Purpose: Treatment resistance is the main cause of adverse disease outcome in breast cancer patients. Here, we aimed to investigate common features in tamoxifen-resistant and radioresistant breast cancer, as tamoxifen-resistant breast cancer cells are cross-resistant to irradiation in vitro.
Experimental Design: RNA sequencing of tamoxifen-resistant and radioresistant breast cancer cells was performed and validated by quantitative PCR. Pathways were further investigated in vitro and in breast cancer patient cohorts to establish their relation with treatment resistance.
Results: Both tamoxifen-resistant and radioresistant breast cancer cells had increased expression levels of genes involved in type I IFN signaling compared with nonresistant cells. IFN-stimulated genes (ISG) were induced in a dose-dependent and time-dependent manner after tamoxifen treatment and irradiation. Tamoxifen treatment also led to ssDNA presence in the cytoplasm, which is known to induce expression of ISGs, a phenomenon that has already been described for irradiation. Moreover, in a breast cancer patient cohort, high expression levels of ISGs were found in the primary tumor in around half of the patients. This was associated with a tumor-infiltrating lymphocyte (TIL) expression signature, although the ISGs were also expressed by the tumor cells themselves. Importantly, the expression of ISGs correlated with outcome in breast cancer patients treated with adjuvant tamoxifen or radiotherapy, but not in systemically untreated patients or chemotherapy-treated patients.
Conclusions: Our data indicate that expression of ISGs by tumor cells is involved in acquired, treatment-induced resistance to tamoxifen and radiotherapy, and might play a role in intrinsic resistance via interaction with TILs. Clin Cancer Res; 24(14); 3397–408. ©2018 AACR.
Investigating features of tamoxifen resistance and radioresistance in breast cancer is imperative to find new targets for treatment-resistant breast cancer. Here, we show that tamoxifen-resistant cells also develop radioresistance in vitro. This presents an interesting view on the new development to implement neoadjuvant tamoxifen treatment in breast cancer patients, who might later also receive radiotherapy. Our research also adds to existing knowledge on the relation between IFN signaling and disease outcome in breast cancer patients, a subject much investigated in light of the immunotherapies making the step to the clinic.
Endocrine treatment and radiotherapy are two mainstays in breast cancer treatment. Although breast cancer survival rates are generally high, in a substantial subset of patients, the development of therapy resistance results in adverse outcome (1, 2). We have previously shown that the unfolded protein response/LAMP3/autophagy pathway is involved in resistance to both tamoxifen treatment and radiotherapy (3, 4). Moreover, cross-resistance has been observed in tamoxifen-resistant breast cancer cells in vitro. These are less sensitive to irradiation (5, 6), suggesting that common mechanisms underlie tamoxifen resistance and radioresistance.
Endocrine therapy targets the estrogen signaling pathway, which is essential for tumor cell proliferation in a large percentage of breast cancers. Two major classes of endocrine therapy can be distinguished: aromatase inhibitors and estrogen receptor (ER) modulators, such as tamoxifen (7, 8). Tamoxifen is an ER antagonist, which has been used in breast cancer treatment for several decades (9). Although tamoxifen is effective in blocking proliferation of breast cancer cells, some tumors demonstrate intrinsic or acquired resistance to its action. Absence of the ER leads to intrinsic resistance to tamoxifen (10), whereas tamoxifen resistance in ER-positive breast cancer is caused by growth factor receptors stimulating estradiol-independent estrogen signaling (11). Acquired resistance to tamoxifen involves several different mechanisms, for example, pharmacologic pathways and mutations in the ER (11, 12).
Radiotherapy aims to induce irreparable DNA damage in tumor cells, leading to tumor cell death. Different tumor types display varying degrees of radiosensitivity, and even within a single tumor, radiosensitivity may vary. Three main mechanisms underlie resistance to radiotherapy: intrinsic resistance, tumor proliferation, and hypoxia (13). The patient's genetic makeup (14) and the type of breast cancer (15) are also known to influence radiosensitivity. Moreover, several pathways such as the DNA damage response, cell-cycle regulation, and apoptosis are involved in radioresistance (2, 16).
Here, we describe the activation of the IFN signaling pathway in tamoxifen-resistant and radioresistant breast cancer. We performed RNA sequencing of tamoxifen-resistant and radioresistant breast cancer cells, followed by pathway analyses, in vitro studies, and clinical validation in three human breast cancer patient cohorts. Expression of IFN-stimulated genes (ISG) by tumor cells is involved in treatment-induced acquired resistance and possibly in intrinsic resistance related to tumor infiltrating lymphocytes (TIL).
