Background
The interferon response can influence the primary and metastatic activity of breast cancers [
1], and can interact with checkpoint immunotherapy to modulate its effects [
2]. Although systemic interferon delivery for cancer produced responses in tumors, it was limited by severe adverse events and has been superseded by newer therapies [
3]. To circumvent some of these challenges, intratumor injection of agents that induce an interferon response are currently being trialed in a large number of cancer settings and among these agents are derivatives of polyinosinic–polycytidylic acid (polyIC) [
4]. In mice, polyIC greatly improved the ability of anti-programmed death-ligand 1 (PD-L1) therapy to clear or reduce melanomas and cancers of the lung and colon, and this effect was dependent on type 1 interferons [
5].
PolyIC and its degradation products mimic viral RNA, key pathogen-associated molecular-patterns displayed during viral replication. PolyIC binds to and activates toll-like receptor 3 (TLR3), a pathogen-associated molecular pattern receptor located on the cell membrane and within endosomes. This generates a number of immune cell effects, such as an interferon response and innate immune cell activation, the latter effect responsible for the use of polyIC as a vaccine adjuvant [
6].
There is, however, another family of pathogen-associated molecular pattern receptors that recognize double-stranded RNA. The OAS family of proteins are potent inducers of the interferon response [
7]. They are pathogen-associated molecular pattern receptors that bind double-stranded RNA in the cytoplasm. These proteins respond by synthesizing 2′–5′-linked RNA from ATP without a template [
8]. This in turn binds to and activates latent ribonuclease (RNaseL), causing it to degrade viral RNA and host ribosomal RNA, so disrupting the viral life cycle. Activation of OAS-RNaseL also promotes apoptosis and provokes a robust interferon response, both via poorly understood mechanisms that involve additional detection of RNaseL-degraded RNA by the RIG1-like pathogen recognition receptors and possibly direct interaction of OAS and RNaseL with the mitochondria [
9]. The OAS family is routinely ignored by most studies using polyIC, which focus on the action of TLR3. The potential of the OAS family as therapeutic targets in oncology has been ignored because the effects of polyIC are attributed to TLR3, and no effective way to separate the actions of OAS and TLR3 was available.
There is some evidence that RNaseL may be involved in the etiology and progression of cancer. Attention has focused on RNaseL, the down-stream effector of the OAS proteins, activation of which predicted the outcomes in virally induced cancers. An association with RNaseL allelic variation and the risk of human papilloma virus-driven cancers of the cervix and head was demonstrated. Unexpectedly, this study also found an association with cancers that do not show a strong viral etiology such as breast [
10]. Activation of RNaseL may occur naturally in cancers, hypothesized to be driven by the recognition of double-stranded RNA species by OAS proteins produced from viral sequences integrated in the genome. These sequences are usually silenced by methylation but are transcribed de novo as a result of the general demethylation of the genome that occurs in cancer cells. This was suggested by the finding that cells deficient in RNaseL were highly resistant to the cytotoxicity of 5-azacytidine, a drug that demethylates the genome. Cells were made resistant to 5-azacytidine via inhibition of RNaseL [
11].
Other cancer studies found that allelic variation in RNaseL was observed to be associated with increased risk of prostate cancer and the increased risk of higher tumor grade, together with an increased level of inflammation markers [
12,
13]. Allelic variation of RNaseL may modify the risk of breast cancer in patients carrying high risk mutations [
14]. The inherited RNaseL alleles responsible for this are proposed to reduce RNaseL activity, and this proposition is supported by observations in lung cancer cells, which increased the levels of the RNaseL inhibitor RL1 to suppress RNaseL-driven apoptosis, with mitochondrial-mediated apoptosis restored by interferon gamma [
15].
Our laboratories conducted an
N-ethyl-
N-nitrosourea mutagenesis project in mice to find new genes driving mammary gland development and function, with the intent of also identifying new genes involved in mammary cancer initiation and metastasis. We discovered a dominant mutation in
Oas2, (referred to as MT, wild type
Oas2 referred to as WT) causing a non-conservative amino acid change that resulted in failed lactation despite normal mammary development during pregnancy. These mice were otherwise completely normal [
16]. We observed a robust interferon response associated with apoptosis in the mammary glands of our MT mice. When MT was expressed in human breast cancer cells we found that the interferon response depended on the presence of RNaseL and interferon regulatory factor 7 (IRF7), the key effector molecules at the proximal and distal ends of the OAS-RNaseL signaling pathway [
16]. This was the first observation of a regulatory pathway linking activation of viral recognition to the control of lactation. The
Oas2 mutant mouse provides a defined and single molecular event that induces an interferon response, with pathological activity only in the mammary gland at the onset of lactation.
The
MMTV-PyMT model is the most widely used mouse model of breast cancer. PyMT protein [
17] acts as a scaffold promoting persistent signaling via binding of SRC family kinases, PI3K, SHC, 14-3-3, PLC-gamma and TAZ. It causes early and multifocal onset of estrogen receptor positive hyperplasia that rapidly progresses through a series of well-defined lesions to estrogen receptor negative invasive carcinoma of the luminal molecular subtype [
18]. The model is weakly sensitive [
19] or refractory [
20] to PD-L1 or PD-1 as monotherapy. Metastasis to lung is frequent and strongly dependent on innate immune cells previously characterized, as macrophages [
21,
22] or myeloid-derived suppressor cells (MDSC) [
23] of monocytic (G-MDSC) or neutrophil (PMN-MDSC) origin, and now thought of as pathologically activated monocytes and neutrophils [
24].
