Differential epigenetic regulation between the alternative promoters, PRDM1α and PRDM1β, of the tumour suppressor gene PRDM1 in human multiple myeloma cells

Multiple myeloma (MM) is a B-cell neoplasm that is characterized by the accumulation of malignant plasma cells in the bone marrow. The transcription factor PRDM1 is a master regulator of plasma cell development and is considered to be an oncosuppressor in several lymphoid neoplasms. The PRDM1β isoform is an alternative promoter of the PRDM1 gene that may interfere with the normal role of the PRDM1α isoform. To explain the induction of the PRDM1β isoform in MM and to offer potential therapeutic strategies to modulate its expression, we characterized the cis regulatory elements and epigenetic status of its promoter. We observed unexpected patterns of hypermethylation and hypomethylation at the PRDM1α and PRDM1β promoters, respectively, and prominent H3K4me1 and H3K9me2 enrichment at the PRDM1β promoter in non-expressing cell lines compared to PRDM1β-expressing cell lines. After treatment with drugs that inhibit DNA methylation, we were able to modify the activity of the PRDM1β promoter but not that of the PRDM1α promoter. Epigenetic drugs may offer the ability to control the expression of the PRDM1α/PRDM1β promoters as components of novel therapeutic approaches.


Overexpression of PRDM1α but not PRDM1β induces the apoptosis of U266 cells. Impaired
PRDM1β ability to repress important genes implicated in the cell cycle (e.g., the MYC gene) has been described 16 . Moreover, we demonstrated that PRDM1β isoform expression is augmented in multiple myeloma cells isolated from patient samples 28 , which was confirmed and correlated with the disease status of myeloma patients in a subsequent study 38 . However, the effect of PRDM1α and/or PRDM1β overexpression on proliferation and apoptosis has not been previously tested in myeloma cells. To this end, U266, NCI-H929 and RPMI-8226 cells were transfected with vectors expressing the PRDM1α and PRDM1β isoforms. Figure 1A shows that the overexpression of PRDM1α, but not PRDM1β, increased the apoptosis of the U266-transfected cells. However, neither PRDM1α nor PRDM1β overexpression affected the proliferation rate (Fig. 1B). Considering these observations and because it is virtually impossible to design specific knockdown assays for both isoforms, as their cDNA coding sequence only differs in 3 codons, we reasoned that the lower PRDM1α/PRDM1β ratio in myeloma cells, compared to that in normal cells, caused the accumulation of malignant cells due to inhibited apoptosis, not an increase in cell proliferation. More sophisticated manipulations would determine whether this reduction in apoptotic events contributes to the development of the myeloma. Therefore, we decided to analyse the unexplored transcriptional regulation of the PRDM1β isoform and the effect of epigenetic regulators on the expression of both isoforms in myeloma cells.
Identification of a repressive regulatory element of PRDM1β expression. To understand the regulatory mechanisms of transcriptional regulation for both isoforms and considering that the upstream regulatory elements and factors controlling the transcription of the human PRDM1β promoter have not been previously described, we decided to identify the putative cis regulatory elements that are critical for the basal expression of PRDM1β. To this end, we cloned and analysed a genomic fragment ~ 2 Kbp upstream of the transcription start site (TSS) in the PRDM1 gene extracted from GenBank, sequence accession number AL022067 ( Fig. 2A). To evaluate the transcriptional mechanisms, the fragment was gradually truncated to create a series of 5′-enddeletion constructs cloned from positions − 2077 to + 41 (pGL3-BamHI), − 1343 to + 41 (pGL3-XbaI), − 678 to + 41 (pGL3-KpnI) − 420 to + 41 (pGL3-S1), − 304 to + 41 (pGL3-S2), − 124 to + 41 (pGL3-S3), and − 49 to + 41 (pGL3-S4) (numbered from the TSS according to the sequence AY198415 16 ). All inserts were cloned upstream of the firefly luciferase reporter gene in a pGL3-basic vector (Fig. 2B). These constructs were transiently transfected into U266 myeloma cells, which expresses endogenous PRDM1 16,28 . As shown