Estrogen-Driven Breast Carcinogenesis in a Rat Model is Linked to Retroelement Activation and Inflammation and is Susceptible to Lamivudine Treatment

Introduction

Although about 60–70% of BC cases are ER-positive (ER⁺), the earliest oncogenic events are still not clearly defined. Progression towards invasive ductal carcinoma (IDC) is usually preceded by ductal hyperplasia and ductal carcinoma in situ (DCIS) before breaching of the ductal myoepithelial layer at the invasive cancer stage.¹ It is well established that ER⁺ BC is initiated by estrogens, sex steroid hormones which bind and activate estrogen receptors, of which ERα is considered the primary oncogenic driver and therapeutic target.² In turn, ERα acts as a transcription factor and epigenetic regulator, which initiates signaling cascades leading to enhanced proliferation, growth and survival of mammary luminal epithelial cells.³ During normal mammary gland development and breast carcinogenesis, ERα exerts its function through the partnership with pioneer transcription factors FOXA1 and GATA3, which are required to activate gene expression programs associated with luminal proliferation and differentiation.^4,5 Although ERα and FOXA1 are considered favorable prognostic markers in BC and early-stage cancers and are efficiently treated with anti-estrogen therapies, ERα and FOXA1 are also highly expressed in metastatic BC resistant to endocrine therapy,^6,7 highlighting the need for new targets and more advanced preventive strategies.

Among the potential BC risk factors, chronic inflammation and immunosuppression are increasingly recognized as important drivers of tumorigenic development and progression.⁸ In many human cancers including breast, tumorigenic progression is also accompanied by activation of endogenous retroviral elements (EREs), including long- and short- interspersed nuclear elements (LINEs and SINEs, respectively) and human endogenous retroviruses (HERVs).^9–13 Most of them are epigenetically silenced in somatic cells, however some, most notably LINE-1 (L1) and HERV-K, have retained the ability to reverse transcribe and retrotranspose, resembling a lifecycle of retroviruses.¹⁴ Both L1 and HERV-K are detected in ER-positive and negative BCs,^15–19 and their association with a particular BC subtype is not well established. Known mechanisms of ERE-driven oncogenic transformation in breast and other types of cancer include regulation of neighboring and distant genes¹⁸ as well as chromosomal aberrations and mutations introduced by retrotransposition of functionally active L1s through their endonuclease (EN) and reverse transcriptase (RT) activities.^9,11–13,20 Furthermore, even truncated and retrotransposition-incapable EREs produce highly immunogenic RNAs and proteins with an oncogenic potential, including HERV-encoded envelope (env), gag, protease (pro) and polymerase (pol), as well as L1-encoded ORF1p and ORF2p proteins,^11,21,22 although their causal role in tumor initiation has not been proven.¹⁴

Due to their viral-like features and the ability to generate double-stranded RNAs and RNA-DNA products of reverse transcription, ERE RNAs may induce a so-called “viral mimicry state”, by increasing immunogenicity and activating antitumor immunity.²³ However, similar to exogenous chronic viral infections, prolonged ERE expression can also lead to overstimulation, immune exhaustion and chronic inflammation, promoting persistent proinflammatory cytokine release and the pro-tumorigenic tumor microenvironment.^14,24 Targeting ERE expression and propagation via reverse transcription may therefore reduce tumor growth by inhibiting ERE induced DNA damage and blocking chronic inflammation. In this regard, repurposing FDA-approved drugs developed against human immunodeficiency virus 1 (HIV-1) is especially promising, given similarities between HIV-1 and L1- and HERV-K-encoded RTs.^25,26 Together with proven long-term efficacy, safety and tolerability of HIV-1 therapeutics,²⁷ their implementation in clinical practice is expected to receive fast track approvals. In particular, the HIV-1 targeting drugs are now tested in preclinical studies and clinical trials to prevent age-related diseases including Alzheimer’s and type 2 diabetes, inflammaging and cancer.^28–30 Among them, several nucleoside RT inhibitors including lamivudine (3TC), abacavir and stavudine showed exceptionally strong antitumor activities in pre-clinical models of aging and cancer.^31–33 Moreover, treatment with 3TC in a recent clinical trial stabilized disease progression in patients with TP53 mutant advanced colorectal cancer.³² RT inhibitors were also shown to mitigate drug resistance caused by ERE-driven inflammatory response in mouse models of neuroblastoma and MMTV-HER2/Neu-induced breast carcinomas.³³ Unlike the existing BC therapies mainly directed at tumor eradication, targeting ERE proteins and RT activities may therefore represent therapeutically novel approach which may simultaneously restrict tumor growth, alleviate cancer-associated inflammation and activate anti-tumor immunity without severe cytotoxic effects associated with conventional anti-cancer modalities. Reliable mouse models of estrogen-driven mammary tumorigenesis adequately modeling the initiation and early progression of human ER⁺ BC are largely lacking. As a result, in our study here we employed the solely estrogen-driven August-Copenhagen-Irish (ACI) rat BC model which is commonly used in mechanistic and prevention studies^34,35 and faithfully recapitulates molecular features and stages of estrogen-driven human BC development from intraductal DCIS-like lesions to invasive cancer.³⁶ Using this model as well as human BC cell lines and tumor specimens, we show here that expression of ERE RNAs and proteins may support breast tumorigenesis by impacting both mammary epithelial cells and PBMCs, and their pro-tumorigenic, pro-inflammatory and immunosuppressive effects can be reversed by treatment with lamivudine/3TC.

Materials and Methods

ACI Rat Housing and Treatments

Female ACI/SegHsd rats, aged 6–8 weeks, were obtained from Envigo (Indianapolis, IN) and housed under standardized conditions of temperature, humidity and a 12-hour light/dark cycle. Following a 1-week acclimatization period, rats were randomly assigned to various experimental groups, with 5 rats per group. Rats were implanted with the slow-release 17β-estradiol (E2) pellets (2.5, 7.5, 10 or 25 mg E2) or placebo pellets (Innovative Research of America Inc). Pellets were implanted under the skin by a small incision made in the shaved dorsal region between the scapulae under halothane anaesthesia following animal care guidelines according to the Canadian Council on Animal Care (CCAC) and the Ontario provincial Animals for Research Act. Lamivudine (3TC) was purchased from Ambeed Inc. (#A107352) and prepared as a stock solution 50 mg/mL in distilled water. 3TC was administered in drinking water at 5 and 15 mg/mL concentration supplemented with 1.5% of sucrose, while the control group received drinking water containing 1.5% sucrose only. Drinking water was replenished every other day to maintain active concentration of 3TC and hygiene standards. Rats were weighed once a month to monitor body weight changes and assess potential physiological effects associated with 3TC and E2 treatments.

