Introduction

Recent advancements in the field of precision oncology have introduced a new class of therapeutic agents known as chromatin remodeler drugs. These constant discoveries are especially significant in difficult-to-treat cancers. In this context, the opposing gene-regulatory function between switch/sucrose nonfermenting (SWI/SNF) and polycomb repressive (PRCs) complexes has been a topic of extensive research. For genes bound by the SWI/SNF complex, inactivation of different subunits has been demonstrated to contribute to elevated levels of histone H3 lysine-27 trimethylation (H3K27me3), which is a PRC2-written repressive chromatin mark and its accumulation leads the chromatin structure become condensed promoting tumor growth and proliferation by suppressing transcription of essential genes1. Furthermore, it was observed that certain SWI/SNF mutant tumor models exhibited increased sensitivity to EZH2 inhibition due to their primary dependency upon the non-catalytic role of EZH2 in stabilization of the PRC2 complex, and partial dependency on EZH2 histone methyltransferase activity which will be annulled by EZH2 inhibitors, ultimately blocking the aberrant PRC2 activity2. Drugs targeting EZH2 for SMARCB1-negative or SMARCA4-negative solid tumors (NCT: 03213665) are in clinical development. Tazemetostat has been approved for treating advanced epithelioid sarcoma, characterized by loss of SMARCB1, and for follicular lymphoma with EZH2 mutation3,4.

Polybromo 1 (PBRM1), a key regulator in the Polybromo-associated BAF (PBAF) subunit, is part of the larger SWI/SNF family of chromatin remodelers and exhibits a high mutation rate across various cancer types. Notably, PBRM1 exhibits the highest mutation frequency in clear cell renal cell carcinoma at (ccRCC) 27.98%. Additionally, its occurrence in biliary tract carcinomas approaches one mutation per five tumors (19.44%), making it the second most common malignancy in terms of mutation incidence5. Enhancer of Zeste Homolog 2 (EZH2) acts as a transcriptional repressor through histone methylation and serves as the functional enzymatic component of the Polycomb Repressive Complex 2 (PRC2).

Interestingly, there is some evidence of synthetic lethality between PBRM1-Mutated Human Chordoma Xenograft and EZH2 inhibition with Tazemetostat resulting in a 100% overall response rate and significant tumor growth delay6. There is also in-vitro data showing that other EZH2 inhibitors (GSK126) have single-agent activity against PBRM1-deficient ccRCC cell lines, showing inhibition of the proliferation, down-regulation of the tri-methylation of histone H3 at Lysine 27 (H3K27me3) as well as apoptotic activities7.

Considering the aforementioned information as well as the unforeseen prolonged survival of our patient with PBRM1 mutant metastatic cholangiocarcinoma, we decided to delve into the in-silico relationship between PBRM1 and EZH2 to comprehend the reason why EZH2 inhibition would elicit a response in this particular patient.

Results

Case description

A 64-year-old man with a smoking index of 60 and a history of Chronic Obstructive Pulmonary Disease (COPD), hypertension, obesity, and Type 2 Diabetes, was diagnosed with a grade 2 multifocal intrahepatic cholangiocarcinoma (Magnetic Resonance of the liver described two lesions located in hepatic segments III and IV) AJCC-TNM 8th Edition cT2Nx.

In August 2018, the patient underwent a partial hepatectomy with resection of segment IV extended posteromedial towards segment III, achieving complete resection and a pathological report showed intrahepatic cholangiocarcinoma AJCC-TNM 8th Edition Stage II (pT2N0). No significant postoperative complications were detected. Seven weeks after surgery the patient received adjuvant treatment with capecitabine for 8 cycles, with relatively good tolerance, experiencing only grade 1 skin and digestive toxicity, ending in February 2019 and initiating follow-up.

In April 2020, the subject presented mesenteric relapse on the Positron Emission Tomography and Computed Tomography (PET-CT) later confirmed by biopsy and started first-line therapy based on Cisplatin-Gemcitabine, completing just six cycles due to poor tolerance secondary to diarrhea and vomiting grade 2, so it was decided to switch to maintenance therapy with gemcitabine until April 2021. At this moment, DNA sequencing (FoundationOne®) was carried out in tumor tissue, showing the presence of a PBRM1 gene alteration in splice site 3617-1 G > T. This is a known pathogenic mutation that affects a splice site. No other pathogenic alterations were detected, and some variants of unknown significance were found in CYP17A1, MERTK, SMAD2, EPHB1, RPTOR, AR, and POLD1 genes. Tumor Mutational Burden Score (TMB) and Microsatellite Instability status were not evaluable.