Materials and Methods
MCF-7 cells obtained in 2011 (LCG Standards) were cultured at 37°C with 5% CO2 in DMEM (Lonza) supplemented with 10% (v/v) fetal bovine serum, 20 mmol/L HEPES, 1× nonessential amino acids, 20 U/mL penicillin, 20 μg/mL streptomycin (all from Gibco), and 2 mmol/L l-glutamine (Lonza). T47D cells purchased in 2015 (LCG Standards) were cultured in RPMI (Lonza) containing identical supplements. All cells were cultured up to 40 passages after thawing. Wild-type cell lines and their resistant progeny were authenticated by short tandem repeat analysis. Mycoplasma tests were performed at regular intervals.
4-hydroxytamoxifen was purchased from Sigma Aldrich (#H7904). Cells were irradiated using an X-Rad 320ix Biological Irradiator (Precision X-Ray) at a dose rate of 3.1 Gy/minute.
Creating resistant cell lines
MCF-7 or T47D cells were cultured with increasing doses of 4-hydroxytamoxifen up to a total dose of 10 μmol/L to acquire tamoxifen-resistant cells (4). To obtain radioresistant cells, wild-type cells were irradiated with 4 Gy every 2 weeks or 4 times 1 Gy a week, in daily fractions, up to total doses of 64 Gy.
Clonogenic cell survival after treatment with 4-hydroxytamoxifen or irradiation was performed as described before (3, 4).
Cell viability assay
MCF-7 cells were seeded in a 96-well plate and allowed to adhere for 24 hours. Then, the cells were treated with different concentrations of doxorubicin (PharmaChemie BV) for 24 hours, after which the cells were allowed to grow for 5 days before analyzing cell viability using Cell Counting Kit 8 (Sigma Aldrich) according to the manufacturer's instructions. Absorption values were measured using an iMark microplate absorbance reader (Bio-Rad Laboratories Inc.).
RNA was isolated using the Total RNA Purification Kit (Norgen Biotek Corp.), with on column DNase treatment (RNase-Free DNase Set #79254), according to the manufacturer's instructions.
RNA from wild-type MCF-7, tamoxifen-resistant MCF-7, and radioresistant MCF-7 cells was used as input for the Illumina TruSeq stranded RNA-seq protocol (Illumina). Subsequently, high-throughput sequencing was performed on an Illumina Next-seq, generating reads of 76 bases, using a paired end protocol. Quality control on the raw fastq files was performed using fastQC. Next, the fastq files were aligned to the human reference genome (GRCh37) using STAR_v2.4.1d (17), and raw read counts per gene were estimated using HTseq (18). Raw read counts were normalized using EdgeR (19).
Reverse transcription of RNA and qRT-PCR was performed as described before (4). The following primers were used: DDX60: FW, 5′-GTCCTTCAAGCAACCCAG-3′; and REV, 5′-ATCTAAATAGCCCTCTTTCACC-3′; STAT1: FW, 5′-CCAAAGGAAGCACCAGAG-3′; and REV, 5′-TCAGACACAGAAATCAACTCAG-3′; OAS1: FW, 5′-ACTATCTCTTGCCAGACACG-3′; and REV, 5′-AGCCACCCTTTACCACCT-3′; IFI6: FW, 5′-CTTGTGCTACCTGCTGCT-3′; and REV, 5′-TTCTTACCTGCCTCCACC-3′; IFI27: FW, 5′-AGTCACTGGGAGCAACTG-3′; and REV, 5′-CTGGCATGGTTCTCTTCTC-3′; HPRT: FW, 5′-TATTGTAATGACCAGTCAACAG-3′; and REV, 5′-GGTCCTTTTCACCAGCAAG-3′ (Biolegio).
Fluorescence microscopy for BrdU staining
MCF-7 cells were seeded in glass chamber slides [Nunc Lab-Tek II (154526); Thermo Fisher Scientific] and incubated with 10 μmol/L 5-Bromo-2′-deoxyuridine [BrdU (#B5002); Sigma Aldrich] for 38 hours (1.5 cell cycles). After washing, cells were treated with 1 μmol/L 4-hydroxytamoxifen. Cells were fixed 10 minutes in acetone at 4°C after 0, 1, and 4 hours. Before staining, cells were rehydrated in PBS for at least 30 minutes, treated with 2 N HCl for 10 minutes, and neutralized with 0.1 mol/L Borax for 10 minutes. Cells were stained with mouse anti-BrdU [B44 (#347580); BD Biosciences], 1:50 in primary antibody diluent (PAD; Bio-Rad Laboratories Inc.). The secondary antibody used was CF-488-conjugated goat anti-mouse IgG [F(ab')2 fragment (#20011); Biotium]. Nuclei were visualized using Hoechst 33342 (1 mg/mL) in a 1:2,000 dilution in PBS. Images were acquired using a Leica DM6000 microscope (Leica).
