The RING finger domain E3 ubiquitin ligases BRCA1 and the RNF20/RNF40 complex in global loss of the chromatin mark histone H2B monoubiquitination (H2Bub1) in cell line models and primary high-grade serous ovarian cancer

Abstract

Enzymatic factors driving cancer-associated chromatin remodelling are of increasing interest as the role of the cancer epigenome in gene expression and DNA repair processes becomes elucidated. Monoubiquitination of histone H2B at lysine 120 (H2Bub1) is a central histone modification that functions in histone cross-talk, transcriptional elongation, DNA repair, maintaining centromeric chromatin and replication-dependent histone mRNA 3'-end processing, as well as being required for the differentiation of stem cells. The loss of global H2Bub1 is seen in a number of aggressive malignancies and has been linked to tumour progression and/or a poorer prognosis in some cancers. Here, we analyse a large cohort of high-grade serous ovarian cancers (HGSOC) and show loss of global H2Bub1 in 77% (313 of 407) of tumours. Loss of H2Bub1 was seen at all stages (I-IV) of HGSOC, indicating it is a relatively early epigenomic event in this aggressive malignancy. Manipulation of key H2Bub1 E3 ubiquitin ligases, RNF20, RNF40 and BRCA1, in ovarian cancer cell line models modulated H2Bub1 levels, indicative of the role of these RING finger ligases in monoubiquitination of H2Bub1 in vitro. However, in primary HGSOC, loss of RNF20 protein expression was identified in just 6% of tumours (26 of 424) and did not correlate with global H2Bub1 loss. Similarly, germline mutation of BRCA1 did not show a correlation with the global H2Bub1 loss. We conclude that the regulation of tumour-associated H2Bub1 levels is complex. Aberrant expression of alternative histone-associated ‘writer’ or ‘eraser’ enzymes are likely responsible for the global loss of H2Bub1 seen in HGSOC.

Introduction

Histones are the most abundant proteins bound to DNA in eukaryotic cells. Post-translational modifications (PTMs) at specific residues of NH₂-terminal histone tails function as drivers of epigenomic gene regulation (1,2). Histone PTMs are key factors in chromatin remodelling, directing chromatin configuration that determines processes such as transcription and DNA repair that rely on access to open chromatin. The balance between open and closed chromatin in response to developmental programming, diseases including cancer and exogenous factors such as DNA damaging agents, is dictated to a large extent by enzymatic complexes. These complexes are made up of chromatin ‘writers’ (methyltransferases, acetyltransferases, ubiquitinases etc.), ‘erasers’ (demethyltransferases, deacetyltransferases, deubiquitinases etc.), and ‘readers’ that associate with histone PTMs, interpreting and responding to signalling information (1,3).

Extensive cross-talk exists between histone PTMs, with specific histone PTMs acting to influence the writing, erasing and/or reading of other histone modifications (4–7). Central to the significant histone cross-talk is monoubiquitination at lysine 120 of the histone H2B tail (H2Bub1) (8, 9). H2Bub1 functions as a recruiting platform for numerous proteins with known roles in histone cross-talk, transcription and DNA damage, including subunits of the BRCA1-A complex (NBA1, RAP80, BRE, CCDC98, and BRCC3) that are recruited to double strand breaks (DSBs) to mediate DNA repair (8). H2Bub1 has been described as a ‘master switch’ for gene regulation (10–13). Important roles for H2Bub1 have been elucidated, including in the maintenance of centromeric chromatin and the regulation of DNA replication (14,15). Furthermore, H2Bub1 is required for the differentiation of stem cells (16,17) and for the maintenance of replication-dependent histone mRNA 3'-end processing (18), as well as enabling accessibility of DNA repair proteins to chromatin (12,19,20).

H2Bub1 is the result of conjugation of a single 8.5 kDa (76 amino acids) ubiquitin molecule to lysine 120 of the histone H2B tail by one of at least four E3 ligase complexes (reviewed in (21)). This dynamic process is further regulated by deubiquitinases (DUBs) that remove ubiquitin, up to eleven of which have been described for H2Bub1 (13,21–23). The RING finger heterodimer RNF20/RNF40 is generally accepted to be the primary E3 ligase complex mediating the H2Bub1 modification (10,24). RNF20 has been described as a putative tumour suppressor, with hypermethylation of its promoter reported in breast cancer (25) and occasional mutation observed in colorectal cancer (26). In addition, RNF20 transcript levels were found to be lower in metastatic prostate cancer relative to benign disease (27). The Cancer Genome Atlas (TCGA)-ovarian data (28) accessed through cBioPortal for Cancer Genomics (29,30) reported only 3 of 316 (< 1%) high-grade serous ovarian cancers (HGSOCs) with RNF20 missense variants of unknown pathogenicity. No mutations were reported in HGSOC in RNF40 in this same series.

