Cancer Stem Cell Phenotypes in ER⁺ Breast Cancer Models Are Promoted by PELP1/AIB1 Complexes Free

Luminal breast cancers account for approximately 75% of newly diagnosed cases of breast cancer and express estrogen receptor (ER) as well as a range of progesterone receptor (PR)-positive cells. Adjuvant hormone therapies targeting ER actions improve overall survival (1). However, approximately 40% of luminal breast tumors eventually progress to ER⁺, endocrine-independent disease (2). Mechanisms of resistance to ER-targeted therapies include upregulation and activation of growth factor receptor (GFR) signaling pathways, ER mutations, and upregulation of ER coactivator proteins (3). GFR signaling enhances phosphorylation of ER and ER pathway components, promotes ER cytoplasmic signaling, and ultimately results in profoundly altered gene expression (4–6). To prevent luminal breast cancer recurrence, we need to understand the molecular mechanisms that drive disease progression and identify new biomarkers that can be targeted in combination with ER-targeted therapies.

A promising biomarker for targeting breast cancer progression is PELP1 (proline, glutamic acid, and leucine rich protein 1; refs. 7, 8). PELP1 is primarily located in the nucleus (9) in mammary epithelial cells where it serves as a coactivator to a number of transcription factors including steroid hormone receptors (SR; e.g., ER) and is involved in chromatin remodeling (7), RNA processing (10), and ribosome biogenesis (10). PELP1 expression is dysregulated in many different cancers (e.g., endometrial, ovarian, prostate, brain) and is overexpressed in over 80% of invasive breast tumors (11). High PELP1 expression is associated with tumor grade, tumor proliferation, node-positive invasive breast cancer and distant metastasis, and decreased breast cancer–specific survival and disease-free survival (11–13). In addition, several studies have shown that PELP1 influences cancer cell biology through mediating changes in proliferation, apoptosis, autophagy, migration, invasion, metastasis, and endocrine resistance (7). Our group demonstrated that both ER and PR form a functional signaling and transcriptional complex with PELP1 to regulate novel estrogen-regulated ER/PR/PELP1 target genes associated with breast cancer progression (14).

PELP1 has also been shown to have cytoplasmic functions. For example, PELP1 acts as a scaffolding protein for growth factors and SRs that modulate cytoplasmic kinase signaling. Altered localization of PELP1 to the cytoplasm was observed in 50% of PELP1-positive breast tumors (9). In preclinical models of breast cancer, overexpression of cytoplasmic PELP1 through mutation of its nuclear localization signal promotes increased activation of cytoplasmic kinase signaling and confers tamoxifen resistance (9). Moreover, expression of cytoplasmic PELP1 in a mammary-specific transgenic mouse model induced mammary gland hyperplasia associated with increased proliferation and pro-survival signaling (i.e., PI3K/Akt and Ras/ERK; refs. 12, 15). Analysis of PELP1 localization from tumor samples revealed that patients with high levels of cytoplasmic PELP1 were less likely to respond to tamoxifen than patients with low cytoplasmic PELP1 levels (12). In addition, our group demonstrated that cytoplasmic PELP1 staining was observed in 36% (4 of 11) of atypical breast needle aspirate samples from women at high risk for developing breast cancer (16). Furthermore, we showed that cytoplasmic PELP1 expression upregulates protumorigenic IKKϵ and inflammatory signals that drive a migratory phenotype associated with breast cancer initiation (17). Collectively, these findings in breast cancer cell models, mammary mouse models, and patient samples demonstrate that altered PELP1 localization to the cytoplasm is an oncogenic event that promotes breast cancer initiation and progression. However, the mechanisms by which cytoplasmic PELP1 promotes oncogenesis are still not clearly defined.

Herein, we sought to identify interacting partners unique to cytoplasmic PELP1 and determine whether they promote oncogenic signaling in breast cancer progression. We identified AIB1 [amplified in breast cancer 1; also known as SRC-3 (steroid receptor co-activator 3) or NCOA3 (nuclear receptor coactivator 3)] as a novel binding partner of cytoplasmic PELP1. We found that cytoplasmic PELP1 expression elevated basal Thr24 phosphorylation levels of AIB1 and increased primary and secondary tumorsphere formation in both the presence and absence of estrogen. Estrogen was not required for upregulation of cytoplasmic PELP1-specific target genes associated with cell survival and cancer stem cell biology. Direct manipulation of AIB1 levels using shRNA downregulated cytoplasmic PELP1-specific target genes and inhibited cytoplasmic PELP1-induced tumorsphere formation. Moreover, knockdown of PELP1 in an AIB1-mouse derived tumor cell line reduced tumorsphere formation in vitro and led to increased survival of recipient tumor-bearing mice as compared with control mice. Taken together, our data demonstrate that cytoplasmic PELP1 interacts with AIB1 to drive estrogen-independent events involved in regulation of breast cancer stem cell biology and disease progression.

Estradiol (E2; Sigma) and hydrocortisone (Sigma) stocks were prepared in ethanol (EtOH). Epidermal growth factor (EGF; Sigma) was prepared in 0.1% BSA containing 10 mmol/L acetic acid. SI-2 stocks were prepared in DMSO (kindly provided by David Lonard, Baylor College of Medicine, Houston, TX).

MCF-7 cells were a gift from Deepali Sachdev (University of Minnesota, Minneapolis, MN). STR authentication was performed January 2012 by the Johns Hopkins CORE facility. MCF-7 cells were cultured in modified IMEM (Life Technologies) containing 5% FBS (Thermo Scientific HyClone), 1% penicillin–streptomycin (Life Technologies), and 67.5 ng/mL insulin (Life Technologies). T47D CO cells were cultured in MEM (Corning) containing 5% FBS, 1% penicillin-streptomycin, 1% MEM nonessential amino acids (Life Technologies), and 6 ng/mL insulin. Mammary epithelial cells (hMEC-hTERT, MCF10A) were cultured as described previously (16, 17). HMECs were purchased from Lonza and MCF10A cells were purchased from ATCC. For experiments with E2, cells were hormone-starved in phenol-free modified Iscove's modified Dulbecco's medium (IMEM) containing 5% DCC (dextran coated charcoal-stripped FBS; HyClone) for 16 hours prior to treatment. J110 cells were a gift from Myles Brown (Dana-Farber Cancer Institute, Boston, MA) and cultured in DMEM/F-12 (Corning) containing 5% FBS, 5 μg/mL insulin, and 0.1 μg/mL hydrocortisone. All cell lines were tested for mycoplasma prior to the initiation of experiments.