in Fig. 2B, the most striking finding was the complete loss of promoter activity observed with the pGL3-S1 construct but not with the pGL3-S2 construct. The transcriptional recovery observed by the removal of the segment − 420/− 304 suggested that a ciselement with repressive actions on PRDM1β was present in this region (Fig. 2B). Similar results were obtained using the PRDM1β-expressing NCI-H929 cell line ( Supplementary Fig. S1). RPMI-8226 was not used in this set of experiments because, as explained below, it did not express the PRDM1β isoform. After a sequence analysis of the S1-S2 fragments using Alibaba2 software based on the TRANSFAC database 39 , we identified several motifs as potential candidates for cis-acting elements ( Supplementary Fig. S2). In addition to general transcription factors such as GATA, SP1, and OCT1, we observed a possible binding site for the transcription factor FOXA1 (also known as HNF3A), which is implicated in the establishment of lineage-specific transcriptional enhancers and programmes 40 . However, this DNA motif is shared among the FOX family members, and therefore, Scientific RepoRtS | (2020) 10:15899 | https://doi.org/10.1038/s41598-020-72946-z www.nature.com/scientificreports/ we referred to this sequence indistinctly as the FOX or FOXA1 motif hereafter. Furthermore, this binding site matched one of the two half-sites in the consensus sequence for TP53 binding previously described by another group 41 ( Fig. 2A). Therefore, we decided to mutate this DNA motif as shown in Fig. 2C. Using luciferase assays, we confirmed the repressive role of this element, as the corresponding mutated FOXA1-site construct, which is homologous to the pGL3-S1 construct (pGL3-S1mutFOXA1), recovered the promoter activity to a level similar to that of the pGL3-S2 and pGL3-S3 constructs (Fig. 2B).
Repressive elements of the PRDM1β promoter do not bind TP53. Next, we further explored the potential role of the FOX-binding site in PRDM1β gene regulation. Thus, we prepared two probes from different regions of the S1-S2 construct that contained either the putative wild-type FOX-binding site (wtFOXA1) or the mutated binding site (mutFOXA1) (see sequences in Fig. 3A). The labelled probes were used in EMSAs with nuclear extract (NE) from U266 cells. As shown in Fig. 3B, different retarded band patterns were observed for the wtFOXA1 (lanes 1-2) and mutFOXA1 (lanes 7-8) probes, indicating that one or more factors were able to bind the wild-type but not the mutated sequences. Then, we investigated whether some of these bands were retarded because of TP53 binding to the wild-type probe. Thus, competition assays were conducted using wtFOXA1 (lane 3), mutFOXA1 (lane 4) and a double-stranded oligonucleotide containing the consensus binding site of TP53 (lane 5) as cold fragments (Fig. 3A). As expected, the retarded bands were abrogated with the wild-type version of the competitor probe, whereas the mutFOXA1 and TP53 cold fragments were unable to effectively compete with the labelled probe, demonstrating the specificity of the protein binding to the FOXA1 motif; in addition, the unidentified factor did not bind the putative TP53-binding site. This negative result was confirmed in a supershift experiment using two different antibodies against human TP53 (Fig. 3A, lane 6 and Supplementary Fig. S3), in which no changes in the band pattern were observed. Other related proteins (TP63 and TP73) were not detected (data not shown). Moreover, the anti-FOXA1 antibody did not lead to any supershift or eliminated retarded bands ( Supplementary Fig. S3), indicating that these bands did not contain any of these transcription factors. To further corroborate that the retarded band corresponded to a factor other than the TP53 protein, additional EMSAs were performed using the wtFOXA1 and TP53 probes with U266 NE and recombinant TP53 protein. Figure 3C shows the different patterns of the bands obtained with the U266 NE and with recombinant TP53 using the wtFOXA1 probe (lane 1 vs. 3). After adding an anti-TP53 antibody to the Under the same experimental conditions using a putative TP53 probe, the same retarded band was observed regardless of whether U266 NE or recombinant