Unraveling the enigmatic complexities of BRMS1-mediated metastasis suppression

Abstract

Expression of BRMS1 causes dramatic suppression of metastasis in multiple in vivo model systems. As we gain further insight into the biochemical mechanisms of BRMS1, we appreciate the importance of both molecular and cellular context for functional metastasis suppression. BRMS1 associates with large chromatin remodeling complexes including SIN3:HDAC which are powerful epigenetic regulators of gene expression. Additionally, BRMS1 inhibits the activity of NFκB, a well-known transcription factor that plays significant roles in tumor progression. Moreover, BRMS1 coordinately regulates the expression of metastasis-associated microRNA known as metastamir. How these biochemical mechanisms and biological pathways are linked, either directly or indirectly, and the influence of molecular and cellular context, are critical considerations for the discovery of novel therapeutic targets for the most deadly aspect of tumor progression - metastasis.

1. Introduction

BRMS1 (BReast cancer Metastasis Suppressor 1) was first described in 2000 as a gene that suppressed metastasis without preventing primary tumor growth [1]. The discovery of BRMS1 was accomplished using a combination of techniques that began with prior karyotypic observations showing that long and short arms of chromosome 11 are often sites of amplification/deletion and are associated with the progression of breast cancer [2]. Phillips and colleagues hypothesized that transferring an intact copy of neomycin-tagged chromosome 11 into metastatic breast cancer cells (MDA-MB-435; neo11/435 with introduced chromosome 11) would alter metastatic potential [3]. Using the technique of microcell-mediated chromosomal transfer (MMCT), they found significantly fewer metastases in the neo11/435 compared to parental cells with no significant changes in orthotopic tumor incidence/growth. Following these experiments, Seraj et al. used differential display RT-PCR (DD-RTPCR) to identify transcripts with ≥5-fold higher expression in the neo11/435 [1]. One of these transcripts was sequenced, cloned, and transfected into MDA-MB-435 and MDA-MB-231 metastatic breast cancer cells. When injected into mammary fat pads of athymic mice, BRMS1-transfected MDA-MB-435 (435-BRMS1) cells demonstrated decreased incidence and number of metastases to lung and lymph nodes. Orthotopic tumor growth rates were similar to parental controls albeit a lag in growth in the 435-BRMS1 cells for approximately 1 week. MDA-MB-231 human breast cancer cells expressing BRMS1 (231-BRMS1) showed a significant expression-dependent decrease in metastatic potential when cells were injected into the lateral tail veins of athymic mice. These cells were still capable of forming tumors in the mammary fat pads with similar growth rates as the parent controls. These results demonstrated that BRMS1 suppresses metastasis without blocking tumorigenicity, satisfying the functional definition of a metastasis suppressor gene.

Since then, many labs using different model systems showed BRMS1 to be a metastasis suppressor for breast carcinoma, melanoma, ovarian carcinoma, and non-small cell lung carcinoma (reviewed in [4]). Although not formally demonstrated to suppress bladder cancer, Seraj et al. found reduced BRMS1 mRNA in a highly metastatic variant of a human bladder carcinoma cell line T24T compared to the less metastatic parental line T24 [5]. This pattern of expression was also noted for one other metastasis suppressor, RhoGDI2. Robertson et al. suggested the presence of a tumor suppressor on the long arm of chromosome 11 in human melanomas and BRMS1 was subsequently shown to be a bona fide melanoma metastasis suppressor gene [6;7]. Quantitative real-time RT-PCR revealed that BRMS1 mRNA expression is high in melanocytes, reduced in early stage melanoma-derived cell lines, and barely detectable in advanced or metastatic melanoma cell lines. Introducing BRMS1 into the highly metastatic melanoma cell line C8161 significantly inhibited metastases in both spontaneous and experimental assays without preventing orthotopic tumor growth. Similar findings were noted for models of ovarian carcinoma [8] and NSCLC [9].

The BRMS1 gene maps to 11q13.1–13.2 [1]. This is within a region that is often lost in late stage breast cancers and is near sites that are among the most common amplifications and deletions associated with breast cancer progression [2]. Within the 5’ upstream region there are several putative regulatory elements including GATA-1, CREB, GATA-2, and CdxA. There is no TATA box suggesting that transcription of BRMS1 proceeds in a TATA-independent mechanism [1]. Two putative CpG islands were identified in the promoter region of BRMS1 (−3477 to −2214 and −531 to +608) and using methylation specific PCR it was determined the distal CpG island (−3477 to −2214) was methylated in tumorigenic and metastatic cell lines leading to the suppression of BRMS1 expression [10].

