BNIP3L in myelodysplastic syndromes and acute myeloid leukemia: impact on disease outcome and cellular response to decitabine

Changes in apoptosis of hematopoietic progenitors are thought to contribute to the progression of myelodysplastic syndromes (MDS) and acute myeloid leukemia (AML).¹ BCL2/adenovirus E1B 19 kDa interacting protein 3 (BNIP3) and BNIP3-like (BNIP3L) are closely related members of the BCL2 family of proteins capable of inducing cell responses that may lead to apoptosis and/or autophagy.² Due to this dual role in cell death and survival, BNIP3 and BNIP3L have been described to execute both oncogenic and tumor suppressor activities.² The downregulation of BNIP3 and BNIP3L are related to their aberrant methylation in acute leukemia and lymphoma,^3,4 respectively. However, they have been poorly investigated in myeloid malignancies. Herein, we evaluated the expression and methylation status of BNIP3 and BNIP3L in MDS and AML, their impact on MDS prognosis, and their participation in the induction of apoptosis by decitabine.

BNIP3 and BNIP3L expression was evaluated by quantitative PCR in bone marrow samples from patients with MDS (n=65), AML with myelodysplasia-related changes (AML-MRC) (n=13), de novo AML (n=61) and from healthy donors (n=26). Comparisons were performed with a Mann-Whitney test. MDS patients were stratified according to the WHO 2008 classification into two groups: RA/RARS/del(5q)/RCMD and RAEB-1/RAEB-2. This study was approved by the local Ethics Committee. All patients were untreated at the time of sample collection and their characteristics are described in the Online Supplementary Table S1.

BNIP3 transcripts were significantly reduced in the RA/RARS/del(5q)/RCMD MDS group and in the RAEB-1/RAEB-2 MDS group in comparison with healthy donors, P=0.013 and P=0.001, respectively. AML-MRC and de novo AML patients presented no changes in BNIP3 expression (Figure 1A). BNIP3L was decreased in the RAEB-1/RAEB-2 MDS group in comparison to healthy donors, P=0.018. A significant reduction of BNIP3L transcripts was also observed in AML-MRC and in de novo AML compared with healthy donors, P=0.005, P=0.0003, respectively (Figure 1B).

Figure 1.

BNIP3 and BNIP3L expression and methylation status in bone marrow cells from MDS and AML patients. BNIP3 (A), BNIP3L (B) and HIF-1α (C) mRNA expression in total bone marrow cells from healthy donors, and from patients with diagnosis of myelodysplastic syndromes (MDS) stratified by ‘WHO (World Health Organization) 2008 classification and acute myeloid leukemia (AML). The “y” axis represents the relative BNIP3, BNIP3L or HIF-1α mRNA expression normalized to HPRT endogenous control; sample from a healthy donor was used as the calibration sample. Each dot represents one subject and horizontal lines indicate medians. The numbers of subjects studied and the P values (Mann–Whitney test) are indicated. (D) Correlation between BNIP3 or BNIP3L and HIF-1α expression in primary bone marrow cells from MDS and AML patients. Spearman correlation test; N, P and r values are indicated in the graph. (E) BNIP3L mRNA levels in total bone marrow cells from five MDS patients at diagnosis and after disease progression. The “y” axis represents the fold change of BNIP3L mRNA expression normalized to HPRT endogenous control at diagnosis (calibration sample) and after disease progression; WHO 2008 classification and bone marrow blast percentage are indicated. Data are presented in arbitrary units (a.u.). (F) Overall survival of MDS patients categorized as median of BNIP3L expression levels (Kaplan-Meier curves). Patients were subgrouped by higher BNIP3L expression (above median; 1.73) and lower BNIP3L expression (below median). P value is indicated. (G) BNIP3 and BNIP3L methylation in bone marrow samples from healthy donors and from MDS and AML patients. Percentage of positive methylation and total number of samples are indicated. RA: refractory anemia; RARS: refractory anemia with ringed sideroblasts; del(5q): MDS with isolated del(5q); RCMD: refractory cytopenia with multilineage dysplasia; RAEB-1: refractory anemia with excess blasts-1; RAEB-2: refractory anemia with excess blasts-2; MDS: myelodysplastic syndromes; MRC; myelodysplastic-related changes.