TP53 was used (lanes [5][6][7][8]. This assay also confirmed the expression of endogenous TP53 in the U266 cells. These results indicated that the retarded band observed with the wtFOXA1 probe was not a consequence of TP53 binding to the probe. Furthermore, HeLa cells, which do not express TP53, were transfected with constructs containing the human TP53 cDNA, in either the wild-type form (wtTP53) or a mutated dominant negative form (TP53DN), to examine their effects on the regulation of the PRDM1 promoters. Thus, the PRDM1α and PRDM1β transcript levels in the HeLa cells transfected with wtTP53, TP53DN or an empty vector were analysed by RT-qPCR, and no difference was observed between them ( Supplementary  Fig. S4A), suggesting that, at least in this cell line, TP53 did not regulate PRDM1 gene transcription. Exogenous TP53 and TP53DN protein expression was confirmed by western blotting (Supplementary Fig. S4B). Therefore, in contrast to what has been described in cell line-derived colon cancer 41 , TP53 does not seem to have a role in PRDM1 transcriptional regulation. Future proteomic approaches should be applied to determine the identity of the regulatory factor that binds the repressive cis-element of the PRDM1β promoter.
Promoter methylation status shows a higher correlation with PRDM1β expression than with PRDM1α expression. In a previous report, we showed that the expression level ratio of PRDM1β/ PRDM1α was altered in plasma cells from bone marrow samples of multiple myeloma patients 28 . Due to the relevant role of epigenetics in cancer and tumorigenesis, we investigated DNA methylation patterns in a more natural environment (as opposed to the in vitro assays described above) in our cellular preparations. To correlate their respective expression levels to the methylation status of their promoters, we examined the methylation pattern in both promoters within the same samples isolated from several cell lines and primary cells from healthy donors and patients (naïve B-cells, PCs and MM-PCs). In total, 35 and 12 CpG dinucleotides in the PRDM1α and PRDM1β promoters, respectively, were selected as potentially methylable for bisulfite sequencing analysis (Fig. 4A). The PRDM1α promoter was hypermethylated compared to the PRDM1β promoter in all tested cell types. Thus, cells well known to be negative for PRDM1α expression, such as the Raji and Daudi lymphoma cell lines, and normal B cells, showed high methylation levels, ~ 68%, ~ 77% and ~ 89%, respectively (Fig. 4A). Strik- Positions are relative to the TSS according to the sequence AY198415. The first exon in the PRDM1β isoform is shown as a filled black box, and exon 4 (in isoform PRDM1α but common to both isoforms) is indicated as an empty box. b Relative luciferase activity levels of the PRDM1β promoter constructs. Human U266 cells were transiently transfected with the indicated reporter constructs. The fold change in luciferase activity was normalized to the signal obtained with a pGL3basic vector (without any regulatory insert). The results represent the mean ± SEM of at least five experiments with triplicate samples. Significant induction (p < 0.05) is indicated by an asterisk. The pGL3-S1 mutFOXA1 construct was identical to pGL3-S1, except for a point mutation in the FOXA1-binding motif (c).
PRDM1β, but not PRDM1α, is regulated by the 5-aza-dC treatment of myeloma cells. The results revealed that, in general, the methylation levels of the PRDM1α promoter are higher than those of the PRDM1β promoter regardless of the expression of PRDM1 transcripts. To corroborate the observation suggesting that promoter methylation status can preferentially affect the transcription of PRDM1β relative to that of PRDM1α, an RT-qPCR assay was designed in myeloma cells treated with the DNA methyltransferase inhibitor 5-aza-dC (Fig. 5A). The results showed that treatment with 5-aza-dC increased the expression of the PRDM1β transcript isoform but had no apparent effect on the PRDM1α transcript level (Fig. 5B). These results supported the observation that promoter methylation is relevant for the expression of PRDM1β but not PRDM1α, as the latter isoform is present in all cells, even those with a hypermethylated promoter.