In August 2021 the CT scan showed new nodal progression, with an increase in the size of abdominopelvic adenopathies, mainly external iliac 17 mm (previous 12 mm) and right inguinal 17 mm (previous 12 mm). After this progression, the patient was referred to our center to evaluate new therapeutic opportunities within a clinical trial. The patient was included in a clinical trial with Tazemetostat (NCT: 04537715).

On first assessment of response, performed in November 18, 2021; the CT was consistent with a partial response (PR) as per RECIST 1.1 criteria with a 33% reduction over baseline (Fig. 1.). This PR was maintained throughout the subsequent evaluations, presenting in the last CT assessment in August 2023 a reduction of 43% over baseline size. During treatment, the patient has presented good tolerance with grade 2 side effects limited to hyporexia related to the study drug and chronic venous insufficiency not related. He also experienced a grade 3 event during cycle eight, when he was hospitalized due to a lung infection caused by Influenza A virus detected by bronchoscopy. The patient recovered quickly after initiation of bronchodilator treatment, with early hospital discharge. At the time of this report, the patient had been on Tazemetostat for over 30 months with sustained partial response and with no dose reduction or treatment interruptions required.

Fig. 1: Course of disease.
Fig. 1: Course of disease.
Full size image

Inguinal lesion (A) 15.21 mm and pelvic lesion (B) 17.04 mm before starting treatment with Tazemetostat. Inguinal lesion (C) 7.83 mm (49% reduction) and pelvic lesion (D) 14.09 mm (18% reduction) in the first reassessment after 3 months of treatment. Inguinal lesion (E) 6.77 mm (55% reduction from baseline) and pelvic lesion (F) 11.55 mm (32% reduction from baseline) after 24 months of treatment.

PBRM1 mutations

PBRM1 mutations are observed across multiple cancer types, with a reported frequency of 3.9% in somatic mutations (Fig. 2B). However, this is descriptive data and does not suggest predictive or diagnostic implications without further validation. The occurrence is almost one mutation per five cholangiocarcinoma (CHOL) tumors (19.44%). However, the highest mutation frequency for PBRM1 is notably observed in clear cell renal cell carcinoma (ccRCC (27.98%). We decided to investigate the distribution of mutations in PBRM1 in a combined study with 10967 samples from different cancer types. As can be seen in Fig. 2B, the widespread mutations along the body entire length of the gene without specific domain enrichment. In the current study, the patient under examination harbors the splicing mutation 3617-1 G > T. It suggests a change that impact the splicing of the mRNA leading to an aberrant mRNA and, consequently, a truncated protein which would disrupt the activity of the SWI/SNF histone remodeling complex.

Fig. 2: Analysis of PBRM1 and EZH2 interactions in Silico.
Fig. 2: Analysis of PBRM1 and EZH2 interactions in Silico.
Full size image

Bar graph showing the mutation frequency of PBRM1 in human pan-cancer (A). Scheme of PBRM1 where all mutations (green for missense, black for truncating, brown for inframe, orange for splice, and purple for fusion) are displayed. The height represents the number of cases (B). Bar graph illustrating the percentage of mutations in PBRM1 alone or combination with EZH1 (i) or EZH2 (ii) in cholangiocarcinoma tumors (C). The dot plot shows the gene expression correlation between PBRM1 and EZH1 in cholangiocarcinoma tumors (D). Box and whiskers graph indicating the sensitivity of ccRCC cell lines to GSK343 (EZH2 inhibitor) depending on PBRM1 WT (grey dots) or mutated (red dots). (E).

PBRM1 and EZH1/2

By examining the relationship between PBRM1 and EZH1/EZH2, particularly CHOL and ccRCC tumors do not show simultaneous mutation in both PBRM1 and EZH1/2 suggesting a potential mutual exclusivity, however it is difficult to extract meaningful conclusions since the number of cell line samples are limited, hence the lack of significance, and the metastatic disease is poorly represented. Performing a broader analysis across all cancer types revealed an important percentage of tumors presenting mutations in both PBRM1 and EZH1 (Supplementary Fig. 1B) or EZH2 (Supplementary Fig. 1C). This significant number of samples with both mutations show that they follow a co-occurrence tendency in most tumors (Supplementary Fig. 1D), with exceptions noted in specific types like CHOL or ccRCC, maybe implying a tissue-specific relationship between PBRM1 and EZH1/2.

To explore potential gene expression correlations, we examined CHOL tumors and found that higher expression levels of PBRM1 correlated with increased transcripts of EZH1, with a Pearson Correlation coefficient of 0.62 (Fig. 2D). Similar results, albeit with a slightly lower correlation (0.5), were observed in ccRCC samples (Supplementary Fig. 2A) and EZH1. This correlation trend extended to EZH2 and PBRM1 in ccRCC cell lines (0.69) (Supplementary Fig. 2C) Unfortunately, there was an insufficient number of CHOL cell lines to confirm PBRM1 and EZH2 association (Supplementary Fig. 2C).