In addition to RNF20/RNF40, the BRCA1/BARD1 complex also functions as an E3 ligase for H2Bub1. In fact, it has been shown to monoubiquitinate all core histones in the nucleosome (31,32). The RING finger consensus motif of BRCA1 is located at the N-terminus from residues 24-64, part of a larger domain comprising residues 1-109 required to form a stable heterodimer with BARD1 (33). Germline and somatic BRCA1 mutations are found in patients with familial breast and ovarian cancer, as well as in sporadic cancers (28,34–37). The potential influence of these mutations on H2Bub1 in these malignancies has not been previously explored. Two other E3 ligases related to H2Bub1 are MDM2 and BAF250B (reviewed in (21)). Based on TCGA-ovarian data, neither MDM2 nor ARID1B (the gene encoding BAF250B) are frequently mutated or genomically rearranged in HGSOC, altered in just 4% and ∼ 2% of these tumours, respectively.

Genome-wide, or global, loss of H2Bub1 has been reported in several malignancies, including breast, colorectal, gastric and lung cancer (38–44). It has been associated with poor patient survival in colorectal cancer (41), and with metastatic breast cancer (39). Strong expression of H2Bub1 has been reported in normal colonic mucosa and in normal breast tissue adjacent to tumours (39,41), as well as in non-tumour-associated gastric mucosa (43). We have shown that loss of H2Bub1 in parathyroid cancer is associated with loss-of-function mutations in the tumour suppressor CDC73, a binding partner of the E3 ligase complex RNF20/RNF40 and a member of the human RNA polymerase II-associated factor 1 (PAF1) transcriptional complex (45). Other links between H2Bub1 and genetic variants associated with cancer remain to be elucidated.

Here we investigate the role of the E3 ubiquitin ligases BRCA1, RNF20 and RNF40 in regulating H2Bub1 levels in cell line models of HGSOC. We contrast those findings with an immunohistochemical analysis of H2Bub1 and RNF20 levels in a large cohort of clinically annotated primary HGSOC, the majority of which have been characterised for germline mutations in BRCA1 and BRCA2.

Results

The E3 ligases BRCA1, RNF20 and RNF40 modulate levels of H2Bub1 in vitro

In order to determine whether there was a difference between the ability of wild-type and RING finger mutant (C61G) BRCA1 to modulate H2Bub1 levels, we transiently expressed wild-type and mutant BRCA1 constructs in the BRCA1-null cell line UWB1.289. While a relatively low level of BRCA1 mRNA was detected in parental empty vector control cells, we were unable to detect BRCA1 protein in these cells by Western blot, consistent with an earlier publication (Fig. 1A) (46). Transient expression of wild-type BRCA1, validated by western blot (Fig. 1A), resulted in a statistically significant increase in H2Bub1 compared to an empty vector transfection (P < 0.02), while total H2B levels remained constant. Expression of the RING finger domain BRCA1 mutant did not significantly increase H2Bub1 levels (Fig. 1A and B). Mutant and wild-type BRCA1 RNA was expressed at similar levels (Fig. 1C). Conversely, we found that down-regulation of endogenous wild-type BRCA1 using two distinct siRNA in the cell lines COV318 and A2780 significantly decreased H2Bub1 levels compared to a non-silencing control in both cell lines (Fig. 2). Taken together, modulation of levels of BRCA1 in cell line models of ovarian cancer had direct effects on H2Bub1 that were not seen for RING finger domain mutant BRCA1.

Figure 1.

Overexpression of wild-type, but not mutant, BRCA1 increases H2Bub1 levels (A) Representative blot of UWB1.289 72 h post transfection. EV (empty vector), BRCA1wt (wild-type BRCA1), BRCA1mut (C61G RING domain mutant). (B) Protein data expressed as fold change relative to EV. ANOVA, mean ± S.E.M. (*P < 0.02, N = 6). (C)BRCA1 mRNA expressed relative to EV. Hydroxymethylbilane Synthase (HMBS) was used as the reference gene. ANOVA, mean ± S.E.M. (P > 0.05, N = 5).

Figure 2.

Down-regulation of BRCA1 decreases H2Bub1 levels. Representative blots shown for down-regulation of wild-type BRCA1 with two independent siRNA (BRCA1-13si and BRCA1-14si). Data is expressed as fold change relative to the non-silencing (n.s.) siRNA. (A) COV318 (shown for siRNA concentrations between 5 – 50 nM) and (B) A2780 (shown for 50 nM siRNA) showed reduction of H2Bub1 in BRCA1 down-regulated cells compared to a n.s. control. H2Bub1 levels are reported relative to total H2B. ANOVA, mean ± S.E.M. (C) COV318 (*P < 0.01; N = 5) and (D) A2780 (*P < 0.01; N = 7).

Given that the RNF20-RNF40 complex is recognized as the main E3 ligase complex enabling H2Bub1 (5), we sought to determine whether decreasing either complex member had an effect on H2Bub1 levels. Independent siRNA down-regulation of RNF20 and RNF40, using two distinct siRNAs in the cell lines COV318 and A2780 significantly decreased H2Bub1 levels relative to a non-silencing control (Fig. 3). As has been previously reported, down-regulation of RNF20 led to down-regulation of RNF40 and vice versa, likely indicating the reliance of each protein on the presence of the other for a stable heterodimeric complex (19).

Figure 3.

