Vieira‑Silva et al. Cancer Cell Int (2019) 19:112 https://doi.org/10.1186/s12935-019-0835-9 Cancer Cell International PRIMARY RESEARCH Open Access Histone variant MacroH2A1 is downregulated in prostate cancer and influences malignant cell phenotype Tânia Soraia Vieira‑Silva1, Sara Monteiro‑Reis1†, Daniela Barros‑Silva1†, João Ramalho‑Carvalho1, Inês Graça1, Isa Carneiro1,2, Ana Teresa Martins1,2, Jorge Oliveira3, Luis Antunes4, Sarah Hurtado‑Bagès5,6, Marcus Buschbeck5,6, Rui Henrique1,2,7 and Carmen Jerónimo1,7* Abstract Background: Prostate cancer (PCa), a major cause of cancer-related morbidity and mortality worldwide and mostly asymptomatic at earliest stages, is characterized by disruption of genetic and epigenetic balance. A better under‑ standing of how those mechanisms orchestrate disease might improve diagnostic and prognostic tools, allowing for improvements in treatment efficacy. Replacement of canonical histones, an epigenetic mechanism, is highly con‑ served among species and altered expression of histones variants (e.g., MacroH2A1) has been associated with tumo‑ rigenesis. H2AFY gene encodes two isoforms of H2A histone variant MacroH2A1: MacroH2A1.1 and MacroH2A1.2. Specifically, MacroH2A1.1 isoform inhibits cell proliferation and promotes cellular differentiation. Because the contri‑ bution of this histone variant to carcinogenesis has been reported in several cancer types, but not for PCa, we aimed to investigate the contribution of MacroH2A1 for prostate carcinogenesis. Methods: MacroH2A1, MacroH2A1.1 and MacroH2A1.2 isoforms and the corresponding splicing regulators transcript levels were evaluated by RT-qPCR, in a tissue cohort composed by PCa, prostatic intraepithelial neoplasia (PIN) and normal prostate cases. Knockdown for MacroH2A1 and MacroH2A1.1 was performed through lentiviral transduction in DU145 cells, and MacroH2A1.1 overexpression was achieved in LNCaP cells by plasmid transfection, followed by functional assays. Biological and/or experimental replicates were performed when necessary, and specific statistical tests were applied to perform data analysis. Results: MacroH2A1.1 transcript levels were downregulated in PIN and primary PCa compared to normal prostate tissues. The same was found for QKI, a MacroH2A1.1’s splicing regulator. Moreover, lower MacroH2A1.1 and QKI expression levels associated with less differentiated tumors (Gleason score ≥ 7). Interestingly, MacroH2A1.1, but more impressively DDX17 (AUC = 0.93; p < 0.0001) and QKI (AUC = 0.94; p < 0.0001), accurately discriminated cancerous from noncancerous prostate tissues. Furthermore, in PCa cell lines, total MacroH2A1 knockdown augmented malig‑ nant features, whereas MacroH2A1.1 overexpression impressively attenuated the malignant phenotype. Conclusions: Overall, our data, derived from primary PCa tissues and cell lines, anticipate a tumor suppressive role for MacroH2A1, particularly for the MacroH2A1.1 isoform, in prostate carcinogenesis. *Correspondence: carmenjeronimo@ipoporto.min‑saude.pt; cljeronimo@icbas.up.pt † Sara Monteiro-Reis and Daniela Barros-Silva contributed equally to this work 1 Cancer Biology & Epigenetics Group, Research Center (CI‑IPOP), Portuguese Oncology Institute of Porto (IPO Porto), Research Center‑LAB 3, F Bdg., 1st Floor, Rua Dr. António Bernardino de Almeida, 4200‑072 Porto, Portugal Full list of author information is available at the end of the article © The Author(s) 2019. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creat​iveco​mmons​.org/licen​ses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creat​iveco​mmons​.org/ publi​cdoma​in/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Vieira‑Silva et al. Cancer Cell Int (2019) 19:112 Page 2 of 13 Keywords: Prostate cancer, Histone variants, MacroH2A1 isoforms, Splicing regulators, Tumor suppressor Background the substitution of canonical histones by variants might Prostate cancer (PCa) is the most common non-cuta- be implicated in prostate carcinogenesis. As opposing neous malignancy in men and a major cause of can- functions for H2AZ and MacroH2A1 in gene regulation cer-related morbidity and mortality worldwide [1]. were reported in cancer cells [29], we investigated the Characteristically asymptomatic at its earliest stages, role of MacroH2A1.1 and MacroH2A1.2 isoforms in PCa when curative-intended therapy is most likely to be suc- development. cessful, PCa is characterized by a broad range of altera- tions in both genetic and epigenetic level [2, 3]. Materials and methods Among epigenetic mechanisms deregulated in cancer, a Patients and clinical samples relevance of changes in histone variants that can replace Tumor samples from 197 Prostate cancer patients and canonical histones (nowadays referred to as replication- 45 PIN lesions were prospectively collected from radical coupled histones) has just started to emerge while still prostatectomy specimens obtained from 2001 to 2006, being poorly understood [4, 5]. Unlike canonical his- at the Portuguese Oncology Institute of Porto, Portu- tones, variants display time-based incorporation at the gal. Among these, 34 patients had both PIN lesions and nucleosomes, are tissue-specific and DNA replication- PCa samples. Immediately after surgery, specimens were independent [6]. The H2A family of histones comprises fully sectioned and “twin fragments” were obtained, one structurally diverse variants, including H2A.X, H2A.Z, for routine histopathological processing and the other MacroH2A1, H2A.Bb, characterized by distinct length, was frozen and stored at − 80 °C. After full mapping of sequence and genome distribution [7, 8]. Deregula- each prostate in FFPE tissue sections, the index tumor tion in H2A variants have been previously implicated in (the tumor nodule in multifocal disease that portrays cancer [4]. The H2AFY gene encodes for MacroH2A1, a the most relevant combination of prognostically relevant histone variant frequently found in repressed chromatin histo-morphological parameters, i.e., stage/local exten- such as the inactive X chromosome in female cells [9, 10]. sion and grade) and PIN lesions were identified. Each Moreover, a subset of genes positively regulated by Mac- PCa case was staged and the corresponding index tumor roH2A1 has also been reported [11–13]. MacroH2A1 was scored according to the Gleason grading system is present in Senescence-Associated Heterochromatic [30]. Then, a frozen fragment corresponding to the index Foci [14], as well as on genes involved in cell cycle [15], tumor and another corresponding to the PIN lesion were pluripotency [16] and development [10, 17]. Addition- selected for RNA extraction. For that purpose, frozen ally, two isoforms of MacroH2A1—MacroH2A1.1 and sections were cut, and H&E stained for allowing for con- MacroH2A1.2—result from the substitution of a single firmation of the presence PIN lesions and index tumor, exon [4, 18]. MacroH2A1.1 