MCF-7 cells were transfected per manufacturer's instructions with human-specific PELP1 double nickase constructs (sc-405376-NIC-2; Santa Cruz Biotechnology) to knockdown endogenous PELP1 (h2 MCF-7). Stable h2 MCF-7 PELP1 models were generated by transducing h2 MCF-7 cells with retroviral LXSN (which also served as vector control) containing WT PELP1 or cytoplasmic PELP1. Cells were selected in and maintained as described above with 0.5 mg/mL G418 sulfate (Corning). Single-cell cloning was used to generate clonal cell lines expressing LXSN, WT PELP1, or cytoplasmic PELP1. Stable shAIB1 (clone TRCN0000365196)-expressing cell lines were created by transducing h2 MCF-7 PELP1 models with pLKO.1 lentiviral vectors containing target gene shRNA sequences from the MISSION TRC library (Sigma). Stable pooled populations were selected in and maintained as described above with 0.5 mg/mL G418 sulfate and 0.5 μg/mL puromycin (MP Biomedicals). To generate J110 PELP1 knockdown cells (PELP1-low), J110 cells were transfected with murine specific PELP1 double nickase constructs (sc-429042-NIC; Santa Cruz Biotechnology) as per manufacturer's instructions. Cells were selected for with puromycin and sorted for GFP positivity by FACS.

Cells were harvested in RIPA-lite lysis buffer [150 mmol/L NaCl, 6 mmol/L Na₂HPO₄, 4 mmol/L NaH₂PO₄, 2 mmol/L EDTA, 100 mmol/L NaF, 1% Triton-X 100, 1X complete mini protease inhibitors (Roche), 1× PhosSTOP (Roche), and supplemented with 1 mmol/L PMSF, 5 mmol/L NaF, 0.05 mmol/L Na₃VO₄, 25 mmol/L beta glycerophosphate (BGP), and 20 μg/mL aprotinin]. For cellular fractionation, cells were harvested using the NE-PER Nuclear and Cytoplasmic Extraction Kit (Thermo Scientific).

Interplay Adenoviral TAP system (Agilent) was used for adenoviral expression of TAP-tagged WT PELP1 and cytoplasmic PELP1. WT PELP1 and cytoplasmic PELP1 (NLS mutation) were subcloned into the pNTAP-A shuttle vector to create N-terminal TAP-PELP1 fusion. The pNTAP shuttle vectors were recombined with pADEasy, and the recombinant DNA was linearized and transfected into AD-293 cells for adenoviral production of TAP-CAT (control), TAP-WT PELP1, and TAP-cytoplasmic PELP1. hMEC-hTERT cells were infected with adenovirus for 48 hours and then subjected to cellular fractionation. Cytoplasmic extract from hMEC-hTERT cells expressing cytoplasmic-PELP1 or TAP-CAT and nuclear extract from hMEC-hTERT cells expressing WT PELP1 or TAP-CAT were processed through the TAP purification protocol per manufacturer's instructions. Purified protein complexes were separated by SDS-PAGE. Unique bands present only in the cytoplasmic PELP1 cytoplasmic extract or the WT PELP1 nuclear extract were identified by mass spectrometry by the University of Minnesota Center for Mass Spectrometry and Proteomics.

Cells were harvested in ELB lysis buffer [50 mmol/L HEPES, 0.1% nonidet P-40 (NP-40), 250 mmol/L NaCl, 5 mmol/L EDTA, 1× complete protease inhibitors (Roche), 1× PhosSTOP (Roche), and supplemented with 1 mmol/L PMSF, 1mmol/L NaF, 0.5 mmol/L Na₃PO₄, 25 mmol/L BGP, and 20 μg/mL aprotinin]. A total of 1,000 μg lysate (1 mg/mL) were incubated with 1 μg of the indicated antibody (e.g., AIB1 or ER) overnight at 4°C. Immunocomplexes were isolated with protein G agarose (Roche) for an additional 2 hours at 4°C. Resin was collected and washed with cold ELB buffer. Immunocomplexes were eluted with sample buffer, resolved by SDS-PAGE, and analyzed by Western blot analysis. For assays with cytoplasmic extract, cells were harvested using the NE-PER Nuclear and Cytoplasmic Extraction Kit. Cytoplasmic extracts were then subjected to coimmunoprecipitation assays as described above.

Immunoblotting was performed with the following antibodies: phospho-AIB1 (Thr24, Cell Signaling Technology), AIB1 (5E11, Cell Signaling Technology), β-actin (AC-40, Sigma), phospho-ER (Ser118, Cell Signaling Technology), ER (F-10, Santa Cruz Biotechnology), GAPDH (0411, Santa Cruz Biotechnology), HDAC2 (H-54, Santa Cruz Biotechnology), p65 (F-6, Santa Cruz Biotechnology), PELP1 (A300-180A, Bethyl Laboratories), goat anti-rabbit IgG-HRP (Bio-Rad), and goat anti-mouse IgG-HRP (Bio-Rad). Blots were developed with ECL using Super Signal West Pico Maximum Sensitivity Substrate (Pierce) and imaged by film.

