Introduction
Recent advances in early detection, and the development of targeted therapies have significantly improved clinical outcome of breast cancer patients [
1‐
3], but metastasis and tumor recurrence remain as major obstacles to successful disease management [
4]. Metastasis requires an ability of tumor cells to survive and expand in alien microenvironments and occurs as a result of a complex set of tumor and host traits that are distinct from those facilitating primary tumor initiation and growth. The metastatic process is still poorly understood at a mechanistic level [
5,
6] although the discovery of promoter and suppressor genes that selectively affect metastases but not primary tumors has facilitated identifying tumor-intrinsic and microenvironmental properties necessary for secondary site colonization [
6,
7].
One oncogenic signaling hub that has been implicated in the metastases of breast and other cancers is the multifunctional, intracellular/extracellular protein, RHAMM (gene name
HMMR) [
8‐
11]. RHAMM expression is heterogeneous in breast cancer, and the presence of strongly RHAMM-positive tumor cell subsets is linked to increased peripheral metastasis and poor clinical outcome [
12]. Experimental models of breast cancer confirm a role for RHAMM in promoting functions associated with breast tumor initiation and metastasis [
13‐
15] predicting that targeting this protein may improve clinical management of breast cancer metastases. Mechanistically, RHAMM performs multiple extracellular and intracellular functions relevant to metastasis [
12,
16] including regulation of mitosis [
17], genomic stability [
18], cell motility, cellular plasticity [
19‐
21], pluripotency of progenitor cells [
20,
22] and oncogenic driver pathway activation [
13]. In addition to tumor-intrinsic functions, RHAMM also regulates host cell responses that can impact tumor cell survival [
8].
To better define the mechanisms by which RHAMM functions affect metastasis, we assessed the consequences of
Rhamm-loss to mammary tumor progression using the MMTV-PyMT transgenic mouse model of breast cancer susceptibility [
23]. This model was chosen for its rapid progression to metastatic disease, and its molecular similarity to both luminal B breast cancer and basal-like breast tumors, which typically express high levels of RHAMM clinically [
24,
25].
Rhamm-loss in this model has no detectable effect on primary tumor initiation or growth but unexpectedly increases, rather than decreases, lung metastasis. This effect is traced to clonal selection of
Rhamm−/− tumor cells with an intrinsic resistance to DNA-damage-induced apoptosis that is sensed by STING/interferon signaling. This mechanism provides a survival advantage in lung but not the mammary microenvironment that is linked to the higher ROS and TGFB levels in lung tissue, which enhance STING-dependent apoptosis of RHAMM
+ve tumor cells but spare RHAMM
−ve comparators. These results identify RHAMM as a novel tissue-specific metastasis regulator and document tumor intrinsic and microenvironmental contexts that trigger its apparent metastasis suppressor functions.
Material and methods
Mouse breeding, genotyping, tumor measurements and whole mount preparations
C57Bl/6
Rhamm−/− and Wildtype mice were crossed to MMTV-PyMT mice on an FVB background (purchased from Jackson Labs) as described by Lopez et al. [
26] to obtain
Rhamm−/−:MMTV-PyMT,
Rhamm±:MMTV-PyMT and
Rhamm+/+:MMTV-PyMT genotypes. The degree of SNP homozygosity, measured using the mouse diversity genotyping array (MDGA), was similar between
Rhamm−/−:MMTV-PyMT and
Rhamm+/+:MMTV-PyMT genotypes. Littermate heterozygotes (
Rhamm±) from the
Rhamm±:MMTV-PyMT x
Rhamm−/−:MMTV-PyMT cross produced a similar tumor profile as the
Rhamm+/+:MMTV-PyMT x C57Bl/6 Wildtype cross mice. The preparation and breeding of
Rhamm−/− mice, tumor measurements and whole mount preparations are described in Additional file
1: Methods.
Methods for IHC, IF, and immunoblot are described in [
27] and in Additional file
1: Methods.