is preferentially expressed including Gleason score representability. Then, each tis- over 1.2 in differentiated cells whereas MacroH2A1.2 is sue fragment was trimmed to maximize the yield of tar- more expressed in proliferative cells [19, 20]. The ratio get cells (> 70% of target cells) through serial cutting of of MacroH2A1 isoforms expression levels is regulated thick (10 µm) sections. This procedure was performed by by pre-mRNA splicing regulators. Indeed, without alter- the same experienced uropathologist (RH) and allowed ing global MacroH2A1 expression, QKI promotes Mac- for confirmation that both PIN lesions and tumor areas roH2A1.1 expression, whilst MacroH2A1.2 expression is identified in the frozen sections were representative of promoted by RNA helicases DDX5/DDX17 [21, 22]. the index lesions identified in the routine assessment of Alterations in the expression of total MacroH2A1 or its the prostatectomy specimen. Relevant clinical data was isoforms [19] has been observed in several cancer types, retrieved from clinical charts. Fifteen samples of MNPT, including breast carcinoma [22, 23], melanoma [24], lung used as controls, were obtained from the peripheral zone carcinoma [25] and colorectal carcinoma [26]. Taken of prostates not harboring PCa, obtained from radi- the bulk literature together the MacroH2A1.1 isoform cal cystoprostatectomies performed for bladder cancer emerges as pleiotropic tumor suppressor, by repressing treatment that were submitted to the same procedure as cell proliferation, migration and invasion, whereas the prostate specimens that harbored cancer. role of MacroH2A1.2 is largely cancer-type dependent [18, 27]. Prostate cancer cell lines We have previously demonstrated that the histone H2A Benign prostate cell line RWPE-1 and PCa cell lines variant H2A.Z is upregulated in PCa [28], suggesting that 22Rv1, LNCaP and VCaP, which are androgen receptor Vieira‑Silva et al. Cancer Cell Int (2019) 19:112 Page 3 of 13 (AR) positive, as well as PCa cell lines DU145 and PC-3, ­Biosystems®), and the mean value was used for data characterized as AR negative, were grown for in vitro analysis. assays. RWPE-1 cells were maintained in K-SFM growth medium supplemented with Bovine pituitary extract + Human recombinant epidermal growth factor Transfection of cell lines and 1% penicillin/streptomycin. 22Rv1 and LNCaP cells DU145 cells, which showed the highest MacroH2A1.1 were grown in RPMI 1640, DU145 and VCaP cells were expression levels, were selected for knocking-down RPMI-1640 + 50% F-12 medium ­(GIBCO®). The culture maintained in MEM and PC-3 cells were grown in 50% assays. Thus, cells were transiently transfected with SMARTpool: siGENOME H2AFY siRNA (Dharma- bovine serum and 1% penicillin/streptomycin ­(GIBCO®). media of PCa cell lines were supplemented with 10% fetal con) to knockdown MacroH2A1, and for MacroH2A1.1, cells were transiently transfected with a previously pub- Cells were grown in an incubator at 37 °C with 5% ­CO2. lished siRNA pool at 25 nM, and a siRNA negative All prostate cell lines were tested for Mycoplasma spp. fectamine™ reagent (Invitrogen, USA) was used for trans- control served as control in all experiments [22]. Oligo- contamination (PCR Mycoplasma Detection Set, Clon- tech Laboratories). Cells later harvested for protein and fection under conditions indicated by the manufacturer. RNA extraction. Cells were seeded 24 h before transfection, according to the purpose: for MTT assay, 4000 cells/well were seeded (96-well plates); apoptosis evaluation 30,000 cells/well RNA extraction and quantitative reverse transcription PCR were plated (24-well plates); and for RNA and protein, Samples were homogenized in ­TRIzol® Reagent (Invitro- (qRT‑PCR) 200,000 cells/well were used (6-well plates). The siRNA gen) and the total RNA was extracted using PureLink™ transfection was performed only once after cell seeding. Overexpression of MacroH2A1.1 was achieved in RNA Mini Kit (Invitrogen). All genomic DNA was elimi- LNCaP cells that depicted the lowest transcript levels. aTM) using ­FuGENE® HD Transfection Reagent (Pro- nated with TURBO DNA-free (Ambion, Applied Bio- This was performed through pEZ-Lv105 (GeneCopoei- ­ ransPlex® systems), according to manufacturer’s instructions. First strand synthesis was performed using the T mega), following manufacturer’s recommendations. After Aldrich®) and QIAquick PCR Purification Kit (QIAGEN) Whole Transcriptome Amplification Kit (Sigma- transfection, stable cell lines transfected with the vector and selected with puromycin dihydrochloride for in vitro for purification. assays. The stable cell lines generated where then use to ­ reen® Gene Expression Assay (Applied Target genes’ expression was determined using test cell viability (MTT assay, 10,000 cells/well), apoptosis ® Fast SYBR G (50,000 cells/well), and for RNA and protein. ­Biosystems ) and normalized to the expression of the In both conditions, protein and RNA extraction was endogenous control β-glucuronidase (GUSB), a house- performed in cells harvested at 72 h. keeping gene. Primers were MacroH2A1.1: forward, 5′-GGC​TTC​ACA​GTC​CTC​TCC​AC-3′, and reverse, 5′-GGT​GAA​CGA​CAG​CAT​CAC​TG-3′; MacroH2A1.2: SDS‑PAGE and western blot forward, 5′- GGC​TTC​ACA​GTC​CTC​TCC​AC-3′, and Total protein was extracted from cell lines using the reverse, 5′-GGA​TTG​ATT​ATG​GCC​TCC​AC-3′; Mac- Kinexus Lysis Buffer with protease inhibitor (Kinexus roH2A1: forward, 5′-TCC​ATT​GCA​TTT​CCA​TCC​ATC​ Bioinformatics Corporation). Protein concentration was GGC​-3′, and reverse, 5′-ACA​CGA​AGT​AAC​TGG​AGA​ determined using Pierce BCA Protein Assay Kit (Thermo TGG​CCT​-3′; QKI: forward, 5′-ATT​AAA​CGG​TCC​CCT​ Scientific) following manufacturer’s instructions. Then, GAA​GC-3′, and reverse, 5′-ATC​AAC​AGC​CCA​AGT​ 30 μg of total protein were separated by SDS-PAGE, blot- GTG​AC-3′; DDX5: forward, 5′-GTA​GCT​CAG​ACT​GGA​ ted in PVDF membranes (BioRad) and incubated in 5% TCT​GG-3′, and reverse, 5′-TCT​CTA​GGA​ATG​GCT​ (ChemCruz™) for 1 h. Membranes were probed with (w/v) bovine serum albumin (BSA) blocking solution GGT​GG-3′DDX17: forward, 5′-AGA​AGT​AGC​AAG​ ACT GAC​TCC​-3′, and reverse, 5′-CCC​CCT​CTC​ACT​ antibodies against MacroH2A1.1 (#12,455, Cell Sign- GTA​ATC​TC-3′; GUSB: forward, 5′-CTC​ATT​TGG​AAT​ aling), MacroH2A1.2 (#4827; Cell Signaling), Cleaved TTT​GCC​GATT-3′, and reverse, 5′-CCG​AGT​GAA​GAT​ endogenous control β-actin (Sigma). Western Bright™ PARP (Asp214) (19F4) (#9546, Cell Signaling) or the CCC​CTT​TTTA-3′; and H2AZ forward, 5′-GGG​AAG​ AAA​GGA​CAA​CAG​-3′, and reverse, 5′-CAC​AGA​GAT​ ECL-spray (Advansta) was used to develop the blots. ACA​GTC​CAC​TGG-3′. RNA levels were determined by Triplicates were performed in all experiments. Rela- the standard curve method. All samples were analyzed tive optical density determination was performed using in triplicate in a 7500 Real-Time PCR system (Applied