Collection of Rat Blood Serum, Plasma, and PBMC

Blood samples (~100–200 µL) were collected from the saphenous vein at 1- and 3- month time points post-implantation. Following collection, blood was allowed to clot at room temperature for 30 min, then centrifuged at 1,500 g, 15 min to separate serum. Serum was subsequently collected and stored at −80°C for downstream analysis. At the 6-month time point, rats were euthanized by CO₂ inhalation, and blood was collected by cardiac puncture into K2EDTA-coated blood collection tubes (#367863, BD Vacutainer) prior to cessation of cardiac activity. Collected blood was transferred to 15-mL centrifuge tubes and centrifuged at 1,200 g, 10 min to separate and collect plasma. The buffy coat layer of white blood cells was collected, mixed 1:1 with phosphate-buffered saline (PBS) and applied over the Lympholyte-Rat density separation medium (#CL5041, Cedarlane Labs) to isolate PBMCs. After centrifugation at 1,500 g, 20 min, cells at the interface were collected, washed 2–3 times in D-PBS and counted using the Beckman Coulter Vi-CELL instrument. At the final step, cell pellets were resuspended in the 10% DMSO-fetal bovine serum (FBS) freezing medium and cryopreserved at −80°C for future applications.

Detection of Serum E2 Levels and Bio-Plex Plasma Cytokine Profiling

Serum or plasma E2 concentrations were determined by HPLC-MS at the St. Michael`s Hospital (Toronto, Ontario) according tо their standard protocols. Plasma cytokine levels were measured using the MilliPlex Rat Cytokine/Chemokine magnetic bead panel (#RECYTMAG-65K, Millipore), including IL-1α, IL-1β, IL-6, IL-13, IL-17A, IL-18 and IFN-γ cytokines. Plasma was diluted 1:2 with the assay buffer prior to loading onto the plate. Assay preparation and procedures were performed in accordance with the manufacturer’s instructions. Detection was performed on the Bio-Plex 200 instrument, using the protocol configured with the kit specifications. Conversions of E2 (MW 272 g/mol) concentrations were done using the following formula: E2 pg/mL × 3.676 = pmol/L.

Harvesting and Processing Rat Mammary Glands

Mammary glands were isolated from sacrificed rats by a small incision across the abdomen and a second incision down the midline. After peeling off the skin, the inguinal mammary glands were removed and either immersed in 10% formalin for histological studies or used fresh for whole mounts or cell isolation. Mammary gland cells were isolated using the Tumor Dissociation Kit (#130-096-730, Miltenyi Biotec) according to the manufacturer’s instructions. Briefly, excised mammary gland tissue was minced into small fragments (2–4 mm diameter), transferred to a gentleMACS™ C Tube containing the enzyme mix provided with the kit and disaggregated and digested using the pre-set program “37C_m_TDK_1”. The resulting cell suspension was filtered through a MACS SmartStrainer (70 µm) and washed with 10 mL of DMEM. After centrifugation, supernatant was carefully aspirated, and cell pellets were resuspended in freezing medium. Cells were counted, aliquoted and kept in the liquid nitrogen tank for future analysis.

ACI Rat Mammary Gland Whole Mounts and Immunohistochemistry (IHC)

For whole mounts, freshly excised mammary glands were spread onto an adherent glass slide without tearing apart the tissue and left for 5 min to allow the tissue to partially dry and stick to the glass surface. The slides were submerged in Carnoy’s solution to completely fix mammary gland and then stained with Carmine Alum solution for 24 h at room temperature, followed by rinsing with the destaining solution and tissue dehydration by successive washes for 5 min in 70%, 80%, 95%, and 100% ethanol, and overnight incubation in xylene at room temperature. The slides were then mounted with Permount mounting medium (#SP15-100, Fisher Scientific) and air-dried at room temperature. Immunohistochemistry was performed on formalin-fixed, paraffin-embedded (FFPE) sections (4 µm thickness), which were deparaffinized and rehydrated prior to staining. Following antigen retrieval with the antigen unmasking solution (#H-3300, Vector), slides were washed three times × 5 min in Tris buffered saline (TBS; #T5912, Sigma), treated with 3% hydrogen peroxide to quench endogenous peroxidase activities, washed in TBS (3 times × 10 min) and blocked with the background sniper solution (#BS966, Biocare). After a brief rinse with TBS, slides were incubated with primary antibodies in the antibody diluent (#003118, Thermo Fisher) overnight at 4°C in a humidified chamber. The slides were then washed with TBS (3 times × 10 min) and incubated with single anti-rabbit or anti-mouse antibody solution (#MP7451, #MP7452, Vector), or MACH2 double stain 1 or 2 (#MRCT523, MRCT525, Biocare) for 30 min at room temperature. HRP and AP detections were performed sequentially using ImmPACT DAB (#SK-4105, Vector) and Fast Red (#FR805, Biocare) according to the manufacturers’ protocols. Slides were counterstained with hematoxylin and blued using Tacha’s Bluing Reagent (#HTBLU-M, Biocare). After rinsing and air-drying at room temperature, sections were dehydrated, mounted, dried at room temperature and placed in 4°C for storage. The antibodies and their dilutions used in this study are listed in Table S1.