Down-regulation of either RNF20 or RNF40 decreases H2Bub1 levels. Representative blots are shown for down-regulation of RNF20 or RNF40 using two separate siRNA at a concentration of 50 nM (for RNF20 - #1 and #3; for RNF40 - #1 and #2). Data are expressed as fold change relative to the non-silencing (n.s.) siRNA. (A) COV318 and (B) A2780 showed reduction of H2Bub1 in RNF20 or RNF40 down-regulated cells compared to a n.s. control. H2Bub1 levels are reported relative to total H2B. ANOVA, mean ± S.E.M. (C) COV318 (*P < 0.001; N = 3) and (D) A2780 (*P < 0.001; N = 4).

H2Bub1 is commonly attenuated in primary HGSOC

With in vitro gain- and loss-of-function data consistent with a role for BRCA1 and RNF20/RNF40 in modulating H2Bub1 levels, we sought to investigate the impact of the loss of these proteins in HGSOC specimens. H2Bub1 IHC data were obtained from an initial cohort of 463 HGSOC that were enriched for known germline BRCA1/2 mutation carriers (Table 1). Fifty-six samples were excluded for technical reasons including TMA core drop-out, insufficient neoplastic material for assessment or absent staining of internal non-neoplastic cells (N = 56, Supplementary Material, Fig. S1) leaving a final cohort of 407 HGSOC with interpretable IHC data. The majority of HGSOC, 77% (313 of 407) showed a total loss of nuclear H2Bub1 (IHC score 0). Low to intermediate nuclear H2Bub1 (IHC score 1) was observed in 19% of HGSOC (78 of 407) and strong nuclear staining (IHC score 2) was observed in only 4% (16 of 407) HGSOC (Fig. 4A, Table 1). Upon further exclusion of 9 cases where FIGO stage was not known (Supplementary Material, Fig. S1), 398 HGSOC were analysed for the distribution of H2Bub1 levels according to FIGO stage (Fig. 4B). No difference was observed in H2Bub1 IHC scores across tumours from patients diagnosed with stages I–IV disease (Kruskal-Wallis H test, χ²(3, N = 398) = 2.338, P = 0.505). When analyses were conducted grouping no staining (IHC = 0) and any staining (IHC = 1 or 2) and compared across tumour stages, differences remained not significant (Kruskal-Wallis H test, χ²(3, N = 398) = 2.408, P = 0.492). Furthermore, no differences were observed between HGSOC from women who received neoadjuvant therapy, compared with those who had not (Kruskal-Wallis H test, χ²(1, N = 407) = 0.384, P = 0.536), suggesting that previous exposure to chemotherapy did not alter levels of H2Bub1 in these tumours.

Figure 4.

Assessment of H2Bub1 levels in HGSOC. (A) i–iii, representative H2Bub1 IHC score 0, 1 and 2 respectively (magnification, x200); iv-vi, H2B IHC of sequential sections for each specimen (magnification, x200) remains constant, IHC score 2. (B) Distribution of H2Bub1 levels within HGSOC tumour stages I-IV shown graphically as a percentage of total tumours within each stage. Actual numbers of samples within each stage are depicted for each IHC score. No difference was observed in H2Bub1 IHC scores across tumour stages (Kruskal-Wallis H test, χ²(3, N = 398) = 2.338, P = 0.505). (C) Kaplan-Meier plots for Progression-free Survival (PFS; diagnosis to first progression or last follow-up) and (D) Overall Survival (OS; diagnosis to death or last follow-up) based on H2Bub1 IHC scores. No significant differences in survival were observed (PFS, P = 0.495; OS, P = 0.772).

Table 1.

H2Bub1 immunostaining in HGSOC

H2Bub1 IHCaScore	HGSOC
0	77% (313 of 407)
1	19% (78 of 407)
2	4% (16 of 407)
Total	407
Unscored	56
overall cohort	463

^aimmunohistochemical score.

Levels of H2Bub1 were not correlated with PFS (diagnosis to first progression or last follow-up) or OS (diagnosis to death or last follow-up) with Hazard Ratios respectively HR = 0.871; 95% CI 0.688–1.104, P = 0.254 and HR = 0.925; 95% CI 0.729–1.173, P = 0.520 (Fig. 4C and D). Survival analyses comparing no H2Bub1 (IHC score 0) to any H2Bub1 (IHC scores 1 and 2) also did not show significant differences (data not shown).

H2Bub1 levels do not correlate with germline BRCA1 and/or BRCA2 mutations in HGSOC

We next sought to determine whether there was a correlation between H2Bub1 levels and BRCA1 mutation status in HGSOC. Germline testing for BRCA1 and BRCA2 mutations had been previously undertaken in 347 women, with 76 samples positive for a BRCA1 mutation, 46 BRCA2 and 209 wild-type for BRCA1 and BRCA2 (Supplementary Material, Table S1). Samples with variants of unknown consequences (N = 16) were excluded from these analyses, as were samples of known BRCA status that were not scored for H2Bub1 (Supplementary Material, Fig S2). This left a final cohort of 283 samples that were characterized for both germline BRCA mutations and H2Bub1 IHC.