ImageJ and normalized for the loading control, β-actin. Vieira‑Silva et al. Cancer Cell Int (2019) 19:112 Page 4 of 13 Cell viability biomarker sensitivity, specificity and accuracy param- Cell viability was assessed in transfected LNCaP and eters were calculated, as well as the positive (LR+) and 2,5-diphenyltetrazolium (MTT; Sigma-Aldrich®) assay. DU145 cells using the 3-(4, 5-dimethylthiazol-2-yl)- negative (LR−) likelihood ratios. Regarding this, as the quantitative value of a calculated likelihood ratio is fur- The viability assay was performed after cells adhered ther away from 1 in either direction (> 1 for LR+ and < 1 to the plate and, for DU145 cell line, right before trans- for LR−), there is increasing utility of a diagnostic test to fection (0 h) and in the subsequent days (24 and 48 h). point toward, or away from, a diagnosis which indicate MTT was added to the cells at 37 °C for 1 h, formazan the value of performing the respective diagnostic tests. DMSO (Sigma-Aldrich®) and plates shaken for 15 min. crystals were dissolved in 100 μL of Dimethyl sulfoxide Correlation between MacroH2A1, respective isoforms and splicing regulators or H2A.Z expression levels were GloMax®-Multi Detection System Colorimetric quantification was performed in an auto- assessed by Spearman’s correlation coefficient (r). mated plate reader ­ In cell lines, fold variation differences in transcript and (Promega) at 560 nm, with a reference filter of 630 nm. protein levels were determined using One-Way Analy- The optical density (OD) was directly proportional to the sis of Variance (ANOVA), followed by Dunnet’s (post number of viable cells. Three biologically independent hoc) test for multiple comparisons, or unpaired t-test, as experiments and methodological triplicates were done appropriate, comparing all PCa cell lines against RWPE-1 for all experiments. or negative control of transfection. All tests were two-sided and statistical significance was set at p < 0.05. Statistical analysis was performed APOPercentage™ kit (Biocolor) was used to evalu- Apoptosis using GraphPad Prism software for Windows version 5.0 ate apoptotic levels using a multi-plate G­ loMax®-Multi (GraphPad Software Inc., La Jolla, CA, USA). Detection System (Promega), for OD measurement of the released dye at 550 nm with a reference filter of 620 nm. Results Isoform MacroH2A1.1 is downregulated in primary PCa To normalize the OD measured in the apoptotic test to the cell number, the OD of apoptosis was normalized to Relevant clinical and pathological data of patients the OD of the viability assay. Three biological independ- included in this study is depicted in Table 1. MacroH2A1, ent experiments were performed with methodological MacroH2A1.1 and MacroH2A1.2 transcript levels were triplicates for each experiment. independently assessed in 15 MNPT, 45 PIN and 197 PCa tissue samples (Fig. 1a, Additional file 2: Table S1). Age was not significantly different among the studied groups. Statistical analysis MacroH2A1 expression levels did not differ signifi- Differences in quantitative expression levels of Mac- cantly between MNPT and PCa, whereas a significant roH2A1 and splicing regulators among MNPT, PIN and PCa were assessed using the non-parametric Kruskal– Wallis test, followed by pairwise comparisons with Mann–Whitney U-test, with Bonferroni’s correction. Table 1 Clinical and pathological features of subjects Differences in gene expression between matched PIN and in each group of tissue samples PCa samples were calculated by Wilcoxon Signed Rank Clinicopathological features MNPT PIN PCa test. Spearman’s correlation test was used to evaluate the Number of subjects, n 15 45 197 association between transcript levels of different genes. The associations between expression levels and standard Age (years) clinicopathological variables (serum PSA levels at diagno- Median (range) 64 (45–80) 64 (51–75) 64 (49–75) sis, Gleason score, histopathological stage) were assessed PSA levels (ng/mL) using the Kruskal–Wallis or Mann–Whitney tests, as Median (range) na na 8.3 (2.9–23) appropriate. A receiver operator characteristic (ROC) Pathological stage curve, its area under the curve and respective confidence pT2, n (%) na na 110 (55.8) intervals, were constructed, as described in [30], to assess pT3, n (%) 87 (44.2) the performance of MacroH2A1.1 and splicing regulators Gleason score [n (%)] expression for discriminating PCa from MNPT [31]. To 6 na na 67 (34.0) demonstrate that the confidence intervals for the AUC 7 115 (58.4) obtained in this series (MNPT = 15 and PCa = 197) are 8 6 (3.0) statistically robust, a simulation study was performed 9 9 (4.6) (Additional file 1: Data S1). Moreover, for the same genes, na not applicable, MNPT morphologically normal prostate tissue, PIN prostatic intraepithelial neoplasia, PCa prostate carcinoma Vieira‑Silva et al. Cancer Cell Int (2019) 19:112 Page 5 of 13 Fig. 1 Transcript levels of MacroH2A1 and splicing regulators in prostate tissue samples. a Transcriptional status of total MacroH2A1, MacroH2A1.1 and MacroH2A1.2 isoforms in MNPT samples (n = 15), PIN lesions (n = 45) and PCa samples (n = 197), assessed by RT-qPCR, and normalized with GUSB gene. b Transcript levels of three splicing regulators of H2AFY mRNA—QKI, DDX5 and DDX17—assessed by RT-qPCR, and normalized with GUSB gene. MNPT morphologically normal prostate tissue, PIN prostatic intraepithelial neoplasia, PCa prostate carcinoma, KW Kruskal–Wallis test. Inter-group analysis with Kruskal–Wallis test, followed by Dunn’s multiple comparisons test: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, ns not significant decrease was observed in PIN (p < 0.001) (Fig. 1a, Addi- displayed the lowest levels of MacroH2A1.2 (0.06) tional file 2: Table S1). MacroH2A1.1 expression levels, (Fig. 1a, Additional file 2: Table S1). however, differed significantly among the three groups [Kruskal–Wallis (KW) test, p < 0.001], with the lowest transcript levels found in PCa (p < 0.001 and p = 0.009, H2AFY splicing regulators and H2AFZ expression compared to MNPT and to PIN, respectively). Indeed, in prostate cancer PIN lesions disclosed higher MacroH2A1.1 transcript To explain the altered ratio between H2AFY gene splicing levels than PCa, in cases in which PIN and matched variants, MacroH2A1 splicing regulators (QKI, DDX5 tumor lesions from the same prostatectomy speci- and DDX17) transcription levels were also assessed in mens were available (Additional file 3: Fig. S1). Regard- the same sample sets (Fig. 1b, Additional file 2: Table S1). ing MacroH2A1.2, significantly lower expression levels A statistically significant downregulation of QKI and were depicted for PIN comparing with MNPT and PCa DDX17 was depicted for PCa (p < 0.001), whereas in PIN, (p < 0.001), whereas no significant differences were appar- DDX17 expression levels were significantly lower com- ent between MNPT and PCa. pared to