IHC Analysis of Human Breast Cancer Tissue Microarrays (TMAs)

BC TMAs containing 300 tumor tissue cores (0.6 mm diameter) from 100 cases in triplicates (50 cases/TMA) were obtained from the Ontario Tumor Bank (OTB). Patient cohort and clinical parameters are listed in Table S3. IHC staining was performed as described above using 5-µm serial sections and antibodies against ERα, FOXA1, L1 ORF1p, HERV-K7 env and HERV-K gag (listed in Table S1). Staining for each marker was done in a single batch and included a no-antibody control. The scoring was performed independently by two qualified researchers, and all discordant cases were re-evaluated to reach a consensus between both observers. Tissues were classified as negative (0), weakly or moderately positive (1+, 2+) and strongly positive (3+), based on protein staining intensity. The percentage of tumour cells present in the core was then multiplied by the intensity number and used to generate a histoscore, ranging from 0 (no staining in 100% of tumour cells) to 300 (3+ staining in 100% of tumour cells). Cores that could not be scored (eg, missing, broken, folded, overlapping with another core) were excluded from further analyses. In total, 115 cores from 65 cases were available for the analysis. The histoscores of all stains attained an intraclass correlation coefficient (ICC) of 0.89 or above for each TMA slide between 2 independent analyzer scores, confirming consistent and reliable scoring. Based on staining intensity for ER and FOXA1 proteins, cores were classified as “high” or “low”, separated by the median histoscore (ie, 50% percentile); these were combined into a singular ER/FOXA1 metric, such as “ER-high, FOXA1-high,” “ER-low, FOXA1-high,” “ER-high, FOXA1-low,” and “ER-low, FOXA1-low”. Likewise, each core was classified as “ERE-high” if 2 or more ERE proteins were highly expressed, otherwise it was classified as “ERE-low.” A two-way contingency table was derived from the core counts from these classifications. Dependence between ERE/ERV protein and ER/FOXA1 expression were evaluated using chi-squared test of independence. This was followed up with post-hoc pairwise chi-square tests to calculate standardized residuals, with corresponding p-values corrected by Benjamini-Hochberg (BH) procedure. Standardized residuals were also visualized as a correlation matrix/Corrplot.

Cell Cultures and E2 Treatments

BC cell lines (MCF-7, T47D, ZR-75-1, MDA-MB231, SK-BR3, BT-549 and MDA-MB-468) and non-transformed breast epithelial cells (MCF-10A) were obtained from the American Type Culture Collection (ATCC) and cultured according to manufacturer’s instructions. For E2 stimulations, MCF7 and T47D cells were cultured for 24–48 h in the phenol red-free MEM (MCF7) or RPMI (T47D) medium supplemented with 10 µg/mL insulin (#12585014, Fisher Scientific) and 10% charcoal stripped fetal bovine serum (FBS) (#A3382101, Fisher Scientific), to deprive cells of the endogenous E2. Cells were then stimulated with 0.5–50 nM E2 in the presence of 100–400 µM of 3TC or PBS control. Following 24 h incubation, cells were washed with PBS and cell pellets were stored at −80°C for downstream analysis.

Immunoblot Analysis of Breast Cancer Cell Lines

Whole cell pellets (1 x 10⁶ cells) were resuspended directly in 200 μL of the SDS sample loading buffer, boiled 5 min at 98°C, resolved on SDS-10% polyacrylamide gel (10–25 μg protein/lane) and transferred onto 0.45 μm nitrocellulose membrane (#162-0115, Bio-Rad). The membrane was incubated for 30 min in the TBST blocking buffer (5% non-fat dry milk, 0.1% Tween-20, TBS) followed by overnight incubation at 4°C with primary antibodies listed in Table S1. The membranes were washed with TBST, incubated for 40 min at room temperature with horseradish peroxidase-conjugated secondary antibodies at 1:5,000 dilution, washed with TBST, developed with ECL reagents and documented by digital imaging using ChemiDoc MP Imaging system (Bio-Rad).

RNA Isolation and Reverse Transcription-Quantitative PCR (RT-qPCR)

Total RNA was extracted from rat mammary epithelial cells and PBMCs or from BC cell line pellets using the mirVana™ miRNA Isolation Kit (#AM1560, Fisher Scientific) following the manufacture`s instructions. Genomic DNA was removed using the TURBO DNA-free kit (#AM1907, Fisher Scientific), by adding 3 µL of TURBO DNase per 100 µL reactions and incubating at 37°C for 20 min. After DNase removal, RNA was precipitated with 75% ethanol and 10 µg/mL linear acrylamide (#AM9520, Fisher Scientific), overnight at −20°C. Samples were centrifuged at 13,000 x g for 15 min at 4°C and the RNA pellets were dissolved in nuclease-free H₂O. RNA concentration and purity were assessed using a NanoDrop spectrophotometer. cDNA was synthesized using SuperScript IV Reverse Transcriptase (#18090200, Fisher Scientific), 100–500 ng of DNase-treated RNA and random hexamer primers. Quantitative PCR (qPCR) was performed on the ViiA 7 Real-Time PCR instrument using the PowerUp SYBR Green Master Mix (#A25777, Fisher Scientific) and the respective primer sets listed in Table S2. RT-negative controls were included in all qPCR reactions. Expression of target transcripts was quantified by ΔΔCt method using normalized Ct values.

Bulk RNAseq and Data Analysis

Total cellular RNAs from rat mammary epithelial cells or PBMCs (100–500 ng) were processed in the Genomics facility (OICR, Toronto, Ontario) using the TruSeq Stranded Total RNA Library Prep Kit with Ribo-Zero Gold (#20020598, Illumina). The quality of libraries was confirmed using the Agilent High Sensitivity DNA assay, and those of sufficient quality and quantity were sequenced to ~100 million paired-end 150-bp reads/sample on the Illumina NovaSeq instrument. Pre-processing and sequence alignments were performed as previously described.³⁷ Briefly, fastq files generated by the Illumina bcl2fastq build 2.20.0.422 were filtered with fastp³⁸ to discard reads of low quality (Q-score < 15) and low complexity, such as polyG and other homopolymer stretches known to be generated on the NovaSeq platform. Repeat/ERE expression was analyzed by mapping paired reads to the Rattus norvegicus reference genome (rn7) using the Burrows-Wheeler Aligner (bwa-mem2 2.2.1). Reads aligned with bwamem2 were assigned to different repeat classes using the bedtools suite 2.29.2 with RepeatMasker annotations (https://www.repeatmasker.org). Analysis of coding gene expression was done using the STAR 2.7.6a aligner with default parameters.³⁹ Gene counts were determined with the SUBREAD suite (2.0.6) featureCount tool⁴⁰ using gene annotations for rn7, Ensembl release 110. For GSEA, RNAseq read counts were normalized with DESeq2⁴¹ followed by calculation of enrichment score using the GSEA v.4.2.3 tools and Molecular Signature Database (MSigDB) version 7.5.1 (http://www.gsea-msigdb.org/gsea/msigdb/collections.jsp) or immune cell annotations from CellMarker 2.0.⁴² Normalized enrichment score (NES) was used as a measure of coordinated changes in hallmark gene pathways with “Gene Set” as the permutation type. The permutation number was set at 1000, FDR values of less than 0.25 and nominal p-value < 0.05 were considered significant.

