Approximately 70% of breast cancers express estrogen receptor α (ERα) and depend on this key transcriptional regulator for proliferation and differentiation. While patients with this disease can be treated with targeted antiendocrine agents, drug resistance remains a significant issue, with almost half of patients ultimately relapsing. Elucidating the mechanisms that control ERα function may further our understanding of breast carcinogenesis and reveal new therapeutic opportunities. Here, we investigated the role of deubiquitinases (DUB) in regulating ERα in breast cancer. An RNAi loss-of-function screen in breast cancer cells targeting all DUBs identified USP11 as a regulator of ERα transcriptional activity, which was further validated by assessment of direct transcriptional targets of ERα. USP11 expression was induced by estradiol, an effect that was blocked by tamoxifen and not observed in ERα-negative cells. Mass spectrometry revealed a significant change to the proteome and ubiquitinome in USP11-knockdown (KD) cells in the presence of estradiol. RNA sequencing in LCC1 USP11-KD cells revealed significant suppression of cell-cycle–associated and ERα target genes, phenotypes that were not observed in LCC9 USP11-KD, antiendocrine-resistant cells. In a breast cancer patient cohort coupled with in silico analysis of publicly available cohorts, high expression of USP11 was significantly associated with poor survival in ERα-positive (ERα+) patients. Overall, this study highlights a novel role for USP11 in the regulation of ERα activity, where USP11 may represent a prognostic marker in ERα+ breast cancer.
A newly identified role for USP11 in ERα transcriptional activity represents a novel mechanism of ERα regulation and a pathway to be exploited for the management of ER-positive breast cancer.
Approximately 70% of breast cancers express estrogen receptor α (ERα) and depend on this key transcriptional regulator for growth and differentiation. Patients with ERα-positive (ERα+) breast cancer are treated with targeted, antiendocrine therapies. Despite advancements in targeted treatment options, drug resistance remains a clinically significant problem, with almost half of patients eventually relapsing on endocrine intervention therapies (1). As such, the discovery of novel mechanisms controlling ERα function in breast cancer represents major advances in our understanding of breast cancer progression and may offer new therapeutic opportunities. In this study, we investigated the role of deubiquitinases (DUB), a class of enzymes that remove ubiquitin moieties from target proteins, in ERα function in breast cancer.
Ubiquitination, a posttranslational modification (PTM) involving the addition of a ubiquitin moiety to a target protein, is the primary mechanism of protein turnover in the cell. This is a complex, multistep biochemical pathway that ultimately leads to a mono- or polyubiquitinated target protein. Polyubiquitin chains are generated using any of the seven lysine residues on ubiquitin itself, with different chain topologies resulting in different functional consequences. Polyubiquitin chains linked at lysine (K) residue 48 (K48) shuttle the tagged protein to the 26S proteasome for degradation (2). On the other hand, monoubiquitination and ubiquitin chains of different topologies result in altered target protein functionality and drive pathways such as endocytosis (3) and DNA damage repair (4).
Like most other PTMs, ubiquitination is a reversible process. By hydrolyzing the isopeptide bond between ubiquitin and the target protein, DUBs remove ubiquitin molecules from ubiquitinated proteins, stabilizing the target and preventing proteasomal degradation. Furthermore, DUBs also play a vital role in ubiquitin recycling, generation of ubiquitin precursors, and protecting the proteasome from free ubiquitin chains (5). The mammalian genome encodes more than 100 DUBs, which are classified into six different groups on the basis of their sequence and structural similarity. Besides the Jabb1/MPN domain-associated metalloproteases (JAMM) DUB family, all DUBs are cysteine proteases, and utilize a catalytic dyad or triad of amino acids to hydrolyze the isopeptide bond between ubiquitin and the target protein. The largest and most diverse DUB family are the ubiquitin-specific proteases (USP); it is predicted that the human genome encodes more than 50 enzymes of this class (6). The USP DUBs encode highly conserved Cys and His box motifs, which contain all catalytic triad residues.
DUBs are often differentially expressed or activated in tumors, and targeting them in the clinic has now become an area of therapeutic interest. Recently, the role of DUBs in nuclear receptor signaling in various cancers has been highlighted in the literature. For example, USP7 regulates the androgen receptor (AR) in prostate cancer by deubiquitinating AR in an androgen-dependent manner, associating with AR at androgen-responsive elements and facilitating chromatin binding (7). Moreover, AR is also regulated by USP26 in a similar manner (8), highlighting an integral role for DUBs in AR signaling. ERα, on the other hand, is deubiquitinated by otubain-1 (OTUB1) in vitro. OTUB1 negatively regulates transcription of the ERα gene itself and can stabilize the receptor in endometrial cancer cells (9). USP9x is also a key ERα interactor. USP9x attenuation was found to render breast cancer cells resistant to tamoxifen, leading to the generation of a gene signature used to define patient outcome following adjuvant tamoxifen treatment (10).
Given the accumulating evidence for nuclear receptor regulation by DUBs, we sought to identify the role of DUBs in ERα function by taking an unbiased, functional genomics approach. An RNAi loss-of-function screen targeting all known or putative DUBs in the human genome was performed. Interestingly, silencing of the BRCA2-associated DUB, USP11 (11), was found to suppress ERα transcriptional activity. USP11 is a protease with multiple cellular functions, including the positive regulation of TGFβ signaling and stabilization of inhibitor of apoptosis proteins (12, 13). Perhaps, the most widely studied function of USP11, however, is its role in homologous recombination (14, 15). Until now, the role of USP11 in ERα function has remained unknown. This study provides strong evidence for the role of USP11 in ERα transcriptional function and identifies USP11 as a marker of poor prognosis in ERα+ breast cancer. We believe that USP11 may be a viable therapeutic target in ERα+ breast cancer, offering treatment options for patients who do not respond to or relapse on currently available therapies.
