Copy number-dependent overexpression of RhoC in human breast cancer

Breast tumor formation and progression are associated with copy number aberrations, single-nucleotide variants and other indels as well as with structural variants. Many of the copy number changes affect Rho signaling [5]. For example, the DLC1 RhoGAP on chromosome 8p, shows hemizygous deletion in 40 to 50% of breast tumors, and homozygous deletion occurs in a small fraction of cases (Additional file 1: Supplementary Figure S1A, S1B). DLC1 is haploinsufficient in the mammary gland [16] and functions as a tumor suppressor through enhanced Rho signaling when deleted [17]. To test for other genomic changes with the potential to increase Rho signaling, we looked for chromosome losses that include genes with RhoGAP-like domains. Indeed, more than 40% of breast tumors in The Cancer Genome Atlas (TCGA) cohort show deletions that included ARHGAP44 and ABR on 17p, ARHGAP20 and ARHGAP32 on 11q, STARD13/DLC2 on 13q, as well as PRR5/ARHGAP8, SH3BP1 and BCR on 22q (Additional file 1: Supplementary Figure S1A, S1B). Copy number gains and structural variants in RhoGEF genes were also evident, many of which have the potential to increase Rho signaling through increased GTP-loading [6]. For example, over 50% of TCGA breast tumors show copy number gains or amplifications involving OBSCN, ARHGEF2 and ARHGEF11 on 1q or PREX2 on 8q (Additional file 1: Supplementary Figure S1C, S1D). Next, we looked for SNV or copy number changes in genes coding for Rho-family proteins. While SNVs were uncommon, copy number gains were seen. For example, RHOC gains were found in 4 (METABRIC) to 16 (TCGA) percent of cases (Fig. 1A and Additional file 1: Supplementary Figure S2A). In comparison to controls, a greater percentage of tumors with increased RHOC gene copies were ERα-negative (Fig. 1A) and, in the case of METABRIC cohort tumors, associated with a significant increase in RHOC gene expression (Fig. 1B). In the TCGA cohort, a trend towards increased expression was seen for tumors with copy number gains or amplifications which included the RHOC gene, although this did not reach significance (Additional file 1: Supplementary Figure S2B). Finally, more tumors with RhoC Gains/Amplifications were of the basal subtype in comparison to tumors without copy number gains for RhoC (Fig. 1C). Also, more were Histological Grade 3 (Fig. 1D).

Fig. 1

The frequency of RHOC copy number gain and amplifications in human breast cancer. A RHOC gains occur in 4% (top) and 16% (bottom) of human breast tumors from METABRIC and TCGA studies, respectively. The ER status of each breast tumor sample is shown. B RhoC Copy number and its association with mRNA expression for this gene—all comparisons in the table below are statistically significant. C, D Breast tumor subtypes (C), and tumor grades (D) are displayed for each group—samples with RHOC gain/amplification vs. those without these alterations. All data are from the METABRIC study

Generation of a cre-conditional RhoC transgenic line

It is well established that RhoGAPs can function as tumor suppressors, whereas RhoGEFs and activated Rho mutants can be oncogenes [2, 18]. The importance of increased Rho protein expression is less clear. Indeed, widespread expression of GEFs and GAPs suggests that RHO proteins are regulated mostly at the level of GTP-loading. Despite this, there are situations where Rho expression is limiting [10, 11]. To study elevated RhoC expression in vivo, we used gene targeting in embryonic stem cells to generate a Rosa26-based transgenic with RhoC linked through IRES sequences to eGFP downstream of a loxP-stop-loxP cassette (Additional file 1: Supplementary Figure S3) [19, 20].