Histone marks associated with PRDM1 gene transcription. In addition to CpG methylation, we examined additional epigenetic mechanisms, such as histone posttranslational modification, as potential contributors to PRDM1 gene regulation. Thus, we obtained cross-linked chromatin from U266 cells (expressing both PRDM1α and PRDM1β isoforms) and RPMI-8226 cells (expressing only the PRDM1α isoform). Quantitative PCR-based ChIP assays with antibodies against mono-or tri-methylated histone H3 Lys4 (H3K4me1 or H3K4me3), di-methylated H3 Lys9 (H3K9me2) and acetylated H3 Lys9 (H3K9ac) revealed some differences in the histones deposited on the PRDM1α and PRDM1β promoters in the U266 and RPMI-8226 cells (Fig. 6). The levels of H3K4me1 (a histone mark associated with enhancer regions but associated with gene repression when located at promoters 42 ) and H3K9me2 (also associated with transcriptional repression) were relatively higher on the PRDM1β promoter in the RPMI cells than on the promoter in the U266 cells, which is consistent with the lack of expression in the former cell line. In agreement with the latter result, the RPMI cells also showed greater enrichment of H3K27me3 (although the difference was not significant). No differences in these histone marks were observed in the PRDM1α promoter. In contrast, the H3K4me3 and H3K9ac levels (both associated with active transcription) were equivalent in both promoters in both cell lines. Overall, the ChIP results suggested that H3K9me2 and/or H3K4me1 were associated with the repression of PRDM1β, but not PRDM1α, in the RPMI cells. In the U266 cells, which exhibited a predominance of active-associated marks, such as H3K4me3 and H3K9ac, both PRDM1α and PRDM1β proteins were expressed.

Discussion
The transcriptional activity of PRDM1 is essential for the differentiation and maintenance of long-lived plasma cells 3,4 , whose downregulation has been associated with several types of cancer. Therefore, PRDM1 is considered a tumour suppressor gene 23,[43][44][45][46][47] . The two major isoforms, PRDM1α and PRDM1β, are expressed in myeloma cells, while only the PRDM1α isoform is expressed in non-tumorigenic plasma cells 11,28 . PRDM1β is a truncated  www.nature.com/scientificreports/ isoform that is through to compete with the full-length isoform PRDM1α, which is overexpressed in myeloma cells, leading to the hypothesis that PRDM1β may be functionally relevant in myeloma cells. Nevertheless, there are few studies linking the expression of PRDM1β to myeloma 16,28,38 . In this study, we tested, for the first time, the effect of the overexpression of PRDM1α and PRDM1β isoforms in myeloma cells. We observed that PRDM1α, but not PRDM1β, induced increased apoptosis without apparently affecting cell proliferation in myeloma. These observations led us to consider an alternative explanation for the maintenance of myeloma cells: Rather than an increase in proliferation, expression of the PRDM1β isoform may inhibit apoptosis. However, some contradictory results were obtained in other studies in which PRDM1α altered the proliferation of HCT116 41 and SW620 47 cells but not RKO colon cancer cells 48 , and ectopic expression of PRDM1 in a chicken fibroblast cell line inhibited cell proliferation 49 . Depending on the cellular context, further studies on the mechanisms of action of the PRDM1 www.nature.com/scientificreports/ isoforms will be important for understanding their roles in the regulation of apoptosis, cell proliferation and the cell cycle.