Patients: Rotterdam cohort
Expression levels of relevant genes were investigated in a cohort of 155 estrogen receptor–positive breast cancer patients, all of whom were treated with first-line tamoxifen after disease recurrence. Additional details and clinical characteristics were described previously (20). Microarray expression data of this cohort were used for a pathway analysis, using BioCarta as pathway database (https://cgap.nci.nih.gov/Pathways/BioCarta_Pathways) and the R-package “globaltest.” Multiple testing correction was performed using the Bonferroni–Holm method. Furthermore, a 1,000× resampling was performed to determine the number of times a randomly selected group of genes of equal size was at least as significant as the true set of genes assigned to a pathway. Pathways were considered significant when both the corrected P value and the resampling probability were below 0.05.
Patients: Kaplan–Meier plotter
To assess the relation between expression of genes of interest and outcome in adjuvant tamoxifen-treated breast cancer patients, we used the Kaplan–Meier plotter at http://kmplot.com/analysis/. The characteristics of the included patients have been described previously (21). Analysis was restricted on the basis of received therapy, and patients were assessed for relapse-free survival. We selected patients receiving endocrine treatment alone (n = 867, 85.6% ER+), (neo-)adjuvant chemotherapy treatment alone (n = 602, 25.6% ER+) or no systemic treatment (n = 1010, 49.8% ER+). Patient samples were grouped on the basis of the expression of the desired gene of interest (Jetset best probeset) using the median cut-off value.
Patients: Radboudumc cohort
In this cohort, the correlation between genes of interest and sensitivity to radiotherapy was assessed. Patients with resectable breast cancer that did not receive adjuvant systemic therapy and had a follow-up of at least 5 years or an event before that time, were selected from a previously described cohort (22, 23). Comparison of the patients that were excluded from the original cohort by this latter criterion indicated no significant differences in stage, grade, or histology. Postoperative radiotherapy was given to the thoracic wall after an incomplete resection, infiltration of the chest or skin wall, or in case of nodal involvement. Parasternal irradiation was applied when the tumor was medially localized. Finally, whole breast irradiation was performed as part of breast-conserving treatment. Patients who received radiotherapy as part of their primary treatment (n = 243) were compared with patients who did not (n = 123).
Statistical analysis of clonogenic survival assays, cell survival assays, and qRT-PCR experiments were performed using GraphPad Prism software with two-sided Student t test or ANOVA where appropriate. In particular, to analyze colony-forming assays for radiotherapy response, a linear–quadratic model for curve fitting [S = exp (αD-βD2)] was used and α and β compared between the curves. Enriched pathways in the gene sets found after RNA sequencing were analyzed using the PANTHER overrepresentation test (24, 25), including a Bonferroni correction for multiple testing. Pathways with a corrected P < 0.05 were considered significant. SPSS software was used to carry out all further statistical tests (SPSS Inc.). Mann–Whitney U and χ2 tests were used to analyze correlations between genes of interest and patient characteristics in the Rotterdam cohort. Relations between genes of interest and outcome in this cohort were assessed by Cox regression analysis, using the Breslow method for ties. In the Radboudumc cohort, survival curves were generated using the Kaplan–Meier method and differences in survival were tested with a Breslow test.
Tamoxifen-resistant breast cancer cells are cross-resistant to irradiation
ER-positive MCF-7 breast cancer cells that were cultured to resistance to 10 μmol/L 4-hydroxytamoxifen (MCF-7TAM; ref. 4) did not show any decrease in survival when treated with up to 10 μmol/L of 4-hydroxytamoxifen (Fig. 1A). Wild-type MCF-7 cells (MCF-7WT) showed a significant decrease in survival, with a surviving fraction of 0.21 after treatment with 1 μmol/L of 4-hydroxytamoxifen (P = 0.02 compared with MCF-7TAM), and surviving fractions of 0.26 and 0.15 after 5 and 10 μmol/L of 4-hydroxytamoxifen, respectively (P = 0.006 and P = 0.002).