Total RNA was extracted from cell samples using TriPure Isolation Reagent (Roche) and isopropanol precipitation. RNA (1,000 ng) was reverse transcribed to cDNA according to manufacturer's instructions using qScript cDNA SuperMix (Quanta BioSciences). qPCR was performed using Light Cycler FastStart DNA Master SYBR Green I (Roche) on a Light Cycler 480 II Real-Time PCR System (Roche). qPCR cycling conditions were: initial denaturation at 95°C (10 minutes), denature at 95°C (10 seconds), anneal at 60°C (10 seconds), and extension at 72°C (5 seconds) for 45 cycles. Target gene levels were normalized to standard housekeeper genes (TBP or 18s) and represent the average of three independent measurements (mean ± SD).

h2 MCF-7 PELP1 cells expressing LXSN, WT PELP1, or cytoplasmic PELP1 were hormone-starved in modified IMEM containing 5% DCC for 16 hours prior to hormone treatment. Cells (in triplicate) were treated with vehicle or E2 (10⁻⁹ mol/L) for 6 hours, followed by RNA extraction using an RNeasy Mini Kit (Qiagen). 50 base pair paired-end sequencing was performed at the University of Minnesota Genomics Center using Illumina HiSeq 2500 system. On average over 12 million reads were sequenced per sample. Each sample was aligned using the Tophat aligner (version 2.0.13). Samtools software (version 1.0_BCFTools_HTSLib) was used to sort and index bam files. Cuffquant (Cufflinks version 2.2.1) was used to generate transcript abundance files. Once all samples within each species were mapped and abundance estimate files were completed, Cuffnorm (Cufflinks version 2.2.1) was used to generate a table of Fragments Per Kilobase Of Exon Per Million Fragments Mapped (FPKM) values for genes within each sample. For each gene, fpkm values were log transformed and mean centered. Genes that varied less than 0.5 SD within the sample set were removed from further analysis. A simple t test with Benjamini–Hochberg correction and a minimum of 1.5-fold change cutoff was used to identify differentially expressed genes between various conditions.

Cells were fixed with cold methanol for 10 minutes at −20°C, washed with PBS, permeabilized for 15 minutes in PBS (0.05% Triton X-100; permeabilization solution), and blocked for 30 minutes in PBS [0.5% Triton X-100, 5% BSA, normal goat IgG (1:500); blocking solution]. Cells were then incubated with rabbit anti-PELP1 primary antibody in blocking solution for 1 hour. Cells were washed with PBS and incubated with AlexaFluor488–conjugated goat anti-rabbit (Life Technologies) secondary antibody in blocking solution for 1 hour. Cells were washed with PBS, dehydrated using ethanol, and mounted with DAPI containing ProLong Gold anti-fade reagent (Life Technologies).

Cells were seeded (4 × 10⁴ cells/well) in 1× sterile low melt agarose (Life Technologies) containing 5% DCC and the appropriate treatment [vehicle (EtOH) or E2 (1 nmol/L)]. Soft agar assays were allowed to proceed for 21 days at 37°C. Afterwards, cell colonies were stained with 0.1% crystal violet for 1 hour and washed with PBS. Data are presented as the average ± SD of three independent measurements.

Adherent cells were washed with PBS and dissociated enzymatically in 0.25% trypsin-EDTA (Invitrogen). Cells were sieved through a 40-μm sieve (BD Falcon). Single cells were plated in ultra-low attachment plates (Corning). Cells were grown in a serum-free mammary epithelial basal medium (MEBM; Lonza) containing 1% methylcellulose (Sigma), 1% B27 supplement (Invitrogen), 1% penicillin–streptomycin, 5 μg/mL insulin, 20 ng/mL EGF, 1 ng/mL hydrocortisone, and 100 μmol/L β-mercaptoethanol. For secondary tumorspheres, primary tumorspheres were collected and dissociated enzymatically in 0.25% trypsin-EDTA. Cells were plated as described above in conditioned media, which consisted of a 1:1 mixture of MEBM media (described above) and media from cultured parental cells. Tumorspheres were allowed to grow for 12 days. Tumorspheres were analyzed by total number, and scored by manual counting using a uniformly scaled grid. Data are presented as the average ± SD of three independent measurements.

An ALDEFLUOR assay kit (Stemcell Technologies) was used per manufacturer's instructions to assay for aldehyde dehydrogenase (ALDH) activity. Cells were plated in ultra-low attachment plates and grown in MEBM media (described above) to generate tumorspheres. Tumorspheres were collected and dissociated enzymatically in 0.25% trypsin-EDTA. Cells were resuspended (5 × 10⁵) in ALDEFLUOR assay buffer (1 mL). ALDEFLUOR reagent was added to this cell suspension, mixed, and then half of the cell suspension was transferred to a separate tube containing DEAB. The cells were incubated at 37°C for 45 minutes, washed, and then subjected to flow cytometry using the BD LSRII H4760 flow cytometer (BD Biosciences).

Ten mice per group were injected with 5 × 10³ J110 (parental or PELP1-low) cells into the right inguinal mammary fat pad of 3-4-week-old female FVB mice (Harlan Laboratories). Once palpable tumors were detected, tumors were measured by caliper every 2–3 days. Mice were euthanized once tumors reached 1 cm³ in size. Differences in survival were analyzed using Kaplan–Meier methodology and curves compared using the Mantel–Cox log-rank test. Animal studies were reviewed and approved by the University of Minnesota Institutional Animal Care and Use Committee (IACUC).

Data were tested for normal distribution using Shapiro–Wilk normality test and homogeneity of variances using Bartlett test. Once data were determined to meet these two requirements, statistical analyses were performed using one-way ANOVA in conjunction with Tukey multiple comparison test for means between more than two groups or Student t test for means between two groups, where significance was determined with 95% confidence.