While down-regulation of BRCA1 in vitro resulted in loss of H2Bub1, tumours harbouring mutant BRCA1 did not show decreased H2Bub1 compared with either BRCA2 mutant tumours, or wild-type BRCA tumours (Kruskal-Wallis H test, χ²(2, N = 283) = 1.410, P = 0.494; Fig. 5A). Additional comparisons assessing H2Bub1 staining for BRCA1 versus BRCA2 and BRCA wild-type tumours grouped together, or total loss of H2Bub1 staining (IHC = 0) with ‘any’ H2Bub1 staining (IHC = 1 or 2) against BRCA status were not significant (data not shown).

Figure 5.

H2Bub1 levels in HGSOC of known BRCA status. (A) Distribution of H2Bub1 levels within HGSOC of known BRCA1/2 status shown graphically as a percentage of total tumours within mutant or wild-type categories. Actual numbers of samples within each mutant category are depicted for each IHC score. BRCA1 mutant tumours did not show decreased H2Bub1 compared with either BRCA2 mutant tumours, or wild-type BRCA tumours (Kruskal-Wallis H test, χ²(2, N = 283) = 1.410, P = 0.494). (B) Distribution of H2Bub1 IHC levels of BRCA1 mutants based on location within and outside of the BRCA1 RING domain shown graphically as a percentage of total tumours within each category. Actual numbers of samples within each mutant category are depicted for each IHC score (N = 66 due to core drop-out and variants of unknown significance). No difference in H2Bub1 levels due to position of mutants within or external to the RING domain were observed (Kruskal-Wallis H test, χ²(1, N = 66) = 2.163, P = 0.141). (C) Kaplan-Meier plots for PFS, and (D) OS comparing BRCA1 RING domain mutants, BRCA1 non-RING domain mutants, BRCA2 and wild-type tumours. Significant differences in survival were observed (PFS, P = 0.005; OS, P < 0.0001).

Of the 76 samples with BRCA1 mutations, 15 (∼20%) had a mutation predicted to affect the E3 RING finger structural domain of BRCA1, comprising residues 1 – 109 with the consensus RING finger domain motif encoded by residues 24-64 (33). Differences in H2Bub1 levels based on location of mutation within or outside of the RING domain were not significant (Fig. 5B; Kruskal-Wallis H test, χ²(1, N = 66) = 2.163, P = 0.141; 10 BRCA1 mutants were not scored for H2Bub1). The inability to show a significant difference may be due to relatively small numbers of HGSOC with BRCA1 mutations within the RING domain. Patients with either a BRCA1 RING finger domain mutant or non-RING finger domain mutant, as well as patients with a BRCA2 mutation, showed increased PFS and OS compared to patients who were wild-type for BRCA (Kaplan-Meier analyses, P = 0.005 and P < 0.001, respectively; Fig. 5C and D), with Hazard Ratios respectively HR = 1.291; 95% CI 1.108 – 1.504, P = 0.001 and HR = 1.383; 95% CI 1.184 – 1.616, p < 0.0001. This is consistent with previous reports showing increased survival for BRCA mutation carriers (28).

RNF20 and CDC73 levels in primary HGSOC

RNF20 IHC data were obtained from the same cohort of patients which had undergone H2Bub1 IHC, with 39 samples excluded (Supplementary Material, Fig. S3) due to core drop-out, insufficient neoplastic material or absent internal positive controls, leaving a final cohort of 424 samples with RNF20 IHC data (Table 2). The majority of HGSOC, 87% (370 of 424) showed high levels of RNF20 (IHC score 2). Only 6% of tumours (26 of 424) showed a total loss of RNF20 (IHC score 0), while an additional 7% (28 of 424) showed intermediate staining levels for RNF20 (IHC score 1) (Fig. 6A and B). Upon further exclusion of samples of unknown stage (an additional 7 samples; Supplementary Material, Fig. S3), 417 HGSOC were analysed for the distribution of H2Bub1 levels within tumour stage (Fig. 6B). No difference was observed in RNF20 IHC scores across tumour stages I – IV (Kruskal-Wallis H test, χ²(3, N = 417) = 1.727, P = 0.631). Furthermore, no differences in RNF20 IHC scores were observed between samples from women subjected to neoadjuvant therapy, compared with those who were not (Kruskal-Wallis H test, χ² (1, N = 424) = 0.88, P = 0.767).

Figure 6.

Assessment of RNF20 levels in HGSOC. (A) i–iii, representative RNF20 IHC scores 0, 1 and 2 respectively (magnification, x200) (B) Distribution of RNF20 levels within tumour stages I-IV shown graphically as a percentage of total tumours within each stage. Actual numbers of samples are depicted for each column. No difference was observed in RNF20 IHC scores across tumour stages (Kruskal-Wallis H test, χ²(3, N = 417) = 1.727, P = 0.631).

Table 2.

RNF20 immunohistochemistry in HGSOC

RNF20 IHCaScore	HGSOC
0	6% (26 of 424)
1	7% (28 of 424)
2	87% (370 of 424)
Total	424
Unscored	39
overall cohort	463

^aimmunohistochemical score.