Materials and Methods
Cell lines and culture
ZR-75-1, T47D, SUM44, MDA-MD-134VI, and CAMA-1 cells were purchased from the ATCC (www.atcc.org). MCF7, LCC1, and LCC9 cells were a kind gift from Prof. Robert Clarke (Georgetown University Medical School, Washington D.C.). HEK293T wild-type and USP11-knockout cell lines were a kind gift from Dr. Daniel Durocher (The Lunenfeld-Tanenbaum Research Institute, Mount Sinai Hospital, Toronto, Ontario, Canada). ZR-75-1, MCF7, and HEK293T cells were cultured in DMEM (Sigma-Aldrich) supplemented with 10% FBS [volume for volume (v/v), Gibco, Invitrogen], 1% penicillin/streptomycin (v/v, Gibco, Invitrogen), and 1% l-Glutamine (Gibco, Invitrogen). ZR-75-1 cell culture media were also supplemented with 1 nmol/L estradiol (Sigma-Aldrich). T47D, SUM44, MDA-MB-134VI, and CAMA-1 were cultured in RPMI1640 (Sigma-Aldrich) supplemented with 10% FBS (v/v, Gibco, Invitrogen), 1% penicillin/streptomycin (v/v, Gibco, Invitrogen), and 1% l-glutamine (Gibco, Invitrogen).
LCC1 and LCC9 cells were cultured in phenol red-free DMEM (Sigma-Aldrich) supplemented with 5% charcoal/dextran-treated FBS (v/v, Gibco, Invitrogen), 1% Penicillin/Streptomycin (v/v, Sigma-Aldrich), and 1% l-Glutamine (Gibco, Invitrogen). Experiments that required hormone depletion were also cultured in phenol red-free media.
Subculturing of all cells took place in a Class II laminar flow hood under sterile conditions. ZR-75-1 stable USP11-knockdown (KD) cell lines were generated following transduction of two independent lentiviral short hairpin RNAs (shRNA) and a nontargeting control (NTC) shRNA. To achieve transient USP11 KD, two independent ON-TARGETplus siRNAs (Dharmacon) targeted to USP11 (30 nmol/L) were transfected into cells using Lipofectamine 2000 (Invitrogen, all sequences are provided in Supplementary Table S1). An ON-TARGETplus SMARTpool Control (30 nmol/L; Invitrogen) was used a negative control. KD was assessed 72–120 hours posttransfection, depending on the assay performed. To examine ERα function in HEK293T cells, cells were transfected with HA-ERα (Addgene, #49498) using Lipofectamine 3000 (Invitrogen).
ZR-75-1 cells were seeded in 24-well plates in antibiotic-free media, 24 hours prior to transfection. The shRNA DUB library used, generated at the Netherlands Cancer Institute (NKI, Amsterdam, the Netherlands) and described previously in the literature by Brummelkamp and colleagues (16), contained four nonoverlapping shRNAs targeted to each DUB (108 DUBs and 432 shRNAs in total). shRNAs were transfected with estrogen response element luciferase (ERE-luc) and cytomegalovirus (CMV) Renilla reporters as described previously (10), the latter representing an internal control. Cells were transferred to hormone-depleted media for 48 hours before stimulation with 1 nmol/L estradiol for 24 hours. Cells were harvested and luciferase activity was measured using the Dual Luciferase Reporter Assay System (Promega) as per the manufacturer's instructions. Hairpins demonstrating no net effect served as internal controls. Results were confirmed in biological triplicate in the presence and absence of estradiol.
Protein samples were diluted in an appropriate volume of 4× NuPAGE Lithium Dodecyl Sulfate Buffer (Invitrogen), supplemented with 2.5% β-Mercaptoethanol (Sigma-Aldrich). Samples were boiled at 100°C for 5 minutes to allow for protein denaturation before loading onto the gel. SDS-PAGE was performed in 1× Tris Glycine running buffer (25 mmol/L Tris, 250 mmol/L glycine, and 0.1% SDS) using a Bio-Rad Mini Protean III Gel System (Bio-Rad Laboratories). Gels were run for approximately 20 minutes at 90 V through the stacking gel, followed by approximately 60 minutes at 120 V through the resolving gel.
Resolved proteins were transferred to a nitrocellulose membrane in 1× transfer buffer (25 mmol/L Tris, 190 mmol/L glycine, and 20% methanol) using a Bio-Rad Mini-Protean III Electrophoretic Transfer Cell (Bio-Rad Laboratories) at 300 mA for 90 minutes at 4°C. Following transfer, membranes were blocked with either 5% (w/v) nonfat dried milk (Sigma-Aldrich) or 5% (w/v) BSA (Sigma-Aldrich), both prepared in 100 mL of 1× TBS buffer containing 0.1% (v/v) Tween 20 (Sigma-Aldrich), and incubated with gentle agitation for 1 hour at room temperature. Membranes were then incubated in blocking buffer containing the primary antibody of choice overnight at 4°C.
Following primary antibody incubation, membranes were washed with TBS-T and subsequently incubated for 1 hour at room temperature with horseradish peroxidase (HRP)-conjugated secondary antibody (anti-mouse/anti-rabbit, Dako) contained in blocking buffer. Membranes were again washed in TBS-T and detection of HRP complexes was achieved by exposing the membranes to Enhanced Chemiluminescence Substrate (Pierce). Membranes were imaged using the Amersham Imager 600 (GE Healthcare Life Sciences).
ZR-75-1 cells were seeded and transferred to phenol red-free DMEM containing 5% charcoal/dextran-treated FBS. After 48 hours, cells were stimulated with either vehicle (EtOH) or 1 nmol/L estradiol. Cellular fractions were generated using Qiagen's Cell Compartment Kit, as per the manufacturer's instructions. Extracts were analyzed using Western blotting, with anti-GAPDH, anti-trimethyl histone H3, and anti-cytochrome c antibodies used as cytosolic, nuclear, and membrane markers, respectively.
ZR-75-1 cells were seeded in to 8-well chamber slides in hormone-depleted media at a density of 100,000 cells per well. After 48 hours, cells were stimulated with 1 nmol/L estradiol for either 4 or 24 hours, or left in hormone-depleted conditions.
Media were aspirated, cells were washed in PBS, and subsequently fixed in 100% methanol for 20 minutes. The cells were washed 2 × 5 minutes in PBS before blocking in 10% goat serum in 5% w/v BSA/PBS for 60 minutes. The cells were washed in PBS for 5 minutes before a 90-minute incubation in the primary antibody (USP11, 1:250) diluted in 10% human serum in 5% w/v BSA/PBS. The cells were washed 2 × 5 minutes in PBS to remove excess primary antibody. The cells were then incubated in secondary antibody (Alex Fluor 594, 1:200) diluted in 10% human serum in 5% w/v BSA/PBS, for 60 minutes, and the slide was covered in tin foil to protect from light. The cells were washed 2 × 5 minutes in PBS before mounting the cover slip on to the chamber slide. A small volume of VECTASHIELD mounting medium containing DAPI (Vector Laboratories) was placed on the slide and the coverslip was gently lowered on to the slide. The coverslip was sealed using clear nail polish.