CRISPR cell line generation
The MDA-MB-231 RHAMM CRISPR cell line was generated by transfection with paired guide RNA’s (5′–3′) GTATTGTATTTGATTAGAAT (within exon 3 of the RHAMM gene) and GAATTTGAGAATTCTAAGCT (within exon 6) in plasmid pCR4-TOPO-U6-HPRT-gRNA. Guide RNA’s were co-transfected with plasmid expressing the CAS9 enzyme (pT3.5 Caggs-FLAG-hCas9) as well as plasmids for puromycin and GFP selection, pcDNA-PB7 and pPB SB-CG-LUC-GFP (Puro)(+CRE), using Lipofectamine 2000 reagent (Invitrogen, cat#11668-019) following the manufacturer’s suggested protocol. Mock cell lines were generated by transfection of parent cells with selection plasmids only and selected as a pool by culture in puromycin containing medium (0.6 µg/ml). RHAMM-CRISPR knockout cell lines were selected by clonal plating in puromycin containing media (0.6 µg/ml). Single cell derived colonies were expanded and screened by genomic PCR for the corresponding deletion within the RHAMM gene (primers 5′–3′ AGATACTACCTTGCCTGCTTCA and ACCTGCAGCTTCATCTCCAT), and by immunoblot for loss of RHAMM protein.
Primary cultures of tumor cells
Tumor cell isolation is described in Additional file
1: Methods.
Cultured cell treatment
For quantification of H
2O
2-induced apoptosis and STAT1 activation,
RHAMM-CRISPR knockout and mock transfected cells were plated on cover slips in DMEM medium containing 10% FBS, resulting in sub-confluent cultures after 24 h incubation. STAT1 activation and apoptosis was induced by incubation in culture medium containing 50–200 µM H
2O
2 for either 4 (STAT1 activation) or 48–72 h (apoptosis). Cells were stained for either STAT1 or cleaved CASPASE 3 as described in Additional file
1: Methods. Staining quantification by ImageJ used confocal images.
To induce DNA double strand breaks or apoptosis in tumor cells that were isolated from primary tumors, cells were plated at low cell density on fibronectin coated coverslips and cultured in growth medium for 24–48 h. Cultures were treated with either Cis-platin or 300 µM H
2O
2 at the indicated concentrations and durations. Cells were stained for either γH2AX or ApopTag as described in Additional file
1: Methods. Staining quantification by ImageJ used confocal images.
Py8119 cells were obtained from the ATCC (ATCC CRL-3278) and cultured in F12K medium (Wisent) supplemented with 5% FBS and Mito + Serum Extender. These cells were originally isolated and cloned from tumors that arose in C57Bl/6 MMTV-PyMT mice and therefore do not contain genomic sequences from FVB mice [
28].
siRNA transfection
Sub-confluent Py8119 tumor cells cultures were transfected using Lipofectamine RNAiMAX (Invitrogen) following the manufacturer’s instructions. Culture medium was changed to CTS Opti-MEM (Gibco). Lipofectamine RNAiMAX reagent was diluted 1:50 with CTS Opti-MEM. siRNA (ID: 151008, 159287, s279, negative control siRNA#1, Ambion) was diluted to a concentration of 500 nM in CTS Opti-MEM. Diluted transfection reagent and siRNA were mixed 1:1, incubated at RT for 20 min and then, added to the cell cultures. After 4–5 h incubation at 37 °C, medium was changed to culture medium containing STING agonists (Vadimezan (DMXAA), 33 µM, G10 for MDA-MB-231 tumor cells, 40 µM), H2O2 (300 µM), DMSO and/or TGFB1 (5 ng/ml). STAT1, CASPASE 3 staining or cell survival were analyzed 20–72 h later.
AlamarBlue assay
Py8119 cells were plated at a density of 3000 cells/well of a 96 well plate using complete culture medium. After ON incubation at 37 °C, cells were transfected with Stat1 siRNA, Rhamm siRNA or negative control siRNA. After transfection, cells were treated with F12K culture medium containing STING Agonist Vadimezan (33 µM), DMSO, H2O2 (300 µM), and/or TGFB1 (5 ng/ml). After 72 h incubation at 37 °C, the number of surviving cells was quantified by adding AlamarBlue reagent (1/10 Vol.) followed by 1–2 h incubation at 37 °C. Fluorescence was measured using a plate reader.
DNA and RNA isolation are described in Additional file
1: Methods.