MNPT (p < 0.001). Concerning DDX5 expres- Overall, PCa tissues showed higher MacroH2A1.1 sion, PIN demonstrated the lowest levels (p < 0.001, both (1.33) than MacroH2A1.2 (0.59) relative expression for MNPT and PCa) but no significant differences were levels, although the highest MacroH2A1.1 levels were apparent between MNPT and PCa samples (Fig. 1b, observed in PIN lesions (2.39). Conversely, PIN lesions Additional file 2: Table S1). Vieira‑Silva et al. Cancer Cell Int (2019) 19:112 Page 6 of 13 DDX5 transcript levels showed a stronger positive (p < 0.01) and QKI (p < 0.001) expression levels. Consid- correlation with MacroH2A1 total expression (ρ = 0.51, ering a four-tier system, corresponding to the Gleason p < 0.001), whereas a significant correlation was also scores 6, 7 (3 + 4), 7 (4 + 3) and ≥ 8, differences among found between QKI and MacroH2A1.1 splice variant the four categories were found both for MacroH2A1.1 expression levels (ρ = 0.56, p < 0.001). Splicing regulator and QKI expression (p = 0.0185 and p < 0.0001, respec- DDX17 expression levels did not impressively correlate tively; Additional file 3: Fig. S4). Pairwise analysis dis- with any of the MacroH2A1 transcripts (Additional file 2: closed significant differences between Gleason scores Table S2). Furthermore, no inverse correlation was found 6 and 7 (3 + 4) for MacroH2A1.1 and QKI expression between H2AFY and H2AFZ transcript levels (Additional (p = 0.0179 and p = 0.0082, respectively), and between file 3: Fig. S2). Gleason scores 6 and ≥ 8 (p < 0.0001). Receiver operator characteristic (ROC) curve Diagnostic performance of MacroH2A1.1 and splicing analysis was performed to assess the ability of Mac- regulators transcript levels in prostate tissues roH2A1.1, DDX5, DDX17 and QKI expression levels Regarding associations with clinical-pathological vari- in discriminating PCa from non-cancerous prostate ables, high MacroH2A1.1 expression levels signifi- tissues (Fig. 2b and Table 2). The empirical cut-off cantly associated with serum PSA levels above 10 ng/ values were set to maximize sensitivity and specific- mL (p < 0.01, (Additional file 3: Fig. S3). Moreover, Mac- ity. Remarkably, QKI outperformed MacroH2A1.1 and roH2A1.1 and QKI expression levels significantly associ- DDX5 and DDX17, displaying 88.8% sensitivity and ated with Gleason score (Fig. 2a), a clinical parameter of 93.3% specificity, corresponding to an area under the disease progression and reduced differentiation. Indeed, curve (AUC) of 0.94 (95% IC: 0.897-0.987); p < 0.0001 considering a two-tier categorization (Gleason score = 6 (Table 2). Moreover, QKI presented a LR+ of 13.25 and vs. Gleason score > 6), higher grade (less differentiated) a LR− of 0.12 (Table 2). Notwithstanding the limited tumors displayed significantly reduced MacroH2A1.1 number of normal tissue samples, a statistical model Fig. 2 Association of MacroH2A1.1 and QKI transcript levels with clinicopathological parameters and diagnostic performance. a Distribution of MacroH2A1.1 and QKI transcript levels in PCa cases, determined by RT-qPCR and normalized with GUSB mRNA levels, according to categorized < 7 (n = 68) and ≥ 7 (n = 129) Gleason score; Mann–Whitney U test: **p < 0.01, ****p < 0.0001. b ROC curves showing biomarker performance evaluation of MacroH2A1.1 (AUC = 0.92, 95% CI 0.876–0.965, p < 0.0001), DDX17 (AUC = 0.93, 95% CI 0.872–0.980, p < 0.0001), and QKI (AUC = 0.94, 95% CI 0.897–0.987, p < 0.0001) transcript levels as discriminators of prostate cancer (PCa, n = 195) from morphological normal prostate tissue (MNPT, n = 15). AUC​ area under curve, CI confidence interval Vieira‑Silva et al. Cancer Cell Int (2019) 19:112 Page 7 of 13 Table 2 Validity estimates for MacroH2A1.1 and QKI expression levels as diagnostic biomarkers for prostate cancer identification Parameter MacroH2A1.1 performance DDX5 performance DDX17 performance QKI performance Sensitivity (%) 80.7 82.7 47.3 88.8 Specificity (%) 93.3 93.3 93.3 93.3 Accuracy (%) 81.6 83.5 50.5 88.7 Positive likelihood ratio (LR+) 12.04 12.34 7.06 13.25 Negative likelihood ratio (LR−) 0.21 0.19 0.56 0.12 AUC (95% IC) 0.92 (0.876–0.965) 0.65 (0.549–0.751) 0.93 (0.872–0.980) 0.94 (0.897–0.987) p value p < 0.0001 p = 0.0533 p < 0.0001 p < 0.0001 Fig. 3 Distribution of MacroH2A1 transcript and protein levels in prostate cancer cell lines. a Transcript levels of total MacroH2A1, MacroH2A1.1 and MacroH2A1.2, assessed by RT-qPCR and normalized to GUSB mRNA levels, in androgen-receptor positive RWPE-1 benign prostate and 22Rv1, LNCaP and VCaP cancer cell lines, and in androgen-receptor negative DU145 and PC-3 prostate cancer cell lines. b Transcript levels of splicing regulators DDX5, DDX17 and QKI, assessed by RT-qPCR and normalized to GUSB, in androgen positive and negative prostate cell lines. c Illustrative images of MacroH2A1 isoforms protein levels in prostate cancer cell lines and in non-malignant prostate cell line RWPE-1. d Fold variation of MacroH2A1.1 and MacroH2A1.2 protein levels, normalized to β-actin, and directly compared to benign prostate cell line RWPE-1. (mean ± SD, n = 3). Dunnet’s pairwise multiple comparisons test: *p < 0.05, **p < 0.01, ***p < 0.001 (1000 simulations) disclosed mean confidence interval and normalized for a benign prostate cell line (RWPE-1) range of 0.101, with 25% and 75% percentiles of 0.079 (Fig. 3a, b). and 0.119. Since the value obtained with the real data- Although 22Rv1 showed the highest MacroH2A1 set was 0.089, we may consider that is within the range transcript levels, no significant differences were appar- of expected values (Additional file 1: Data S1). ent among PCa cell lines, comparing with RWPE-1. AR negative (AR−) PCa cell line DU145 displayed MacroH2A1 phenotypic impact in PCa cell lines significantly higher MacroH2A1.1 expression levels Transcript levels of MacroH2A1, respective isoforms than RWPE-1 (p < 0.05), whereas the lowest transcript and splicing regulators were assessed by RT-qPCR in five levels of this variant were depicted for LNCaP, an AR PCa cell lines (22Rv1, LNCaP, VCaP, DU145 and PC-3) positive (AR+) PCa cell line (p < 0.05). Concerning Vieira‑Silva et al. Cancer Cell Int (2019) 19:112 Page 8 of 13 MacroH2A1.2, all AR+ PCa cell lines (22Rv1, LNCaP Concerning protein expression, MacroH2A1.1 and and VCaP) displayed significantly higher expression MacroH2A1.2 protein levels were also rather variable levels than the benign prostate cell line (p < 0.01, for among cell lines (Fig. 3c, d). MacroH2A1.1 protein all) whereas AR− cancer lines did not disclose signifi- expression pattern only differed from transcript lev- cant differences compared to RWPE-1 (Fig. 3a). As to els in VCaP and DU145 cell lines, whereas for Mac- splicing regulators, among AR+ cell lines, only LNCaP roH2A1.2, protein levels in PCa cell lines followed depicted significantly higher DDX5 and DDX17 tran- the same trend of the