Next, the radiosensitivity of MCF-7TAM was examined. In a clonogenic survival assay, MCF-7TAM displayed increased survival after irradiation compared with MCF-7WT (Fig. 1B). The surviving fraction after 6 Gy was 0.01 for MCF-7WT compared with 0.09 for MCF-7TAM, and the surviving fraction after 8 Gy was 0.001 for MCF-7WT compared with 0.02 for MCF-7TAM. There was a significant difference between the curves fitted according to the linear–quadratic model (P < 0.0001). The β-component significantly differed: for the curve fitted for MCF-7WT, it was 0.08 [95% confidence interval (CI), 0.06–0.09], whereas for the curve fitted for MCF-7TAM, the β-component was 0.02 (95% CI: 0.003–0.04).
MCF-7 cells that were independently cultured to tamoxifen resistance also survived doses up to 10 μmol/L 4-hydroxytamoxifen in a clonogenic survival assay (Supplementary Fig. S1A). These tamoxifen-resistant cells similarly exhibited increased survival after irradiation compared with wild-type MCF-7 cells (Supplementary Fig. S1B).
Tamoxifen-resistant and radioresistant breast cancer cells are not resistant to doxorubicin chemotherapy
In a clonogenic survival assay, MCF-7 cells that were cultured to radioresistance using 13 fractions of 4 Gy (MCF-7RT) showed an increased survival after treatment with increasing radiation doses compared with MCF-7WT (Fig. 1C). The surviving fraction after 4 Gy was 0.04 for MCF-7WT compared with 0.10 for MCF-7RT, and the surviving fraction after 6 Gy was 0.003 for MCF-7WT compared with 0.02 for MCF-7RT. Curves fitted according to the linear–quadratic model differed significantly (P < 0.0001), although this could not be attributed to either the α- or β-component. This phenotype was replicated in MCF-7 cells that were independently cultured to radioresistance (Supplementary Fig. S1C).
Both MCF-7TAM and MCF-7RT did not show an altered response to doxorubicin treatment compared with MCF-7WT (Fig. 1D). Because MCF-7TAM and MCF-7RT were not broadly resistant, it is more likely that a specific mechanism causes tamoxifen resistance and radioresistance, which is not involved in resistance to chemotherapy. This observation was repeated with MCF-7 cells independently cultured with up to 5 μmol/L 4-hydroxytamoxifen and MCF-7 cells that had been irradiated 16 times with 4 Gy. Again, no altered responses were observed in the resistant cell lines compared with wild-type MCF-7 cells (Supplementary Fig. S1D).
Tamoxifen-resistant and radioresistant breast cancer cells show increased expression of ISGs
To evaluate common features of resistance in tamoxifen resistance and radioresistance, gene expression analysis of MCF-7WT, MCF-7TAM, and MCF-7RT was performed using RNA sequencing. We focused on genes that had at least 2-fold increased expression levels in either of the resistant cells compared with MCF-7WT. To avoid false positive hits due to low read coverage, genes that had less than 10 normalized read counts in all three samples (MCF-7WT, MCF-7TAM, and MCF-7RT) were excluded. Following these guidelines, expression levels of 584 genes were increased in MCF-7TAM compared with MCF-7WT, whereas in MCF-7RT, expression levels of 302 genes were increased (Fig. 2A). An overrepresentation test was performed using the PANTHER Classification system (24, 25) to elucidate biological processes (Gene Ontology, GO) that were enriched within the gene set with increased expression levels in MCF-7TAM or MCF-7RT. In both types of resistant cells, the significantly enriched pathways (corrected P < 0.05) were related to the GO terms “response to IFNα/β,” “type I IFN signaling pathway,” “IFNγ-mediated signaling pathway,” “negative regulation of viral genome replication,” and “defense response to virus” (Fig. 2A). Furthermore, 94 genes were increased in both MCF-7TAM and MCF-7RT (Supplementary Table S1). The only pathways that were significantly enriched in these 94 genes (corrected P < 0.05) were related to the GO terms “response to IFNα/β/γ,” “type I IFN signaling pathway,” “negative regulation of viral genome replication,” “defense response to virus,” and “negative regulation of cellular process” (Fig. 2A).
To further investigate the role of IFN-related signaling in tamoxifen resistance and radioresistance, we chose five ISGs, based on literature (26, 27), to assess in in vitro experiments: nucleic acid recognition protein DDX60, signal transducer STAT1, and downstream effectors OAS1, IFI6, and IFI27. These five genes had increased mRNA expression levels in both types of resistant cells (Supplementary Table S1) and represent various steps of the IFN signaling pathway.