In the present study, luciferase promoter analysis identified a putative cis-element in the promoter region -304 to -420 relative to the TSS. The in silico analysis identified a possible binding site for the FOX transcription factors in positions -345 to -334 within that region. Moreover, this site partially overlapped with a previously proposed DNA binding motif for TP53 41 ; in the previous work, the authors described a transcriptional regulation loop through which PRDM1 is activated by TP53, which is then repressed by PRDM1. However, we were unable to reproduce TP53 involvement in the regulation of PRDM1 in the EMSAs, an in vitro assay in which proteins bind DNA motifs solely by sequence affinity, or in the overexpression experiments, in which we did not observe any increase/decrease in PRDM1α or PRDM1β levels in HeLa or U266 cell cultures with western blotting or RT-qPCR ( Supplementary Fig. S3 and data not shown, respectively). This lack of reproducibility may be due to, at least, two reasons: (1) the PRDM1/TP53 autoregulatory loop may be dependent on the cell line studied (HTC116 cells in the case of the original publication) 41 , and (2) the previous study referred to PRDM1 activation or repression without clearly distinguishing between the two isoforms PRDM1α and PRDM1β, despite the identification of the TP53 binding site in the PRDM1β promoter. However, we failed to identify a transcription factor bound to this promoter in the supershift assays (TP53, Fig. 3 and Supplementary Fig. S2) or in a preliminary mass spectrometry analysis (not shown).
In addition, the previous experiments were based on exogenous DNA sequences that did not properly reproduce the cellular chromatin context. Here, we report, for the first time, the methylation status of alternative promoters of the PRDM1 gene from isolated myeloma cells. We found that the PRDM1α promoter was relatively hypermethylated (range 40-90%) in all the cell lines tested regardless of the expression of the PRDM1α transcript or protein. However, the PRDM1β promoter generally presented with a much lower methylation level (10-25%) with significant differences in the methylation of specific CpG positions in the cells expressing and not expressing PRDM1β. CpG-5, CpG-9 and CpG-12 clearly had greater methylation enrichment in the PRDM1β-negative cells than in PRDM1β-positive cells. No difference in the methylation of any specific CpG position was observed in the PRDM1α promoter. In summary, the following conclusions are drawn based on our methylation analysis. First, the overall level of methylation at the PRDM1α promoter does not determine its expression in all cells tested; for instance, the Raji cell line (~ 68% methylated) does not express PRDM1α, while the NCI-H929 cell line (~ 68%) and MM cells (~ 74%) both highly express PRDM1α. Second, even though the global methylation level of the PRDM1β promoter (< 25%) is much lower than that of the PRDM1α promoter (> 40%), the methylation level does not appear to be decisive of the relative expression levels of both isoforms, although one or both promoters may be activated by 5-aza-dC (this article and previous works 24,25,27 ). In addition to methylation, other elements, such as specific transcription factors and miRNAs, may be critical for the imbalanced expression of PRDM1α/ PRDM1β in tumours 50,51 . Nevertheless, differing results of the methylation status of the PRDM1α promoter have been reported. The hypermethylation of the PRDM1α promoter in EBV + BL cell lines, including the Daudi cell line, which is hypermethylated as described in Fig. 4A, is associated with low PRDM1 expression 25 . Nonetheless, in contrast to our work, the authors observed only trace or no methylation in normal B-cell subsets (i.e., GC cells and memory and naïve B cells), although it is widely reported that these cell subsets do not express PRDM1α, which is a marker of differentiated PCs. In other reports, the low PRDM1α transcript levels were not the result of promoter hypermethylation in diffuse large B-cell lymphoma 51 or chronic lymphocytic leukaemia 52 , but the opposite results have also been reported in NK-cell malignancies 24,53 . In conclusion, the general lack of consistency between promoter methylation and the gene expression of PRDM1α in previous works does not support a relevant role for CpG methylation in the regulation of this isoform. This conclusion is supported by the results of the cell culture treatment with the methyltransferase inhibitor 5-aza-dC, which significantly increased the expression of PRDM1β but not PRDM1α (Fig. 5).