In addition to the wild-type transcript, the mRNA for several splice variants was identified. Wild-type BRMS1 has 10 exons spanning 741 nucleotides. BRMS1.v2 contains an alternative splice site in exon 10, BRMS1.v3 has an alternative splice site in exon 5, lacks exons 6–9, and alternatively splices exon 10, and BRMS1.v4 lacks exon 9 [11]. These splice variants are differentially expressed in metastatic compared to non-metastatic breast cancer cell lines. However, the functional relevance of each variant is presently unknown since translation to protein has not yet been verified.

The BRMS1 protein is 246 amino acids and is predicted to be 28.5 kDa [1]; however, it migrates closer to 35 kDa on SDS-PAGE [11]. This is presumably due to the highly charged N-terminal glutamate rich region within the first 50 amino acids. There are two coiled-coil regions between AA51-81 and AA147-180 and two predicted nuclear localization sequences between AA198-205 and AA239-245. The first NLS is required for localization to the nucleus [12;13]. It is not yet clear if the second NLS has any functional relevance to localization and/or trafficking although we have preliminary data suggesting roles in nuclear retention (D.R. Hurst and D.R. Welch, unpublished). Several putative phosphorylation sites exist however there have been no reports of phosphorylation. Other members of the SIN3 complex including BRMS1 family members have been shown to be phosphorylated by cell cycle regulatory kinases [14;15]. However, whether BRMS1 phosphorylation occurs and whether the modifications by cell cycle machinery are responsible for, or a consequence of, different growth in orthotopic versus ectopic tissues remains to be determined. There are no predicted glycosylation sites. The protein is stabilized by HSP90 and turnover is proteasome dependent [16].

BRMS1 is part of a family that includes suppressor of defective silencing 3 (SUDS3 or mSds3) and BRMS1-like (BRMS1L or p40) [17;18]. A conserved region known as the Sds3-like domain is present in each of these proteins. This region has not yet been functionally defined although it is likely important for protein-protein interactions. BRMS1 is 23% identical and 49% similar in amino acid sequence to SUDS3 and 57% identical and 79% similar to BRMS1L. Expression of SUDS3 in metastatic breast cancer cell lines does not significantly suppress metastasis as would be predicted, suggesting that, although similar, these proteins have distinct functions and do not functionally compensate for one another [17].

High sequence homology at both the DNA and amino acid levels of BRMS1 is found in multiple species. In particular, the murine ortholog (Brms1) is 85% homologous at the DNA level and the amino acid sequence is 95% identical [19]. In related subpopulations of tumor cells derived from a single mammary carcinoma, but varying in metastatic potential, Brms1 expression was inversely correlated with increased aggressiveness. Importantly, Brms1 functions as a metastasis suppressor in murine mammary carcinomas. Upon transfection of Brms1 into 66c14 or 4T1, metastasis was significantly suppressed while orthotopic tumor growth was indistinguishable from parental or vector-only transfectants [19;20]. Interaction with HDAC1 was demonstrated by co-immunoprecipitation suggesting an association with similar chromatin remodeling complexes. Altogether, these results suggest that the mouse ortholog has similar functionalities as human BRMS1. Interestingly, a genetic mapping study showed that metastatic ability of mammary tumors using inbred mouse strains was linked to allelic variations of Brms1 [21;22]. Recent, as yet unpublished data from our laboratories shows that transgenic expression of Brms1 in F1 generation MMTV-PyMT crosses yields significantly reduced lung metastases while not affecting primary tumor development (L.M. Cook, D.R. Hurst, and D.R. Welch, in preparation).

Although it has been clearly demonstrated that BRMS1 suppresses metastasis, determining specific functions and delineating mechanisms of suppression have proven to be difficult. BRMS1 does not completely block any single step in the metastatic cascade. Rather, it inhibits several hallmarks of metastasis leading to 80–90% inhibition in mouse models of metastasis. Proteomic and gene array studies have identified multiple changes in protein coding and non-coding genes when BRMS1 is expressed [23–26]. These expression changes may be responsible for driving the multiple phenotypic changes observed including invasion and migration, restoration of gap junctional intercellular communication (GJIC), and anoikis. Two mechanisms have been identified that could lead to gene expression changes that may be linked including interaction with chromatin remodeling complexes and inhibition of NFκB activity. These are discussed in greater detail below.