First, the increased expression levels of DDX60, STAT1, OAS1, IFI6, and IFI27 in MCF-7TAM and MCF-7RT were validated by qRT-PCR (Fig. 2B). In independent wild-type MCF-7 cells treated with increasing doses of 4-hydroxytamoxifen up to 5 μmol/L or weekly irradiated with 4 Gy to a total of 20 Gy, the phenotype was reproduced (Supplementary Fig. S1E). The different dosing regimens used here might explain the quantitative differences with Fig. 2B. Similarly, another ER-positive breast cancer cell line, T47D, was cultured with increasing doses of 4-hydroxytamoxifen up to 5 μmol/L, or daily irradiated with 1 Gy to a total of 38 Gy. These cells also showed increased expression levels of DDX60, STAT1, OAS1, IFI6, and IFI27 compared with wild-type T47D cells (Fig. 2C).
Tamoxifen treatment and irradiation induce ISG expression in vitro
The increased expression levels of DDX60, STAT1, OAS1, IFI6, and IFI27 in MCF-7TAM and MCF-7RT suggest that ISGs could be part of a survival mechanism induced after treatment, ultimately leading to a resistant phenotype. To investigate this, expression levels of DDX60, STAT1, OAS1, IFI6, and IFI27 were measured after a single dose of tamoxifen or irradiation. MCF-7WT was treated with 1 or 10 μmol/L 4-hydroxytamoxifen, and gene expression levels were measured after 24 hours. Expression levels of all five genes were increased 24 hours after tamoxifen treatment in a dose-dependent manner (Fig. 3A). Expression levels of DDX60, STAT1, OAS1, IFI6, and IFI27 were also measured 24 and 48 hours after irradiating MCF-7WT with 4 Gy. DDX60, OAS1, IFI6, and IFI27 showed a time-dependent increase after irradiation, whereas STAT1 was only slightly increased 48 hours after irradiation (Fig. 3B).
Tamoxifen treatment induces cytoplasmic ssDNA
An earlier study showed that increased ISG expression levels after irradiation were caused by ssDNA in the cytoplasm (28). We hypothesized that the increased expression levels of ISGs after tamoxifen treatment might also be caused by cytoplasmic ssDNA. MCF-7WT was labeled with BrdU for 38 hours (1.5 cell cycle), and then treated with 1 μmol/L 4-hydroxytamoxifen. After 0, 1, and 4 hours of tamoxifen treatment, BrdU was visualized. One hour after 4-hydroxytamoxifen treatment, we observed BrdU foci in the cytoplasm, which were more profound 4 hours after treatment (Fig. 3C). We semiquantitatively scored the cells as containing “low,” “medium,” or “high” levels of cytoplasmic ssDNA (Supplementary Fig. S2). The majority of cells in the untreated sample contain hardly any foci, whereas after 1 or 4 hours of tamoxifen treatment, more foci are observed in the cytoplasm of the cells (Fig. 3D). Because this particular BrdU antibody only recognizes BrdU incorporated in ssDNA (29), and the signal of the dsDNA intercalator Hoechst was absent in the cytoplasm, the foci represent ssDNA. This indicates that tamoxifen treatment, like irradiation (28), can induce cytoplasmic ssDNA, which in turn is known to induce ISG expression.
Expression levels of ISGs are correlated in patients with breast cancer and are associated with a TIL signature
Several cohorts and specific subgroups of patients were analyzed to investigate the relation of ISG expression with treatment sensitivity. In the Rotterdam cohort, consisting of ER-positive breast cancer patients (n = 155) who were treated with first-line tamoxifen for advanced breast cancer (Supplementary Table S2; ref. 20), we analyzed the relation of ISG expression with tamoxifen sensitivity.