transcript. Interestingly, Mac- script levels than RWPE-1, whereas both AR− cells roH2A1.2 proteins levels were significantly higher in disclosed significantly higher DDX5, DDX17 and QKI AR+ vs. AR− PCa cell lines (paralleling the findings at mRNA level, cells except for QKI in DU145 (p < 0.05, transcript level), whereas no differences were found for Fig. 3b). Although no significant differences in global MacroH2A1.1 (Additional file 3: Fig. S5c). MacroH2A1 transcript levels between AR+ and AR− To uncover the biological role of MacroH2A1, DU145 cancer cell lines were found, AR− cells depicted sig- cells were knockdown for this protein (Fig. 4a, b). nificantly higher MacroH2A1.1 expression levels than Although reduced protein levels of both MacroH2A1 AR+ cells, whereas the opposite was found for Mac- variants (p < 0.05) was achieved, a more impressive roH2A1.2 (Additional file 3: Fig. S5a). Nonetheless, effect was observed in MacroH2A1.2 (Fig. 4a, b). Phe- among splicing regulators, significant differences in notypically, increased cell viability was observed in splicing regulators expression levels between AR+ and MacroH2A1 knockdown DU145 cells at 48 h (p < 0.01) AR− were only found for QKI (Additional file 3: Fig. (Fig. 4c), whereas a significant decrease in apoptosis was S5b). apparent after 72 h, in the same cells (p < 0.05) (Fig. 4d). Fig. 4 Knockdown of MacroH2A1 in DU145 cell line. a Illustrative images of MacroH2A1 isoforms protein after MacroH2A1 knockdown, carried out by si-RNA, in DU145 prostate cancer cell line at 72 h’ post-transfection, obtained by Western-Blot. b Fold variation of MacroH2A1.1 and MacroH2A1.2 protein levels in MacroH2A1-knockdown DU145 cell line, in comparison to si-scramble DU145 cell line (mean ± SD, n = 3), assessed by western-blot. Unpaired t test: *p < 0.05. Impact of MacroH2A1 knockdown in cell viability (c) and apoptosis levels (d) after 72 h (mean ± SD, n = 3). Mann–Whitney U-test: *p < 0.05, ***p < 0.001 Vieira‑Silva et al. Cancer Cell Int (2019) 19:112 Page 9 of 13 Additionally, MacroH2A1.1 isoform was ectopically We found that MacroH2A1.1 transcript levels were overexpressed in LNCaP (cell line with the lowest levels) downregulated in PIN and primary PCa, compared to and specifically silenced in DU145 cells (cell line with normal prostate tissues. Notwithstanding the reduced the highest levels) (Fig. 5). Effective stable MacroH2A1.1 number of normal prostate tissues included in this overexpression (MacroH2A1.1OE) (Fig. 5a, b) and silenc- study, which should be construed as a limitation, our ing (si-MacroH2A1.1) (Fig. 5e, f ) were confirmed at pro- results are in line with previous observations on Mac- tein level. Indeed, MacroH2A1.1 protein levels almost roH2A1.1 expression in other primary cancers [26, 32] doubled in MacroH2A1.1OE LNCaP cells (p < 0.05) and the intermediate expression levels depicted in PIN and significantly associated with decreased cell viability is consistent with its PCa precursor condition. How- (p < 0.001) (Fig. 5c), whereas apoptosis was significantly ever, MacroH2A1.2 expression levels did not parallel the increased 72 h after transfection (p < 0.05) (Fig. 5d). upregulation reported for other tumors [25, 26]., as no Paradoxically, MacroH2A1.1 knockdown in DU145 also significant differences were observed between PCa and resulted in decreased cell viability (p < 0.001), with a normal tissues. Contrarily, prostate cancer precursor concomitant increase in apoptosis (p < 0.001) (Fig. 5g, h, lesion (PIN), displayed significantly lower MacroH2A1.2 respectively). transcript levels than those of MNPT and PCa. Interest- As expected, both MacroH2A1.1 overexpressing and ingly, it was recently demonstrated that MacroH2A1.2 knockdown PCa cells did not show any effect on Mac- attenuates osteoclastogenesis in a PCa in vitro model roH2A1.2 protein levels (Fig. 5a, e). [33], but no direct comparisons with our results can be made as we only studied organ confined PCa. Regarding, MacroH2A1 levels, albeit PIN displayed the Discussion lowest compared to MNPT and PCa, this was mostly due PCa is the most common malignancy in men and a lead- to MacroH2A1.2 downregulation. Indeed, MacroH2A1.1 ing cause of mortality and morbidity, worldwide. Both expression levels were lower in PCa comparing to MNPT. genetic and epigenetic disruption has been implicated in Thus, our results suggest that sustained MacroH2A1.1 its initiation and progression. Unravel the mechanisms downregulation is associated with the emergence of the underlying tumor development are key to provide a more malignant phenotype in the prostate, whereas concomi- profound insight into PCa biology, which might translate tant MacroH2A1.2 decreased expression might be rel- into better diagnostic and prognostic tools, and the iden- evant for the development of PIN lesions only, which tification of novel therapeutic targets [2]. frequently do not seem to progress into invasive car- Among epigenetic mechanisms, the replacement of cinoma. Nevertheless, MacroH2A1 downregulation is canonical histones has been recently implicated in tum- likely to play a role along prostate tumorigenesis, sugges- origenesis [7]. Indeed, the role of the two MacroH2A1 tive of a tumor suppressive effect. isoforms have been investigated in cancer [22–26]. Because an inverse correlation between MacroH2A1 In particular, MacroH2A1.1 is mostly considered a and H2A.Z has been suggested in previous studies [29], tumor suppressor, inhibiting stem cell-like properties we assessed the transcript levels of the latter in our data- and counteracting the functions of MacroH2A1.2 [16, set, as well, but no correlation was found between tran- 17]. Although its role has been previously investigated scripts levels of those two histone variants. Considering in other tumor models, no data is available for PCa, to these findings, we hypothesized that differential expres- the best of our knowledge. Thus, we aimed to assess the sion of MacroH2A1 isoforms may be related with altered putative role of MacroH2A1 isoforms in PCa initiation expression of its splicing regulators, specifically QKI, and progression. DDX5 and DDX17. Remarkably, in primary tumors, (See figure on next page.) Fig. 5 Expression modulation of MacroH2A1.1 isoform in PCa cell lines. a Illustrative images of MacroH2A1 isoforms protein and β-actin protein in MacroH2A1.1 transfected LNCaP cell line ­(LNCaPMacroH2A1.1OE) and transfected control cell line ­(LNCaPempty vector) at 72 h post-transfection. b Fold variation of MacroH2A1.1 and MacroH2A1.2 protein levels in ­LNCaPMacroH2A1.1OE, compared to ­LNCaPempty vector (mean ± SD, n = 3). Unpaired t test: *p < 0.05, ns = not significant. Impact of MacroH2A1.1 overexpression in cell viability (c) and apoptosis levels (d) after 72 h (mean ± SD, n = 3). Mann–Whitney U-test: *p < 0.05, ***p < 0.001. e Illustrative images of MacroH2A1 isoforms protein after MacroH2A1.1 