Statistical Analyses

Data analysis was performed using GraphPad Prism 9 software, Microsoft Excel or Rstudio R 4.2.0 (https://www.R-project.org/). Additional packages used for data analysis and visualization were downloaded from Bioconductor (https://www.bioconductor.org) and CRAN (https://cran.r-project.org/web/packages/). Statistical comparisons between groups were conducted using two-tailed Student’s t-tests, Wilcoxon rank-sum tests or chi-squared tests of independence, as indicated in the Figure Legends. Statistical significance was conditioned on the adjusted p value < 0.05 (Holm, Bonferroni or Benjamini-Hochberg (BH) corrections). Normal distribution of RNAseq data was confirmed with Shapiro–Wilk test. To correct for multiple comparisons, the unpaired t-test on log-transformed values were adjusted using Benjamini-Hochberg (BH) test. For data that failed a normality test, the Wilcoxon signed-rank test was applied. Visualization of the results was done in Rstudio using packages ggplot2 and pheatmap. DeSeq2 library was used for pre-processing (normalization) of gene expression data before GSEA analysis and visualization with heatmaps. No statistical method was used to predetermine sample size. No data were excluded from the analyses if the library and/or sequencing quality passed our criteria.

Results

Estrogen Drives Early Pre-Invasive Mammary Ductal Lesions in Peripubertal ACI Rats

To study the earliest stages of BC development we employed the ACI strain of rats, the only existing immunocompetent model of the ER-positive BC which closely recapitulates human breast tumorigenesis and is driven solely by estrogen without any chemical induction or gene manipulation.³⁶ In this model, elevated estrogen levels in blood circulation are maintained by subcutaneously implanting slow-release pellets of 17β-Estradiol (E2; a predominant circulating estrogen in humans and rats) into the shoulders of peripubertal (5–6-week-old) female rats (Figure 1A). The timing for E2 implantation was chosen based on the enhanced mammary gland susceptibility to estrogen during puberty, which heightens the risk of BC.⁴³ We first tested 2.5, 7.5, 10 mg and 25 mg E2-containing pellets (Figure 1A) to establish a minimal dose required to sustain elevated E2 levels over the 6-month treatment period. High-performance liquid chromatography-mass spectrometry (HPLC-MS) analysis of rat serum at 1 month after pellet implantation showed that 7.5 mg and higher-dose E2 pellets increased serum E2 levels by ~50–100-fold compared to the placebo rats implanted with empty pellets (p < 0.01, paired two-tailed t-test; Figure 1B). Even at 6 months post-implantation, plasma E2 levels remained elevated, from ~50 pmol/L (13.6 pg/mL) in the placebo cohort to 500 pmol/L (136 pg/mL) in the rats implanted with 7.5 mg pellets, and up to 2,000 pmol/L in rats implanted with 10 and 25 mg pellets (Figure 1C). Notably, E2 concentrations achieved with 7.5 mg pellets were within the range of serum E2 concentration in prepubertal girls and reproductive-age women, or in postmenopausal women treated with E2 as hormone replacement therapy,^44–46 and 7.5 mg pellets were thus chosen as a dose for further testing.

Figure 1 Estrogen drives early pre-invasive mammary ductal lesions in peripubertal ACI rats. (A) Experimental outline of the rat model and treatments described in this study. The ACI rats (n = 5 per group) were implanted with slow-release E2 or placebo pellets, treated with 3TC in drinking water or left untreated, and monitored for 6 months. Blood and tissue collection and analysis was done as indicated. (B and C) HPLC-MS analysis of E2 concentration in serum after 1 month of pellet implantation (B) or in plasma of rats at the 6-month endpoint (C). Data presented as box plots with mean ± SD. Comparisons made to the placebo control rats. *p < 0.05, **p < 0.01, ***p < 0.001, unpaired two-tailed t-test. (D and E) Hematoxylin-eosin (H&E) staining of FFPE sections of mammary glands from rats implanted with placebo control pellets (D) or with 7.5 mg E2 (E) for 6 months. Arrows indicate areas of ductal hyperplasia in mammary glands of E2-treated rats.

As expected, by 6 months, all rats implanted with 7.5 mg E2 pellets developed pre-invasive intraductal lesions reminiscent of ductal carcinoma in situ consisting of proliferating ductal epithelial cells (Figure 1D and E) with intense ER and FOXA1 staining by IHC (Figure 2D), while the control group of rats had no abnormalities in mammary glands (Figure 1D and E). Whole mount mammary gland staining showed extensive swelling due to tumor cell proliferation in the lobular-alveolar structures (Figure 2A, lower left panel). No undue morbidity effects were observed, and tumor penetrance was 100% in E2-treated rats versus none in the placebo control group. Similar morphological changes in mammary epithelium were also induced by implanting 10 and 25 mg E2 pellets (data not shown), however higher doses (25 mg) caused increasing comorbidities, including weight loss, extensive fatigue, ruffling and loss of fur, and were thus omitted from further testing. These data establish that 7.5 mg E2 in slow-release pellets is a minimal dose sufficient to induce mammary tumorigenesis in the ACI rats.

Figure 2 3TC blocks E2-induced mammary tumorigenesis. (A) Whole mount analysis of mammary glands of ACI rats implanted with 7.5 mg E2 or placebo pellets and treated for 6 months with 3TC or no-3TC control in drinking water, as indicated. (B) HPLC-MS analysis of plasma E2 concentrations from rats in (A). Data presented as boxplots with mean ± SD. Comparisons made to the placebo control rats. *p < 0.05, **p < 0.01, unpaired two-tailed t-test. Note that E2 levels were elevated in rats implanted with E2 pellets compared to placebo and were not changed upon 3TC treatment. (C) H&E staining of mammary gland FFPE sections from rats implanted with placebo or E2 pellets, treated or untreated with 3TC, as indicated. (D) Immunohistochemistry (IHC) of mammary glands from (C) using antibodies against ER and SMA (top panels) or FOXA1 and SMA (bottom panels). Scale bars in (C and D) are 100 and 60 µm, respectively. Note that mammary glands from rats implanted with E2 exhibit extensive areas of ductal hyperplasia and high positivity for ER and FOXA1, and treatment with 3TC completely cleared the ducts, leaving some milky liquid inside.