Widefield fluorescence microscopy was carried out using a Nikon Eclipse 90i equipped with a DS-Ri1 camera and Plan Fluor 20× (N.A 0.5) objective paired with DAPI and TRITC filtersets. NIS-Elements BR 3.10 was used to capture images with fixed acquisition settings.
A Carl Zeiss LSM 710 equipped with a W Plan-Apochromat 20× objective (N.A 1.0) was used to capture confocal images with fixed acquisition settings. Samples were simultaneously excited with 405 and 594 nm lasers and the resulting emissions were captured using spectral detectors over the range of 409–495 nm and 598–726 nm, respectively. 4× averaging and a spacing of 0.806 μm was used when capturing z stacks in the Zen 2008 software.
FIJI (17) was used for the preparation of both widefield and confocal images. All z stacks are presented as maximum image projections.
RNA was extracted in Tri-Reagent (Sigma-Aldrich) and quantified using a NanoDrop UV-Vis Spectrophotometer (Thermo Fisher Scientific) according to the manufacturer's instructions. RNA samples were diluted in nuclease-free water to yield the same concentration per sample and were treated with DNase I (Invitrogen) to remove genomic DNA from the samples. cDNA was synthesized using the high capacity cDNA Reverse Transcription Kit (Invitrogen) as per the manufacturer's instructions. Samples were incubated in the Bio-Rad T100 Thermal Cycler (Bio-Rad Laboratories) at 25°C for 10 minutes, followed by 37°C for 60 minutes, and 85°C for 5 minutes.
cDNA was diluted in 2× SYBR Green (Promega), forward and reverse primers (Eurofins MWG), and dH2O in a 96-well plate. qRT-PCR was performed using the Applied Biosystems 7500 Real-Time PCR System set to the following temperature cycles: 2 minutes at 50°C, 10 minutes at 95°C, followed by 40 cycles oscillating between 15 seconds at 95°C and 1 minute at 60°C. A melting curve was generated after each run to ensure primer specificity: 15 seconds at 95°C, 15 seconds at 60°C, and 15 seconds at 95°C. Primer sequences can be found in Supplementary Table S1.
Both LCC1 and LCC9 cell lines were transfected with two independent siRNAs targeted to USP11 and an siRNA NTC using lipofectamine 2000, as described above. Cells were incubated for 72 hours before extraction of RNA. Samples were prepared in biological triplicate.
cDNA libraries were prepared from RNA samples using Illumina's NeoPrep Library Card, as per the manufacturer's instructions. Samples were prepared for sequencing on the NextSeq 500 (Illumina) as per the manufacturer's instructions. Two × 75-bp paired-end reading was performed.
For bioinformatic analysis, Fastq files were downloaded from Illumina BaseSpace using the BaseSpace download tool (https://github.com/ReddyLab/BaseSpaceFastqDownload).
The quality of Fastq files was determined using FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/).
Reads were trimmed to remove poor quality base calls (Phred score < 20) and sequencing adaptors using BBDuk tool in the BBMap package (https://jgi.doe.gov/data-and-tools/bbtools/bb-tools-user-guide/bbmap-guide/).
Sequencing reads were aligned to the human hg19/GRCh37 genome reference using the STAR alignment algorithm, version 2.5.2a (18). This produced a BAM file that was sorted by coordinate. Duplicate reads were marked in the BAM using Picard-Tools “MarkDuplicates” call (https://broadinstitute.github.io/picard/). Read counts were produced by the featureCounts tool from the SubRead package (19). These counts were combined for all samples and used as input for differential gene expression analysis.
Differential expression analysis of genes was carried out using the DESeq2 package (20) in the R statistical environment (R Development Core Team, 2012). The data.frame of counts had all genes with a sum of zero across all samples removed. A “conditions” data.frame was created on the basis of the sample names, their group (i.e., LCC1_control, LCC1_si7, LCC1_si8, LCC9_control, LCC9_si7, or LCC9_si8) and their biological replicate number. The counts and conditions data.frames were loaded into a DESeq2DataSet class object using the DESeqDataSetFromMatrix() call, with the design variable set as “∼ group.” The DESeq() call produced two sets of results, based on LCC1 or LCC9 cells, comparing the two individual siRNA KDs with the control for each cell line. Four text files resulted, containing each gene expressed, in the log2 fold change value and the FDR adjusted P value. Principal component analysis plots were produced for the full dataset to determine the quality of the count data and similarity of samples from the six conditions. These followed the standard protocol from the DESeq2 guide (https://bioconductor.org/packages/release/bioc/vignettes/DESeq2/inst/doc/DESeq2.html). Fragments-per-kilobase per million reads were produced using the edgeR package (21), rpkm() call.
RNA sequencing (RNA-seq) was validated using qRT-PCR as described above, using the same sequenced RNA samples.
Differentially expressed (DE) genes were subject to gene ontology (GO) enrichment analysis to detect altered cellular pathways following USP11 KD. The Database for Annotation, Visualization, and Integrated Discovery (DAVID; http://david.abcc.ncifcrf.gov/) was used for GO and Kyoto Encyclopedia of Genes and Genomes pathway enrichment (22). Significantly DE genes, grouped according to whether they were up- or downregulated with USP11 KD, were input into DAVID for enrichment analysis. An FDR < 5% was considered statistically significant.
Gene set enrichment analysis (GSEA) is an improved method of analysis to define the biological functions of gene sets (23). The method uses weighted genes according to their correlation with phenotype rather than equal weights for each gene. GSEA 3.0 and gene sets were downloaded from the molecular signatures database (MsigDBv3.1). Phenotypes were assigned to both KD and control samples and enrichment analysis was carried out against a defined gene set. The GSEA procedure was performed as follows: an enrichment score (ES) was calculated ranking genes according to their differential expression and a nominal P value was obtained as an estimate of the significance level of the ES. Subsequently, the ES was normalized (NES) and FDR was calculated to adjust for multiple hypothesis testing.