Levels of RNF20 were not correlated with patient progression free survival or overall survival with Hazard Ratios respectively HR = 1.022; 95% CI 0.828 - 1.262, P = 0.837 and HR = 1.043; 95% CI 0.842 - 1.291, P = 0.702 (Supplementary Material, Fig S4). Further, patient survival was not different dependent upon any loss of RNF20 (IHC scores 0 or 1) compared to higher levels of RNF20 (IHC score 2; data not shown).

It is noted that no tumour had both a BRCA1 mutation and total loss of RNF20 expression, suggesting that loss of function of both E3 ligases that work to monoubiquitinate histone H2B at lysine 120 may negatively impact upon cell survival (Table 3). It cannot be excluded however, that expansion of this study to a larger cohort would identify HGSOC where both BRCA1 and RNF20 were affected. Lastly, CDC73 stained strongly in all tumours, excluding loss of CDC73 as a mechanism of loss of global H2Bub1. We have previously defined strong CDC73 staining as diffuse nuclear staining in > 95% of all tumour cells (data not shown) (47).

Table 3.

BRCA mutation and RNF20 immunostaining in HGSOC (N = 298)

	BRCA1 mutant	BRCA2 mutant	BRCA wild-type
RNF20 IHC
0	0	1	9
1	5	3	8
2	68	39	165

Discussion

Chromatin remodelling contributes to aberrant gene expression associated with cancer. We have analysed a large HGSOC cohort and shown that the vast majority, ∼77%, of these tumours have lost global H2Bub1, synonymous with a more condensed chromatin configuration. Our data show that loss of global levels of H2Bub1 is a relatively early event in HGSOC, given that it occurs in Stage I HGSOC, and at a similar frequency across all HGSOC Stages (I–IV). Another example of an event occurring at an early stage of HGSOC tumour progression is a mutation of TP53, in fact TP53 mutation is almost a sine qua non of HGSOC (reviewed in (48)).

The discovery that the global loss of H2Bub1 is associated with HGSOC is consistent with reports of H2Bub1 loss in other aggressive malignancies. We have previously shown a loss of global H2Bub1 in the rare aggressive tumour parathyroid carcinoma that occurs either sporadically, or as part of the familial syndrome Hyperparathyroidism Jaw Tumour syndrome (45). Within this syndrome, CDC73 mutated parathyroid lesions diagnosed as adenomas are viewed as having ‘malignant potential’. Parathyroid tumours with global loss of H2Bub1 demonstrated nuclear loss of the tumour suppressor CDC73, most often due to a germline or somatic loss-of-function CDC73 mutation (45). CDC73 is a member of the RNA polymerase II PAFI complex and binding partner of the E3 ubiquitin ligases RNF20 and RNF40 (45). In the current study, nuclear CDC73 was expressed at high and consistent levels across all tumours, excluding loss of CDC73 as a mechanism of global H2Bub1 loss in HGSOC. Further, analysis of the TCGA-ovarian dataset does not reveal mutation of CDC73. In breast cancer, loss of H2Bub1 is reported in malignant and metastatic tumours, but not in normal mammary epithelium or benign tumours (39,44). No genetic events have been proposed for loss of global H2Bub1 in breast cancer. The loss of global H2Bub1 in tumour tissue with retention in normal tissue has also been reported in colon, gastric and lung tumours (40,41,43). Furthermore, Urasaki and colleagues suggest loss of H2Bub1 may be linked to aberrant glucose metabolism in cancer cells that is not seen in normal tissue (40).

H2Bub1 levels did not correlate with patient outcomes in HGSOC. Poorer survival outcomes associated with loss of H2Bub1 have recently been reported for high grade colorectal cancers (41). Further, positive H2Bub1 staining in gastric carcinoma has been associated with higher 5 year survival relative to patients whose tumours stained negatively for H2Bub1 (43). Our cohort was relatively large; however, given that most women with HGSOC present with Stage III or IV disease, it was skewed towards more advanced tumours. While this may have impacted upon our ability to detect differences in survival outcomes based on H2Bub1 levels, the fact that loss of global H2Bub1 occurs as a relatively early event in HGSOC suggests that these tumours are a more homogeneous group in relation to H2Bub1 than other tumour types such as colorectal or gastric cancer where survival differences have been reported (41,43).

While we show that down-regulation of RNF20 and RNF40 decreased the level of H2Bub1 in HGSOC cell line models, loss of RNF20 was an infrequent event in primary HGSOCs. Only 6% of HGSOC showed a total loss of this protein, and a further 7% showed intermediate staining in our cohort. Furthermore, loss of RNF20 did not correlate with reduced or absent H2Bub1 in HGSOC, suggesting that other E3 ubiquitin ligases can monoubiquitinate H2Bub1 in vivo in the absence of RNF20. Given strong staining for RNF20 in the vast majority of HGSOC, it is unlikely that RNF20 is hypermethylated as has been reported for breast cancer (25). TCGA-ovarian data also shows that RNF20 is rarely mutated in HGSOC. These findings appear to be in contrast to recent reports in colorectal cancer showing reduced expression of RNF20 and RNF40 mRNA, although a direct comparison between H2Bub1 and RNF20/RNF40 was not undertaken in human tumours in this study (42). Furthermore, it is interesting to note that colons from patients with ulcerative colitis have shown lower levels of RNF20/RNF40 and that levels of these E3 ubiquitin ligases were inversely correlated with levels of inflammatory cytokines such as IL-6 and IL-8 (42). Both RNF20 and RNF40 have also been shown to be reduced in testicular seminoma compared to normal tissue (49).