Analysis of de novo mutation genotypes
The somatic (de novo) mutation burden in Wildtype and
Rhamm−/− MMTV-PyMT tumors were compared by identifying germline and mammary tumor mutations using a mouse genomic diversity array (MGDA) unbiased platform [
29,
30]. The MDGA detects large, de novo postzygotic deletions and duplications as CNVs using close to 900,000 markers, and also de novo postzygotic base substitutions at single nucleotide polymorphic loci as SNVs, assayed at close to 500,000 SNV loci distributed across the mouse genome. Mouse diversity genotyping array (MDGA) hybridization was performed at the London Regional Genomics Centre (Robarts Research Institute, London, ON) according to instructions in the Affymetrix® Genome-Wide Human single nucleotide polymorphism (SNP) Nsp/Sty 6.0 manual (Affymetrix 2007;
https://assests.thermofisher.com/TFS-Assets/LSG/manuals/snp6_atp_userguide.pdf). The resulting CEL files were then used for single nucleotide variant (SNV) genotyping and copy number variant (CNV) identification as described (Additional file
1: Methods).
Ilk/
Rrp8, Taf10 genes (Mm00232271_cn), which overlap the same CNV region, were used for confirmation by ddPCR. The transferrin receptor gene (
Tfrc) was used as the diploid copy number reference for the
Ilk,
Taf10 and
Rhamm assays. No-template controls and two technical replicates were used in all assays. DdPCR procedures are described in Additional file
1: Methods. Recurrent and unique CNVs were determined using HD-CNV (Additional file
1: Methods) [
31]. To visualize CNV occurrence across the genome for individual mouse samples, a timeline-style plot was generated in R using ggplot2 (v3.2.1). Genic annotation used to identify the genic content of CNVs was obtained from Ensembl’s BioMart (Ensembl genes 67, NCBIM37). Protein-coding genes, non-coding genes, and pseudogenes that completely overlapped CNVs of the same state, in all three samples of a group (shared
Rhamm genotype and tumor type), were considered recurrent within that group.
Phenogram construction was done as previously described [
29] and is described in Additional file
1: Methods. Filtering procedures are described in Additional file
1: methods. Identification of candidate de novo mutations is described in Additional file
1: Methods. The spatial distribution of candidate de novo mutations across the genome was visualized using rainfall plots (Additional file
1: Methods) [
32]. Chromosomes were treated linearly, and the genomic position of de novo mutations was assigned in an additive manner based on the position of the locus in relation to the whole GRCm38.p4 genome.
RNA sequencing is described in Additional file
1: Methods.
Pathway analysis
Gene expression differences between
Rhamm−/− and Wildtype tumors of 1.5-fold with a p-value of less than 0.05 were analyzed for enriched Gene Ontology (GO) and KEGG pathways. The functions of differentially expressed genes were further probed using the hallmark gene set from the Molecular Signatures Database v 7.1 (
https://www.gsea-msigdb.org/gsea/msigdb/index.jsp), used in the GSEA analysis, and Metascape (
https://metascape.org/gp/index.html#/main/step1). Significantly down-regulated genes that grouped into the top hallmark gene sets in the GSEA analysis were assessed for mRNA co-expression (
p < 0.05) with
RHAMM (
HMMR) using breast cancer data sets in cBioPortal (cbioportal.org). Invasive breast cancer subtypes that express the highest
RHAMM mRNA levels were identified using the molecular subtypes in the TCGA PanCancer breast invasive carcinoma data set in cBioPortal (cbioportal.org).
Apoptag® staining, ROS/NOS and 8-oxodG ELISA are described in Additional file
1: Methods.
Discussion
Our results identify a novel tumor-intrinsic mechanism resulting from Rhamm-loss that provides a survival advantage in the lung microenvironment and that associates with clonal dominance. We identify specific properties of the lung microenvironment (high ROS, TGFB levels), which combine with a Rhamm-dependent tumor-intrinsic impairment of DNA damage sensing to provide a selective survival/growth advantage in this tissue but not in the mammary gland microenvironment.
Advanced genomic and diagnostic technologies identify metastasis as an extremely complex process that can result from single or multiple clones arising early or late in the genetic evolution of primary tumors [
53]. Tumor clonal genetic heterogeneity is considered to be a driving force in both successful metastatic colonization and treatment resistance. Each step in the colonization of extraneous tissues exerts selective pressure, which results in a genetic and epigenetic heterogeneity that is distinct from the primary tumor. However, the factors driving this selection process are still poorly understood. Therapeutic strategies remain largely based upon analyses of primary tumors [
53], and this knowledge gap contributes to the present restricted ability to eradicate and/or manage metastases. Our results identify one mechanism, loss of RHAMM signaling, that associates with reduced genetic diversity and, paradoxically, clonal amplification of lung colonies that are resistant to ROS-mediated DNA damage.