EZH2 inhibitors

Given the positive response of the patient to Tazemetostat, we sought to determine if this response was a general trend in this tumor type. To explore this, we conducted an analysis using ANOVA to examine the association between a genomic marker and drug sensitivity in KIRC cell lines (Supplementary Fig. 3A). Interestingly, we observed heightened sensitivity to another EZH2 inhibitor, GSK343. Anticipating that the responsive cell lines might possess mutations in PBRM1, we segregated the cell lines based on this factor. Our analysis showed a trend towards increased sensitivity to GSK343 among cell lines with PBRM1 mutations (Fig. 2E). However, this finding did not reach statistical significance, and further studies are required to confirm the association.

To understand the molecular mechanisms underlying the sensitivity to EZH2 inhibitors, we analyzed the transcriptomic signature associated with PBRM1 mutations in ccRCC tumors (Supplementary Fig. 3B). Notably, we observed that most dysregulated transcripts were downregulated. Yet, our focus turned to the significantly upregulated genes (149 in total) and the biological processes in which they are enriched (Supplementary Fig. 3C). The most significant biological process identified was regulation of transcription (GO:1903506) and chromatin organization (GO:0006325). This supports our search since they are functions where PBRM1 has a role and is expected to appear. Additionally, functions linked to the regulation of protein levels, such as protein modification processes (GO:0036211) and protein degradation (Proteasome-mediated ubiquitin-dependent protein catabolic process, GO:0043161), emerged from our analysis. Finally, two different biological processes, transmembrane receptor protein tyrosine kinase signaling pathway (GO:0007169) and regulation of apoptotic process (GO:0042981) were significantly enriched. These processes may be involved in the increased sensitivity of PBRM1-mutated tumors to treatment with EZH2 inhibitors.

Genetic interactions and coessentiality

Upon comparing the pBAF subunit and PRC2 complex across several networks (Table 1) (Fig. 3A), we found interesting findings. PBRM1 and EZH2 showed negative interactions (pi score −0.24) in one of the CRISPR screen networks (Fig. 3B). Bromodomain 7 (BRD7), another pBAF subunit and relevant partner to PBRM1, also exhibited negative interactions with EZH2 (pi score −0.48). Unfortunately, both interactions did not reach the statistical significance (p value, 0.264 and 0.093 respectively) BRD7 also demonstrated interactions with two different members of the PRC2 complex, JARID2 and SUZ12. Finally, exploring related connections with the BAF subunit, a different subunit in the SWI/SNF family, significant findings were found concerning ARID1A and ARID1B, interconnected respectively with EED and SUZ12.

Fig. 3: Protein-protein genetic interactions network.
Fig. 3: Protein-protein genetic interactions network.
Full size image

Proteins within each complex, SWI/SNF (blue) and PRC2 (orange) are connected with dashed edges. Reported genetic interactions between proteins from the two complexes are drawn with continuous edges. Edge colors represent the source from the reported genetic interaction as i) extracted using RNAi experiments from Pan et al. 18 (green); ii) extracted using CRISPR experiments from Pan et al. 18 (yellow); iii) extracted using CRISPR experiments from Rauscher et al. 19 (red). A Edge width represents the score of the interaction provided by the source and node size corresponds with the number of interactions of the protein, bigger nodes are proteins with more interactions. B Set of genetic interactions extracted from Rauscher et al. 19 for EZH2. P-score and log10 (pvalue) are represented as provided in the original dataset. P-score > 0.2 represents positive genetic interactions and < −0.2 represent negative genetic interactions. PBRM1 and BRD7 are highlighted in red for visualization purposes. Each dot corresponds to one gene.

Table 1 Genetic interactions BAF/pBAF subunit and PRC2 complex
Full size table

Discussion

Following first-line chemotherapy, just 15–25% of metastatic cholangiocarcinoma patients receive second-line therapy due to a rapid deterioration in functional status. The reported median overall survival and progression-free survival for second-line therapy are 7.2 and 3.2 months, respectively8. The case report highlights a sustained partial response to Tazemetostat ( > 30 months) in a single patient with metastatic cholangiocarcinoma harboring a PBRM1 splice site mutation. While promising, these findings are preliminary and should be interpreted with caution until validated in larger cohorts.

Consistently with previous literature, our analysis confirmed the high mutation burden of PBRM1 across various cancer types, notably in ccRCC and cholangiocarcinoma, highlighting the importance of understanding the functional consequences of PBRM1 alterations in cancer development and progression. The widespread distribution of mutations along the PBRM1 gene suggests a lack of domain specificity, possibly reflecting its broad impact on chromatin remodeling.

