Type I interferon/IRF7 axis instigates chemotherapy-induced immunological dormancy in breast cancer

Neoadjuvant and adjuvant chemotherapies provide survival benefits to breast cancer patients, in particular in estrogen receptor negative (ER−) cancers, by reducing rates of recurrences. It is assumed that the benefits of (neo)adjuvant chemotherapy are due to the killing of disseminated, residual cancer cells, however, there is no formal evidence for it. Here, we provide experimental evidence that ER− breast cancer cells that survived high-dose Doxorubicin and Methotrexate based chemotherapies elicit a state of immunological dormancy. Hallmark of this dormant phenotype is the sustained activation of the IRF7/IFN-β/IFNAR axis subsisting beyond chemotherapy treatment. Upregulation of IRF7 in treated cancer cells promoted resistance to chemotherapy, reduced cell growth and induced switching of the response from a myeloid derived suppressor cell-dominated immune response to a CD4+/CD8+ T cell-dependent anti-tumor response. IRF7 silencing in tumor cells or systemic blocking of IFNAR reversed the state of dormancy, while spontaneous escape from dormancy was associated with loss of IFN-β production. Presence of IFN-β in the circulation of ER− breast cancer patients treated with neoadjuvant Epirubicin chemotherapy correlated with a significantly longer distant metastasis-free survival. These findings establish chemotherapy-induced immunological dormancy in ER− breast cancer as a novel concept for (neo)adjuvant chemotherapy activity, and implicate sustained activation of the IRF7/IFN-β/IFNAR pathway in this effect. Further, IFN-β emerges as a potential predictive biomarker and therapeutic molecule to improve outcome of ER− breast cancer patients treated with (neo)adjuvant chemotherapy.

In vitro cytotoxic assay. Tumor cells were plated at a concentration of 1000 cell/well into 96-wells plate. The following day, a series of concentrations of the different drugs were supplemented to the culture medium. Untreated control cells were kept in normal culture medium. Cell viability of each well was assessed with crystal violet staining 48 hours after treatment, as described above. Results were analyzed by Prism software by a non-linear regression analysis and expressed as relative cell viability compared with non-treated control. The 50% or 85% maximum inhibition concentrations (IC50, IC85) were used to determine the drug-resistant ability of treated cells. infection and selection with puromycin (5 µg/ml) were performed as previously described [1]. Gene knocking down efficiency were vilified by RT-qPCR, the most effective one (TRCN0000077289) were used for further experiment.
Briefly, triplicates wells of 4T1 and MR20 or DR500 cells grown in culture were used for RNA extraction using RNeasy kit (QIAGEN). Probe synthesis and GeneChip Mouse Gene Exon 1.0 ST Array (Affymetrix Ltd) hybridization were performed at the Genomic Technologies Facility (GTF, Lausanne, Switzerland) at UNIL (Lausanne).

Microarray data analysis. Statistical analysis.
Microarray analyses were carried out with R, a free software environment available at http://www.r-project.org/. After quantification of gene expression with robust multi-array normalization [2] using the BioConductor package Affy, (http://www.bioconductor.org/) significance of differential gene expression was determined by computing moderated t-statistics and false discovery rates with the limma package [3]. Annotation was based on the genome version NCBI Build 36 (Feb. 2006).
The obtained p-values were corrected for multiple testing by calculating estimated false discovery rates (FDR) using the method of Benjamini-Hochberg. The transcriptional factors enrichment analysis was then performed by using MetaCore from Thomson Reuters (version 6.11) by filter with thresholds 1.5 and p-value 0.01. Heatmaps were produced by color-coding gene-wise standardized log gene expression levels (mean zero standard deviation one). Probe-sets were shown hierarchically clustered by similarity based on Euclidean distance and the ward aggregation algorithm. The data have been deposited in GEO (GSE100973). The interferon alpha gene signature was taken from MSigDB [4] and the interferon heatmap was drawn using the R package heatmap.

Expansion of relapsed tumor cells from MR20 bearing mice.
To obtain the MR20 tumor cells that relapse from dormancy in vivo, relapsed tumors were processed as described for Flow cytometry analysis. After enzymatic treatment and filtering through a 100 µm and a 70 µm sterile nylon gauzes, tumor suspensions were cultured in 15 cm plate.

T cell depletion.
In vivo depletion of CD4-and CD8-specific T cells was performed as descripted previously [5]. Briefly, 0.5 mg antibodies per mouse were injected intraperitoneally for 3 consecutive days. 6 days after the first injection, depletion efficiency was examined by flow cytometry from blood samples before tumor implantation. To maintain the depletion condition, same quantity of antibodies was injected twice per week.
Anti-CD4 (clone GK1.5), anti-CD8 (clone 2.43) antibodies and Rat IgG control (clone LTF-2) were purchased from BioXCell. The staining procedure and the flow cytometry acquisition were the same as outlined above for blood. Data acquisition was performed using the FACSCalibur (BD Biosciences) or MACSQuant flow cytometer from MiltenyiBiotec and data analyzed by FlowJo v10.0.7.
Tissue morphology. Tumors and lungs were harvested at the end of the experiment, fixed in formalin and embedded in paraffin. 5 µm thick serial sections were cut from the tissue blocks. 3-4 sections that with 100 µm distance were stained with hematoxylin and eosin (H&E) and used to assess tumor morphology and quantify lung metastasis. Slides were scanned by Nanozoomer (Hamamatsu Photonics) and metastasis are counted manually using NDP.viewer2 software (Hamamatsu Photonics).
Patients, clinical study and data analysis. The prospective multicentric TOP trial enrolled 149 ER-negative breast cancer patients between January 2003 and June 2008 that were treated with anthracycline monotherapy [6]. The present retrospective evaluation of the samples from this study was approved by the central ethics committees of the Institute Jules Bordet (20042017, CE2688). Blood samples were taken before therapy, at the end of the cycle one, and before surgery and were available for 51 patients.
Chi-squared tests were performed using chisq.test function, from package stats version Gene expression data for TOP study were retrieved from GEO database under the id GSE16446. Analysis were performed within R version 3.3.3. Each immune signature [7][8][9] was computed as a weighted average of the gene expressions included in the signature.
Weights were set to -1 or +1 according to the sign of the original coefficients of the published signature. Signatures were scaled to insure reliable comparisons with the rescale function, from the package genefu version 2.6.0, applied with default parameters.
This ensure that quantiles at 2.5% and 97.5% equal -1 and +1 respectively in each signature. Correlation matrix was built by computing Pearson coefficients for each pair of variables, using cor function from the package stats version 3.3.3. Only correlations having a p-value<0.05 were kept.