3TC Blocks E2-Induced Mammary Tumorigenesis and Systemic Inflammation

Given the reported anti-cancer and anti-inflammatory effects of lamivudine (3TC) and other nucleoside RT inhibitors in breast and other types of cancer,^31,32,47,48 we tested if 3TC can prevent the incidence of these early proliferative ductal lesions in the ACI rats. We thus treated groups of rats implanted with placebo or 7.5 mg E2 pellets with 3TC administered over the 6-month period in drinking water at 5 or 15 mg/mL concentrations, which approximately corresponds to human therapeutic doses used for treatment of HIV-infected patients (300 mg 3TC per day, 5–8 μM in plasma).⁴⁹ Whole mount analysis of mammary gland morphology showed a striking difference in rats implanted with 7.5 mg pellets and treated with 15mg/mL 3TC compared to no-3TC treatment controls. While the latter displayed extensive swelling of the gland and distortion of the lobular-ductal architecture, treatment with 15 mg/mL 3TC substantially reduced the size of the inflamed lesions (Figure 2A). Similar but less pronounced effects were found with 5 mg/mL 3TC (data not shown). 3TC at 15 mg/mL (Figure 2A) or at 5 mg/mL (data not shown) had no morphological effects on the mammary glands of the rats implanted with placebo (no E2) pellets. Also, 3TC showed no toxicity or behavioral changes in treated rats, nor did it affect plasma E2 levels (Figure 2B), indicating that its effects were not due to the reduction of plasma E2 levels at 6 months. These findings were further corroborated by hematoxylin-eosin (H&E) staining of formalin-fixed mammary tissue sections, which showed extensive hyperproliferative pre-invasive intraductal lesions in untreated rats implanted with 7.5 mg E2 pellets (Figure 2C, left panel). In contrast, rats treated with 3TC displayed a clear lumen with an intact outer myoepithelial layer and no sign of intraductal hyperproliferative cells (Figure 2C, right panel). Furthermore, examination of mammary glands by IHC showed that both ER and FOXA1, two major determinants of the estrogen pathway and markers for ER-associated DCIS and invasive cancer,^4,5 were highly upregulated in the ductal lesions of rats implanted with 7.5 mg E2 pellets, with inner luminal mammary epithelial cells displaying mixed cytosolic and nuclear staining for ER (α and β forms) and purely nuclear staining for FOXA1, in contrast to the outer layer of myoepithelial cells positive for smooth muscle actin (SMA) only (Figure 2D, left panels). Most importantly, ER and FOXA1 were almost undetectable in mammary glands of the 3TC-treated rats, although the ducts themselves were enlarged and filled with milky liquid (Figure 2D, right panels).

Considering that estrogens are also known to influence immune and inflammatory processes in health and disease including breast cancer,^50–52 we also tested the effects of E2 and 3TC at the systemic level. Bio-Plex cytokine profiling of key proinflammatory cytokines in plasma of rats that were exposed to E2 for 6 months showed a significant upregulation of IL-1α, IL-1β, IL-6, IL-13, IL-17A and IL-18, compared to the rats implanted with placebo pellets (p < 0.01, paired two-tailed t-test; Figure 3). In contrast, treatment with 3TC reduced all cytokines back to the placebo levels, except for IFN-γ which was upregulated by more than 10-fold compared to no-3TC rats implanted with 7.5 mg E2 pellets (p < 0.001, paired two-tailed t-test; Figure 3), suggesting potential activation of a functional T cell response. Together, these results indicate that 3TC counteracts E2-driven inflammation of the mammary gland, restores close-to-normal lobular-ductal architecture, reduces ER and FOXA1 expression and diminishes plasma levels of proinflammatory cytokines while upregulating IFN-γ.

Figure 3 Treatment with 3TC blocks systemic inflammation. Bio-Plex cytokine profiling of plasma from rats implanted with placebo or 7.5 mg E2 pellets and treated with 3TC or no 3TC in drinking water for 6 months. Data presented as boxplots with mean ± SD. Comparisons made to the placebo control rats. *p < 0.05, **p < 0.01, unpaired two-tailed t-test.

ERE RNA Expression in Rat Mammary Glands and PBMCs is Activated by E2 and Blocked by 3TC

Given a reported efficacy of 3TC against L1- and HERV-K-encoded RTs,^25,26,31,53 we next examined a possibility that its anti-tumorigenic effects may be associated with disrupting expression or activities of L1, HERV-K or other EREs. Indeed, whole transcriptome sequencing (RNAseq) of mammary epithelial cells from rats implanted with 7.5 mg E2 pellets compared to placebo control rats showed upregulation of multiple ERE families, including SINEs (Alu, B2 and B4), L1, ERVL-MaLR and ERV-K (Figure 4A), whose activation in other rodent models was associated with tumorigenesis, inflammation, drug resistance, senescence and aging.^31,33,54 Strikingly, treatment with 3TC diminished their expression levels (Figure 4A), most likely as a result of 3TC blocking RT activity and ERE propagation via reverse transcription, although indirect effects caused by intrinsic anti-inflammatory properties of 3TC cannot be excluded (see also Discussion). Coincident with the reduced ERE expression in mammary epithelial cells from 3TC-treated vs no-3TC treatment control rats implanted with E2 pellets, gene set enrichment analysis (GSEA) revealed downregulation of numerous proinflammatory pathways including IL2-STAT5, IL-6-JAK-STAT3, reactive oxygen species and allograft rejection (normalized enrichment score (NES) > 1.5, false discovery rate [FDR]-adjusted p value < 0.05), while UV response/DNA repair, epithelial-to-mesenchymal transition (EMT) and bile acid metabolism pathways were upregulated (Figure 4B), indicative of reduced inflammation and morphological and metabolic shifts in mammary glands from rats implanted with E2 pellets and treated with 3TC.