Immunoaffinity purification and mass spectrometry
ZR-75-1 USP11-KD and NTC cells were processed as described previously (24). In brief, cells were lysed in urea lysis buffer, and proteins were digested by sonication. DTT and iodoacetamide were used to reduce and alkylate the samples, respectively. Samples were further digested overnight at room temperature using trypsin, and protein digestion was confirmed using SDS-PAGE. To remove fatty acids, acidification of samples was performed using trifluoroacetic acid (TFA). Samples were treated with a final concentration of 1% TFA, and centrifuged at 1,780 × g for 15 minutes at room temperature to isolate fatty acid precipitates. Proteins were purified using the Sep-Pak C18 system, as per the manufacturer's instructions.
The lyophilized peptide was centrifuged at 2,000 × g for 5 minutes at room temperature. All samples were resuspended in 1.4 mL immunoaffinity purification (IAP) buffer and transferred to a 1.7 mL reaction tube. A neutral pH was confirmed and samples were centrifuged at 10,000 × g for 15 minutes at 4°C. UbiScan beads were washed four times with PBS, centrifuged at 2,000 × g for 30 seconds between each wash. The beads were resuspended in 40 μL PBS following the final wash. The supernatant from the peptide solution was added to the beads and all samples were incubated at 4°C for 2 hours with constant agitation.
The flow-through was transferred to a new reaction tube and stored. The beads were washed with 1× IAP buffer two times, centrifuged at 2,000 × g for 30 seconds at 4°C between each wash. The beads were then washed with water three times, centrifuged at 2,000 × g for 30 seconds at 4°C between each wash. Fifty-five microliters of 0.15 % TFA were added to each tube of beads and mixed gently. Samples were left to incubate at room temperature for 10 minutes, mixing gently every 2 minutes. The samples were centrifuged at 2,000 × g for 30 seconds and the supernatant was collected in a new tube. This elution step was repeated. Peptides were concentrated and purified using the StageTip protocol, as per the manufacturer's instructions, and prepared for LC/MS.
Mass spectrometry and proteomics data analysis
Samples prepared as described above were analyzed by LC/MS-MS as described previously (24). To convert MS spectra generated from the Q Exactive Hybrid Quadrupole-Orbitrap Mass Spectrometer (Thermo Fisher Scientific) to protein identifications, the MaxQuant (25) computational platform (version 1.5.2) was used. Label-free quantification (LFQ) values were extracted from the protein groups (proteome) and peptides (ubiquitome) previously generated by MaxQuant, and uploaded to the Perseus computational platform (version 18.104.22.168; ref. 26). Statistical analysis of the identified and quantified proteins was performed using the Perseus software package. A histogram showing counts versus LFQ intensity was generated for each sample to highlight the distribution of counts, and the imputation performed previously. Principle component analysis (PCA) was performed on samples to demonstrate how each sample grouped respective to all other samples. To highlight differentially expressed proteins between any two groups, a volcano plot was generated using a two-sided t test (unpaired). The FDR (permutation FDR for multiple testing) for all volcano plots was set to 5% (0.05), and the s0 LFQ difference value was kept at the default 0.1.
IHC was performed using the Ultravision LP Large Volume Detection System HPR Polymer (Labvision) and DAB Plus Substrate System (Labvision). Slides were incubated for 10 minutes in 3% H2O2 to block endogenous peroxidases, and then washed with PBS 0.1% v/v Tween (PBST). Protein block was applied for 5 minutes before incubation with anti-USP11 primary antibody (Bethyl Laboratories) for 60 minutes. Slides were then washed and incubated with secondary antibody solution (primary antibody enhancer) for 10 minutes and washed again with PBST. Finally, slides were incubated in HRP polymer for 15 minutes, washed with PBST, and incubated in 3, 3′-diaminobenzidinetrahydrochloride (DAB) for 10 minutes. Slides were then counterstained by a 3-minute exposure to hematoxylin and subsequent agitation in tepid dH2O. Slides were dehydrated in ascending concentrations of ethanol in Coplin jars as follows: 3 minutes in 80% v/v ethanol, 3 minutes in 95% v/v ethanol, 2 × 3 minutes in 100% v/v ethanol, and 2 × 3 minutes in xylene. Slides were then mounted in Permount Mounting Medium (Thermo Fisher Scientific) using an automated cover-slipper (Leica Microsystems). Slides were scanned using a ScanScope XT Digital Slide Scanner (Aperio Technologies) and analyzed using ImageScope Software (Aperio Technologies).
The tissue microarray (TMA) cohort used in this study, described previously in the literature by DeNardo and colleagues (27), was generated from a consecutive cohort of 144 patients diagnosed with invasive breast cancer at Malmö University Hospital (Malmö, Sweden), in 2001 and 2002. These patients did not receive any neoadjuvant treatment prior to surgery. A total of 103 (72%) patients had ER-positive tumors, and complete data on antiendocrine therapy were available for 95 patients; 77 of whom received tamoxifen, three received an aromatase inhibitor, and 25 received a combination of both. Thirty patients received adjuvant chemotherapy and 83 received radiotherapy. At the time of the last follow-up, 41 patients had died, 22 of whom as a direct result of breast cancer. USP11 expression and its association with overall, recurrence-free, and breast cancer–specific survival (BCSS) were the primary endpoints of focus.
In silico analysis of publicly available breast cancer datasets was carried out using Gene expression-based Outcome (GOBO; ref. 28), where differences were outlined in overall survival (OS), distant metastasis-free survival, and recurrence-free survival (RFS) according to USP11 expression. A P < 0.05 was considered statistically significant.
For TMA analysis, biopsy cores were manually scored (0–3) on the basis of the degree of DAB-positive staining. Kaplan–Meier and Cox regression analysis were then used to determine differences in OS, RFS, and BCSS according to USP11 expression. Pearson χ2 test was used to evaluate associations between USP11 expression and clinicopathologic characteristics. SPSS version 20.0 (SPSS Inc) was used to carry out statistical analysis of TMA-derived data, and a P < 0.05 was considered statistically significant.
GraphPad Prism 5 was used to carry out statistical analysis on all in vitro work. Student t test/one-way ANOVA statistical tests were used, with a P < 0.05 considered statistically significant.