knockdown, carried out by si-RNA, in DU145 prostate cancer cell line at 72 h’ post-transfection, obtained by Western-Blot. f Fold variation of MacroH2A1.1 and MacroH2A1.2 protein levels in MacroH2A1.1-knockdown DU145 cell line, in comparison to si-scramble DU145 cell line (mean ± SD, n = 3), assessed by western-blot. Unpaired t test: *p < 0.05. Impact of MacroH2A1.1 knockdown in cell viability (g) and apoptosis levels (h) after 72 h (mean ± SD, n = 3). Mann–Whitney U-test: *p < 0.05, **p < 0.01, ***p < 0.001 Vieira‑Silva et al. Cancer Cell Int (2019) 19:112 Page 10 of 13 Vieira‑Silva et al. Cancer Cell Int (2019) 19:112 Page 11 of 13 variations in QKI and DDX17 expression levels paral- in PCa samples compared to matched PIN, a finding that leled those of MacroH2A1.1, whereas DDX5 transcript parallels the observed variations in the whole case series. levels followed the same pattern of MacroH2A1 and Moreover, this result further supports a causal role for MacroH2A1.2. Statistical analysis showed a moderately QKI downregulation in MacroH2A1.1 decreased expres- positive correlation between QKI and MacroH2A1.1, sion along prostate tumorigenesis. Indeed, decreased as well as between DDX5 and MacroH2A1 and its iso- QKI and MacroH2A1.1 expression levels are clearly forms. Thus, we might speculate whether variations in associated with PCa, as demonstrated by its ability to MacroH2A1 isoforms expression in PCa is due to altered discriminate cancerous from non-cancerous prostate tis- expression of its splicing regulators, although other, yet sues, notwithstanding the limited number of the latter unidentified, factors might be involved, as well. In sup- samples (n = 15). port of this hypothesis, QKI has been considered a tumor Some interesting associations between QKI and Mac- suppressor in various cancers and frequently associated roH2A1.1 expression and clinicopathological parameters with MacroH2A1.1 downregulation [21]. Concerning were depicted. Specifically, PCa with higher Gleason DDX5 and DDX17 expression, our results are some- score (i.e., less differentiated) displayed lower QKI and what unexpected as both are considered highly homol- MacroH2A1.1 levels. Concerning MacroH2A1.1 expres- ogous oncogenic RNA-helicases [34]. Nevertheless, sion, the same has been reported for other carcinomas, lower expression of both DDX5 and DDX17 has been being its loss associated with worse outcome in colon reported in in situ breast cancer, along with increased cancer patients [26]. Conversely, MacroH2A1.1 expres- MacroH2A1.1/MacroH2A1.2 ratios [22]. Remarkably, sion was higher in PCa patients with higher serum this parallels our observations in PIN lesions, which are PSA levels at diagnosis. Eventually, a comparison based considered pre-invasive forms of PCa. Interestingly, on the basic Gleason patterns (3, 4 or 5) instead of the DDX5, but not DDX17, affect key cellular pathways, Gleason score might be considered more biologically including upregulation of AR in PCa and induction of relevant. However, we should emphasize that statistical epithelial-mesenchymal transition (EMT), a feature that associations were evaluated between clinicopathologi- is associated with tumor invasion capabilities [22, 35, 36]. cal parameters and molecular data, and, in this regard, Interestingly, we found significant differences in Mac- the meaningful comparison is among Gleason scores roH2A1.1, MacroH2A1.2 and QKI, but not DDX5 and and not Gleason grades, as the latter are only used as a DDX17, transcript levels between AR+ and AR− PCa basis for Gleason score [30]. Moreover, Gleason grade 3, cell lines. Indeed, whereas the association between Mac- 4 or 5 tumor areas are frequently intermingled and are roH2A1.1 and QKI expression found in primary PCa tis- not easily discriminated in tumors with mixed grades. sues seems to hold true for PCa cell lines, the same was Thus, an attempt to selectively collect those areas would not evident for DDX5 and DDX17. Importantly, at pro- most likely result in “contamination” and consequent tein level, significant differences between AR+ and AR− analysis bias. On the other hand, if only 3 + 3, 4 + 4 and were found only for MacroH2A1.2, suggesting that other 5 + 5 tumors would be analyzed, most of tumors, which factors are involved in expression regulation of the iso- are Gleason score 7 (3 + 4 or 4 + 3) would be excluded. forms. Nonetheless, because our series of primary PCa Moreover, the procedure used for selectively identify- only represent hormone-therapy-naïve tumors, no defin- ing the index tumor assured that the tumor area selected itive conclusions can be made and a comparative study for molecular analyses was representative of the index with castration-resistant PCa cases may further elucidate tumor. Although the association with the Gleason score these findings. seems intuitive as higher scores correspond to less differ- Subsequently, we focused our attention on QKI and entiated and more aggressive PCa, the association with MacroH2A1.1 expression in a subset of cases with serum PSA levels, on the contrary, is almost counterin- matched PIN and PCa tissues. Although paired lesions tuitive. Nevertheless, it should be recalled that cells from were found in the same gland, a direct causal link less differentiated PCa produce less PSA, which might between them should not be construed. Nevertheless, have a negative impact on global serum PSA levels not- this analysis might elucidate how QKI and MacroH2A1.1 withstanding heavy disease burden and corresponding expression is altered along the carcinogenic process in poor outcome [37]. Overall, these findings suggest that the prostate gland. In approximately two-thirds of these lower QKI and MacroH2A1.1 expression levels might be cases, both QKI and MacroH2A1.1 expression was lower associated with worse PCa-related survival, a hypothesis Vieira‑Silva et al. Cancer Cell Int (2019) 19:112 Page 12 of 13 that follows the same reported for other cancer mod- Additional files els, but which requires further investigation in a larger cohort of PCa patients. Additional file 1: Data S1. To illustrate the confidence intervals for the To further understand MacroH2A1 biological role in AUC that can be obtained with these data, a simulation study has been performed. PCa, the phenotypic effect of MacroH2A1 silencing was evaluated in DU145 PCa cell line, which resulted in sig- Additional file 2: Table S1. Distribution of expression levels (assessed by RT-qPCR) for total MacroH2A1, splice variants and regulators among differ‑ nificantly increased cell viability, paralleling previous ent prostate tissue samples. Table S2. Spearman’s ρ correlations among observations in bladder cancer cells [17, 32]. Because total MacroH2A1 and splice variants with three splicing regulators. in primary PCa MacroH2A1.1 levels predominate over Additional file 3: Fig. S1. Scatter blots representation of transcript