To identify proteins that may participate in oncogenic PELP1 signaling (i.e., cytoplasmic PELP1) during breast cancer progression, we expressed tandem affinity purification (TAP)-tagged PELP1 constructs in hMEC-hTERT cells, a model of normal human mammary epithelial cells (hMEC). Following adenovirus transduction of hMEC-hTERT cells with TAP-WT PELP1, TAP-cyto PELP1, or TAP-CAT (control), cell lysates were subjected to cellular fractionation, processed through the TAP protocol to purify cytoplasmic or nuclear TAP–PELP1 complexes, and separated using SDS-PAGE; bands of interest were analyzed by mass spectrometry. The mass spectrometry results are summarized in Supplementary Table S1. Pathway analysis (IPA and KEGG) was performed for both cytoplasmic (Supplementary Table S2) and nuclear extracts (Supplementary Table S3). Pathways of interest that were identified from the cytoplasmic interacting proteins by IPA included “Cancer,” “Cellular Movement,” and “Cell Cycle” pathways. The top pathway identified by KEGG analysis of PELP1 nuclear interacting proteins was ‘Ribosome biogenesis in eukaryotes’, a known PELP1 nuclear function (10). Interestingly, AIB1 (also known as SRC-3 or NCOA3) was identified only in the cytoplasmic protein extract. The AIB1 gene locus is amplified approximately 10% in breast and ovarian cancer and overexpressed in approximately 60% in breast tumors (18). Similar to PELP1, AIB1 functions as a SR coactivator, is a critical member of hormone-responsive transcriptional machinery (19), and is a known driver of oncogenic phenotypes associated with luminal breast cancer progression (20, 21).

Given the clear relevance of AIB1 in breast cancer biology (22) and functional overlap between AIB1 and PELP1, we chose to further probe the relationship between these two coactivators. We first confirmed the results from our TAP-tagged mass spectrometry analysis by performing the TAP pull-down of cytoplasmic PELP1 from whole-cell extracts and then immunoblotting for copurified AIB1 (Fig. 1A); AIB1 and cytoplasmic PELP1 readily interact in TAP pull-down samples. In addition, we assessed AIB1′s interaction with cytoplasmic PELP1 (i.e., untagged) in previously described MCF10A cells (17), an hMEC model engineered to stably express LXSN (vector control) or cytoplasmic PELP1. Immunoprecipitation of AIB1 confirmed increased association between AIB1 and PELP1 in cells expressing cytoplasmic PELP1 relative to LXSN (Fig. 1B). These results confirm the findings from our TAP-tag mass spectrometry approach and demonstrate that AIB1 interacts with cytoplasmic PELP1 independently of the TAP-tag in hMECs.

Models of normal mammary epithelial cells contain low SR levels and, in general, are SR negative. Our prior work demonstrated that both ERα and PR-B form a functional signaling and transcriptional complex with PELP1 to regulate novel ER/PR/PELP1 target genes associated with endocrine resistance (14). Growth factor–induced activation of cytoplasmic kinases enhances SR phosphorylation and transcriptional activation (5, 6, 23). Namely, cytoplasmic PELP1 links growth factor signaling to highly efficient ER phosphorylation and activation (9). To study the impact of cytoplasmic PELP1 and AIB1 in the context of SR function, we sought to generate PELP1 knock-down and add-back models of SR⁺ breast cancer cells. We first screened a panel of known SR⁺ luminal breast cancer cells for high expression of AIB1 and PELP1. MCF-7, BT-474, ZR-75, and T47D CO breast cancer lines were chosen on the basis of their known increased SR expression (i.e., ER and PR; ref. 24). In addition, MCF-7 and BT-474 cells are known to have AIB1 gene amplification (18). T47D CO is a T47D variant cell line that contains both ER and PR, wherein PR is constitutively expressed (25). Notably, basal mRNA (Fig. 2A) and protein (Supplementary Fig. S1A) expression levels suggest that PELP1 and AIB1 may be coexpressed in ER⁺ breast cancer cells. MCF-7 cells were chosen for further study based on their high expression of ER, PR, AIB1, and PELP1 (24). To reduce endogenous PELP1 expression in parental MCF-7 cells, we employed a double nickase approach to target endogenous PELP1 (Fig. 2B; h2 MCF-7). These PELP1 knockdown cells will be referred to as h2 MCF-7 cells. h2 MCF-7 cells were then engineered to stably express LXSN (vector control), WT PELP1, or cytoplasmic PELP1 by retroviral transduction (Fig. 2C; h2 MCF-7 PELP1). To generate cells stably expressing cytoplasmic PELP1, we mutated the PELP1 nuclear localization signal (KKLK → EELE; cytoplasmic PELP1 NLS mutant) as reported previously (9, 16). Analysis of protein (Fig. 2C) and mRNA (Supplementary Fig. S1B) levels demonstrated successful overexpression of both WT and cytoplasmic PELP1. Of note, despite significant PELP1 mRNA and protein knockdown, we were unable to completely knockout PELP1 expression; hence, a slight band representing endogenous PELP1 is observed in LXSN whole-cell lysates (Fig. 2C) and nuclear extracts (Fig. 2D). Similar results were observed upon PELP1 knockout via CRISPR-based technology in additional breast cancer cell lines, suggesting that PELP1 nuclear functions are required for cell viability. Importantly, PELP1 dynamically shuttles between the cytoplasm and nucleus. The NLS mutation decreases the rate of nuclear translocation and thus increases steady-state levels of PELP1 in the cytoplasm. We further characterized h2 MCF-7 PELP1 cells via cellular fractionation (Fig. 2D) and immunofluorescence (Fig. 2E). As we predicted, signal is observed in both the nuclear and cytoplasmic fractions in cells stably overexpressing either WT or cytoplasmic PELP1. Western blotting of cytoplasmic and nuclear extracts confirmed that WT PELP1 is primarily located in the nucleus, whereas a larger fraction of cytoplasmic PELP1 is localized to the cytoplasm in the respective models (Fig. 2D). Similar results were observed in intact cells stained for PELP1 via immunofluorescence (Fig. 2E). In addition, we performed coimmunoprecipitation experiments in h2 MCF-7 PELP1 cells. We consistently observed increased interaction between AIB1 and PELP1 in cells expressing cytoplasmic PELP1 relative to either WT PELP1 or LXSN (Fig. 2F, from whole-cell lysates; Fig. 2G, from cytoplasmic extracts) under basal conditions. Similarly, enhanced interaction between ER and PELP1 occurred in cells expressing cytoplasmic PELP1 (Fig. 2H; basal conditions). The interaction between ER and AIB1 as detected in co-IPs was also increased in cells expressing cytoplasmic PELP1 (Fig. 2I; basal conditions). Collectively, characterization of h2 MCF-7 PELP1 cells confirms overexpression of predominately nuclear (WT PELP1) or cytoplasmic (cytoplasmic PELP1 NLS mutant) PELP1, and corroborates our finding that AIB1 is a novel preferential binding partner of cytoplasmic PELP1 in multiple cell lines. Moreover, expression of cytoplasmic PELP1 has a direct effect on other protein–protein interactions such as AIB1 co-association with ER in SR⁺ breast cancer models.