Another important finding in this study is the differential posttranslational modifications in the histones at PRDM1 promoters in the PRDM1β-expressing and non-expressing cell lines. Indicative of promoter activation, similar levels of H3K4me3 and H3K9ac were found in both promoters in U266 cells (PRDM1α positive and PRDM1β positive) and RPMI-8226 cells (PRDM1α positive and PRDM1β negative). However, the enrichment of H3K4me1, H3K9me2, and, to a lesser extent, H3K27me3 was much higher at the PRDM1β promoter in the RPMI-8226 cells, suggesting that the abundance of these modifications may regulate the transcriptional activity of the PRDM1β promoter towards repression. In general, the H3K4me1 histone mark has been associated with enhancer activity, particularly in B-cell fate 54 , and it may also be possible that the PRDM1β promoter acts as an enhancer/promoter region "poised" to be activated while awaiting additional signalling (e.g., acetylation of H3K9). PRDM1 expression has been considered an oncogene in MM 55 , but in-depth analysis of the differential expression of its isoforms has been lacking, until the present work was performed. We and other researchers have analysed public microarray data comparing normal plasma cells and samples from multiple myeloma, and no correlations were observed for low or high PRDM1 expression levels with the prognosis of the disease 56 . Unfortunately, these analyses were mainly based on microarray data, which were not designed to discriminate between PRDM1α or PRDM1β isoforms, and new analysis data sets (RNA-seq, new microarrays, etc.) comprising possible alternative isoforms are necessary to discern whether differential PRDM1 expression may be related to the pathophysiology of the disease. In conjunction with the transcriptional regulation described in this study, other post-transcriptional regulatory mechanisms involving miRNAs may be key to regulating the expression of PRDM1, as altered expression of some miRNAs targeting PRDM1 mRNA has been described during plasma cell differentiation and in malignant plasma cells 55,57 . Whether these miRNAs specifically modulate the PRDM1α or the PRDM1β isoform remains to be explored.
In summary, our results indicated that PRDM1 gene promoters are modulated through a broad range of genetic and epigenetic control mechanisms that depend on the tumour cell line, supporting further investigation In the case of novel treatments of multiple myeloma (Fig. 7), the different transcriptional activity found in the PRDM1α and PRDM1β promoters in response to treatment with epigenetic modifiers may pave the way to the generation of more specific and tailored epigenetic-based therapeutic strategies for this disease by favouring the expression of PRDM1α over that of PRDM1β.

Materials and methods
Human cell lines. The human cell lines U266, NCI-H929, RPMI-8226, HeLa, Daudi and Raji were obtained from The European Collection of Authenticated Cell Cultures (ECACC, Sigma-Aldrich, Madrid, Spain) and maintained in RPMI 1640 medium (Gibco Life Technologies, Carlsbad, CA) supplemented with 10% heat-inactivated foetal bovine serum (FBS), 2 mM glutamine, 100 IU/ml penicillin and 100 μg/ml streptomycin (Gibco) at 37 °C in a humidified atmosphere containing 5% CO 2 and were routinely verified by morphology and periodically checked for Mycoplasma contamination.

Patient tissue samples and normal cells.
Multiple myeloma PCs (MM-PCs) were purified from donated bone marrow aspirates from patients diagnosed with MM (excluding MGUS or SMM) 58 using monoclonal antibody anti-CD138-coated microbeads (Miltenyi-Biotec, Auburn, CA, USA). Normal PCs were obtained from discarded bone marrow specimens with no abnormalities at diagnosis after FAC sorting CD138 + CD19 +/− CD38 +++ cells. Naïve human B-lymphocytes were isolated from peripheral blood samples from normal volunteers after T-cell depletion by a rosette technique and FAC-sorted to obtain CD19 + CD27 − IgD + cells.  (Table 1), were selected on the basis of the sequence of the human PRDM1 genomic DNA accession number AL022067. The resulting PCR fragment was digested with BamHI/BglII and cloned into a pGL3-basic firefly luciferase reporter vector (Promega, Madison, WI) at the BglII site to obtain the pGL3-BamHI construct. This construct was digested with the NheI/XbaI restriction enzymes to obtain the pGL3-XbaI construct, which was subsequently digested with the KpnI restriction enzyme to obtain the pGL3-KpnI construct. The pGL3-S1, pGL3-S2, pGL3-S3 and pGL3-S4 constructs were generated by PCR using the following forward oligonucleotide primers, respectively, in Table 1: PRDM1β-S1-F, PRDM1β-  www.nature.com/scientificreports/ S2-F, PRDM1β-S3-F and PRDM1β-S4-F (all including a restriction site for the enzyme SmaI at the 5′ ends) together with the PRDM1β-BglII-R oligonucleotide as the reverse primer ( Table 1). The resulting PCR products were digested with SmaI/BglII restriction enzymes and cloned into a pGL3-basic plasmid. Primary sequences of the inserts were validated by Sanger sequencing. All PRDM1β promoter constructs were transfected into U266 cells using the Nucleofector Kit C Program X-005 with an Amaxa nucleofector system following the manufacturer's instructions (Lonza, Barcelona, Spain) or into NCI-H929 cells using a Gene Pulser II apparatus (Bio-Rad, Madrid, Spain) set at 400 µF and 240 V, with a time constant of t ~ 20 ms. The cells were harvested after 48 h, and luciferase activity was measured using a dual luciferase reporter assay kit (Promega). Each transfection experiment was performed in triplicate with at least two different plasmid preparations.