Loss of USP11 negatively regulates ERα transcriptional activity
To determine the role of DUBs in ERα transcriptional function in breast cancer cells, an unbiased RNAi loss-of-function screen was performed, targeting all 108 known or putative DUBs in the human genome. The library contained four nonoverlapping shRNAs targeted to each DUB, 432 in total. Each pool was cotransfected into ZR-75-1 ERα+ breast cancer cells along with ERE-luc and CMV Renilla reporters, to detect transcriptional activity of ERα following estradiol stimulation with suppression of each DUB (Fig. 1A). DUB KD led to repressed, enhanced, or unchanged activity at the ERE-luc reporter (Fig. 1B). A selection of DUBs was chosen for triplicate validation of the RNAi screen in both the presence and absence of estradiol. The DUBs selected included five that repressed activity, five that enhanced activity, and five that unchanged activity (serving as baseline controls) at the ERE-luc reporter from the RNAi screen. This validation experiment supported the results obtained from the RNAi screen (Fig. 1C) with all selected DUBs displaying the same phenotype as observed previously. Interestingly, the two DUBs whose KD suppressed ERα transcriptional activity to the greatest extent were OTUB-1 and USP11. As mentioned earlier, a previous study has demonstrated that ERα is deubiquitinated by OTUB1, which regulates its transcriptional activity in endometrial cancer cells (9). USP11 has been previously associated with the BRCA proteins (11) and is a key enzyme in the DNA damage response (14, 15); however, its role in ERα function has not been reported previously. We hypothesized from the loss-of-function DUB screen that USP11 is a novel regulator of ERα transcription and may represent a target of interest in ERα+ breast cancer.
USP11 silencing abrogates ERα target gene expression
To validate the results obtained from the RNAi screen, HEK293T USP11-knockout cells were used to examine the impact of USP11 function on ERα activity. Cells were transfected with an ERα expression vector (HA-ERα; Fig. 1D) and ERE-luc and CMV Renilla reporters. ERα activity was significantly suppressed in USP11-knockout cells following stimulation with estradiol (Fig. 1E), further supporting a role for USP11 in regulating ERα transcriptional activity, as first demonstrated in the RNAi screen. The induction of two ERα target genes, progesterone receptor (PgR) and tre-foil factor 1 (TFF1), following estradiol stimulation was also suppressed in USP11-knockout cells (Fig. 1F and G). To validate this in a breast cancer model, USP11 was knocked down in ZR-75-1 breast cancer cells using two individual siRNAs targeted to USP11. KD was confirmed 72 hours posttransfection using qRT-PCR (Fig. 1H). The mRNA expression of ERα target genes was subsequently examined and the expression of PgR and TFF1 was notably suppressed following USP11 silencing (Fig. 1I and J).
Estradiol increases the expression of USP11
Next, the expression of USP11 was examined in a panel of ERα+ breast cancer cell lines (ZR-75-1, T47D, MCF7, LCC1, LCC9, CAMA-1, MDA-MB-134VI, and SUM44). Basal expression of USP11 varied from low to high across all cell lines, however, expression levels correlated with ERα expression (Fig. 2A). This was confirmed by densitometric and statistical analysis (Pearson r, 0.76; P = 0.027; Fig. 2B). Subsequently, we investigated whether USP11 expression was regulated by estradiol and if any changes were affected by the presence of 4-hydoxytamoxifen (4-HT). ZR-75-1 and MCF7 cells were stimulated with either 1 nmol/L estradiol alone or 1 nmol/L estradiol combined with 1 μmol/L 4-HT over a time course (4, 24, and 48 hours). USP11 protein expression was increased following estradiol exposure in a time-dependent manner in both cell lines, as determined by Western blotting. Remarkably, this change in expression was blocked in the presence of 4-HT (Fig. 2C). As expected, ERα protein expression was initially repressed following estradiol exposure (4 hours) and reexpressed later (24 and 48 hours); an effect that was antagonized by 4-HT. ERα protein expression, therefore, serves as a valuable control to estradiol/4-HT treatment. Furthermore, MDA-MB-231, ERα-negative breast cancer cells, were also examined for change in USP11 expression following estradiol/4-HT exposure. Neither compounds had an effect on USP11 expression (Fig. 2C).
To assess localization of USP11 following estradiol treatment, protein subcellular fractions were generated, allowing for analysis of the nuclear, cytoplasmic, and membrane proteins within the cell. USP11 expression was upregulated in the nucleus of ZR-75-1 cells following estradiol stimulation in a time-dependent manner (Fig. 2D). This result was supported using immunocytochemistry and confocal microscopy, where expression of USP11 was found to be upregulated in the nucleus following estradiol treatment (Fig. 2E).
This evidence strongly suggests that USP11 is regulated by estradiol, via ERα, perhaps in a positive feedback loop to enhance ERα transcriptional activity. The exact mechanism, and whether this occurs via a direct or indirect effect of ERα, warrants further investigation.
USP11 suppression significantly alters the proteome and ubiquitinome in breast cancer cells in the presence of estradiol
The data presented above support a key role for USP11 in ERα transcriptional function, and demonstrate that suppression of USP11 can abrogate the action of the receptor. To further explore this role, we sought to examine how USP11 modulation effects the proteome and ubiquitinome of ERα+ cells, and how these modulations are affected in the presence and absence of estradiol. To do so, we used mass spectrometry to analyze the proteome and ubiquitinome of USP11-KD (shUSP11_1 and shUSP11_4) and control (NTC) ZR-75-1 cells, in both the presence and absence of estradiol (Fig. 3A). Hierarchical clustering of proteomic data demonstrated a unique proteomic pattern in the presence and absence of estrogen (Fig. 3B). Next, both KD cell lines were each compared with NTC cells to examine individual protein changes among samples. A number of proteins were significantly upregulated in KD and control samples in the presence of estradiol (Fig. 3C and D). Strikingly, in the absence of estradiol, modulation of USP11 with either hairpin had no significant effect on the proteome of ZR-75-1 cells (Fig. 3E and F). This suggests that the role of USP11 in ZR-75-1 cells is highly dependent on estradiol.
We next examined differentially expressed proteins common to both KD samples. Proteins depleted in USP11-KD cells included macrophage migration inhibitory factor (MIF), a protein involved in the inflammatory pathway and cancer pathogenesis (29), and fatty acid binding protein 5 (FABP5), which has been implicated in several cancer types, including breast cancer (30).