RhoC overexpression cooperates with ErbB2/Neu to induce mammary tumor formation

RhoC overexpression and ErbB2/Neu status are positively correlated in invasive carcinoma [21]. However, potential cooperation between RhoC overexpression and ErbB2/Neu has never been directly studied in vivo. To test for this, we crossed our RhoC transgenics to two different models for ErbB2/Neu+ breast cancer: FloxNeoNeuNT (with an activated NeuNT cDNA targeted to the mouse ErbB2 locus but preceded by loxP-stop-loxP sequences) [22] and NIC (where a NeuNDL2-5-IRES-Cre transcript is regulated by the MMTV LTR) [23]. Previous work has shown that Cre-dependent FloxNeoNeuNT mice develop mammary tumors at a mean age of 15 months [22], whereas MMTV-NIC mice develop tumors as early as 4 months [23]. FloxNeoNeuNT;MMTV-Cre mammary tumors, for the most part, do not metastasize [22]. For our experiments, we used MMTV-CreNLST to activate NeuNT expression in FloxNeoNeuNT mice [24]. This Cre transgenic line is mammary-specific but appears to express in fewer mammary epithelial cells or at a lower level in mammary epithelium than other MMTV-Cre delete strains, including MMTV-CreLine7, which was used previously to activate NeuNT expression in FloxNeoNeuNT mice [22]. Indeed, only 4/30 FloxNeoNeuNT;MMTV-CreNLST mice even developed mammary tumors, and all of these occurred in very old animals (Additional file 1: Supplementary Figure S4A and S4B). Mammary tumors in these mice were predominantly squamous (Additional file 1: Supplementary Figure S4C).

Many mammary tumors that form in FloxNeoNeuNT model mice select for amplification of the ErbB2/NeuNT locus [22]. In fact, it has been suggested that amplification of ErbB2 is a mechanism to circumvent repression of the ErbB2 promoter by Gata4 and other DNA-binding proteins [25, 26]. We tested for this by deleting one copy of Gata4 in this model. While trending towards decreased latency, mammary tumor formation in Gata4loxP/+;FloxNeoNeuNT;MMTV-CreNLST mice was not significantly different than seen in FloxNeoNeuNT;MMTV-CreNLST controls (Additional file 1: Supplementary Figure S4A and S4B). Most tumors that formed in Gata4loxP/+;FloxNeoNeuNT;MMTV-CreNLST mice were either poorly differentiated adenocarcinomas or solid nodular carcinomas (a histology commonly associated with transformation by activated Neu [23, 27]) (Additional file 1: Supplementary Figure S4C). As tumor latency was not significantly affected by heterozygous deletion of Gata4, both cohorts (FloxNeoNeuNT;MMTV-CreNLST and Gata4loxP/+;FloxNeoNeuNT;MMTV-CreNLST) were combined and used as controls for the effect of RhoC in a greater number of animals (see “NeuNT controls” below and in Fig. 2A).

Fig. 2

RhoC overexpression cooperates with endogenously driven activated Neu (FloxNeoNeuNT) to enhance tumor formation. A Kaplan Meier survival curve showing cooperation between FloxNeoNeuNT and RhoC overexpression. Death due to mammary tumor end-point was compared between cohorts. Statistical analysis for KM survival curves was calculated using Log-rank (Mantel-Cox) test via GraphPad Prism (shown in the table below) and p-values of less than 0.05 are considered significant (red text). B Graph comparing the ages between cohorts at the end-point due to mammary gland tumors. C The column graph shows the mammary tumor histology. Tumor types are represented with different colors. Mammary tumor histotypes are divided as below. ASC, Adenosquamous carcinoma; SC, Squamous cyst; SCC, Squamous cell carcinoma; PDA, Poorly differentiated adenocarcinoma; Pap, Papillary adenocarcinoma; SNC, Solid nodular carcinoma; AME, Adenomyoepithelioma; CAC, Complex adenocarcinoma; STC, Scirrhous tubular carcinoma; SCT, Spindle cell tumor. FloxNeoNeuNT controls (or NeuNT controls) contain data from FloxNeoNeuNT;MMTV-CreNLST and Gata4loxP/+;FloxNeoNeuNT;MMTV-CreNLST cohorts (for separate analysis of these cohorts, see Additional file 1: Supplementary Figure S4). Note: CreT = CreNLST

Ectopic expression of RhoC dramatically reduced tumor-free survival (Fig. 2A). In addition, R26RhoC/+;FloxNeoNeuNT;MMTV-CreNLST mice developed mammary tumors much faster than controls: as early as 4.5 months, whereas the average age at which tumors formed in control mice was close to a year and a half (Fig. 2B). While not significant, a trend towards an increased number of R26RhoC/+;FloxNeoNeuNT;MMTV-CreNLST mice with metastasis was also seen (Additional file 1: Supplementary Figure S5A, S5B). On a NIC background, RhoC did not significantly alter mammary tumor-free survival curves, although RhoC-NIC model mice did die from mammary tumors at a significantly younger age than NIC controls (Additional file 1: Supplementary Figure S6A and S6B). This relatively subtle effect is likely related to the short latency for tumor formation in this model. Tumors in NIC model mice, with or without ectopic RhoC expression were almost exclusively solid nodular carcinomas (Additional file 1: Supplementary Figure S6C).