Positive selection, detected by SNV similarities, is clearly evident in
Rhamm−/− primary mammary tumors even though phenotypic properties are not detectably altered. Selection is strongly amplified in lung tissue metastatic tumors, as indicated by an almost 100-fold enrichment in shared SNVs, which predicts that the lung microenvironment creates a bottleneck limiting expansion of metastatic clones. The enhanced tumor-intrinsic survival capability in the lung is linked to loss of RHAMM expression and blunting of STING/IFN signaling in both
Rhamm−/− and Wildtype genotypes. Interestingly, RHAMM expression-loss in Wildtype lung metastatic colonies is not associated with the increased SNV homogeneity observed in
Rhamm−/− lung tumor colonies, suggesting that aberrant STING-IFN survival responses alone are not responsible for clonal homogeneity. We analyzed genic SNVs for clues as to mechanisms that might complement this
Rhamm−/− intrinsic tumor cell phenotype. It is remarkable that while Wildtype lung metastases share few genic SNVs (Additional file
2: Table S3), and these mutant genes do not group onto the same signalling pathways (data not shown),
Rhamm−/− lung metastases collectively share an extraordinary 125 genic SNVs. None of these mutations associate with IFN or TGFB-regulated signalling as queried by KEGG, IPA or GSEA data sets (IPA shown, Additional file
2: Table S6), but in silico analyses predict a role for these mutations in cancer and wound repair. Wound and cancer microenvironments share many immune and stromal fingerprints, suggesting that critical changes in the microenvironment of
Rhamm−/− primary tumors may contribute to positive selection of homogeneous clones. Another consideration is evidence that RHAMM expression contributes to maintenance of embryonic stem cell pluripotency [
20] and is a marker for a subpopulation of renewing tumor stem cells [
54‐
56], which are thought to drive tumor cell heterogeneity.
Rhamm-loss may increase clonal homogeneity, in part, by restricting pluripotency of these tumor cell subsets. Further multiplexed and spatial platform analyses are required to identify the tumor-intrinsic and host mechanisms that are regulated by RHAMM, and that underlie the clonal selection and expansion of mammary tumor cells.
Evidence is emerging that the oncogenicity of many dominant oncogenic driver genes is context- and tissue-dependent [
36,
48,
49]. For example, high expression of CD44, like RHAMM, is linked to breast cancer initiation yet both proteins can perform metastasis suppressor functions in experimental models of breast cancer susceptibility [
26]. Both genes are subject to alternative splicing of mRNA, and selective expression of isoforms can have different tumorigenic effects. For example, expression of one of two RHAMM isoforms is oncogenic in an experimental model of pancreatic cancer [
9]. Our results probe the duality in oncogenic functions of RHAMM by providing mechanistic insight into the conditions favoring the metastasis suppressor over promoter functions using RHAMM as an example. This type of contextual and mechanistic information is critical for successful targeting of many multifunctional proteins typified by context-dependent oncogenic and suppressor properties [
6,
57]. Thus, while a large body of data predict that RHAMM acts as a tumor promoter in breast cancer [
13,
15], our results uncover an unexpected role for RHAMM in suppressing metastasis under the very specific conditions of whole animal
Rhamm-loss. In addition to identifying and providing context for the dual oncogenic and tumor suppressor functions of RHAMM, our results raise the possibility of using RHAMM expression as a biomarker for sensitivity to interferon therapy, which, as an example, can impact immune therapy effectiveness in tumor management [
58,
59].
In summary, our conclusions from the present study are that RHAMM contributes to primary tumor and metastasis clonal heterogeneity by regulating tumor cell survival via STING/IFNG/STAT1 signaling in microenvironments characterized by a high potential for DNA damage. When RHAMM is lost, these properties do not detectably impact primary tumor initiation or growth but favor survival and growth of homogeneous metastatic clones in lung tissue that are under positive selection as a result of host Rhamm-loss.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.