Chromatin immunoprecipitation assays. The chromatin immunoprecipitation (ChIP) procedure was performed using a SimpleChIP Enzymatic Chromatin IP Kit as recommended in the manufacturer's instructions (Cell Signaling Technologies, Barcelona, Spain). Briefly, RPMI-8226 cells or U266 cells (40 × 10 6 ) were incubated with 1% formaldehyde for 10 min at room temperature. Chromatin was sheared to an average size of 200 to 1000 bp by enzymatic digestion and sonication with a Sonifier-150 microtip probe (Branson Ultrasonics, Danbury, CT) using the following parameters: 1 pulse × 15 s, output control 2.5 for U266 cells and 2 pulses × 15 s, output control 2 for RPMI-8226 cells maintained on ice for 30 s between pulses. After a spinning step to reduce debris, immunoprecipitated cross-linked complexes were prepared using 3 µg of the following antibodies specific for histone modifications: anti-H3K9me2 and anti-H3K9ac (EpiGentek, Farmingdale, NY), anti-H3K4me3 and anti-H3K4me1 (Abcam, Cambridge, UK) and negative rabbit IgG (Cell Signaling Technologies). After an incubation overnight at 4 °C on a rocking platform, 30 μl of ChIP-grade protein G agarose beads were added and incubated for 2 h at 4 °C with rotation. Then, the bound antibody-protein-DNA complexes were washed, and after cross-linking reversal and proteinase K digestion, each individual immunoprecipitated complex was purified using spin columns. Real-time PCR was performed with the corresponding specific primers indicated in Table 1.
Promoter methylation analysis. The methylation of the PRDM1α and PRDM1β promoters was assessed by bisulfite sequencing. Genomic DNA was extracted from all cell types analysed using a ZR-Duet DNA/RNA Miniprep kit (Zymo Research, Irvine CA), and then, 500 ng of DNA was used for the bisulfite modification with a premium bisulfite kit according to the manufacturer's instructions (Diagenode, Belgium). Bisulfite sequencing nested-PCR was performed to assess the methylation of the CpG islands in the PRDM1α and PRDM1β promoters using PRDM1α and PRDM1β bisulfite-specific primers (Table 1) and Q5 High-Fidelity 2X Master Mix (New England Biolabs, Ipswich, MA). The amplified DNA was cloned into a pSpark I vector (Canvax, Córdoba, Spain), and more than 10 individual clones were Sanger sequenced. The obtained sequences were analysed by the online BISMA (bisulfite sequencing DNA methylation analysis) program 59 .

5-Aza-dC treatment.
RPMI-8226, U266, NCI-H929 or MM-isolated cells were cultured with or without the inhibitor 5-aza-2′-deoxycytidine (5-aza-dC; 0.5, 1, or 2 µM) (Sigma), which was initially dissolved in DMSO. The cells were cultured at 37 °C in a 5% CO 2 incubator for 48 h. After incubation, the viability of the cells was measured using Trypan blue, and the variations in the expression of the PRDM1 transcript isoforms were studied by RT-qPCR. All reactions were performed with identical conditions in triplicate.