Figure 4 ERE RNA expression and inflammatory responses in rat mammary glands and PBMCs are activated by E2 and blocked by 3TC. (A) Representation of ERE families expressed in mammary glands of rats implanted with E2 or placebo pellets and treated with 3TC or left untreated. Comparisons made to the placebo rats; p < 0.05, Wilcoxon rank sum test. (B) GSEA of hallmark pathways in E2+3TC vs E2 samples from (A). Upregulated (red) and downregulated (blue) pathways are ranked by normalized enrichment score; FDR < 0.25, adjusted p < 0.05. (C) Representation of ERE families expressed in PBMCs from rats implanted with E2 or placebo pellets and treated with 3TC or left untreated. Comparisons made to the placebo rats; p < 0.05, Wilcoxon rank sum test. (D) GSEA of hallmark pathways in E2+3TC vs E2 samples from (C). Upregulated (red) and downregulated (blue) pathways are ranked by normalized enrichment score; FDR < 0.25, adjusted p < 0.05. (E) GSEA of immune cell marker signatures in PBMCs from (C). (F) Heatmap and hierarchical clustering of marker genes from the selected immune cell populations from (E). Expression is plotted as library-size normalized counts calculated with DESeq2 Bioconductor software. Scale bar on the heatmap indicates the range of normalized values.

We then tested if E2 and 3TC treatments may affect blood-circulating immune cells, considering that ERα and ERβ are expressed by virtually all types of immune cells, including T cells, B cells, natural killer (NK) cells, monocytes and dendritic cells.⁵⁵ Similar to changes observed in mammary glands, RNAseq analysis revealed upregulation of SINEs, L1, ERVL-MaLR and ERV-K in PBMCs from rats implanted with E2 pellets, and their downregulation upon treatment with 3TC (Figure 4C). Furthermore, GSEA analysis of these PBMCs showed that transcripts associated with inflammatory response, TNFα and NFkB signaling and early estrogen response were highly upregulated in PBMCs from rats implanted with E2 pellets compared to placebo control rats, and their expression levels were reduced or completely abolished by treatment with 3TC (Figure 4D and Figure S1A and B). Downregulation of proinflammatory and estrogen-induced pathways in 3TC-treated rats was also associated with upregulation of gene expression programs associated with proliferation (MYC targets, E2F targets, mitotic spindle) and PI3K-AKT and mTOR signaling (Figure 4D and Figure S1C), suggesting an intriguing possibility that 3TC may reverse the E2 effects by reducing ERE expression and the associated inflammation, which, in turn, may boost proliferation in PBMCs from E2-implanted rats that will require further functional studies to verify.

To identify the most affected immune cell populations, we performed immune cell deconvolution analysis of placebo vs E2 and placebo vs E2+3TC PBMCs using the GSEA software with the CellMarker signature annotations for selective immune populations.⁴² This revealed substantial transcriptional reprogramming of multiple immune cell populations in 3TC-treated rats compared to the no-3TC treatment control, including a reduction of dendritic cell signatures and an increase in signatures of CD4⁺ and CD8⁺ T cells (NES > 1.5, FDR p adj < 0.05), including activated and effector memory T cell subpopulations (Figure 4E and F). Given the observed increase in plasma IFN-γ levels in 3TC-treated rats (Figure 3) and upregulation of hallmark genes critical for T cell activation (eg, IL2RB, LCK, LCP2 and CCL5; Figure S1A), these results suggest that directly or indirectly 3TC treatment may enhance T cell activation, albeit small size of animal cohorts precludes definitive conclusions and needs further validation. Overall, these findings demonstrate that ERE RNA expression in rat mammary glands and PBMCs is induced by E2 and mitigated by 3TC, paralleled by suppression of inflammatory and estrogen-driven gene expression programs and by transcriptional reprogramming of multiple PBMC populations.

ER-Dependent and ER-Independent Expression of L1 and HERV-K in Human BC Cell Lines

To examine if L1 and HERVs are expressed in human BC, we used a panel of human cell lines representative of the main BC subtypes, including the ER and progesterone receptor (PR)-positive (MCF7, T47D and ZR-75-1), human epidermal growth factor receptor 2 (HER2)-positive (SK-BR3) and triple-negative (MDA-MB-231, MDA-MB-468 and BT549). Compared to non-transformed mammary epithelial MCF10A cells, RT-PCR analysis revealed elevated expression of L1 RNAs encoding L1 ORF1 and ORF2 proteins in all BC cell lines, with MCF7, T47D and MDA-MB-231 cells exhibiting the highest expression levels (Figure 5A). In contrast, HERV-K RNAs encoding envelope (env), gag and protease (pro) proteins were highly expressed in MDA-MB-231 and MDA-MB-468 cell lines but less so in MCF10A, MCF7, T47D and SK-BR3 cells (Figure 5B). At the protein level, immunoblotting analysis revealed that L1 ORF1p was exclusively expressed in the ER- and PR-positive MCF7 and T47D cell lines, in contrast to HERV-K gag and env proteins which were variably expressed across different cell lines (Figure 5C).

Figure 5 ER-dependent and -independent expression of L1 and HERV-K in human breast cancer cell lines. (A and B) RT-PCR detection of L1 ORF1 and L1 ORF2 (A) and HERV-K (B). Values normalized to 18S RNA ±SD. Representative results from one of three experiments performed in triplicate are shown. Fold change relative to the respective RNAs in MCF10A cells; *p < 0.05, **p < 0.01, ***p < 0.001, unpaired two-tailed t-test. (C) Immunoblotting analysis of L1 ORF1p, HERV-K gag and HERV-K7 envelope protein expression in a panel of breast cancer cell lines. (D) Immunoblotting analysis of L1 ORF1p and other ERE protein expression in T47D and MCF7 cells starved for E2 for 24 h (-) and then stimulated with increasing concentrations of E2 in the absence or presence of 3TC, as indicated. GAPDH was used as a loading control in (C and D). (E and F) RT-PCR detection of L1 ORF1 and HERV-K RNAs in T47D (E) and MCF7 (F) cell lines starved for E2 for 24 h and then stimulated with E2 (1 nM) and progesterone (PG; 1 nM) in the absence or presence of 3TC (600 µM), as indicated. Values normalized to 18S RNA ±SD. Representative results from one of three experiments performed in triplicate are shown. Fold change relative to the respective RNAs in E2-depleted cells; *p < 0.05, **p < 0.01, ***p < 0.001 unpaired two-tailed t-test.

