were decreased when coculturing BMSCs with BACH2OE cells (Supplementary Fig. 3B). These results were further validated by ELISA assays respectively for single cytokine from independent experiments (Supplementary Fig. 3C). As a result, the coculture media showed protective effects against Ara-C with much higher IC50 values, no matter in control or BACH2KD cells (Fig. 5C), suggesting that BMSCs-secreted cytokines are very likely involved in Ara-C resistance of BM-resident leukemic cells.
To extend our findings to primary cells, we performed experiments with BM cells from two p-ALL patients using a similar coculture setting. Primary cells or drug-resistant BACH2KD cells did get great benefit from these secreted cytokines, because neutralization of IL-8 in coculture media led to decreased cell adhesion to BMSCs (Fig. 5D), whereas GM-CSF- or IL-6-neutralizing antibodies increased Ara-C-derived cytotoxicity (Fig. 5E). These results indicated that stromal microenvironmental alterations have many tumor-promoting effects that not only enhance cell adhesion, but also protect leukemic cells from chemotherapeutic-derived cytotoxicity in ALL.
Proto-oncoprotein FOS is a novel downstream target repressed by BACH2 in pre-B leukemic cells
BACH2 is a basic region leucine zipper (bZIP) protein that forms heterodimers with the small Maf proteins (38). The BACH2-Maf heterodimers repress transcription by binding to DNA sequences termed Maf recognition elements [MARE, 5’-TGCTGA(G/C)TCAGCA-3’) (10, 38). Interestingly, MARE includes a core TPA response element [TRE, 5’-TGA(G/C)TCA-3’] that can be bound by activator protein 1 (AP-1), a dimeric transcriptional activator formed by JUN and FOS (6). Therefore, BACH2 regulates gene expression by competing with AP-1 complex (6). Despite well-established oncogenic functions of AP-1 signaling (39–42), each component of AP-1 complex plays independent roles. JUN involves in tumorigenesis by regulating malignant transformation, apoptosis, angiogenesis and DNA methylation (43–47), whereas FOS, in addition to a similar role as JUN (48–50), plays an extra role in regulating bone cell differentiation and osteoimmunology (51). Mice lacking Fos develop severe osteopetrosis with deficiencies in bone remodeling (52) and exhibit altered B-cell differentiation due to an impaired BM microenvironment (53). In BCR-ABL1-induced leukemia, FOS was recently identified as one of critical mediators for leukemogenesis and imatinib resistance (26). Given a similar effect of BACH2 and FOS on BM microenvironmental regulation, we hypothesized that FOS might correlate with BACH2 in p-ALL.
To test this hypothesis, we started with the correlation analysis of BACH2 expression and FOS expression using published microarray data (n = 238), which showed a significant inverse correlation between them in B-ALL at ND (Fig. 6A). However, we did not find an inverse, but observed a positive correlation between BACH2 and FOS in T-ALL (n = 46) (Supplementary Fig. 4A), implying a totally different regulatory network of BACH2 in T cells.
Next, we detected the mRNA levels of BACH2 and FOS in clinical p-ALL samples, and the same inverse correlation was found in B-ALL at ND, except T-ALL and B/T MPAL cases (Supplementary Fig. 4B and Fig. 6B). Further immunoblots showed differential expression of FOS protein among different subtypes (Fig. 6C), exactly in contrary to the corresponding BACH2 levels in B-ALL group (Fig. 2E). In pre-B leukemic cells, FOS levels were also increased after BACH2 silencing while reduced in BACH2OE cells (Supplementary Fig. 4C). These results indicated that BACH2 is very likely a potential suppressing regulator of FOS in pre-B leukemic cells.
We next wondered whether the FOS gene is a transcriptional target repressed by BACH2 protein. Based on sequence alignment methods, we identified three potential MARE binding sites of the FOS gene within ± 1000 bp, which are located at the proximal promoter (MARE1, -212/-202), the 5’ untranslated region (MARE2, + 32/+43) and the proximal Exon 1 (MARE3, + 181/+192), respectively (Fig. 6D). Truncated luciferase reporters containing different numbers of putative MAREs were constructed respectively. The luciferase activity of pGL3-MARE3 was significantly decreased in cells when co-transfected with BACH2 expression plasmids compared with the controls (Fig. 6D). Further sequencing using CUT&Tag, a next-generation technique to investigate interactions between proteins and DNA instead of a chromatin immunoprecipitation (ChIP) assay, in pre-B Nalm-6 cells confirmed BACH2-FOS interaction (Fig. 6E). Indeed, two most enriched regions (fragments b and c) of the FOS gene containing BACH2-binding sites were further validated by PCR amplification (Supplementary Fig. 5 and Fig. 6F). These findings support that BACH2 suppresses FOS transcription by binding to MARE sites on proximal regions of the FOS gene.
Blocking FOS by small molecule inhibitors sensitizes leukemic cells to Ara-C in xenografts
Given that the FOS gene is a downstream target repressed by BACH2, we then reasoned that FOS might be functionally involved in BACH2-induced BM microenvironmental alterations and chemoresistance. We firstly tested the effects of two chemical compounds targeting FOS, nordihydroguaiaretic acid (NDGA) and curcumin (26), on microenvironmental secretion of cytokines in a coculture setting of BACH2KD cells and BMSCs. Both NDGA and curcumin were effective in suppressing the secretion of GM-CSF, IL-6 and IL-8 in coculture media (Supplementary Fig. 6A). In particular, blocking FOS with NDGA or curcumin obviously sensitized leukemic cells to Ara-C treatment in coculture media, no matter in Ara-C-resistant BACH2KD cells (Supplementary Fig. 6B) or in primary cells (Fig. 7A), suggesting that FOS is a very important mediator responsible for BACH2-induced microenvironmental changes and chemoresistance. The synergistic cytotoxic effects of Ara-C and NDGA/curcumin were further analyzed by combination index (CI) plots, which were all less than 1, indicating dramatic synergistic responses (Supplementary Fig. 7), in which, more synergistic efficacy of Ara-C and NDGA was observed than the combination of Ara-C with curcumin.
Next, we tested whether blocking FOS function is also effective in leukemia xenograft models. Ten days after intravenously transplantation with Ara-C-resistant BACH2KD cells, mice were treated with Ara-C alone or in combination with NDGA (Ara-C + N), curcumin (Ara-C + C) or both (Ara-C + both) for 3 days and euthanized 2 days later (Fig. 7B). Mice treated with Ara-C + N or Ara-C + both significantly reduced splenomegaly after one course of treatment (Fig. 7C). Combined treatment with Ara-C + N or Ara-C + both also let to a more effective inhibition of leukemia burden in SP and BM compared to single-agent Ara-C group (Fig. 7D-E). In a separate cohort, mice treated with Ara-C + N or Ara-C + both showed improved survival compared to other groups (Fig. 7F), indicating that the combined treatment was effective to prolong survival of tumor-bearing animals.