Retrotranscription and RT-qPCR.
For the PRDM1α and PRDM1β transcript analysis, the total RNA from the cultured cells was purified using NucleoSpin RNA XS (Macherey-Nagel, Düren, Germany) and reverse transcribed using random hexamers with a Maxima H Minus First Strand cDNA synthesis kit (Thermo Fisher, Waltham, MA). Then, the cDNAs were assessed with real-time PCR TaqMan probes for PRDM1α, PRDM1β and β-actin (Table 1) and a SensiFAST Probe No-ROX kit (Bioline, London, UK) in a RotorGene 6000 (Corbett, Sidney, Australia). The relative quantitative values were calculated using the 2 −ΔΔCT method 60 .
Western blotting. Western blotting was performed previously described 61 . Fifty micrograms of total protein lysates were added to each well. The following antibodies were used at the indicated dilutions: anti-PRDM1 (sc-66015 from Santa Cruz Biotechnologies, Dallas, TX), anti-β-actin (ab6276 from Abcam), anti-TP53 (P6874) and anti-mouse HRP-conjugated (A2554 from Sigma). The bound antibodies were detected using a Western-Bright Sirius kit (Advansta, Menlo Park, CA) and a ChemiDoc XRS system (Bio-Rad, Madrid, Spain). As indicated, the blots were reprobed with the β-actin or α-tubulin antibody as a loading control.
Electrophoretic mobility shift assays (EMSAs). Nuclear extracts (NEs) were prepared from the cell by hypotonic lysis, followed by high salt extraction of nuclei as previously described 62 . All probes and cold fragments used for the competition assays (Table 1) were obtained by annealing synthesized complementary oligonucleotides with protruding ends and labelled by Klenow filling with [α-32 P]-dCTP (Perkin-Elmer, Madrid, Spain). The DNA-protein-binding assays were carried out as previously described 37 . For the competition studies, 100-fold molar excess competitor DNA was added to the reaction mixture 10 min prior to the labelled probe. Supershift assays were performed by preincubating the NEs with an antibody against TP53 purchased from Sigma (P6874) or Santa Cruz (sc-98) as indicated or anti-FOXA1 purchased from Santa Cruz (sc-6553) for 1 h on ice before the probe addition. As indicated, recombinant human TP53 protein (ref. 506165) from Calbiochem (San Diego, CA) was used instead of the nuclear extracts.
Scientific RepoRtS | (2020) 10:15899 | https://doi.org/10.1038/s41598-020-72946-z www.nature.com/scientificreports/ Apoptosis and cell proliferation assays. U266, RPMI-8226 and NCI-H929 cells were transfected with a pIRES2-EGFP empty vector (Clontech, Takara Bio USA Inc, USA) or a vector containing the cDNAs for PRDM1α or PRDM1β, both with an HA epitope. The cells were harvested after 48 h and treated for apoptosis or proliferation analysis. Apoptosis was determined by Annexin V-allophycocyanin (APC) staining according to the protocol of the manufacturer (ImmunoTools, Friesoythe; Germany). Cell proliferation was determined with a Click-iT plus Edu Alexa 647 flow cytometry assay kit following the manufacturer's instructions (Invitrogen, Carlsbad, CA). Only EGFP + -transfected cells were selected for apoptosis and proliferation analyses by a FACSCalibur flow cytometer with CellQuest software (Becton Dickinson, Franklin Lakes, NJ). To validate the transfections, Western blotting was performed as described above using a conjugate HRP antibody against the HA epitope (ref. A00169, GenScript, Piscataway, NJ).
Statistical analysis. The statistical package SPSS version 21 was used. All variables exhibited a nonparametric distribution after applying the Kolmogorov-Smirnov test; thus, the data were analysed using the nonparametric Mann-Whitney U test (two sided) to determine significant differences between two experimental groups. The sample size (n) for each experimental group/condition is given in the corresponding figure legend.
The results are generally expressed as the means ± SEM, and p < 0.05 was considered statistically significant.