Elevated wildtype RhoC enhances EMT signaling in FloxNeoNeu NT model tumors

As noted above, high-level expression of activated ErbB2/Neu induces solid nodular carcinomas (SNC) in the mouse mammary gland [23, 27]. In contrast, activated ErbB2/Neu when expressed at a lower level in FloxNeoNeuNT:MMTV-CreNLST mice resulted in tumors with multiple different histologies [22]. Enhanced mammary tumor formation in R26RhoC/+;FloxNeoNeuNT;MMTV-CreNLST mice raises the possibility that RhoC-expression could alleviate a requirement for transgene amplification, at least not to the same extent as seen in our combined control cohort tumors. Therefore to assess amplification of the ErbB2/NeuNT locus in R26RhoC/+;FloxNeoNeuNT;MMTV-CreNLST and controls, we used digital droplet PCR-based copy number analysis for ErbB2/NeuNT and Grb7 (the neighboring gene). Indeed, R26RhoC/+;FloxNeoNeuNT;MMTV-CreNLST tumors had a mean of 5.5 and 3.8 copies of ErbB2/NeuNT and Grb7, respectively. In contrast, NeuNT controls showed an average of 618 and 676 copies. While these mean values appear very different, due to the wide variation seen for copy number changes at the ErbB2/NeuNT locus in controls, these differences are not significant (Additional file 1: Supplementary Figure S7).

Most mammary tumors in FloxNeoNeuNT control mice were poorly differentiated adenocarcinomas, solid nodular carcinomas, or tumors with squamous differentiation (Fig. 2C). A similar mix was seen in R26RhoC/+;FloxNeoNeuNT;MMTV-CreNLST mice, although many tumors in this cohort showed a heterogeneous or complex histological pattern (Fig. 2C). Next, to identify transcriptional changes linked to RhoC-mediated accelerated mammary tumor formation, we performed bulk RNA-seq analysis on tumors from R26RhoC/+;FloxNeoNeuNT;MMTV-CreNLST and control cohorts (Additional file 1: Supplementary Table S1). Differential gene expression analysis was then performed using the DESeq2 tool within R. Tumors from the same cohorts clustered together by principal component analysis (PCA). Next, we performed pathway enrichment analysis using GSEA (Additional file 1: Supplementary Tables S2 and S3) and gProfiler (Additional file 1: Supplementary Tables S4). GSEA does not require a threshold to categorize differentially and non-differentially expressed genes. Therefore, the complete gene list identified from DESeq2 analysis was used. EMT, p53, Notch and WNT/β-catenin pathway signatures were increased in RhoC cohort (R) tumors, while Interferon α/Immune responses, E2F targets, Myc targets and G2M checkpoint pathways were decreased (Fig. 3A as well as Additional file 1: Supplementary Tables S2 and S3). EMT signature changes included significantly altered expression of Dst, Msx1, P3h1, Notch2, Magee1, Tgfb1, Serpinh1, Tnc, Fbln2 and Bmp1 (Fig. 3B and Additional file 1: Supplementary Table S2). Consistent with the trend towards lower-level ErbB2/Grb7 copy number gains/amplification in R26RhoC/+;FloxNeoNeuNT;MMTV-CreNLST tumors, many of the genes near ErbB2 were expressed at a lower level in RhoC tumors as compared to controls, while ErbB2/Neu mRNA levels were similar to what was seen in control tumors (Additional file 1: Supplementary Figure S8).

Fig. 3

Gene Set Enrichment Analysis (GSEA) of RNA sequencing data from mammary tumor samples. A Enrichment plots profiling GSEA analysis based on mouse hallmark gene sets using differential gene expression data from R26RhoC/+; FloxNeoNeuNT; MMTV-CreNLST (experimental) mammary tumors compared to NeuNT controls. Gene expression associated with activation of p53, EMT, Notch and WNT/β-Catenin pathways was increased in the experimental group, while Interferon α response, E2F targets, Myc targets and G2M checkpoint pathways were increased in the control group. Only the top 4 up-/down-regulated pathways were shown and the rest can be found in Additional file 1: Supplementary Table 3. B Enrichment map visualization of the enriched pathways in mammary tumors from experimental and control mice. Nodes in the network represent pathways (Reactome, Biocarta, Wiki Pathways) and similar pathways with many common genes are connected. Node size is proportional to the number of genes in each node and colors indicate whether the member genes of a set are up (red) or down (blue) regulated in the experimental group compared to controls