Our in vitro models of HGSOC clearly show that H2Bub1 levels are decreased through down-regulation of BRCA1. Furthermore, overexpression of BRCA1 in the BRCA1 null cell line UWB1.289 increased levels of H2Bub1, whereas overexpression of a BRCA1 RING finger domain mutant did not. We were, however, unable to show a correlation between BRCA1 mutation and loss of H2Bub1 in primary HGSOC. It is possible that we have not detected all BRCA1 dysfunction in this study, given that we have not screened for somatic BRCA1 mutation or hypermethylation of the BRCA1 promoter. Clearly however, numerous BRCA1 germline mutants showed intermediate (IHC score of 1) or high (IHC score of 2) levels of H2Bub1, suggesting that lack of these additional analyses has not been a confounding factor for our overall conclusions. Consistent with previous reports, patients with a germline BRCA1 or BRCA2 mutation showed improved PFS and OS compared to those that were wild-type for BRCA (28).

In summary, regulation of H2Bub1 in primary tumours is clearly complex and relies on multiple enzymatic processes. The importance of RNF20 in regulating global H2Bub1 in primary tumours appears to differ between tumour types given that this E3 ubiquitin ligase is decreased in colorectal cancers and testicular seminomas, yet is infrequently lost in HGSOC and does not always correlate with loss of H2Bub1. The key enzymatic factor(s) that are aberrantly regulated and responsible for the extensive loss of global H2Bub1 seen in HGSOC remain to be elucidated. As a potential therapeutic target, however, H2Bub1 depletion via inhibition of RNF20 warrants further attention given the role of H2Bub1 in facilitating DNA repair that may assist tumours to overcome therapeutic DNA damage.

Materials and Methods

Cell lines

The human HGSOC cell line COV318 (50) was purchased from Sigma-Aldrich European Collection of Authenticated Cell Cultures (ECACC) Oceania Inventory (Castle Hill, NSW, Australia). UWB1.289 is a BRCA1-null human papillary serous ovarian carcinoma, carrying an exon 11 BRCA1 mutation (c.2594delC resulting in a premature STOP at codon 845 of BRCA1) and loss of the corresponding wild-type allele, obtained from the American Type Culture Collection (ATCC; Manassas, VA, USA) (46). A2780 cells (51) were also obtained from ATCC. We note that A2780 cells have recently been queried as a strong model for HGSOC based on genomic data; however, they were used here as an in vitro model to test the effects of exogenous alteration in gene expression (50).

COV318 cells were maintained in DMEM (HyClone #SH30243, GE Healthcare Life Sciences) and A2780 cells in RPMI 1640 (HyClone #SH30027, GE Healthcare Life Sciences), both supplemented with 10% FCS (AusGeneX, Molendinar, QLD, Australia) and incubated at 37 °C in a humidified 5% CO₂ atmosphere. The UWB1.289 cell line was cultured in 50% RPMI 1640 (HyClone #SH30027, GE Healthcare Life Sciences), 50% MEGM (Clonetics™ MEBM supplemented with SingleQuot additives #CC-3150 from Lonza, Walkersville, MD, USA), supplemented with 3% FCS (AusGeneX, Molendinar, QLD, Australia) and incubated at 37 °C in a humidified 5% CO₂ atmosphere.

Authentication of the cell lines was performed by CellBank Australia (Westmead, New South Wales, Australia) by short tandem repeat profiling using the AMPFLSTR^® Identifiler^® PCR Amplification Kit (Applied Biosystems), a 16 loci (15 STR loci plus Amelogenin) STR multiplex kit.

Expression of wild-type and RING finger domain mutant BRCA1

Twenty four hours prior to transfection, 2x10⁵ UWB1.289 cells were seeded into single wells of 6 well culture dishes. Cells were transfected with 1-2 μg of empty vector (EV) or wild-type BRCA1 or BRCA1 RING finger domain mutant (C61G) constructs, BRCA1wt or BRCA1mut respectively, in pYFP-CMV (52) using Xtreme Gene 9 according to the manufacturer’s instructions (Roche Diagnostics Australia Pty Ltd, NSW, Australia). Cells were harvested 72h post transfection and underwent immunoblotting for BRCA1, total histone H2B and H2Bub1, and GAPDH, as well as being analysed for BRCA1 mRNA by quantitative real-time reverse transcription PCR (qRT-PCR).

siRNA down-regulation of BRCA1, RNF20 and RNF40

The following siRNAs were purchased from Qiagen (Chadstone, VIC, Australia); BRCA1 siRNA #13 (SI02654575), BRCA1 siRNA #14 (SI02664361); RNF20 siRNA#1 (SI00124173), RNF20 siRNA #3 (SI00124187); RNF40 siRNA #1 (SI00108045), RNF40 siRNA #2 (SI00108052). 2x10⁵ COV318 or 3.5 x10⁵ A2780 cells were seeded into single wells of 6 well culture dishes and allowed to attach overnight. Cells were transfected with 5-50 nM of the gene of interest siRNA or a non-silencing negative control (Allstars) using HiPerfect transfection reagent (Qiagen). Cells were harvested 48h post transfection.