Previous studies have shown that cytoplasmic phosphorylation of AIB1 is required for its nuclear translocation and function as a coactivator of transcription (19, 26). We predicted that AIB1 phosphorylation levels may be altered as a result of its enhanced interaction with cytoplasmic PELP1. Therefore, we analyzed basal levels of AIB1 Thr24 phosphorylation in multiple cell models (normal and cancer) expressing cytoplasmic PELP1. Consistent with the role of cytoplasmic AIB1 phosphorylation as a required step for nuclear activation, cellular fractionation studies revealed that Thr24 phosphorylation levels of AIB1 were increased in the nuclear fractions of both MCF10A (Fig. 3A) and T47D CO (Fig. 3B) cells overexpressing cytoplasmic PELP1 relative to LXSN or WT PELP1. Notably, basal levels of AIB1 Thr24 phosphorylation in h2 MCF-7 cytoplasmic PELP1 cells were increased in the cytoplasmic fraction relative to LXSN and WT PELP1 cells (Fig. 3C; left); however, due to high levels of phosphorylated Thr24 AIB1 in the nuclear fractions of all h2 MCF-7 PELP1 cell lines, we were unable to detect PELP1-induced changes in nuclear AIB1 Thr24 phosphorylation. Interestingly, estrogen (E2) treatment had no effect on AIB1 Thr24 phosphorylation over vehicle control (Supplementary Fig. S2). These results indicate that AIB1 Thr24 phosphorylation (i.e., a molecular marker of nuclear activation) is increased in multiple models expressing cytoplasmic PELP1.

Previous global gene expression analysis in SR-negative hMEC-hTERT PELP1 models revealed upregulation of NFκB-regulated genes involved in inflammation (e.g., IL8, CXCL1, IL1β; ref. 17). However, we did not observe similar cytoplasmic PELP1-induced changes in the expression of inflammatory mediators in SR⁺ h2 MCF-7 PELP1 models. These results suggest that cytoplasmic PELP1 signaling is significantly different in SR⁺ versus SR⁻ cell lines and that other pathways are modulated by cytoplasmic PELP1 in SR⁺ breast cancer. To further elucidate pathways altered by cytoplasmic PELP1 in SR⁺ breast cancer cells, we performed an RNA-Seq analysis in h2 MCF-7 PELP1 models. h2 MCF-7 PELP1 cells stably expressing LXSN, WT PELP1, or cytoplasmic PELP1 were treated with vehicle (EtOH) or E2 for 6 hours, and RNA was isolated and subjected to RNA-Seq analysis as described in Materials and Methods.

Differentially expressed genes were identified between WT PELP1 or cyto PELP1 h2 MCF-7 cells (≥ 1.5-fold, Benjamini–Hochberg q < 0.05, hierarchical clustering; Fig. 4A). Cytoplasmic PELP1-expressing cells had 39 genes differentially expressed genes compared with WT PELP1 in vehicle conditions (17 upregulated, 22 downregulated) and 57 differentially expressed genes compared with WT PELP1 in E2 conditions (23 upregulated, 34 downregulated). Ingenuity Pathway analysis (IPA) was performed for cyto PELP1 versus WT PELP1 for vehicle (Supplementary Table S4) and E2 conditions (Supplementary Table S5). Genes known to be involved in cancer biology were pursued further. AMIGO2, NELL2, and HAPLN1 were chosen for their involvement in cell survival and stem/progenitor cell formation. Cytoplasmic PELP1-induced expression of these genes was confirmed by quantitative (q)RT-PCR validation (Fig. 4B). In addition, SERPINE1, which encodes for PAI-1, is elevated in breast tumor tissues and is correlated with reduced response to tamoxifen (27, 28), was found to be upregulated by cytoplasmic PELP1 but failed to meet the 1.5-fold cutoff for inclusion in the heatmap (Fig. 4A). Upon query of the Kaplan–Meier Plotter breast cancer database (29), we found that AMIGO2 expression is associated with a decrease in progression-free survival (PFS) in patients with ER⁺ breast cancer (Supplementary Fig. S3A), and that SERPINE1 expression is associated with a decrease in PFS regardless of ER status (Supplementary Fig. S3B and S3C). In summary, genes associated with advanced breast cancer biology were upregulated in SR⁺ breast cancer cells expressing cytoplasmic PELP1 as compared with cells expressing LXSN or WT PELP1, as demonstrated by RNA-Seq analysis and qRT-PCR validation experiments.

On the basis of the results of the RNA-Seq studies indicating that cytoplasmic PELP1 is promoting advanced breast cancer biology, we initially performed soft agar colony formation assays to examine effects on cell survival and proliferation in an anchorage-independent environment. h2 MCF-7 PELP1 cells were cultured in hormone-free starvation media in the presence of vehicle or E2 (1 nmol/L). Soft agar analysis revealed no differences between LXSN, WT PELP1, or cytoplasmic PELP1 with respect to total number of colonies (Fig. 5A); however, there was a approximately 2-fold decrease in average colony size among cytoplasmic PELP1-expressing cells as compared with cells expressing either LXSN or WT PELP1 (Fig. 5B; P < 0.0001; Supplementary Fig. S4). Furthermore, E2 treatment had no effect on total number of colonies (Fig. 5A) or colony size (Fig. 5B). These results indicate that cytoplasmic PELP1 expression attenuates proliferation as indicated by colony size, but does not alter cell survival as indicated by colony number in anchorage-independent growth assays.