To analyze the USP11 ubiquitinome, we applied UbiScan, a technique that uses a ubiquitin remnant motif (K-ϵ-GG) antibody–bead conjugate to isolate ubiquitinated peptides (24, 31). Digestion of proteins with trypsin during mass spectrometry sample preparation cleaves ubiquitin at the C-terminus, leaving a Gly-Gly residue (K-ϵ-GG) that is still attached to a lysine on the target protein, thus providing evidence of a ubiquitinated protein.
The USP11 ubiquitinome was analyzed using a stable KD cell line (shUSP11_1) and control cells (NTC). Hierarchical clustering highlighted unique changes in ubiquitinated proteins in USP11-KD cells, predominantly in the presence of estradiol (Fig. 3G). This was supported with further analysis of individual ubiquitinated proteins, where volcano plots highlighted a number of significant changes in ubiquitinated proteins in the presence of estradiol following USP11 modulation (Fig. 3H). Again, very few significant changes occurred in the absence of estradiol (Fig. 3I), further supporting the observation that USP11 function in ERα+ cells is highly regulated by estradiol. Interestingly, ubiquitination of γH2AX, a known substrate of USP11 (32), was increased following USP11 suppression, supporting the reliability of these results.
Proteins ubiquitinated following USP11 silencing may represent novel substrates and, in the presence of estradiol, may contribute to the role of USP11 in ERα transcriptional activity (Fig. 3H, red dots). To decipher a link between USP11 and the receptor, the proteins significantly upregulated in USP11-KD cells when compared with control cells (USP11 ubiquitinome) were uploaded to STRING (version 10.5) to detect known and putative protein–protein interactions with ERα. The STRING search revealed several known, putative, direct, and indirect ERα interactors (Supplementary Fig. S1).
USP11 is upregulated in LCC1 cells and regulates ERα function
Our data above demonstrate that USP11 is a key regulator of ERα transcriptional activity and that its effects are highly dependent on estradiol and ERα itself. To further explore this, we chose to examine the effects of USP11 modulation in the LCC isogenic ERα+ cell line series (33, 34). LCC1 cells remained responsive to antiendocrine agents, suggesting that ERα still drives cellular growth, while LCC9 cells were antiendocrine cross-resistant (fulvestrant and tamoxifen) and, therefore, no longer depend on ERα or estradiol for growth and survival. USP11 was significantly upregulated in both cell lines when compared with their parental MCF7 cells, at both the protein and mRNA level (Fig. 4A and B).
To investigate the role of USP11 on ERα transcriptional activity in each cell line, we examined the mRNA expression of the ERα+ target genes PgR, PKIB, and GREB1. USP11 was knocked down using two independent siRNAs and KD was confirmed 72 hours posttransfection. USP11 silencing significantly decreased the mRNA expression of ERα target genes in LCC1 cells (Fig. 4C). With the exception of moderately suppressed PKIB expression with one siRNA, no significant changes in expression occurred in LCC9 USP11-KD cells (Fig. 4D).
To further elucidate the role of USP11 on ERα transcriptional activity in the LCC cell line model, RNA-seq was performed on LCC1 and LCC9 USP11-KD cells (LCC1 siUSP11_1, siUSP11_2, and siControl; LCC9 siUSP11_1, siUSP11_2, and siControl; biological triplicate). First, PCA was carried out and values were plotted to visualize any variance between groups and replicates. Principle components were corrected for batch effect (Supplementary Fig. S2A and S2B). Next, log2 fold change values of siControl samples were compared with each individual siUSP11 sample to determine genes that were differentially expressed following USP11 silencing. Both siRNA lists were compared and the common DE genes were subjected to further analysis, to minimize any off-target DE genes. In LCC1 cells, 278 DE genes were common to both siRNAs and in LCC9 cells, only 29 DE genes were common to both siRNAs (Fig. 4E). DE genes in both cell lines were also compared. Only nine genes were common to both cell lines (Fig. 4F), demonstrating a differential role for USP11 in cells that depend on ERα for growth and survival and those that depend on other mechanisms.
We examined genes that were both up- and downregulated with KD in both cell lines (Fig. 4G). In LCC1 cells, 188 genes were downregulated (blue) and 90 were upregulated (red) with USP11 suppression. In LCC9 cells, only three genes were downregulated (blue) and 26 genes were upregulated (red) with USP11 suppression. To further understand the effect of USP11 silencing on cellular function, these DE genes were subjected to GO analysis, a method that allows for the query of genes on the basis of their shared biology (35). This allowed for the identification of key cellular pathways that were altered with USP11 silencing in LCC1 and LCC9 cells. Up- and downregulated genes in each cell line were grouped for GO pathway analysis, completed using the GSEA 3.0 (23) bioinformatic tool. ENSEMBL gene IDs were uploaded and all GO biological concepts were analyzed, including biological processes, molecular function, and cellular component. Interestingly, KD of USP11 in LCC1 cells resulted in a significant decrease in cell-cycle- and division-associated genes, consistent with the fact that LCC1 cells were still ERα driven. Key cellular processes, such as mitosis, chromosome segregation, and organelle fission, were all significantly downregulated with USP11 KD (Fig. 4H). This phenotype was unique to LCC1 cells, suggesting that USP11 has a major role in controlling the cell cycle in cells dependent on ERα for growth (Fig. 4I).
Conversely, KD of USP11 resulted in a significant increase in inflammatory-associated genes in both cell lines, identifying a common attribute of USP11 silencing in both LCC1 and LCC9 cells (Supplementary Fig. S2C and S2D). In LCC1 USP11-KD cells, upregulated genes were associated with an innate immune response and INF signaling, while the genes upregulated in LCC9 USP11-KD cells were primarily associated with viral response pathways. This is perhaps due to the role of USP11 in NFκB regulation (36).