Immunoblotting and gene expression

At the time of harvest, cells were washed with cold PBS, lysed in urea buffer (20 mM NaH₂PO4, 0.5 M NaCl, 20 mM imidazole, 8 M urea, 0.5% triton, 20 mM Tris, pH 8, 0.5 mM DTT, 0.5 mM iodacetamide), removed from dishes by scraping and protein-loading buffer (6% w/v SDS, 40% w/v sucrose, 20 mM Tris, pH 6.8, 0.15% w/v bromophenol blue) added at a 1:3 ratio to urea buffer. Extracts were sonicated for 30 s and incubated at 95 °C for 5 min before being separated on 4-12% Bis-Tris gels (Life Technologies, Thornton, NSW, Australia) at 180 V for 1 h followed by transfer to a nitrocellulose membrane (Amersham™ Protran® Supported Western Blotting Membrane, GE Healthcare, Sigma-Aldrich) using a wet transfer system at 100 V for 1.5–2 h (Bio-Rad Laboratories, Gladesville, NSW, Australia). Membranes were blocked with 5% skim milk for 1 h at ambient temperature, and probed overnight at 4 °C with the following primary antibodies: anti-monoubiquitinated H2B-K120 antibody (Medimabs, Montreal, Canada); anti-Histone H2B, anti-RNF20 or anti-RNF40 antibody (Abcam, Cambridge, MA, USA); GAPDH [14C10], or anti-BRCA1 (Cell Signaling Technology, Danvers, MA, USA). Membranes were probed with peroxidase labelled secondary antibodies (anti-rabbit IgG HRP-linked whole antibody or anti mouse IgG-HRP, both from GE Healthcare, VWR International, Murarrie, Queensland, Australia) for 1 h. Chemiluminescent signal was detected by SuperSignal ECL Dura reagent (Pierce, Rockford, IL, USA) and visualized using the Fujifilm LAS-4000 imaging system (Berthold Australia Pty. Ltd., Bundoora, VIC, Australia). Quantitation was undertaken using Multi Gauge 3.0 software (Fujifilm Australia Pty. Ltd., Brookvale, NSW, Australia).

Total RNA was extracted from all cell lines using the RNeasy mini kit automated on a QIACube (Qiagen, Doncaster, VIC, Australia). RNA (500 μg) was reverse transcribed using Superscript III reverse transcriptase (Life Technologies). Quantitative real-time RT-PCR was performed in triplicate using a TaqMan Gene Expression Assay for BRCA1 HS01556193_m1; Life Technologies) and an endogenous reference gene (hydroxymethylbilane synthase (HMBS); Hs00609297_m1, Life Technologies), using iTaq Universal Probes Supermix (Bio-Rad Laboratories) on a 7900HT Fast Real-Time PCR System (Applied Biosystems). Reagents were aliquoted using an epMotion 5070 robot (Eppendorf South Pacific Pty Ltd., North Ryde, NSW, Australia).

Ovarian tumour cohort and clinical data

Tissue microarrays (TMAs) representing 463 HGSOCs were obtained from the Australian Ovarian Cancer Study (AOCS, N = 358) and the Gynecological Oncology Biobank at Westmead (N = 105). The median age at diagnosis was 59 years (range 29 to 84 years; Table 4). Surgical stage was assessed in accordance with the International Federation of Gynecology and Obstetrics (FIGO) classification. Of the total cohort, 12 cases were Stage I, 23 cases were Stage II, 348 cases Stage III, 70 cases Stage IV and 10 cases did not have tumour stage recorded. At the time of censoring for this analysis, 28% (131 of 463) of women were alive and 72% (332 of 463) were known to be deceased (Table 4). Data were recorded on overall survival (diagnosis to death or last follow-up), progression free survival (diagnosis to first progression or last follow-up) and whether the patient had received neoadjuvant treatment (Supplementary Material, Table S1).

Table 4.

Clinico-pathologic characteristics of the HGSOC cohort

Clinico-pathologic characteristics	Cohort, N=463
	Years
Age at diagnosis:
median	59
range	29–84
FIGO Stage:	Cases
I	12 (2.6%)
II	23 (5.0%)
III	348 (75.2%)
IV	70 (15.1%)
unknown	10 (2.2%)
BRCA germline mutation:	cases
BRCA1	76 (16.4%)
BRCA2	46 (9.9%)
unknown variant (UV)	16 (3.5%)
wild-type	209 (45.1%)
not tested	116 (25.1%)
Neoadjuvant chemotherapy:	cases
Yes	45 (9.7%)
No	418 (90.3%)
Mortality at censoring:	cases
Alive	131 (28.3%)
Dead	332 (71.7%)

The majority of the cohort, 347 women, had previously undergone germline testing for mutations in BRCA1 and/or BRCA2, either clinical cancer genetic testing or as previously reported (34), with 76 samples found to be BRCA1 mutant, 46 BRCA2 mutant and 209 wild-type for BRCA1/2. A further 16 samples had variants of unknown pathological consequence. Germline BRCA mutations were not tested for in 116 women (Supplementary Materials, Table S1, Fig. S2).