2. Regulation of gene expression…but not a transcription factor

The coordinate expression of genetic programs is necessary to enable a cancer cell to complete all the required steps of the metastatic cascade [27–29]. Molecules that regulate gene transcription could dramatically impact the metastatic process. There have been multiple reports showing that BRMS1 alters the expression level of metastasis associated genes. The questions that have not yet been fully answered are how? and which are the most relevant complexes? There is no evidence for BRMS1 functioning as a transcription factor. However, there is concrete evidence that it associates with transcriptional repressive chromatin remodeling complexes. BRMS1 presumably regulates the transcription of genes by interaction with large SIN3:HDAC chromatin remodeling complexes through the direct interaction with AT rich interacting domain 4A (ARID4A) and suppressor of silencing 3 (SUDS3) leading to the suppression of basal transcription [30;31]. These findings have been confirmed by protein-protein interaction studies of other proteins known to be a part of these complexes in addition to BRMS1 [18;32–36].

Chromatin remodeling complexes are diverse and incompletely characterized. Much more is known regarding lower species and, although there are many similarities with the complexes in mammals, they are much more complex and exist in many forms compared to yeast or lower organisms. For example, in S. cerevisiae, two distinct SIN3 complexes have been isolated known as Rpd3L and Rpd3S (note: a yeast ortholog of BRMS1 has not been identified) [37]. However, we and others have demonstrated that SIN3 complexes in mammalian cells exist in many sizes and are not limited to only two. The mammalian core components have been detected in multiple size exclusion FPLC fractions from cell lysates in sizes ranging from several hundred kDa to those exceeding 2 MDa. Because these components are found in a range of sizes demonstrates the existence of multiple complexes. Understanding the composition of these differently sized complexes will be important to determine how they function.

Many components of the SIN3 complexes play tumor suppressive roles in cancer. Typical studies have been done by ectopically expressing one gene and characterizing the nature of tumor inhibition. Often, clinical studies of expression of a protein in cancer progression do not correlate because it is important to define the nature of the complex as a whole and not an individual protein. How the total composition of SIN3 complexes is playing a role in cancer is currently poorly-defined. However, it has been shown that altering the composition of SIN3 complexes by expressing BRMS1 or BRMS1 mutants results in ~90% suppression of metastasis [30]. The function of these SIN3 complexes is dependent upon the specific modification of the protein composition and further emphasizes the need to characterize these complexes as a whole rather than by individual protein expression.

Gene expression regulated through chromatin remodeling complexes occurs by multiple mechanisms [38]. In simple terms, although it is much more complicated, histone acetyltransferase (HAT) containing complexes add acetyl moieties to specific histone lysine side chains resulting in an open conformation of DNA (euchromatin) that is more accessible to transcription factors. Histone deacetylase (HDAC) containing complexes remove acetyl moieties from the lysine side chains resulting in a more compact conformation of DNA (heterochromatin) causing a repression of transcription. SIN3 (switch-independent 3) is a scaffold that is well-known for recruitment of HDACs capable of transcriptional silencing [39;40]. While the majority of current studies with chromatin remodeling complexes are focused on HAT or HDAC activity, these complexes recruit many other protein and DNA modifying enzymes in addition to deacetylases including glycosyl transferases and methyl transferases. It is now realized that many of these enzymatic modifications are linked and influence the function of protein and DNA interacting molecules. The functional specificity (i.e., which genes they regulate and, as a result, the phenotypes that are altered) of these complexes is dictated by their protein composition and resulting enzyme activity.

HDAC inhibitors are currently in clinical trials since they promote apoptosis, inhibit invasion and angiogenesis, and cause cell cycle arrest [41;42]. However, the mechanisms for how HDAC inhibitors are functioning are not clear. There are currently 11 zinc containing HDACs and 7 NAD⁺ - dependent HDACs that are found in nearly every cell compartment. They have many diverse functions since they are not limited to deacetylation of chromatin modifying histones in the nucleus. It is important to understand how HDACs and other enzymes are functioning in cancer progression so that inhibitors can be used appropriately. Matrix metalloproteinase (MMP) inhibitors were rapidly advanced into clinical trials because they very effectively inhibited invasion [43]. Unfortunately, clinical trials with MMP inhibitors failed. The failure was in part due to the fact that the biology of MMPs was not well understood and their activities were much more diverse than originally thought. A similar situation could occur with the HDAC inhibitors if we fail to understand the biology and fail to identify proper targets specifically for metastasis.