To determine if activation of OAS2 by a future breast cancer therapeutic could provide a new therapeutic strategy, and to distinguish OAS activation from that of TLR3, we combined the
Oas2 mutation with expression of the polyoma middle T oncogene (
PyMT) driven by the mouse mammary tumor virus promoter (
MMTV) [
25] and asked if this altered the initiation or progression of mammary cancer, and the interaction with checkpoint immunotherapy.
Methods
Mice
The
Oas2 colony [
16], on an inbred FVBN background, was crossed with the polyoma middle T antigen mouse (
MMTV-PyMT) also on an FVBN background, a kind gift from Dr. William Muller [
25]. The colony was then maintained by breeding compound heterozygous males (MT
Oas2/WT
Oas2/
MMTV-PyMT/non-transgenic) with WT
Oas2/WT
Oas2 and non-transgenic FVBN females. All animals were housed with food and water ad libitum with a 12-h day/night cycle at 22 °C and 80% relative humidity.
Timed mating
One or two 7–9-week-old females were housed with single males overnight and monitored daily for a vaginal plug, and pregnancy was confirmed by weight gain. Females were single housed for parturition and pups culled at birth. Mammary gland tumors were measured with calipers throughout pregnancy and the postpartum period. Tumor volume (mm3) was calculated as (minimum measurement2) * (maximum measurement)/2. Mice were euthanized at specified endpoints, via anesthesia (5% isoflurane, 1 L/min O2) followed by cervical dislocation. At autopsy 4th and combined 2nd/3rd glands or tumor weights were recorded.
PDL-1 antibody treatment in parous mice
Parous mice received 10 µl per gram body weight of a solution of PDL-1 (cat BP0101) or IgG control (cat BP0090) at 1 mg/ml via I.P. injection for six treatments, delivered twice weekly.
Mammary gland whole mounts and lung collection
Mouse mammary glands, tumors and lungs were harvested at specified timepoints, and fixed in 10% buffered formalin for 4 h at room temperature. Tissues were changed into 70% ethanol prior to embedding in paraffin. Mammary glands were defatted in 3–4 changes of acetone before being dehydrated and stained in carmine alum for whole mount photography. Glands were then passed through a series of graded alcohols and embedded in paraffin for sectioning and immunohistochemistry.
Immuno-histochemistry
Tissue sections were either stained with hematoxylin and eosin for routine histochemistry or stained with antibodies to the following antigens using immunohistochemistry protocols as detailed in Additional file
1: Table S1. For analysis of lung metastasis, two or more sections at 100 µm intervals were cut for staining.
Image analysis
Whole mammary gland or lung sections on glass slides were scanned in Aperio ScanScope XT at Katharina Gaus Light Microscopy Facility, and Garvan Microscope Facility, Sydney, Australia, and analyzed by QuPath [
26]. A set of multiple reference images was used to train the DAB pixel detection algorithm. Quantification of phospho-Stat positive nuclei used the positive cell detection QuPath algorithm. Regions that contain positive 3,3′-diaminobenzidine (DAB), hematoxylin or no-stain were classified by batch processing and reported as total tissue area (µm
2), DAB-stained area (µm
2), non-DAB-stained area (µm
2) and non-stained area (µm
2). DAB stain area percentage without lumen non-stained area was calculated as;
$$\begin{aligned} & {\text{PYMT stain area percentage with non-stained area}} \\ & \quad = \left( {\frac{{\text{PYMT stain area}}}{{{\text{PYMT stain area}} + {\text{non - PYMT stain area}}}}} \right)*100 \\ \end{aligned}$$
Kaplan–Meier survival analysis, graph drawing and statistical tests were performed in GraphPad Prism 8.0.2. Strong evidence against the observation being made by chance was accepted at the 90% level.
Immune cell profiling by flow cytometry
Mouse mammary glands (2/3rd & 4th) were crudely macerated and cells disassociated in 1% hyaluronidase, 37 °C shaking, 20 min. Digestion was neutralized with 2% fetal calf serum (FCS)/phosphate-buffered saline (PBS) prior to centrifugation at 1200 rpm, 4 °C. Cell pellets were resuspended (25 µg/ml DNAse1 in 2% FCS/PBS) and strained through 40 µm mesh and the supernatant concentrated by centrifugation. Cell pellets were resuspended 0.8% NH
4Cl, DNase1 (25 µg/ml) in PBS and incubated in 37 °C water bath, 1 min. Red blood cells (RBC)/DNA lysis was neutralized in 7 × volume 2% FCS/PBS prior to centrifugation and resuspension in 2% FCS/PBS for immune cell staining using fluorophore conjugated antibodies (refer to antibody panels in Additional file
2: Table S2). Mouse spleen, thymus and lymph nodes (from 4th mammary glands) were collected separately into RPMI media (+ 10% FCS) and strained through 70 µm mesh directly prior to RBC/DNA lysis, cell resuspension and immune cell staining steps (as detailed above). Flow cytometry was performed using BD LSR II SORP and data exported to the FlowJo software (Tree Star Inc.) for data analysis. Routine T [
27,
28] and innate immune cell protocols [
29] were followed using antibodies listed in Additional file
2: Table S2.