First, the 25 genes that showed the highest increase in expression levels in both MCF-7TAM and MCF-7RT, according to RNA sequencing (Supplementary Table S1), were analyzed in primary tumors of these patients. A correlation analysis revealed that expression levels of specifically those 18 genes that play a role in antiviral type I IFN signaling were highly correlated in the primary tumors from these patients (Fig. 4A; Supplementary Table S1). On the basis of the relative expression of those 18 ISGs, the patients could be clustered in a group with high expression of ISGs (ISGHIGH) and a group with low expression of ISGs (ISGLOW, Fig. 4B). To investigate mechanisms responsible for this a priori increase in expression of ISGs, we performed a Biocarta pathway analysis. Fourteen pathways were differentially expressed in the ISGHIGH patients compared with the ISGLOW patients (Supplementary Fig. S3). In 5 of these 14 pathways, there is a remarkably large contribution of STAT1 alone. Because STAT1 is a major regulator in the IFN signaling pathway, we investigated the expression of STAT1 in this cohort. The ISGHIGH patients displayed a higher expression of STAT1, compared with ISGLOW patients (P < 0.0001; Fig. 4C).
There was no difference in clinical parameters (age, T/N/M staging, grade, histology, and subtype), or in response to tamoxifen treatment for advanced breast cancer (Supplementary Table S2) between the ISGHIGH and ISGLOW patients, and only in the case of P2RY6 was there a correlation between gene expression and treatment response in this cohort (P = 0.03; Supplementary Fig. S4).
Because several of these patients showed high ISG expression levels in their primary tumor prior to any treatment, the increased expression of ISGs could not be caused by the treatment itself, as we showed in vitro. We hypothesized that the high expression of ISGs might be related to immune infiltration of the tumor. A previously identified TIL expression signature (30) was investigated in the ISGHIGH and ISGLOW patients. Indeed, an association was found, showing an increased TIL level in the ISGHIGH patients (P < 0.0001; Fig. 4D).
Increased ISG expression in the bulk tumor tissues of ISGHIGH patients could be due to expression of these genes in TILs, instead of in tumor cells as we showed in vitro. Therefore, IHC expression patterns of DDX60, STAT1, OAS1, IFI6, and IFI27 were investigated in breast cancer tissue via the publicly available database Protein Atlas (31). Besides expression in the TILs, identified based on their morphologic characteristics, these five ISGs were found to be highly expressed by the tumor cells themselves (Supplementary Fig. S5), which is in line with our in vitro data.
High ISG expression levels are correlated with worse outcome in breast cancer patients treated with adjuvant tamoxifen
The online available database Kaplan–Meier plotter (21) was used to identify the relation between ISG expression and relapse-free survival in patients treated with adjuvant tamoxifen. The correlation between the five ISGs that were investigated in vitro and outcome was first assessed. High expression levels of DDX60 and STAT1 were significantly associated with a shorter relapse-free survival in tamoxifen-treated patients only (P = 0.03 and P = 0.002), but not in patients who did not receive adjuvant systemic treatment or who received chemotherapy (Fig. 5A). In addition, the effects of the 14 other ISGs that showed correlation in patients with advanced breast cancer (Supplementary Table S1) were investigated. Seven of these (IFI44L, IFI44, P2RY6, MX1, IFIT1, OAS3, and LGALS3BP) were also significantly correlated with relapse-free survival in breast cancer patients treated with adjuvant tamoxifen (P = 0.03, P = 0.03, P = 0.02, P = 0.001, P = 0.008, and P = 0.04), but not in patients who did not receive any adjuvant systemic treatment or who received chemotherapy (Fig. 5A).
High expression levels of ISGs are correlated with worse outcome in breast cancer patients treated with radiotherapy
Finally, we investigated the association of DDX60, STAT1, OAS1, IFI6, and IFI27 expression with outcome in a cohort of breast cancer patients (n = 366) who had not received systemic therapy and had at least 5-year follow-up or an event before that. To assess the association of ISGs with radiosensitivity, the cohort was analyzed separately for those who did or did not receive radiotherapy, as reported earlier in a smaller cohort (22). In patients who did not receive radiotherapy as part of their primary treatment (n = 123), none of the five tested ISGs exhibited any relation with disease-free survival (Fig. 5B). However, patients treated with radiotherapy (n = 243) and with upper tertile expression levels of DDX60, STAT1, OAS1, IFI6, and IFI27 had worse disease-free survival, which reached significance in three of five genes (P = 0.08, P = 0.04, P = 0.05, P = 0.06, and P = 0.01, respectively).