levels those of MacroH2A1.2, we further assessed the phe- of MacroH2A1.1 and QKI in matched PIN and PCa lesions from the same patients, assessed by RT-qPCR and normalized to GUSB mRNA levels. Fig. notypic impact of its overexpression and silencing. In S2. Correlation between MacroH2A1 and H2A.Z transcript levels in pros‑ LNCaP cells, MacroH2A1.1 forced expression signifi- tate cancer samples, assessed by RT-qPCR and normalized to GUSB. Fig. cantly decreased cell viability and increased apoptosis, S3. Scatter blots representation of distribution of MacroH2A1.1 mRNA lev‑ els in prostate cancer tissue samples, assessed by RT-qPCR and normalized which is accordance with a putative tumor suppressive to GUSB, according to categorized ≤ 10 ng/mL and > 10 ng/mL PSA levels. role [25, 26]. However, in DU145 cells, MacroH2A1.1 Fig. S4. Scatter blots representation of distribution of MacroH2A1.1 and knockdown disclosed opposite results to those of Mac- QKI mRNA levels assessed by RT-qPCR and normalized to GUSB, among Gleason scores ≤ 6, 7 (3 + 4), 7 (4 + 3) and ≥ 8 prostate cancer tissue roH2A1 knockdown in the same cells (which affected samples. Fig. S5. (a) Distribution of transcript levels of total MacroH2A1, both isoforms, MacroH2A1.1 and MacroH2A1.2). Inter- MacroH2A1.1 and MacroH2A1.2, assessed by RT-qPCR and normalized estingly, it was suggested that MacroH2A1 isoforms to GUSB mRNA levels, in androgen-receptor positive prostate cancer cell lines (22Rv1, LNCaP and VCaP), and in androgen-receptor negative may have different effects [18, 27], since reduced Mac- prostate cancer cell lines (DU145 and PC-3). (b) Distribution of transcript roH2A1.1 expression has been associated with a more levels of splicing regulators DDX5, DDX17 and QKI, assessed by RT-qPCR aggressive phenotype [25, 26], whereas increased Mac- and normalized to GUSB, in androgen positive and negative prostate cancer cell lines. (c) Distribution of MacroH2A1.1 and MacroH2A1.2 protein roH2A1.1 correlated with poor prognosis in triple-neg- levels, normalized to β-actin, in androgen positive and negative prostate ative breast cancer patients [23]. Thus, its putative role cancer cell lines. Mann–Whitney U-test: **p < 0.01, ***p < 0.001. ns—non might be tumor model-dependent, since MacroH2A1.1 significant. might have an activating or repressive function depend- ing on external cellular signals [38]. This may explain, Abbreviations at the least partially, our paradoxical finding in Mac- AR: androgen receptor; BSA: bovine serum albumin; DDX5: deadbox 5 (also known as p68); DDX17: deadbox 17 (also known as p72); EMT: epithelial roH2A1.1 silenced DU145 cells. Furthermore, the mod- mesenchymal transition; FFPE: formalin-fixed, paraffin embedded; GUSβ: est silencing achieved for MacroH2A1.1 (only about β-glucuronidase; H3K27me3: tri-methylated histone H3 at lysine 27 (repress 50%), might have also contributed to that result. transcription); MNPT: morphologic normal prostate tissue; mRNA: messenger ribonucleic acid; PCa: prostate cancer; PIN: prostate intraepithelial neoplasia; PSA: prostate-specific antigen; qRT-PCR: quantitative reverse transcriptase polymerase chain reaction; QKI: quaking, KH domain containing RNA binding. Conclusions In conclusion, this study is the first to report variations Authors’ contributions in expression of MacroH2A1 and its isoforms in prostate Study concepts and design: RH, CJ; Data acquisition: TSV-S, JR-C, IG, SM-R, DB-S, AR, IC, ATM, JO; Data analysis and interpretation: SH-B, MB, TSV-S, tissues, encompassing morphologically normal and neo- JR-C, RH, CJ; Manuscript preparation: TSV-S and CJ; Manuscript review: MB, RH, plastic (both pre-invasive and invasive) lesions. Globally, CJ. All authors read and approved the final manuscript. we found that MacroH2A1.1 transcript levels gradually Author details decrease during tumorigenesis, whereas MacroH2A1 and 1 Cancer Biology & Epigenetics Group, Research Center (CI‑IPOP), Portuguese MacroH2A1.2 were downregulated only in PIN lesions. Oncology Institute of Porto (IPO Porto), Research Center‑LAB 3, F Bdg., 1st Floor, Interestingly, variations in MacroH2A1 are mostly Rua Dr. António Bernardino de Almeida, 4200‑072 Porto, Portugal. 2 Depart‑ ment of Pathology, Portuguese Oncology Institute of Porto (IPO Porto), Rua Dr. affected by MacroH2A1.2 isoform and these alterations António Bernardino de Almeida, 4200‑072 Porto, Portugal. 3 Department of Urol‑ are associated with altered expression of splicing regula- ogy, Portuguese Oncology Institute of Porto (IPO Porto), Rua Dr. António Ber‑ tors, specifically QKI for MacroH2A1.1, as well as DDX5 nardino de Almeida, 4200‑072 Porto, Portugal. 4 Department of Epidemiology, Portuguese Oncology Institute of Porto (IPO Porto), Rua Dr. António Bernardino for MacroH2A1 and isoforms ratios. Moreover, less dif- de Almeida, 4200‑072 Porto, Portugal. 5 Josep Carreras Leukaemia Research Insti‑ ferentiated and more aggressive PCa cases display lower tute (IJC), Campus ICO‑Germans Trias i Pujol, Campus Can Ruti, 08916 Badalona, QKI and MacroH2A1.1 transcripts levels, as expected Spain. 6 Program for Predictive and Personalized Medicine of Cancer, Germans Trias i Pujol Research Institute (PMPPC-IGTP), Barcelona, Spain. 7 Department for putative tumor suppressors. Finally, the attenuation of Pathology and Molecular Immunology, Institute of Biomedical Sciences Abel of malignant phenotype of PCa cell lines after manipula- Salazar (ICBAS), University of Porto, Porto, Portugal. tion of Macro H2A1 expression, further suggest a tumor Acknowledgements suppressor role for this histone variant, although Mac- The authors are grateful to the patients that volunteered to provide samples roH2A1.1 and MacroH2A1.2’s role in PCa require fur- and to all the personnel of the Department of Urology of the Portuguese ther investigation. Oncology Institute—Porto that kindly collaborated in this study. Vieira‑Silva et al. Cancer Cell Int (2019) 19:112 Page 13 of 13 Competing interests 14. Zhang R, Poustovoitov MV, Ye X, Santos HA, Chen W, Daganzo SM, et al. For‑ The authors declare that they have no competing interests. mation of MacroH2A-containing senescence-associated heterochromatin foci and senescence driven by ASF1a and HIRA. Dev Cell. 2005;8(1):19–30. Availability of data and materials 15. Gamble MJ, Kraus WL. Multiple facets of the unique histone variant Mac‑ The datasets used and/or analysed during the current study are available from roH2A: from genomics to cell biology. Cell Cycle. 2014;9(13):2568–74. the corresponding author on reasonable request. 16. Gaspar-Maia A, Qadeer ZA, Hasson D, Ratnakumar K, Leu NA, Leroy G, et al. MacroH2A histone variants act as a barrier upon reprogramming towards Consent for publication pluripotency. Nat Commun. 2013;4:1565. Not applicable. 