BNIP3 and BNIP3L are regulated by HIF-1α,⁵ which has been proposed to function as a tumor suppressor in AML.^6,7 Therefore, HIF-1α transcript levels were also evaluated in our cohort of patients. HIF-1α expression was reduced in the RA/RARS/del(5q)/RCMD and in the RAEB-1/RAEB-2 MDS groups compared to healthy donors, P=0.007 and P=0.014, respectively. Patients with de novo AML also presented decreased levels of HIF-1α in comparison with healthy donors (P=0.0007) (Figure 1C). Spearman’s rank correlation coefficient indicated a moderate correlation between HIF-1α and BNIP3L (r=0.41, P<0.0001), but not with BNIP3, in patient bone marrow samples (Figure 1D). Bone marrow samples from five MDS patients from our cohort who progressed according to the WHO 2008 classification were analyzed before and after progression. Interestingly, BNIP3L expression decreased after the progression of four of these patients (Figure 1E). BNIP3 and HIF-1α expression varied after progression of these patients (data not shown).

A Cox regression model was used to estimate overall survival (OS) and event-free survival (EFS) of MDS patients. OS and EFS were defined as previously described.^8,9 With an average follow-up of 37 months, univariate analysis showed that lower levels of BNIP3L (below the median) negatively impacted OS, along with IPSS and advanced age (all P<0.05). Multivariate analysis showed that lower BNIP3L expression was an independent prognostic factor for worse OS (all P<0.05). As expected, IPSS and advanced age remained as independent predictors for OS. Lower levels of BNIP3L and IPSS also appeared as independent prognostic factors for worse EFS (all P<0.05). Kaplan-Meier analysis indicated a 5-year OS of 66% versus 32% for MDS patients with a higher BNIP3L expression versus a lower BNIP3L expression, respectively (Figure 1F). Similar results were observed for 5-year EFS (data not shown). Neither BNIP3 nor HIF-1α expression affected the survival outcomes of MDS patients (Table 1).

Table 1.

Univariate and Multivariate analyses of survival outcomes for MDS patients.

Investigation of the methylation status of BNIP3 and BNIP3L promoter regions was performed as previously described.^3,4 BNIP3 methylation was observed in 9% (3/33) RA/RARS/del(5q)/RCMD MDS, 17% (2/12) RAEB-1/RAEB-2 MDS, 17% (1/6) AML-MRC and 18% (7/39) de novo AML patients, and 0/13 healthy donors. BNIP3L methylation was investigated in a total of 7 heathy donors and 37 patients (RA/RARS/del(5q)/RCMD MDS =12, RAEB-1/RAEB-2 MDS =3, AML-MRC =2 and de novo AML =20), and was detected in 6 de novo AML patients, 1 RCMD and 1 RAEB MDS patient (Figure 1G). Part of these samples were randomly analyzed for BNIP3 (n=20) and BNIP3L (n=17) methylation by high resolution melting, and the same results were observed (data not shown).

Despite the low frequency of BNIP3 and BNIP3L methylation in MDS and AML, we evaluated their gene expression in leukemia cells treated with the hypomethylating agent decitabine. Decitabine is one of the standard treatments for high-risk MDS¹ and has been considered for some cases of AML,¹⁰ but the hematological response is still limited.¹ Decitabine is known to act by sequestering DNA methyltransferase 1 (DNMT1), leading to the restoration of silenced genes. However, its direct effects upon cell death may also be important in order to achieve clinical efficacy.¹¹ In this regard, interventions that potentiate decitabine-induced apoptosis may be beneficial to increase clinical response. CD34⁺ bone marrow cells of 4 de novo AML patients from our cohort were treated with decitabine (0.1, 0.5, 1 and 5μM) for 48 hours. Among these patients, only 1 patient presented BNIP3 and BNIP3L methylation, whereas the others presented no methylation of either gene. Apoptotic cells were measured using annexin V and propidium iodide. Apoptosis was mainly induced with the highest doses of decitabine (1μM and 5μM) and positively correlated with BNIP3L expression (r=0.59, P=0.006). In contrast, BNIP3 expression was induced by decitabine only in the methylated sample (Online Supplementary Figure S1).

Among a panel of myeloid cell lines, a strong correlation was observed between BNIP3 and BNIP3L expression (r=0.78, P=0.01), but only U937 cells presented BNIP3 hemi-methylation, whereas BNIP3L methylation was not detected (Figure 2A,B). BNIP3 and BNIP3L expression was also evaluated in HL60, K562, U937 and HEL cell lines treated with 1mM or 5mM decitabine. BNIP3L expression was particularly increased by decitabine treatment in U937 and HEL cells, which presented higher apoptotic levels. BNIP3 upregulation was observed in U937 cells only, concomitantly with partial demethylation of its promoter region. Autophagy was evaluated by acridine orange staining and was induced by decitabine in all tested cell lines (Online Supplementary Figure S2). Taken together, these results suggest that BNIP3L expression is increased by decitabine treatment in myeloid cells that undergo apoptosis, even in the absence of gene promoter methylation.

Figure 2.