Proliferative pathways often oppose pathways involved in stem cell expansion or EMT programs (30–32). On the basis of our observation that cytoplasmic PELP1 inhibited proliferation, but not survival, in anchorage-independent growth assays (Fig. 5B), we tested whether cytoplasmic PELP1 expression affected cancer stem cell phenotypes. Tumorsphere assays assess the ability of a minority population of breast cancer stem cells to expand in suspension culture (33). We performed tumorsphere assays (primary and secondary) using h2 MCF-7 PELP1 cells stably expressing LXSN (control), WT PELP1, or cytoplasmic PELP1. Interestingly, cells expressing cytoplasmic PELP1 exhibited a basal increase in primary tumorsphere formation as compared with cells expressing either LXSN or WT PELP1 (Fig. 5C; P = 0.001). This trend was also observed in E2-treated cells expressing cytoplasmic PELP1; E2 treatment increased the number of primary tumorspheres in all cell lines (Fig. 5C). We then performed secondary tumorsphere assays as a functional readout for breast cancer stem cell expansion. Cells expressing cytoplasmic PELP1 cells exhibited increased secondary tumorsphere formation in both the absence and presence of estrogen (Fig. 5D). In addition, enhanced intrinsic aldehyde dehydrogenase (ALDH) activity is a hallmark of cancer stem cells (34). AldeFluor assays confirmed that cells expressing cytoplasmic PELP1 (11.4% ± 1.9%) were enriched for ALDH⁺ activity relative to cells expressing LXSN (6.3% ± 0.56%; P = 0.014) or WT PELP1 (2.7% ± 0.64%; P = 0.0008; Fig. 5E; Supplementary Fig. S5). These results demonstrate that cytoplasmic PELP1 expression inhibits proliferation as measured by anchorage-independent growth assays, but drives breast cancer stem cell expansion, as measured using secondary tumorsphere assays and ALDH⁺ activity.

Cytoplasmic PELP1 complexes with AIB1 and mediates increased AIB1 Thr24 phosphorylation/activation. To test the requirement for AIB1 in cytoplasmic PELP1-induced changes in gene expression and associated breast cancer phenotypes, we first transduced h2 MCF-7 PELP1 cells with shRNA (shControl or shAIB1) to knockdown endogenous AIB1. Analysis of AIB1 mRNA (Fig. 6A) and protein (Fig. 6B) levels demonstrates successful knockdown in h2 MCF-7 PELP1 cells with no effect on PELP1 expression. Cytoplasmic PELP1-regulated genes identified by RNA-Seq (e.g., AMIGO2, NELL2, SERPINE1) were downregulated upon AIB1 knockdown in cells expressing cytoplasmic PELP1 (Fig. 6C). Tumorsphere assays were performed using shAIB1 knockdown cells. Similar to previous results (Fig. 5C and D), shControl cytoplasmic PELP1 cells exhibited a basal increase in both primary and secondary tumorspheres (Fig. 6D and E). However, tumorsphere formation was markedly decreased in both primary and secondary tumorspheres derived from shAIB1 knockdown cells relative to shRNA control cells (Fig. 6D and E). Again, E2 treatment increased primary tumorsphere formation in all cell lines (shControl and shAIB1). However, the E2-induced proliferative effect was greatly attenuated in secondary tumorspheres derived from shAIB1 cell lines. In addition to AIB1 shRNA knockdown, we tested whether pharmacologic inhibition of AIB1 would disrupt the cytoplasmic PELP1/AIB1 interaction and block tumorsphere formation. SI-2 is a small-molecule inhibitor of AIB1 that has been shown to reduce AIB1 protein levels and thereby inhibit primary tumor growth in mouse models of breast cancer (35). Notably, treatment of SR⁺ h2-MCF-7-PELP1 cells with 100 nmol/L SI-2 inhibited the cytoplasmic PELP1/AIB1 interaction in coimmunoprecipitation experiments (Fig. 6F). To test the effect of SI-2 on tumorsphere formation, h2 MCF-7 PELP1 cells were treated with either SI-1 (100 nmol/L) or vehicle. Interestingly, treatment with SI-2 (Fig. 6G) decreased tumorsphere formation in cytoplasmic PELP1-expressing cells, but had little effect on LXSN and WT PELP1 h2 MCF-7 cells. Collectively, these results demonstrate that AIB1 is a key mediator of cytoplasmic PELP1-induced changes in gene expression and associated breast cancer stem cell–like phenotypes.

The mouse mammary tumor virus-AIB1 (MMTV-AIB1) model exhibits mammary gland hyperplasia that progresses to invasive ER⁺/ER⁻ tumors (36). To evaluate the role of AIB1/PELP1 signaling in an AIB1 tumor model in vitro and in vivo, we utilized a mouse tumor cell line (J110) established from the MMTV-AIB1 mouse model (37) and employed a double nickase approach to knockdown endogenous PELP1. Although we were unable to completely knockout PELP1, we achieved significant PELP1 knockdown (Fig. 7A; PELP-low). Tumorsphere assays demonstrated that J110 PELP1-low cells exhibited reduced tumorsphere formation compared with J110 parental cells (Fig. 7B). In addition, similar to the h2 MCF-7 PELP1 model, SI-2 inhibited tumorsphere formation in J110 parental cells (Fig. 7C).

To test the effect of PELP1 knockdown in vivo, J110 cells (parental or PELP1-low) were injected into the mammary fat pads of syngeneic FVB mice (10 mice/group). Mice injected with J110 PELP1-low cells exhibited increased median survival relative to mice injected with J110 parental cells (34 vs. 24 days, P < 0.0001; Fig. 7D). In sum, these results demonstrate that directly targeting PELP1 has the potential to inhibit tumor growth, particularly in the context of AIB1-induced tumorigenesis.