17. Creppe C, Janich P, Cantarino N, Noguera M, Valero V, Musulen E, et al. Mac‑ roH2A1 regulates the balance between self-renewal and differentiation com‑ Ethics approval and consent to participate mitment in embryonic and adult stem cells. Mol Cell Biol. 2012;32(8):1442–52. This study was approved by the institutional review board of the Portuguese 18. Cantarino N, Douet J, Buschbeck M. MacroH2A—an epigenetic regulator of Oncology Institute of Porto, Portugal (Comissão de Ética para a Saúde–CES cancer. Cancer Lett. 2013;336(2):247–52. 019/2008). 19. Chen H, Ruiz PD, Novikov L, Casill AD, Park JW, Gamble MJ. MacroH2A1.1 and PARP-1 cooperate to regulate transcription by promoting CBP-medi‑ Funding ated H2B acetylation. Nat Struct Mol Biol. 2014;21(11):981–9. This study was supported by research grants from Research Center of the 20. Pasque V, Radzisheuskaya A, Gillich A, Halley-Stott RP, Panamarova M, Portuguese Oncology Institute—Porto (CI-IPOP 4-2012 and CI-IPOP-17-2015) Zernicka-Goetz M, et al. Histone variant MacroH2A marks embryonic differ‑ and COST ACTION CM1406. JRC, IG, SM-R and D B-S are research fellows entiation in vivo and acts as an epigenetic barrier to induced pluripotency. J from Federal funds through Programa Operacional Temático Factores de Cell Sci. 2012;125(Pt 24):6094–104. Competitividade (COMPETE) with co-participation from the European Com‑ 21. Novikov L, Park JW, Chen H, Klerman H, Jalloh AS, Gamble MJ. QKI-mediated munity Fund (FEDER) and by national funds through Fundação para a Ciência alternative splicing of the histone variant MacroH2A1 regulates cancer cell e Tecnología (FCT) (SFRH/BD/71293/2010, CI-IPOP-BPD/UID/DTP/00776/2013, proliferation. Mol Cell Biol. 2011;31(20):4244–55. SFRH/BD/112673/2015 and FCT-SFRH/BD/136007/2018, respectively). 22. Dardenne E, Pierredon S, Driouch K, Gratadou L, Lacroix-Triki M, Espinoza MP, et al. Splicing switch of an epigenetic regulator by RNA helicases promotes tumor-cell invasiveness. Nat Struct Mol Biol. 2012;19(11):1139–46. Publisher’s Note 23. Lavigne AC, Castells M, Mermet J, Kocanova S, Dalvai M, Bystricky K. Springer Nature remains neutral with regard to jurisdictional claims in pub‑ Increased MacroH2A1.1 expression correlates with poor survival of triple- lished maps and institutional affiliations. negative breast cancer patients. PLoS ONE. 2014;9(6):e98930. 24. Kapoor A, Goldberg MS, Cumberland LK, Ratnakumar K, Segura MF, Received: 16 November 2018 Accepted: 23 April 2019 Emanuel PO, et al. The histone variant MacroH2A suppresses melanoma progression through regulation of CDK8. Nature. 2010;468(7327):1105–9. 25. Sporn JC, Kustatscher G, Hothorn T, Collado M, Serrano M, Muley T, et al. Histone MacroH2A isoforms predict the risk of lung cancer recurrence. Oncogene. 2009;28(38):3423–8. References 26. Sporn JC, Jung B. Differential regulation and predictive potential of Mac‑ 1. Siegel RL, Miller KD, Jemal A. Cancer statistics, 2015. CA Cancer J Clin. roH2A1 isoforms in colon cancer. Am J Pathol. 2012;180(6):2516–26. 2015;65:25. 27. Corujo D, Buschbeck M. Post-translational modifications of H2A histone 2. Shen MM, Abate-Shen C. Molecular genetics of prostate cancer: new pros‑ variants and their role in cancer. Cancers. 2018;10(3):59. pects for old challenges. Genes Dev. 2010;24(18):1967–2000. 28. Baptista T, Graça I, Sousa EJ, et al. Regulation of histone H2A.Z expression is 3. Jeronimo C, Bastian PJ, Bjartell A, Carbone GM, Catto JW, Clark SJ, et al. mediated by sirtuin 1 in prostate cancer. Oncotarget. 2013;4(10):1673–85. Epigenetics in prostate cancer: biologic and clinical relevance. Eur Urol. 29. Barzily-Rokni M, Friedman N, Ron-Bigger S, Isaac S, Michlin D, Eden A. Syner‑ 2011;60(4):753–66. gism between DNA methylation and MacroH2A1 occupancy in epigenetic 4. Vardabasso C, Hasson D, Ratnakumar K, Chung CY, Duarte LF, Bernstein silencing of the tumor suppressor gene p16(CDKN2A). Nucleic Acids Res. E. Histone variants: emerging players in cancer biology. Cell Mol Life Sci. 2011;39(4):1326–35. 2014;71(3):379–404. 30. Eble JN, Organization WH, Cancer IAfRo. Pathology and genetics of tumours 5. Biterge B, Schneider R. Histone variants: key players of chromatin. Cell Tissue of the urinary system and male genital organs. Lyon: IARC Press; 2004. Res. 2014;356(3):457–66. 31. Hanley JA, McNeil BJ. The meaning and use of the area under a receiver 6. Volle C, Dalal Y. Histone variants: the tricksters of the chromatin world. Curr operating characteristic (ROC) curve. Radiology. 1982;143(1):29–36. Opin Genet Dev. 2014;25(8–14):138. 32. Park SJ, Shim JW, Park HS, et al. MacroH2A1 downregulation enhances the 7. Monteiro FL, Baptista T, Amado F, Vitorino R, Jerónimo C, Helguero LA. stem-like properties of bladder cancer cells by transactivation of Lin28B. Expression and functionality of histone H2A variants in cancer. Oncotarget. Oncogene. 2015;35(10):1292–301. 2014;5(11):3428–43. 33. Kim JM, Shin Y, Lee S, et al. MacroH2A1.2 inhibits prostate cancer-induced 8. Kamakaka RT, Biggins S. Histone variants: deviants? Genes Dev. osteoclastogenesis through cooperation with HP1α and H1.2. Oncogene. 2005;19:295–310. 2018;. https​://doi.org/10.1038/s4138​8-018-0356-3. 9. Costanzi C, Pehrson JR. Histone MacroH2A1 is concentrated in the inactive 34. Shin S, Rossow KL, Grande JP, Janknecht R. Involvement of RNA helicases X chromosome of female mammals. Nature. 1998;393(6685):599–601. p68 and p72 in colon cancer. Cancer Res. 2007;67(16):7572–8. 10. Buschbeck M, Uribesalgo I, Wibowo I, Rue P, Martin D, Gutierrez A, et al. The 35. Clark EL, Coulson A, Dalgliesh C, Rajan P, Nicol SM, Fleming S, et al. The RNA histone variant MacroH2A is an epigenetic regulator of key developmental helicase p68 is a novel androgen receptor coactivator involved in splicing genes. Nat Struct Mol Biol. 2009;16(10):1074–9. and is overexpressed in prostate cancer. Cancer Res. 2008;68(19):7938–46. 11. Gamble MJ, Frizzell KM, Yang C, Krishnakumar R, Kraus WL. The histone 36. Fuller-Pace FV. DEAD box RNA helicase functions in cancer. RNA Biol. variant MacroH2A1 marks repressed autosomal chromatin, but protects a 2013;10(1):121–32. subset of its target genes from silencing. Genes Dev. 2010;24(1):21–32. 37. McGuire BB, Helfand BT, Loeb S, Hu Q, O’Brien D, Cooper P, et al. Outcomes 12. Borghesan M, Fusilli C, Rappa F, Panebianco C, Rizzo G, Oben JA, et al. DNA in patients with Gleason score 8-10 prostate cancer: relation to preoperative Hypomethylation and histone variant MacroH2A1 synergistically attenuate PSA level. BJU Int. 2012;109(12):1764–9. chemotherapy-induced senescence to promote hepatocellular carcinoma 38. Hurtado-Bages S, Guberovic I, Buschbeck M. The MacroH2A11–PARP1 progression. Cancer Res. 2016;76(3):594–606. axis at the intersection between stress response and metabolism. Front 13. Chen H, Ruiz PD, McKimpson WM, Novikov L, Kitsis RN, Gamble MJ. Mac‑ Genet. 2018;9:417. roH2A1 and ATM play opposing roles in paracrine senescence and the senescence-associated secretory phenotype. Mol Cell. 2015;59(5):719–31.