A number of well-known ERα target genes were found to be significantly altered in LCC1 USP11-KD cells. For example, TOP2A, BLM, and BRCA1 were significantly downregulated according to RNA-seq data and confirmed by qRT-PCR analysis (Fig. 4J). TRIM22, a target of p53 and a tumor suppressor gene (37), was upregulated in both LCC1 and LCC9 USP11-KD cells (Fig. 4J). LCC1 USP11 KD downregulated genes were then compared against GSEA curated gene sets, an online collection of datasets obtained from various sources, such as the scientific literature and online pathway databases. Remarkably, one of the most significant overlapping genes sets was that published by Dutertre and colleagues (38), which is composed of 324 genes upregulated in MCF7 cells treated with estradiol for 24 hours. Of the 188 downregulated genes in the LCC1 USP11-KD list, 95 overlapped with the Dutertre dataset (P = 8.69 e-162; FDR q value = 2.06 e-158; Fig. 4K), suggesting that these are ERα target genes downregulated by USP11 silencing. We further examined patients with ERα+ breast cancer in The Cancer Genome Atlas (TCGA) dataset and found that one-third (n = 102) of genes in the Dutertre dataset significantly correlated with USP11 in these patients. Of the 95 genes associated with the Dutertre gene set in LCC1 USP11-KD samples described above, 26 positively correlated with USP11 in TCGA samples, suggesting that expression of a subset of ERα target genes is significantly correlated with USP11 expression in patients with breast cancer (Supplementary Fig. S3). These findings further support the pivotal role USP11 plays in controlling ERα transcriptional activity.
USP11 is a marker of poor prognosis in ERα+ breast cancer
Finally, the prognostic relevance of USP11 in ERα+ breast cancer was assessed to determine the validity of this protein as a marker in the clinic. Initially, an in silico analysis was performed using GOBO online (http://co.bmc.lu.se/gobo/gsa.pl; ref. 28). The online database, developed at Lund University (Lund, Sweden), consists of 1,881 cases pooled from 11 public microarray datasets analyzed using Affymetrix U133A. For survival analysis, two groups were selected: those with high expression of USP11 (red) and those with low expression (gray). Full censoring (years) was applied and OS was selected as the endpoint. High expression of USP11 was significantly associated with poor OS in ERα+ patients (P = 0.032; Fig. 5A), while no significant association between USP11 expression and survival was made in ERα-negative patients (P = 0.51; Fig. 5B). Multivariate analysis of the same datasets indicated that high USP11 expression was associated with tumor size, grade, and patients diagnosed over 50 years of age in the ERα+ cohort only (Fig. 5C). These associations were not observed in patients with ERα-negative breast cancer (Fig. 5D), suggesting that USP11 may be a useful prognostic marker in the ERα+ setting.
To support these findings, IHC staining of a TMA, described previously in the literature (27, 39), with tumor samples from 144 cases of invasive breast cancer, diagnosed at Malmö University Hospital (Malmö, Sweden) between 2001 and 2002, was performed. The TMA consisted of 103 patients with ERα+ tumors for final analysis. Manual scoring was carried out by two independent researchers using the displayed images as guidance (Fig. 5E). USP11 expression was grouped into low (score 0 and 1) and high (score 2 and 3) expression, and outcome based on each group was determined following statistical analysis of patient data. Kaplan–Meier analysis revealed a significant association between high USP11 expression and poor OS (P = 0.003; Fig. 5F) and BCSS (P = 0.041; Fig. 5G). A similar trend was observed between high USP11 expression and poor RFS (P = 0.066; Fig. 5H). Cross-tabulation analysis was also carried out to correlate USP11 staining with clinicopathologic features such as tumor grade, histopathologic subtype, and hormone receptor status. High USP11 expression was significantly associated with positive lymph node status (P = 0.009), but not with any other clinicopathologic features in this cohort.
While our understanding and management of breast cancer is now better than ever, more than 600,000 women worldwide succumb to this disease every year (40). Targeted, antiendocrine therapies have been widely effective in treating patients with ERα+ breast cancer, however, drug resistance and relapse remain a significant issue in the clinic. Given that the tumorigenic properties of ERα primarily lie in its function as a growth-controlling transcription factor, we sought to identify novel modulators of ERα transcriptional activity. Further understanding of the mechanisms that control ERα function is critical to design effective new targeted therapies, and while our knowledge of ERα function in breast cancer has vastly improved in recent years, translating this knowledge into the clinical setting has proved challenging thus far.
The PTMs of the receptor represent one such area of interest, with previous studies highlighting prominent roles for these modifications in determining receptor function in oncogenic pathways (10, 41). Here, we focus on DUBs, the enzymes that reverse ubiquitination of proteins and their role in ERα transcriptional activity. The oncogenic and tumor suppressive roles of these enzymes have come to light in recent years, and as a result, interest is growing in their therapeutic targeting (42). We highlight for the first time, evidence for the role of the BRCA-associated (11, 15) DUB USP11 in controlling ERα transcriptional activity. This represents a novel mechanism of ERα regulation in breast cancer and a pathway that could be exploited for the management of ERα+ breast cancer.
In this study, functional genomic screening, examining the effect of DUB KD on ERα transcriptional activity in an unbiased fashion, identified a role for USP11 in ERα transcriptional regulation in breast cancer cells (Fig. 1B). This finding was further validated in USP11-silenced cells, where ERα activity was significantly abrogated (Fig. 1E–J).
Furthermore, USP11 expression was regulated by estradiol in ERα+ cells (Fig. 2C), and the examination of subcellular fractions indicated that this upregulation is occurring in the cell nucleus (Fig. 2D). As estradiol treatment had no effect on USP11 levels in ERα-negative MDA-MB-231 cells (Fig. 2C), we hypothesize that estradiol mediates its effect on USP11 via ERα. The exact mechanism of interaction, however, warrants further investigation. In silico chromatin immunoprecipitation analysis of ERα binding following estradiol stimulation in a series of ERα+ breast cancer cell lines failed to demonstrate direct binding to either the USP11 promotor or gene body, suggesting either distal regulation or an indirect mechanism (Supplementary Fig. S4).
To further elucidate the relationship between USP11 and ERα, mass spectrometry was used to examine global changes to the proteome and ubiquitinome in breast cancer cells following USP11 suppression. Interestingly, USP11 KD resulted in significant changes to the proteome only in the presence of estrogen (Fig. 3A–C). This suggests that the role of USP11 in ZR-75-1 breast cancer cells is highly dependent on estradiol, and perhaps ERα. Further to this, USP11 suppression induced significant changes in the ubiquitinome in the presence of estradiol, and revealed novel putative substrates of USP11 (Fig. 3D–F). Intriguingly, cyclin D1, a well-described ERα target gene, was ubiquitinated following USP11 suppression and may represent a novel substrate. Cyclin D1 is upregulated in more than half of all diagnosed breast cancers, and can regulate ERα in a positive feedback manner. Interestingly, cyclin D1 expression was also positively correlated with USP11 expression in TCGA breast cancer patients. This observation may be of interest in the link between USP11 and ERα, and should be further explored.