Recently Smith et al. showed that select HDAC inhibitors cause changes to the composition of SIN3 complexes independent of HDAC inhibition that could lead to significant changes in complex specificity [34]. This may lead to differential efficacy of HDAC inhibitors in the treatment of cancer caused by these secondary effects. Since many of these inhibitors are being tested in clinical trials, it is important to understand how SIN3 complexes are functioning and how the composition could be affecting enzyme activity. It is not completely clear how (or even if) BRMS1 alters HDAC activity although some reports suggest it functions by recruiting HDAC1 to enhance HDAC activity at specific locations [44;45].

The recruitment of HDAC1, and presumably SIN3 complexes, by BRMS1 leads to inhibition of NFκB activity. BRMS1 has been shown to recruit HDAC1 to NFκB consensus binding regions using ChIP assays [44;45]. It has been suggested that HDAC1 leads to H3 deacetylation and reduced binding of p65 at the NFκB site of specific promoters. These data imply that BRMS1 is recruited, possibly via SIN3 complexes, to promoter binding regions where it inhibits HDAC1 activity leading to deacetylation of p65 and reduced transactivation by NFκB. Acetylation is an important regulatory mechanism in BRMS1-dependent gene regulation of NFκB transactivation. The p65 subunit of NFκB is acetylated at lysine residues by several co-activators including p300, CBP and p/CAF. K310 is a prominent residue playing a key role in p65 transcriptional regulation. Interestingly, BRMS1 abolishes TNF-dependent acetylation of p65 on K310 and results in a substantial decrease in the transactivation potential of p65 [45]. In that same study, BRMS1 promoted HDAC1 binding to the RelA/p65 subunit of NFκB where HDAC1 promoted the deacetylation of K310 on p65 at NFκB binding sites at the promoter regions of cIAP2 and Bfl-1/A1 leading to the loss of NFκB -dependent transcriptional activation.

An interesting regulatory feedback loop with microRNA-146a (miR-146a) has been described for NFκB by Baltimore’s group [46]. Upon stimulation by LPS, NFκB upregulates miR-146a which in turn targets the translation of the upstream signaling molecules IRAK1 and TRAF6 [47]. BRMS1 also upregulates miR-146a and expression of miR-146a or miR-146b in MDA-MB-231 cells is sufficient to suppress metastasis [48]. It is not clear how BRMS1 upregulates miR-146a since it negatively regulates NFκB although it could be dependent on the particular environmental stimulus. Other metastasis-associated microRNAs (mestamir) have been found to be coordinately regulated by BRMS1 [23]. Molecular pathways, including metastamir, regulated by BRMS1 (and presumably SIN3 complexes) are beginning to be elucidated that significantly impact hallmarks of metastasis [48–51](FIGURE 1). Understanding the biochemical mechanisms of BRMS1:SIN3:NFκB and how these molecules are affected by the microenvironment will prove important to map key pathological pathways.

Figure 1.

Pathways regulated by BRMS1. The composition of the SIN3 complex dictates which genes are regulated. Multiple metastasis associated genes are regulated by BRMS1 including metastamir (green). The dotted lines indicate either indirect mechanisms or multiple steps not shown. References are provided in the text.

3. BRMS1 itself is epigenetically regulated

As described above it has become increasingly clear that epigenetic modifications play significant roles in the regulation of genes that drive cancer progression. In fact, the metastasis suppressor gene nm23 is silenced through promoter methylation and re-expression following treatment with 5-aza-2’-deoxycytidine was found to cause a decrease in motility of the breast cancer cells [52]. Breast cancer cell lines with differing tumorigenic and metastatic potential have been utilized to demonstrate decreased levels of BRMS1 with increasing aggressiveness [10]. The promoter region of BRMS1 contains two putative CpG islands leading to the hypothesis that BRMS1 expression is downregulated by promoter hypermethylation in advanced stages of cancer. Methylation specific PCR (MSP) was used to show that metastatic cells are hypermethylated in the distal CpG island (−3477 to −2214) of the BRMS1 promoter and treatment with 5-aza-2’-deoxycytidine reversed the methylation pattern leading to an increase in BRMS1 expression. However, one must be cautious since the expression level of BRMS1 in the cell lines tested has not been noted to be directly proportional to the CpG methylation pattern. At this time it remains to be determined if there are feedback or feedforward regulatory loops involving the expression level of BRMS1 and the interconnectivity of HDAC and methyltransferase activity.