Finally, Rho mutant oncogenes have been identified in some human tumors. To test for the selection of activating mutations within the RhoC transgene, we used PCR-sequencing. No such mutations could be identified in 14 tumors from R26RhoC/+;FloxNeoNeuNT;MMTV-CreNLST mice, indicating that wildtype RhoC was responsible for accelerating mammary tumor formation (Additional file 1: Supplementary Table S5).

RhoC overexpression cooperates with Pik3ca H1047R

PIK3CA mutations are frequently seen in breast cancer. Therefore to test for the effect of RhoC overexpression on mammary tumor induction by a different oncogenic driver, we also crossed R26RhoC mice to our model for PIK3CA-mutant breast cancer (R26-Pik3caH1047R;MMTV-CreNLST) [28]. 55% of R26-Pik3caH1047R/RhoC;MMTV-CreNLST mice developed mammary tumors, a similar proportion to that seen in R26-Pik3caH1047R;MMTV-CreNLST controls (45%). However, Pik3caH1047R/RhoC mice reached endpoint with mammary tumors, on average, 100 days earlier than seen in Pik3caH1047R mice (Fig. 4A and B). Most mutant Pik3ca tumors were adenosquamous carcinomas (42%), Adenomyoepitheliomas (AMEs) (43%), or Squamous Cysts (SCs)(6%) (Fig. 4C). In contrast, R26-Pik3caH1047R/RhoC;MMTV-CreNLST mice developed more spindle-family tumors (Fig. 4C). A coincidental reduction in the percentage of AMEs was evident (Fig. 4C). This result is also consistent with induction of EMT signature gene expression as seen in FloxNeoNeuNT model tumors discussed above.

Fig. 4

RhoC overexpression cooperates with activated Pik3ca (H1047R mutant) to enhance tumor formation. A Kaplan Meier survival curve showing cooperation between Pik3caH1047R and RhoC overexpression. Death due to mammary tumor progression was compared between cohorts. Statistical analysis for KM survival curves were calculated using Log-rank (Mantel-Cox) test via GrapPad Prism (shown in the table below) and p-values of less than 0.05 are considered significant. B Graph comparing the ages between cohorts when mice died due to mammary gland tumors. C The column graph shows mammary tumor histology for each cohort. Tumor types are represented with different colors. Mammary tumor histotypes are divided as below. ASC, Adenosquamous carcinoma; SC, Squamous cyst; SCC, Squamous cell carcinoma; PDA, Poorly differentiated adenocarcinoma; Pap, Papillary adenocarcinoma; AME, Adenomyoepithelioma; STC, Scirrhous tubular carcinoma; SCT, Spindle cell tumor; ST, Scirrhous tumor; CAC, Complex adenocarcinoma. Note: CreT = CreNLST

RhoC overexpression does not enhance ErbB2 gene expression or PI3K/Akt signaling in T47D cells

One possible explanation for RhoC overexpression cooperating with ErbB2/Neu and Pik3ca oncogenes in transformation of mammary epithelium could involve RhoC-mediated enhancement of ErbB2 expression and/or PI3K to Akt signaling. To test for this, we assessed the effect of RhoC on both parameters in transiently transfected T47D breast cancer cells. This cell line was chosen since it expresses ErbB2 [29] and has an H1047R mutation in PIK3CA [30]. Despite overexpression of ErbB2/NeuNT in R26RhoC/+; FloxNeoNeuNT;MMTV-CreNLST tumors without apparent selection for high-level amplification of ErbB2/NeuNT (see Additional file 1: Supplementary Figure S9)), overexpression of RhoC did not enhance ERBB2 protein accumulation in transfected cells (Additional file 1: Supplementary Figure S9). Similarly, based on Threonine 308 or Serine 473 phosphorylation of Akt proteins (Additional file 1: Supplementary Figure S9), overexpression of RhoC also did not significantly enhance PI3K to Akt signaling. Thus, while RhoC overexpression cooperates with both oncogenic proteins/pathways, this effect is not easily modeled in vitro and may well relate to non-cell-autonomous effects of RhoC in the tumor microenvironment.