Given selective L1 ORF1p expression in the ER⁺ BC cell lines, we next examined whether its expression could be estrogen-dependent and/or responsive to 3TC. Consistent with activation of the ER signaling, immunoblotting analysis showed upregulation of phosphorylated/activated form of ERα and FOXA1 in the E2-treated MCF7 and T47D cells, however the effect on L1 ORF1p, HERV-K env and HERV-W1 (Syncytin-1) protein levels was negligible (Figure 5D), suggesting that their expression is largely independent of acute E2 stimulation. There was also no considerable change in the L1 ORF1p or HERV proteins in 3TC-treated MCF7 and T47D cells (Figure S2), concordant with the 3TC’s mode of action directed at blocking ERE-driven reverse transcription but not gene expression or protein synthesis. At the transcriptional level, RT-PCR showed that expression of HERV-K RNAs (but not L1) in T47D was induced by E2 and reduced upon 3TC treatment (Figure 5E), consistent with previously reported HERV-K expression in E2-stimulated T47D cells.^56,57 This phenomenon can also be dependent on the p53 status since no such effect was observed in MCF7 cells (Figure 5F). In contrast to T47D, MCF7 cells express the wild-type p53 protein which is known to restrict ERE expression.^32,58 These results show that L1 and HERV-K RNAs are expressed at different levels in BC cell lines, and HERV-K expression can be induced by E2 and suppressed by 3TC in T47D cells.

HERV and L1 ORF1 Proteins are Highly Expressed in FOXA1+ and ER+ Human BCs

Having observed expression of various ERE RNAs and proteins in ER⁺ and FOXA1⁺ preinvasive ductal lesions in ACI rat mammary glands and in human BC cell lines, we next employed IHC to determine if ERE protein expression is associated with ER and FOXA1 in human BC. Using tissue microarrays (TMAs) constructed from a cohort of primary tumors representative of ER⁺, PR⁺, HER2⁺ and triple-negative BC (Table S3), we found that L1 ORF1p, HERV-K gag, HERV-K7 env and HERV-W1 proteins were co-expressed with ERα and FOXA1, ranging from negative (scored “0”) to strongly positive (“3+”; Figure 6A and Figure S3A). In total, 115 high-quality tumor cores from 65 cases were analyzed and scored for ERα and FOXA1 protein staining intensities, yielding four major subgroups, ER-low/FOXA1-low (n=59), ER-high/FOXA1-low (n=2), ER-low/FOXA1-high (n=21) and ER-high/FOXA1-high (n=33) (Figure S3B). In line with well-documented FOXA1 expression in ER⁺ BC,^4,5 linear regression analysis revealed a significant positive correlation between ER and FOXA1 levels (p = 2.5 x 10⁻¹⁶; with ~94% ER-positive tumors expressing FOXA1 (Figure S3B, C and Table S3). We also noted that ERE proteins including L1 ORF1p, HERV-K gag, HERV-K7 env and HERV-W1 were co-expressed and preferentially detected in tumors with high ER and FOXA1 levels (Figure 6A, B and Figure S3A), albeit none of them alone reached statistical significance (Figure S3D). We thus generated histoscores for the combined ERE protein frequencies and intensities, classifying each core as “ERE-high” or “ERE-low” based on the above or below median histoscores for two or more ERE proteins, respectively. Pairwise comparisons with the ER/FOXA1 subgroups revealed the strongest correlation between ERE-high and ER-high/FOXA1-high tumors (p= 0.03, pairwise post-hoc chi-squared test of independence), while ERE-low tumors correlated with the ER-low/FOXA1-low tumors (p = 0.07; Figure 6C and Table S3). Interestingly, ERE proteins were also highly expressed in ~62% of ER-low/FOXA1-high tumors (Figure 6B), suggesting that FOXA1 may contribute to ERE expression independent of ER. Furthermore, we also found that expression of FOXA1 (but not ER), HERV-K7 env and HERV-K gag was significantly elevated in tumors with progressively increasing lymph node involvement, from N0 to N3 (Figure 6D and Table S3), suggesting a potential association with metastatic cancer. However, we found no significant correlation with patient outcome or clinical characteristics including tumor stage and histological subtype, which could also be due to a small size of this patient cohort. Together, these findings indicate that L1 ORF1p and HERV proteins are largely co-expressed in ER-positive BC, especially in those with high FOXA1 levels, and may thus represent potential targets for therapeutic interventions.

Figure 6 HERV and L1 ORF1 proteins are highly expressed in FOXA1 and ER-positive human breast cancers. (A) Representative IHC staining for L1/ORF1p, HERV-K Gag, HERV-K7, HERV-W-1, FOXA1 and ER across breast TMA cores grouped by the ER/FOXA1 status indicated below each column. (B) Mosaic plot, distribution of tumors categorized as ERE -high or -low within each ER/FOXA1 group. (C) Correlation matrix plot of pairwise comparisons between ERE expression groups and ER/FOXA1 expression rankings (low to high), quantified using standardized residuals from chi-squared test of Independence. Note that ERE-high tumors are enriched in the ER-high/FOXA1-high category, whereas ERE-low tumors are predominantly ER-low/FOXA1-low. (D) Boxplots of FOXA1, HERV-K gag and HERV-K7 env protein expression (histoscore) in tumors with increasing lymph node involvement (from N0 to N3; NX and NA- undetermined). Comparisons made to N0 stage (no node involvement), with N3 vs N0 showing significance; *p < 0.05, pairwise Wilcoxon rank-sum test with Benjamini-Hochberg correction.

Discussion

Despite well-known BC promoting effects of estrogens and the effectiveness of tamoxifen and other antiestrogens, the primary triggering events of tumor initiation and mechanisms of treatment resistance are not fully understood. The role of estrogens in cancer development has been traditionally viewed in the context of their ability to enhance mammary epithelial cell proliferation and survival through the interactions with the estrogen receptors and downstream enhancers such as FOXA1, leading to transcriptional and epigenomic reprogramming.^2–5 Aberrant estrogen signaling has also been shown to promote DNA damage and genomic instability by generating genotoxic estrogen metabolites and reactive oxygen species.⁵⁹ Additional mechanisms of estrogen-driven breast tumorigenesis include hormonal regulation of the tumor microenvironment and the immune system causing inflammation and immunosuppression.^52,60,61 Using the ACI rat model of spontaneous ER-positive BC as well as human BC cell lines and tissue microarrays, we demonstrate here that at least some of these effects could be attributed to the activation of ERE RNA and protein expression in estrogen-driven BC.