Immunohistochemical analyses of HGSOC

Immunohistochemistry was performed on tissue microarrays containing formalin fixed paraffin-embedded tumour tissue for H2Bub1 using a mouse monoclonal antibody (Medimabs, MM-0029) at a dilution of 1:2000. Heat-induced epitope retrieval was performed at 97 °C for 30 min in the manufacturer’s (Leica Microsystems) acidic retrieval solution ER1. A total H2B mouse monoclonal antibody (Abcam, Clone 52484) was used at a dilution of 1 in 500. Heat-induced epitope retrieval was performed at 97 °C for 30 min in the manufacturer’s alkaline retrieval solution ER2. For RNF20, a rabbit polyclonal antibody (Abcam) was used at a dilution of 1:100. Heat-induced epitope retrieval was performed as for histone H2B. Methodology for CDC73 immunohistochemistry, previously known as parafibromin, was as we have previously published (47). All immunohistochemistry was automated using a Leica BOND-III™ autostainer (Leica Biosystems, Mount Waverley, Victoria, Australia). Scoring was performed in a blinded fashion by experienced surgical pathologists [AJG or AC]: 0 (no staining), 1 (weak to moderate nuclear staining), or 2 (dark nuclear staining). In all cases, where multiple cores from the same sample were present, the highest immunohistochemical (IHC) score was used in analyses.

Statistical analyses

Data analyses were undertaken using IBM SPSS software version 22.0 (SPSS Australasia Pty Ltd., Chatswood, NSW, Australia). Western blot and qRT-PCR data from analyses of ovarian cancer cell lines are expressed as the mean ± S.E.M. from a minimum of three independent experiments. Statistical significance was determined by one-way ANOVA. The potential correlation between H2Bub1 levels and tumour stage or BRCA mutation status was determined by the Kruskal-Wallis H test or Mann Whitney U test as indicated. Differences in overall or progression free survival, OS and PFS respectively, based on H2Bub1 or RNF20 levels, or BRCA mutation, were analysed using the Kaplan-Meier method and compared by the log-rank (Mantel-Cox) test. The Cox proportional hazards model was used to determine Hazard Ratios examining the relationship between H2Bub1 or RNF20 levels and overall or progression free survival. For all analyses, P < 0.05 was considered to be statistically significant.

Supplementary Material

Supplementary Material is available at HMG online.

Acknowledgements

The authors wish to acknowledge the Australian Ovarian Cancer Study and the Gynecological Oncology Biobank at Westmead. We acknowledge the contribution of the AOCS study nurses and research assistants and would like to thank all of the women who participated in the study. We gratefully acknowledge other members of the AOCS Management Group, G Chenevix-Trench, A Green, P Webb, D Gertig, the AOCS collaborators (see http://www.aocstudy.org; date last accessed November 2, 2016), and the co-operation of the following institutions: New South Wales: John Hunter Hospital, North Shore Private Hospital, Royal Hospital for Women, Royal North Shore Hospital, Royal Prince Alfred Hospital, Westmead Hospital; Queensland: Mater Misericordiae Hospital, Royal Brisbane and Women’s Hospital, Townsville Hospital, Wesley Hospital; South Australia: Flinders Medical Centre, Queen Elizabeth II, Royal Adelaide Hospital; Tasmania: Royal Hobart Hospital; Victoria: Freemasons Hospital, Mercy Hospital for Women, Monash Medical Centre, Royal Women’s Hospital; Western Australia: King Edward Memorial Hospital, St John of God Hospitals Subiaco, Sir Charles Gairdner Hospital, Western Australia Research Tissue Network (WARTN).

Conflict of Interest statement. None declared.

Funding

This work was supported by the Cancer Council NSW, Australia [RG13-10 to D.J.M., A.J.G. and G.B.G.]; Australian Research Council [FT100100489 to D.J.M]; National Health and Medical Research Council [APP1004799 to D.J.M], Australia Postgraduate Award and Sydney Medical School – Northern scholarship, Australia [to A.J.C.]. The Australian Ovarian Cancer Study was supported by the U.S. Army Medical Research and Materiel Command under DAMD17-01-1-0729, The Cancer Council Victoria, Queensland Cancer Fund, The Cancer Council New South Wales, The Cancer Council South Australia, The Cancer Foundation of Western Australia, The Cancer Council Tasmania and the National Health and Medical Research Council of Australia (NHMRC; ID400413, ID400281). The Gynecological Oncology Biobank at Westmead was funded by NHMRC Enabling Grants ID310670 & ID628903 and Cancer Institute NSW Grants 12/RIG/1-17 & 15/RIG/1-16. AdeF is funded by the University of Sydney Cancer Research Fund and the Cancer Institute NSW through the Sydney-West Translational Cancer Research Centre.