4. Clinical relevance of BRMS1 expression

Promoter hypermethylation of BRMS1 in aggressive tumor cells might suggest that decreased expression of BRMS1 could serve as a clinically relevant biomarker. Indeed, a significant number of lymph node metastases in multiple matched patient tissues were hypermethylated in the distal CpG island (−3477 to −2214) [10]. Hypermethylation occurred in 60% of the lymph node metastases and 45% of the primary tumors, suggesting that an increase in methylation coincides with cancer development and progression. A recent study examining circulating tumor cells isolated from the peripheral blood of operable breast cancer patients identified hypermethylation of BRMS1 promoter sequences in patients with metastases compared to healthy control population [53]. Other studies found that loss of BRMS1 protein has been correlated with reduced disease-free survival when patient samples were stratified by loss of estrogen or progesterone receptor (ER, PR) or expression of HER2 [54]. Laser capture microdissection is necessary to ensure purity of the material from a heterogeneous mass and was used to show that BRMS1 localization shifting from nuclear to cytoplasmic is associated with highly proliferative ER-negative breast cancers [55]. More recently, decreased BRMS1 expression was identified in metastatic melanomas that is thought to contribute to angiogenesis and melanoma progression [56].

Consistent with the protein data, BRMS1 mRNA correlated with PR and loss of HER2 and loss of BRMS1 mRNA correlated with poor prognosis [57]. However, not all studies identified a correlation of decreased BRMS1 expression with cancer progression. In fact, clinical studies with BRMS1 have been relatively inconsistent with regard to patient survival and likelihood of metastasis. The inconsistencies are thought to occur mainly because most clinical studies were performed with measurements of mRNA; and it has been observed that BRMS1 protein and mRNA do not necessarily correlate [11]. As techniques are more refined and the target populations are more defined, studies with BRMS1 are starting to be stratified. Moreover, considering the biochemical mechanisms of BRMS1 function, it is unlikely that BRMS1 expression alone will be useful as a prognostic indicator of metastatic potential in cancer patients. As is the case with the majority of cancer biomarkers, expression of sets of genes rather than individual genes is more useful to predict disease progression. This further supports the need to understand the nature of the protein complexes with which BRMS1 is interacting and functioning so that clinically relevant biomarkers may be identified. Future studies may show identification of specific ‘complexome’ markers associated with cancer progression and metastasis.

5. Concluding thoughts

It has become increasingly recognized that metastasis is a relevant therapeutic target for cancer patients. Many genes involved in metastasis are regulated epigenetically and are therefore influenced by their surrounding environments. Understanding how metastasis regulatory genes are regulated by the microenvironment in which tumor cells find themselves will be a significant advancement to the understanding of the mechanisms of metastasis. By definition, metastasis suppressors are involved in differential responses to exogenous signals since cells grow in orthotopic but not ectopic locations. BRMS1 exemplifies how expressing- and non-expressing cells respond differentially to the same environmental cues [49]. Further dissection of the mechanisms by which the metastasis suppressors control cellular responses to signals (growth factors, matrix molecules, cell adhesion proteins, etc.) might allow design of new therapeutic strategies targeting metastasis by controlling those interactions.

BRMS1 and SIN3 chromatin remodeling complexes regulate specific coding and non-coding genes, depending on the composition of the complex. These complexes are likely regulated by signals from the environment surrounding the cell. We hypothesize that complex composition is likely regulated by signals from the environments surrounding tumor cells. As metastatic cells travel to distant sites, they continuously receive different signals from the changing microenvironments. Depending upon the signal(s) received, different complexes would be assembled. Then, depending on the complexes and the genes with which they interact, pro-survival, pro-apoptotic, or senescence “operons” would be initiated. BRMS1 may, in fact, alter chromatin remodeling complex assembly thereby leading to different cellular responses to the changing microenvironment. Experiments are currently underway to test these hypotheses and corollaries.