According to prior research, it is well known that the mutual exclusivity of oncogenic mutations may reveal unexpected vulnerabilities and therapeutic possibilities9. In line with this, our analysis suggested a potential trend of mutual exclusivity between mutations in PBRM1 and EZH1/2 in cholangiocarcinoma and ccRCC. However, given the limited sample size and lack of statistical significance, these observations require further investigation to confirm their relevance. While this exclusivity was not observed across all cancer types, it may highlight tissue-specific relationships that merit further investigation. We also found a positive correlation between PBRM1 expression and EZH1/2 transcripts in both cholangiocarcinoma and ccRCC tumor samples, proposing potential regulatory crosstalk between these entities. However, the absence of sufficient cholangiocarcinoma cell line data would require additional studies to validate this correlation comprehensively.

In terms of sensitivity to EZH2 inhibitors, the clinical response of our patient with PBRM1-mutated cholangiocarcinoma to Tazemetostat alongside with the previous positive results reported in-vivo and in-vitro, prompted an exploration into the broader sensitivity of PBRM1-mutated tumors to EZH2 inhibitors7,10. Our analysis revealed sensitivity to GSK343 among cell lines with PBRM1 mutations and transcriptomic analysis unveiled dysregulated genes primarily involved in transcriptional regulation and chromatin organization, corroborating the functional relevance of PBRM1 mutations. Notably, upregulated genes implicated in protein modification and apoptotic processes offer insights into the molecular mechanisms driving the enhanced sensitivity of PBRM1-mutated tumors to EZH2 inhibition.

Finally, despite the caveat’s coessentiality works may represent in terms of ubiquitous false positives, we show a previously unreported genetic interactions between PBRM1, EZH2, and other complex subunits, particularly BRD7, a pBAF subunit. These interactions, combined with the emerging evidence of PBRM1 mutations co-occurrence with additional loss of function alterations in chromatin remodeling genes11 and the oncogenic dependency on EZH2 for those tumors with mutations in SWI/SNF genes12, underscore the intricate network of regulatory proteins orchestrating chromatin dynamics and gene expression.

In conclusion, our study provides valuable insights into the challenging interplay between SWI/SNF and PRC2 chromatin remodeling protein complexes and their role in cancer biology and treatment response. These findings suggest a possible therapeutic potential of targeting EZH2 in PBRM1-mutated cancers. However, additional preclinical and clinical studies are needed to establish robust evidence for this approach and its applicability across cancer types. Further research is warranted to elucidate the underlying mechanisms and optimize therapeutic strategies for patients with PBRM1-altered tumors.

Methods

Identification of PBRM1 mutations in cancer patients, data collection, and processing

We used data contained at cBioportal (www.cbioportal.org) (accessed in October 2023)5,13 to explore the mutations of the PBRM1 gene in patients with any type of cancer. We used a combination of 32 studies that contained 10,967 samples from 10,953 patients (TCGA data). This web resource also provides mutated variants mapped to genomic domains.

Correlation between gene mutation and drug sensitivity

To explore the associations between gene mutation and drug sensitivity we used Genomics of Drug Sensitivity in Cancer (https://www.cancerrxgene.org) (accessed in October 2023)14. This database provides drug response data and genomic markers of sensitivity.

Gene Expression correlation

We explored the University of Alabama at Birmingham Cancer data Analysis Portal (https://ualcan.path.uab.edu/index.html) (accessed in October 2023)15,16 to extract cell line data from TCGA about gene expression correlation.

Functional annotation of de-regulated genes

The EnrichR online platform (https://maayanlab.cloud/Enrichr/) (accessed on December 2023)17 was used to address the Gene Ontology Biological process related to each gene set. We represented the most relevant pathways according to their ranking.

Graphical design

Bars, heatmaps, dot plots, and volcano plots were represented using GraphPad Prism software (GraphPad Software, San Diego, CA, USA) in terms of absolute counts, relative frequencies, and hazard ratios.

Co-essentiality and genetic interactions

As a complementary source of phenotypic and molecular information, we extracted genetic interactions from public resources and published articles18,19. We searched all possible combinations between proteins from complex pBAF and proteins from complex PRC2 in 5 coessentiality networks20. We used Cytoscape v10 for network visualization.

Ethics statement

The patient provided written informed consent for participation in the clinical trial (NCT: 04537715), for tumor biopsy and sequencing procedures, and for publication of clinical data and anonymized images. The protocol was reviewed and approved by the Institutional Review Board (IRB) of Fundación Jiménez Díaz Hospital in accordance with the Declaration of Helsinki.