Accumulating evidence demonstrates an important role of EREs, including L1, SINE/Alu and LTR-HERV families, in initiation and progression of BC and other human cancers.^9,12 Taking advantage of a close resemblance of estrogen-dependent mammary tumorigenesis in rats to human BC,³⁶ we show here that prolonged exposure to E2 induced ER⁺ and FOXA1⁺ pre-invasive intraductal hyperproliferations in mammary glands of peripubertal rats, accompanied by significantly increased plasma levels of key proinflammatory cytokines, including IL-1α, IL-1β and IL-6. Furthermore, RNAseq analysis showed upregulation of RNAs from multiple ERE families, including SINEs, L1, ERVL-MaLR, ERV-K and ERV-L, in mammary glands from the E2-treated rats. Surprisingly, these RNAs were also upregulated in PBMCs, which may be due to direct effects of E2 on innate (monocytes, macrophages and NK cells) and adaptive (T and B cells) immune cells, all of which express estrogen receptors.^55,60,62

Although further mechanistic studies are needed, these findings provide support for a model whereby exposure to elevated E2 results in reactivation of diverse ERE species in both mammary epithelial cells and immune cells that, in turn, activates innate immune responses and inflammatory cytokine release, reinforcing ERE expression and inflammation-driven tumorigenesis. Similar to other cancers,^12,13 ERE expression in BC and propagation through reverse transcription and retrotransposition may drive oncogenic transformation by inducing DNA damage, replication stress, mitotic errors and insertional mutagenesis. Moreover, EREs, particularly L1 and HERV-K, have also been implicated in cancer stem cell maintenance,^63,64 epithelial-to mesenchymal transition⁶⁵ and chemoresistance,³³ and may therefore enhance tumorigenic potential and metastatic ability of cancer cells. In immune and non-immune cells, EREs also act as potent modulators of host innate immunity, augmenting immune responsiveness to pathogen infections and environmental stressors, but their prolonged expression leads to chronic inflammation and immunosuppression.^14,24,54,66 In our rat model of estrogen-driven BC, we also saw this association between ERE RNA induction and proinflammatory cytokine production.

The most intriguing part of our study is the ability of lamivudine/3TC to reverse the E2 effects, including local and systemic effects. Specifically, 3TC attenuated E2-stimulated growth of ER⁺ and FOXA1⁺ intraductal hyperplasias in mammary glands, reduced plasma levels of proinflammatory cytokines while inducing IFN-γ and activated T cell signatures in PBMCs of 3TC-treated rats. This was also accompanied by downregulation of ERE RNA levels in both mammary glands and PBMCs from 3TC-treated rats. 3TC, an FDA-approved drug with excellent safety profile and demonstrated efficacy against HIV-1 and hepatitis B infections, was also shown to exhibit anti-neoplastic and anti-inflammatory properties in different cell and mouse models of cancer including breast.^33,47 Furthermore, 3TC and other HIV RT inhibitors demonstrated efficacy as a preventive and adjuvant therapy for treatment of HIV-positive cancer patients⁶⁷ and patients with TP53 mutant colorectal cancer.³² Aside from inhibiting reverse transcription and propagation of L1 and HERV-K in cancerous and senescent cells,^25,26,31 nucleoside RT inhibitors including 3TC were shown to block SINE/Alu-mediated NLRP3 inflammasome activation,⁴⁸ senescence-associated secretory phenotype (SASP) as well as L1-driven DNA damage and age-associated inflammation,^31,68 while having off-target effect on mitochondrial RNA polymerases.⁶⁹ Most likely, both RT-dependent and independent activities of 3TC may be important for counteracting pro-inflammatory and pro-tumorigenic effects of E2 in our rat model. Overall, the protective activity of 3TC against breast tumorigenesis and systemic inflammation in a rat model may provide an incentive for testing as a monotherapy or adjuvant therapy in BC susceptible patients.

The applicability of our findings to human BC was further supported by analyzing ERE RNA and protein expression in BC cell lines and tumor specimens. Importantly, L1 and HERV RNAs were detected in all cell lines analyzed, including ER⁺PR⁺, HER2⁺ and triple-negative TNBCs, and their expression was elevated in cancer cell lines compared to MCF10A non-tumorigenic mammary epithelial cells. At the protein level, L1 ORF1p and HERV-K7 env were preferentially detected in the ER⁺PR⁺ cell lines and were not affected by 3TC, which is expected given selective inhibitory activity of 3TC against enzymatic function of RT. The ongoing reverse transcription and retrotransposition may however be important for ERE RNA expression, given that HERV-K RNA (but not L1) levels were induced by E2 and inhibited by 3TC in T47D cells. This effect was not observed in MCF7 cells, potentially due to expression of the wild type p53 tumor suppressor protein capable of restricting ERE expression.^32,58 Similar findings showing activation of HERV-K expression in T47D cells in response to treatment with E2 and progesterone were first reported 40 years ago,⁵⁶ and were later supported by discovery of OCT4 as a downstream transcriptional factor required for HERV-K expression.⁵⁷ Additional mechanisms supporting ERE expression may include FOXA1, which was previously shown to activate L1 expression and innate immunity in senescent cells.³¹ In our study, FOXA1 levels in BC tissues positively correlated with elevated expression of ERE proteins, including L1 ORF1p, HERV-K gag, HERV-K7 env and HERV-W1/Syncytin-1, and were also associated with more aggressive disease that spread to lymph nodes. In summary, our study demonstrates that activation of ERE expression is one of the earliest events triggered by estrogen during mammary tumor development and provides preclinical evidence that 3TC has a potential to block estrogen-dependent breast tumorigenesis and systemic inflammation and to reactivate adaptive immunity, especially in the ER- and FOXA1- positive breast cancers. A limitation of this study is that it is based on a small sample size, including the rat model (3–5 rats per group) and human BC TMA (65 cases), which may limit statistical power and introduce bias. Future studies should thus involve larger size of clinical specimens and rat cohorts to confirm our findings. Alongside 3TC, other drugs of this class and their combinations with standard chemotherapeutic agents or tamoxifen should also be tested to define their clinical utility.

Conclusions

Our study implicates EREs, especially LINE-1 and HERV-K, in estrogen-driven mammary tumorigenesis and systemic inflammation, and provides evidence for anti-cancer, anti-inflammatory and immunostimulatory activities of 3TC, albeit a contribution of ERE-independent mechanisms to the observed E2 and 3TC effects cannot be excluded. In addition to RNAs, LINE-1 and HERV-K encoded proteins are highly expressed in FOXA1 and ER-positive tumors, supporting their important roles in breast tumorigenesis at both, RNA and protein levels, and their potential use as markers of progressive disease and therapeutic targets.