USP11 was upregulated in LCC1 and LCC9 breast cancer cells, when compared with their parental MCF7 cells (Fig. 4A). LCC1 cells, derived from MCF7 breast cancer cells (33), are estrogen independent, but are responsive to antiendocrine agents, suggesting that ERα continues to drive cellular growth, but in a ligand-independent manner. LCC9 cells, following exposure to increasing concentrations of fulvestrant (34), are antiendocrine resistant and depend on other mechanisms for growth and survival. Following KD of USP11, ERα transcriptional function was abrogated in LCC1 cells only (Fig. 4C). RNA-seq revealed a further role for USP11 in LCC1 cells, with many cell cycle and ERα target genes suppressed with USP11 KD (Fig. 4H and I). USP11 suppression had little effect on LCC9 cells (Fig. 4D), which are antiendocrine therapy resistant and rely on alternative growth mechanisms. This strongly supports a role for USP11 in ERα function and suggests that USP11 silencing can interfere with ERα-dependent growth mechanisms. From the perspective of future therapeutic targeting of USP11, our results suggest that patients whose tumors remain driven by ERα would be the most appropriate candidates, for example, those with activating mutations in ERα in recurrent tumors, rendering them less responsive to tamoxifen, but susceptible to abrogation of ERα transcriptional activity.
Finally, high expression of USP11 was significantly associated with poor survival in patients with ERα+, but not ERα-negative tumors (Fig. 5A and B), further supporting a role for USP11 in this subtype of breast cancer. Analysis of a USP11-stained TMA further supported this finding at the protein expression level (Fig. 5E–H). Many previous studies have outlined the role of DUBs in cancer prognosis, for example, USP22 has been associated with poor prognosis in breast, colorectal, and liver cancer, while OTUB1 expression has been associated with a poor outcome in colorectal and ovarian cancer (43). With further studies to support these findings, USP11 may represent a novel marker to be utilized in the clinic to predict patient outcome and response to therapeutic intervention.
With the success of proteasome inhibition in the clinic (44, 45), the ubiquitin-proteasome system is becoming an attractive area of therapeutic intervention, and as DUBs are often differentially expressed or activated in tumors, much of the focus around the ubiquitin system has shifted in this direction (46). As most DUBs are cysteine proteases (47), a well-researched class of pharmacologic targets, it is feasible to construct specific inhibitors of these enzymes. There are a number of preclinical DUB inhibitors in development, for example, WP1130, a nonspecific inhibitor of USP9x, USP5, and USP14, which induces apoptosis and enhances response to chemotherapy (48), and FT671, a newly developed, specific inhibitor of USP7, which results in stabilization of p53 and induction of apoptosis (49). As we further interpret the role of DUBs in both physiologic and oncogenic pathways, the coming years may see rise some exciting advances in DUB drug discovery (50).
This research demonstrates, for the first time, a key role for USP11 in ERα activity in breast cancer. With almost half of patients not responding to current antiendocrine intervention therapies, it is vital to explore and elucidate novel mechanisms that control ERα function. USP11 may represent a promising route of therapeutic intervention and with further research, may be considered in the DUB drug discovery pipeline for the treatment of ERα+ breast cancer.
Disclosure of Potential Conflicts of Interest
A.E. O'Connor reports grants from Irish Research Council (IRCSET EMPOWER Postdoctoral Fellowship) during the conduct of the study. L. Mulrane reports grants from Irish Cancer Society (employed through BREAST-PREDICT grant) during the conduct of the study and other compensation from Novartis/Advanced Accelerator Applications, a Novartis company (employee of Novartis/Advanced Accelerator Applications after the time of contribution to this research) outside the submitted work. B.M. Kessler reports a relationship with the Innovative Technology Enabling Network (ITEN) funded by Pfizer. T. Ní Chonghaile reports other compensation from AbbVie (financial support) outside the submitted work. W.M. Gallagher reports grants from Science Foundation Ireland, Irish Cancer Society, and European Commission during the conduct of the study and personal fees and other compensation from OncoMark Limited (stock ownership) outside the submitted work. No potential conflicts of interest were disclosed by the other authors.
L. Dwane: Conceptualization, formal analysis, investigation, writing-original draft, writing-review and editing. A.E. O'Connor: Conceptualization, formal analysis, writing-original draft. S. Das: Formal analysis, supervision, writing-original draft, writing-review and editing. B. Moran: Formal analysis, investigation, writing-original draft. L. Mulrane: Conceptualization, formal analysis, investigation, writing-original draft. A. Pinto-Fernandez: Formal analysis, investigation, writing-original draft. E. Ward: Formal analysis, investigation, writing-review and editing. A.M. Blümel: Data curation, software, formal analysis, investigation, writing-review and editing. B.L. Cavanagh: Formal analysis, investigation. B. Mooney: Formal analysis, investigation. A.M. Dirac: Conceptualization, resources. K. Jirström: Resources, formal analysis, writing-original draft, resources, formal analysis, writing-original draft. B.M. Kessler: Resources, formal analysis, supervision. T. Ní Chonghaile: Supervision, writing-original draft. R. Bernards: Conceptualization, resources, supervision, funding acquisition, writing-original draft, writing-review and editing. W.M. Gallagher: Conceptualization, supervision, funding acquisition, writing-original draft, writing-review and editing. D.P. O'Connor: Conceptualization, resources, formal analysis, supervision, funding acquisition, writing-original draft, writing-review and editing.
This research was predominantly funded by BREAST-PREDICT, the Irish Cancer Society's Collaborative Cancer Research Centre (CCRC13GAL). Additional support was received from the European Union Seventh Framework Programme under the RATHER project 258967, Science Foundation Ireland Career Development Award to D.P. O'Connor (15/CDA/3438) and Investigator Programme award OPTi-PREDICT to W.M. Gallagher (15/IA/3104) and the SFI Strategic Partnership “Precision Oncology Ireland” (18/SPP/3522). This work was further supported by a grant from the Dutch Cancer Society through the Oncode Institute to R. Bernards. A.E. O'Connor was funded by the Irish Research Council to support this research. D.P. O'Connor received additional support from the Human Frontiers Science Programme, UICC, and the European Association for Cancer Research.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.