Discussion
The publications on allelic variation in the Oas-RNaseL pathway suggest that inherited alleles that are thought to cause pathway activation produce better control of virally induced tumors, but also of non-viral tumors by providing an enhanced response to de novo dsRNA production from integrated viruses. A separate group of Oas-RNaseL pathway alleles is thought to cause pathway suppression and to be involved in the genesis of prostate cancer. Our results suggest, however, that the same Oas2 alleles may have context-dependent effects. In the nulliparous state, activation of Oas2 caused faster primary tumor progression, but in the parous state this effect was lost, and instead suppression of metastases and sensitization to checkpoint immunotherapy was observed. It is likely that distinct effects of Oas-RNaseL activation within the cells of the epithelium and immune system add complexity to the role of this pathway in cancer.
A fully completed pregnancy remains an independent predictor of breast cancer risk [
30], and pregnancy-associated breast cancer is associated with a 40% increase in mortality due to increased metastasis [
30], with even higher rates in cases occurring during pregnancy [
31]. The period of increased risk caused by pregnancy may last three or more decades for women who have their first child after their mid 30’s, but less than a decade for women who have their first child before their early 20’s [
32,
33]. A degree of life-long protection then follows. This age-dependent period of increased risk defines a large proportion of breast cancer cases as associated with pregnancy; however, current clinical practice uses arbitrary postpartum time periods to define pregnancy associated breast cancer, and does not distinguish pregnancy-associated breast cancer based on the age-dependent period of increased risk. Regardless, this type of breast cancer has no special therapeutic guidelines.
The number of first pregnancies of women over 30 years of age continues to increase especially in developed and developing countries where increasing age at first pregnancy is correlated with increasing female education [
34,
35]. Breast cancer associated with pregnancy [
36] is one of the main drivers of the increased prevalence of breast cancer [
37]. This effect is compounded by an ongoing decrease in breast feeding, as every 12 months decreases breast cancer risk by 4.3%. In developed countries, decreases in childbirth and lactation have doubled the cumulative risk of breast cancer, with two-thirds of this effect attributable to reduced lactation [
38].
The results of this study have direct relevance to the changes in reproduction practices in developed and developing nations. Activation of OAS2 in the MMTV-PyMT model of breast cancer prevented pregnancy from increasing the number of lung metastases, a finding that establishes the OAS family as therapeutic targets for the development of agents designed to prevent pregnancy from driving breast cancer metastasis. A small interaction with PD-L1 therapy was also detected, albeit in a model at best only partially sensitive to PD-L1. The data provide impetus for further study to determine if the combination of OAS activation and checkpoint inhibition is of therapeutic relevance for the treatment of breast cancer during pregnancy, or other cancers where a more robust effect of PD-L1 occurs. A cautionary finding was that in nulliparous mice activation of OAS2 resulted in more rapid primary tumor progression, suggesting that in nulliparous women the use of interferon-inducing agents should be carefully examined, especially for local recurrences. As the mechanisms responsible for the effects of pregnancy and lactation on postpartum breast cancer risk are poorly understood these findings may also have significance for our understanding of pregnancy associated breast cancer. Importantly, the MT mice show no pathology other than failed lactation, a rare observation in the interferon-response field which is beset with serious adverse events, suggesting that such a therapy would have a favorable toxicity profile.
Our findings may have relevance to breast cancer more generally. We have recently described a pro-metastasis pathology we called involution mimicry [
39] that occurs in response to forced expression of the ETS transcription factor ELF5, the master regulator of lobuloalveolar development during pregnancy, which causes a limited lactation and ongoing involution response in an ELF5 mouse model [
23], but is also seen in breast cancer patients. The ELF5 mouse model shows a similar a engorgement pathology to that we saw following parity in the MT mice and an increase in lung metastasis that is abrogated by neutrophil depletion [
23]. We found profound changes in the state of the epithelium, cancer-associated fibroblasts (CAFs), endothelial cells and immune cells [
39]. Among the CAFs we found populations previously associated with mammary gland involution following lactation, and an analysis of ligand-receptor interactions among these cell types established the existence of a complex regulatory network linking the induction of ELF5 in the epithelium to profound transcriptional changes in the CAFs and endothelial cells, and to the recruitment and M2 polarization of MDSC [
39]. Two linked human molecular signatures were discovered in breast cancer patients, a lactation signature and an involution signature. The lactation signature was enriched in the epithelial cell populations and the involution signature was present in both epithelial and stromal populations. The involution signature was prognostic in the METABRIC cohort [
39]. We surmise that elevated ELF5, possibly due to loss of promoter demethylation [
40], drives a limited lactation and involution in these breast cancers.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.