Here, we show that the IFN signaling pathway is activated in tamoxifen-resistant and radioresistant breast cancer. After validating the cross-resistance for irradiation in tamoxifen-resistant breast cancer cells in vitro, we observed increased expression levels of genes related to IFN signaling in both tamoxifen-resistant and radioresistant breast cancer cells. Expression of ISGs was induced by tamoxifen treatment or irradiation in vitro, potentially triggered by the presence of cytoplasmic ssDNA, which was observed after tamoxifen treatment and in an earlier study shown after irradiation (28). We hypothesize that prolonged treatment with tamoxifen or repeated fractions of irradiation leads to the increased ISG expression found in tamoxifen-resistant and radioresistant breast cancer cells. In a breast cancer patient cohort, we investigated the 25 genes whose expression levels were most increased in both tamoxifen-resistant and radioresistant breast cancer cells. Specifically, the expression of genes related to IFN signaling was highly correlated in breast cancer patients. On the basis of this set of correlated genes, the patients could be clustered in groups with high and low ISG expression levels, which in turn significantly correlated with STAT1 expression and a TIL expression signature in these patients. Of note, in tumor samples, ISGs were expressed by tumor cells, instead of being solely expressed by TILs. This suggests that increased ISG expression levels by the tumor itself could influence the antitumor immune response. Finally, high ISG expression in breast cancer patients was correlated to worse outcome in patients treated with adjuvant tamoxifen or radiotherapy. Although our in vitro data point to a role of IFN signaling in acquired treatment resistance, our patient data indicate that the same pathway may also be involved in intrinsic resistance.
Although the decreased radiosensitivity of tamoxifen-resistant breast cancer cells has been observed before (5, 6), and we have previously identified mechanisms that play a role in both tamoxifen resistance and radioresistance (3, 4), this is the first nontargeted approach to find common features of resistance in these two types of treatment resistance in breast cancer. The only pathways that were significantly increased in both tamoxifen-resistant and radioresistant breast cancer cells were related to IFN signaling. Similarly, an immune-associated gene signature is increased in tamoxifen-resistant mice bearing a targeted overexpression of the ER (32), and a comparable signature is a predictive marker for radiation in breast cancer (33). Our data confirm and extend the data on IFN signaling as an important pathway that tamoxifen-resistant and radioresistant breast cancer cells share. However, we have not assessed the direct cause of the cross-resistance for radiation in tamoxifen-resistant breast cancer cells. Previous research (34) showed that there is differential expression of DNA damage (repair) genes in tamoxifen-resistant breast cancer cells after irradiation, which may explain the cross-resistance we and others observe.
In this study, we chose five genes to represent the IFN signaling pathway, spanning the entire IFN response. DDX60 is involved in recognition of viral RNA/DNA (35), STAT1 is an important mediator directly downstream of the IFN receptor (36), and OAS1, IFI6, and IFI27 are all IFN-induced effectors that perform different functions in mediating the IFN response. OAS1 is involved in inhibiting viral replication (37), whereas IFI6 and IFI27 play a role in apoptosis (38). All five were increased after a single dose of tamoxifen treatment or irradiation. It had already been established that tamoxifen induces mRNA expression of ISGs (39), and that DNA damage caused by radiation can induce an IFN response (33, 40). However, this had not yet been investigated in the context of treatment resistance. Earlier studies showed that ISG expression after irradiation is induced by cytoplasmic DNA, mediated by DNA sensors (28, 41). We show here that tamoxifen treatment also induced cytoplasmic ssDNA, which is likely to cause the increased mRNA expression of ISGs after tamoxifen treatment. Although the exact mechanism mediating this is currently unknown, tamoxifen is known to have genotoxic effects (42), which might lead to cytoplasmic DNA. DNA sensors such as STING and DDX60 could subsequently mediate an IFN response. Our data suggest that treatment induction of ISG expression leads to the increased ISG expression levels found in tamoxifen-resistant and radioresistant cells. This is illustrated by the fact that both MCF-7 cells cultured for longer periods of time with 10 μmol/L tamoxifen, as well as those that were cultured with only 5 μmol/L tamoxifen, showed increased mRNA levels of ISGs. Similarly, different irradiation schedules increase ISG expression levels in MCF-7 and T47D cells. Prolonged exposure to tamoxifen or repeated fractions of irradiation, as are applied clinically, could provide a survival advantage for a subset of cells and contribute to the development of therapy resistance.
Cross-resistance for radiotherapy induced by tamoxifen treatment could pose a problem for breast cancer patients, as neoadjuvant endocrine treatment before surgery is now considered a viable treatment option. In several ongoing clinical trials, patients with ER-positive breast cancer are given 3 or 4 months of endocrine treatment before surgical removal of the tumor (43). When radiotherapy is subsequently applied after surgery, these patients could face an increased risk of radioresistance induced by endocrine treatment. In addition, patients that have been treated with adjuvant endocrine treatment might respond poorly to radiotherapy when applied after recurrence to, for instance, painful bone metastases. Our results suggest that these patients might benefit from additional treatment specifically targeting the IFN signaling pathway. This approach has been tested in vitro in the case of aromatase inhibitor resistance. Aromatase inhibitor–resistant breast cancer cells show increased expression levels of ISGs in vitro, and targeting these genes resulted in a sensitization of the resistant cells (27). A similar approach should be further investigated for tamoxifen resistance.