Cytoplasmic PELP1 signaling has been implicated in breast cancer initiation and progression (7, 17). However, the molecular mechanisms of cytoplasmic PELP1-driven changes in breast cancer biology remain largely unknown. Our study identifies a novel role for AIB1 as a preferential binding partner for cytoplasmic PELP1 in hMEC and breast cancer cell models. In particular, our data demonstrate that cytoplasmic PELP1 overexpression increases Thr24 phosphorylation of AIB1 (an event required for nuclear AIB1 function as a coactivator), upregulates specific target genes, and enhances ALDH⁺ tumorsphere formation. Knockdown of AIB1 downregulates cytoplasmic PELP1-specific target genes and inhibits tumorsphere formation. Moreover, knockdown of PELP1 in a tumor cell line derived from AIB1 transgenic mice inhibits in vitro tumorsphere formation and in vivo tumorigenesis. Taken together, our data identify cytoplasmic PELP1-dependent AIB1 signaling as a key mediator of tumor progression and indicate that the PELP1/AIB1 complex is an important driver of breast cancer stem cell expansion and escape from hormonal regulation.

Many similarities exist between PELP1 and AIB1 beyond acting as SR coactivators. For example, AIB1 amplification and overexpression has been shown to promote breast tumor initiation, progression, and metastasis (22). In addition, AIB1 overexpression is linked to endocrine resistance in breast cancer cells and human breast tumors (38–40). Female AIB1 transgenic mice develop mammary hyperplasia, similar to cytoplasmic PELP1 transgenic mice (12), which eventually progresses to invasive SR⁺/SR⁻ tumors (36). Interestingly, increased AIB1 cytoplasmic localization has been reported in breast tumors (36). These diverse studies examined cytoplasmic PELP1 or AIB1 separately, and although there are clear parallels, our findings are the first to mechanistically link PELP1-mediated events to activated AIB1 in breast cancer.

A growing body of evidence implicates AIB1 as a mediator of stem/progenitor cell formation. Relevant to our findings, AIB1 is known to maintain self-renewal in embryonic stem cells (41), drive formation of cancer stem-like cells, and support tumor outgrowth in breast cancer models (42). We show that overexpression of cytoplasmic PELP1 enhances tumorsphere formation in h2 MCF-7 PELP1 models (Fig. 5C and D); however, this effect is abrogated in response to shRNA-mediated AIB1 knockdown (Fig. 6D and E). In addition, Rohira and colleagues demonstrated that SI-2 treatment decreased AIB1-induced cancer stem cell formation in breast cancer and xenograft models (42). We demonstrate that SI-2 treatment in h2 MCF-7 PELP1 models reduced tumorsphere formation in cytoplasmic PELP1 cells (Fig. 6G). Our findings support a growing body of evidence that implicates AIB1 as a mediator of stem/progenitor formation and links this event directly to cytoplasmic PELP1 signaling.

Herein, our findings demonstrate that expression of cytoplasmic PELP1 in h2 MCF-7 PELP1 models enhances SR protein–protein interactions, particularly between cytoplasmic PELP1/ER and AIB1/ER (Fig. 2H and I). Surprisingly, our RNA-Seq studies did not reveal significant changes in E2-induced gene expression in cells expressing cytoplasmic PELP1 relative to WT PELP1. Instead, however, cytoplasmic PELP1 enhanced basal gene expression, which was highly dependent on AIB1 (Fig. 6C). Furthermore, cytoplasmic PELP1 expression enhanced basal AIB1 Thr24 phosphorylation in multiple PELP1 breast cell models (Fig. 3). Studies have shown that AIB1 must be phosphorylated in the cytoplasm to enable AIB1 coactivator function in the nucleus (19, 26). Therefore, the increased binding interaction between cytoplasmic PELP1 and AIB1 observed in our coimmunoprecipitation assays (Figs. 1B and 2F) may serve as a mechanism to augment phospho-Thr24 AIB1 activation levels in breast cancer cells, leading to increased AIB1-dependent gene expression.

Our gene expression analyses are consistent with our finding that E2 treatment had little effect on AIB1 Thr24 phosphorylation relative to vehicle controls in cytoplasmic PELP1-expressing cells (Supplementary Fig. S2). These data suggest that in the context of luminal breast cancer, cytoplasmic PELP1-induced AIB1 Thr24 phosphorylation and activation may promote the emergence of hormone-independent pathways to SR target gene expression. Further studies are needed to understand how oncogenic PELP1/AIB1 complexes may alter hormonal inputs to SR signaling.

Our RNA-Seq analysis identified target genes of interest related to cell survival, breast tumor progression, and stem/progenitor cell formation (AMIGO2, NELL2, SERPINE1, and HAPLN1). AMIGO2, a cell adhesion molecule, has been shown to modulate cell survival in endothelial cells by directly interacting with PDK1 and facilitating Akt signaling to promote tumor growth and angiogenesis (43). NELL2 has been shown to mediate increased cell survival in neural cells through transactivation by ERα and ERβ (44). SERPINE1 encodes for PAI-1 and belongs to the serpin superfamily of protease inhibitors. PAI-1 is emerging as a prognostic biomarker for breast cancer, and a number of studies suggest that PAI-1 may enhance disease progression by blocking apoptotic pathways and thereby enhancing cell survival (45). Thus, cytoplasmic PELP1 expression may upregulate genes involved in mediating cell survival as a means to promote breast cancer progression. Our studies show that AMIGO2, NELL2, and SERPINE1 are upregulated in cytoplasmic PELP1-expressing cells and downregulated in response to AIB1 shRNA-mediated knockdown; these findings suggest these genes are reflective of a cytoplasmic PELP1/AIB1 gene signature.