Effects of BNIP3 and BNIP3L silencing on the apoptosis and autophagy of U937 cells treated with decitabine. (A) BNIP3 and BNIP3L mRNA expression in KG1, HL60, U937, K562, HEL and THP1 myeloid leukemia cells. HPRT was used as the housekeeping gene and a KG1 cell line was used as the calibration sample. Data are presented in arbitrary units (a.u.). (B) Methylation status of BNIP3 promoter region in myeloid cell lines evaluated by methylation-specific PCR. Methylation status of BNIP3L promoter region in myeloid cell lines evaluated by combined bisulfite restriction analysis (COBRA). For all methylation analyses, a universal methylated human DNA was used as positive control. (C–G) U937 cells were silenced for BNIP3 (shBNIP3), BNIP3L (shBNIP3L) or transduced with lentivirus-mediated shRNA control (shControl). Analyses were performed after 48h of decitabine treatment (0, 1 or 5μM). (C) Expression of BNIP3 and BNIP3L was assessed by qPCR (quantitative polymerase chain reaction) and (D) western blot. (E) Apoptosis was detected by flow cytometry using Annexin V/PI staining. (F) Acidic vesicular organelles were detected by flow cytometry using acridine orange staining. All data shown are the mean±SD of 4 independent experiments. *P≤0.05 decitabine-treated cells vs. untreated cells under the same conditions, #P≤0.05 shBNIP3L vs. shControl cells; Student’s t-test. (G) Western blot analysis of BCL2, SQSTM1/p62, Beclin 1, LC3I/II expression, caspase 3 (total and cleaved), and cleaved PARP1. The membranes were reprobed with the specific antibody for detection of the respective protein or actin (loading control). PCR: polymerase chain reaction; mRNA: messenger RNA.

We next tested whether BNIP3 and BNIP3L could participate in the induction of apoptosis by decitabine. For this purpose, the effects of decitabine treatment (1μM or 5μM for 48h) were investigated in U937 cells silenced for BNIP3 or BNIP3L. Gene silencing was performed using lentivirusmediated shRNA, as previously described,⁹ and was confirmed by quantitative PCR and western blot. In addition to gene expression, protein levels of BNIP3 and BNIP3L were increased by decitabine (Figure 2C,D). Surprisingly, the combination of BNIP3L silencing (but not BNIP3) and decitabine treatment slightly induced apoptosis further (Figure 2E). Decitabine-induced autophagy was not altered by BNIP3 or BNIP3L knockdown in U937 cells (Figure 2F). As expected, decitabine treatment reduced BCL2 expression and increased the expression of SQSTM1/p62, LC3I/II and the cleaved forms of caspase 3 and PARP1. In agreement with annexin V staining, higher levels of cleaved caspase 3 and cleaved PARP1 were detected in U937 cells silenced for BNIP3L and treated with 5μM decitabine (Figure 2G), confirming that BNIP3L silencing and decitabine cooperate to sensitize leukemia cells to apoptosis.

Taken together, we provide the first evidence of BNIP3 being downregulated in bone marrow cells from MDS patients. BNIP3L expression was not only decreased in advanced MDS and AML samples, but was also an independent prognostic factor in MDS. These results are in accordance with BNIP3 and BNIP3L proapoptotic functions, and suggest that BNIP3 is most likely to participate in the imbalance of apoptosis found in MDS cells, whereas BNIP3L may be involved in the decreased apoptosis of high-risk MDS or AML. On the other hand, an essential role of BNIP3L was described during erythroid differentiation.^12,13 BNIP3L knockout mice presented anemia, erythroid hyperplasia and reticulocytosis associated with a critical defect of mitochondrial clearance during reticulocyte maturation.¹³ In addition, the failure of mitochondria removal has been proposed to be associated with increased ROS production and subsequent progression to AML.¹⁴ In accordance with this, decreased autophagic flux and increased amounts of defective mitochondria have recently been reported in human AML blasts.¹⁵ Thus, we speculate that the low expression of BNIP3L in advanced MDS and in AML may be associated with deficient mitophagy that could favor the progression of these diseases.

The low frequency of methylation in BNIP3 and BNIP3L promoter regions indicates that this process is not the single reason for their decreased expression in patient bone marrow samples. In fact, histone acetylation of the BNIP3 promoter region was already associated with gene expression in leukemia cell lines.³ Even when unmethylated, BNIP3L expression was increased in leukemia cells that underwent apoptosis induced by decitabine. Surprisingly, BNIP3L silencing slightly amplified the apoptotic effect of decitabine, contradicting the pro-apoptotic function of this protein. Since autophagy was apparently not modulated, this finding cannot yet be explained. In this sense, further studies will be important to elucidate the participation of BNIP3L in the response of MDS and AML patients to hypomethylating agent-based therapy.