The interaction between IFN signaling and breast cancer treatment is not unambiguous. The ISG signature published by Weichselbaum and colleagues (33) is not only predictive for poor outcome after radiotherapy, but also after chemotherapy. However, we show here that the cross-resistance for irradiation in tamoxifen-resistant breast cancer cells did not extend to doxorubicin in vitro and that ISGs were not linked to outcome after chemotherapy in a breast cancer patient cohort. Moreover, in another study, chemoresponsive, patient-derived, ER-negative breast cancer xenografts show activation of the IFN signaling pathway (44). These opposite results found in different studies highlight the diverse effects that IFN signaling can have on treatment outcome. IFN signaling can both stimulate and suppress immune functions, and also has immune-independent effects (45). This is illustrated by the fact that in earlier studies IFN was administered to breast cancer patients treated with tamoxifen to prevent resistance (46–48). However, as IFN induces expression of ISGs (36), our data suggest that administration of IFN could promote resistance. This complex interplay between ISG expression and breast cancer treatment, its effects on tumor cells and immune cells in the tumor microenvironment, and the relation with patient outcome need to be further investigated before translating these data to the clinic.
Patient and tumor characteristics are likely to impact the relation between ISG expression, treatment, and outcome. We have investigated the association between ISG expression and outcome in a set of cohorts, in which patient characteristics differ greatly. The Rotterdam cohort consists of all ER-positive breast cancer patients, who have not been systemically treated for their primary disease (20). They have all had a recurrence of the disease, for which they have been treated with first-line tamoxifen. In another patient set, using an online available database (21), we compared systemically untreated patients with patients receiving adjuvant tamoxifen or (neo-)adjuvant chemotherapy. Although the group that received endocrine treatment was almost completely ER-positive, only 25.6% of the patients receiving chemotherapy and 49.8% of the systemically untreated patients were ER-positive. Moreover, immune infiltration in the tumor microenvironment differs between breast cancer subtypes, and this influences disease outcome (49).
The presence of TILs in the tumor microenvironment is generally associated with a better outcome in breast cancer (50), although this is highly dependent on breast cancer subtype. TILs inducing a suppressive immune phenotype may result in worse disease outcome (49). We show that ISG expression levels in primary breast tumors correlated with a TIL expression signature. An earlier study already found that STAT1 expression is correlated to macrophage infiltration in breast cancers and a worse outcome (51). This also raises the question whether ISG expression by tumor cells increases the presence of TILs in the tumor microenvironment, or whether TILs induce ISG expression in the tumor cells. An interesting approach to investigate this further would be to directly compare TIL percentages in breast tumors to ISG expression of the bulk tumor. The effect that IFN-related signaling by tumor cells has on immune cells in the tumor microenvironment could then be further investigated.
In conclusion, we show here that resistance to tamoxifen treatment and irradiation share a common feature: activation of the IFN signaling pathway. We show that this pathway plays a role in acquired therapy resistance, as it is induced by treatment in vitro. It might also be involved in intrinsic therapy resistance, based on the increased expression of ISGs in a substantial amount of breast cancer patients in their primary tumor before treatment and an association with a TIL expression signature. The interaction between IFN signaling by tumor cells and immune cell infiltration in the tumor microenvironment should be studied in detail before potential targeted therapies toward this pathway can be considered.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: A.E.M. Post, J. Bussink, F.C.G.J. Sweep, P.N. Span
Development of methodology: A.E.M. Post, P.N. Span
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.E.M. Post, A. Nagelkerke, J.W.M. Martens, P.N. Span
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.E.M. Post, M. Smid, A. Nagelkerke, J.W.M. Martens, J. Bussink, F.C.G.J. Sweep, P.N. Span
Writing, review, and/or revision of the manuscript: A.E.M. Post, M. Smid, A. Nagelkerke, J.W.M. Martens, J. Bussink, F.C.G.J. Sweep, P.N. Span
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): P.N. Span
Study supervision: J. Bussink, F.C.G.J. Sweep, P.N. Span