We also identified a cytoplasmic PELP1-induced gene (HAPLN1) relevant to cancer stem cell biology. Mebarki and colleagues demonstrated de novo HAPLN1 expression occurs in aggressive hepatocellular carcinoma tumors expressing stem cell markers and in in vitro models of EMT via the Wnt/β-catenin signaling pathway (46). Considering that cells expressing cytoplasmic PELP1 increased both tumorsphere formation and intrinsic ALDH activity, we were surprised that our RNA-Seq studies did not identify more genes directly related to cancer stem cell biology. However, breast cancer stem cells represent a small minority of the total cell population (1%–5% of MCF-7 cells are cancer stem cells; ref. 47), making it difficult to detect changes in global gene expression by RNA-Seq, particularly when performed on RNA isolated from cells grown in adherent 2D conditions. Future studies will specifically examine gene expression in breast cancer stem cell populations maintained as spheroids (i.e., 3D culture systems) to further identify genes specifically regulated by cytoplasmic PELP1/AIB1 signaling.

PELP1 is emerging as a viable therapeutic target and biomarker for women with breast cancer. Preclinical studies illustrate that peptidomimetics and small-molecule inhibitors can specifically disrupt PELP1 protein–protein interactions. The PELP1 peptidomimetic inhibitor, D2, is designed to specifically inhibit the PELP1/androgen receptor complex via disruption of LXXLL-mediated protein–protein interactions (48). Our study identifies the PELP1/AIB1 complex as a novel protein–protein interaction that could be selectively targeted by potent inhibitors. Small-molecule inhibitors (e.g., SI-2) recently developed for AIB1 have shown preclinical utility in breast cancer cell lines and mouse models (35), and here we show that SI-2 inhibits PELP1-induced tumorsphere formation. Others have also shown that targeting the PELP1-interacting proteins KDM1 (49) and mTOR (50) inhibits PELP1-mediated tumor growth. Furthermore, these reports also demonstrate that PELP1 promotes endocrine therapy resistance and inhibition of KDM1 and mTOR sensitizes cells to PELP1-mediated tamoxifen resistance. Acquired resistance to hormone therapy remains a major clinical problem. Therefore, inhibitors capable of selectively targeting cytoplasmic PELP1-interacting partners that promote therapy resistance would provide a means to target luminal breast cancers and could be used in lieu of (or in conjunction with) ER-targeted therapies.

In addition to therapeutic targeting, our studies suggest that cytoplasmic PELP1 and phospho-Thr24 of AIB1 could be used as biomarkers to identify breast cancer patients likely to respond to therapeutic strategies designed to selectively target PELP1, AIB1, or the oncogenic PELP1/AIB1 complex. Our in vivo studies (Fig. 7) indicate that mice transplanted with AIB1-derived mouse tumor cells with PELP1 knockdown survive longer than mice transplanted with tumor cells expressing PELP1. These studies suggest that manipulating PELP1 location or levels has the potential to mitigate tumor progression, particularly in the context of AIB1-mediated tumorigenesis.

Our data suggests a model (Fig. 7E) in which cytoplasmic PELP1/AIB1 complexes promote persistently elevated cytoplasmic AIB1 Thr24 phosphorylation. We hypothesize this occurs through PELP1 functioning as a cytoplasmic scaffolding protein that brings together signaling molecules and protein kinases that mediate robust AIB1 Thr24 phosphorylation. AIB1 phosphorylation is known to occur on at least 6 different sites in response to different stimuli (hormones, growth factors, and cytokines), but phospho-specific antibodies are not commercially available for all sites. Thus, it is likely that cytoplasmic PELP1 signaling promotes site-specific Thr24 phosphorylation that directs AIB1 to regulate specific target genes. Future studies will be aimed at identifying PELP1-induced AIB1 phosphorylation sites and the kinase signaling complexes recruited by cytoplasmic PELP1.

Our data also suggests that inhibiting the cytoplasmic PELP1/AIB1 signaling complex may specifically target the breast cancer stem cell population. Breast cancer stem cells are believed to play a key role in breast cancer development, disease progression, and therapy resistance (51). Importantly, these minority populations of pluripotent cells are slow growing and not targeted by standard chemotherapies or endocrine therapies. If we can identify intracellular pathways or molecular complexes that drive survival and expansion of breast cancer stem cells, we would be able to target this subpopulation of cancer cells and have the potential to significantly impact on the overall survival of women diagnosed with breast cancer.

No potential conflicts of interest were disclosed.

Conception and design: T.H. Truong, C.A. Lange, J.H. Ostrander

Development of methodology: T.H. Truong, J.H. Ostrander

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T.H. Truong, H. Hu, K. Hagen, B.J. Girard, N.J. Brady, K.L. Schwertfeger, C.A. Lange, J.H. Ostrander

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T.H. Truong, H. Hu, N.A. Temiz, K. Hagen, B.J. Girard, N.J. Brady, J.H. Ostrander

Writing, review, and/or revision of the manuscript: T.H. Truong, N.A. Temiz, N.J. Brady, C.A. Lange, J.H. Ostrander

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T.H. Truong, J.H. Ostrander

Study supervision: K.L. Schwertfeger, C.A. Lange, J.H. Ostrander

This work was supported by the NIH grants R01 CA159712 (to C.A. Lange), K07 CA131501 (to J.H. Ostrander), F32 CA210340 (to T.H. Truong), and T32 HL007741 (to T.H. Truong), ACS Institutional Research Grant #124166-IRG-58-001-52-IRG5 (to J.H. Ostrander), Masonic Cancer Center, University of Minnesota, SP3 Award (to C.A. Lange, J.H. Ostrander, K.L. Schwertfeger), and the National Center for Advancing Translational Sciences of the National Institutes of Health Award UL1TR000114 (to J.H. Ostrander).

The authors thank Kira Downey, Katherine Leehy, and Olivia Wicker for technical assistance. We also thank Michael Franklin for critical reading of this manuscript.

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.