PERSPECTIVES ON CLASSIC ARTICLES: Metastasis Suppressor Genes

The metastasis suppressor gene field was launched in 1988 with the publication of an article in JNCI identifying nm23 (1). Despite the devastating consequences of tumor metastasis, little of its molecular regulation was known in 1988. Most studies at that time focused on the acquisition of traits, principally those involved in invasion. The instability of metastatic cell lines in terms of their behavior, gene expression patterns, and karyotypic abnormalities was thought to preclude the existence of consistent underlying molecular events. Genes up- or down-regulated in metastasis were identified almost exclusively in rodent model systems. We took a different approach and compared seven cell lines derived from a single murine K-1735 melanoma using differential colony hybridization (1). All of the cell lines formed primary tumors on intravenous injection into syngeneic and nude mice, but they varied widely in metastatic potential. Differential colony hybridization identified one cDNA, nm23, whose expression was quantitatively reduced in five highly metastatic cell lines as compared with two related, less metastatic cell lines. A similar pattern was observed in an independent model system, NMU-induced rat mammary carcinomas. The data suggested that consistent changes in gene expression occurr in metastasis, and that the expression of specific genes is reduced in tumors with high metastatic potential. The latter observation, occurring on the heels of the discovery of tumor suppressors (2), prompted the hypothesis that metastasis may also be controlled by suppressor genes. Both concepts, novel at the time, were hotly debated. The nm23 manuscript was rejected three times; one review, which I still have, questioned the entire experimental approach: “This reviewer cannot understand the rationale of the entire experimental plan described here....As we know, when a cancer cell undergoes transformation (or acquires metastatic ability), 152 separate genes get turned on and 378 get turned off.... By studying correlates of phenotypic change, one has little if any chance of understandings its causes.” Time has proven this assertion unfounded. nm23 cDNAs have now been transfected into 10 different tumor cell lines. Injection of control and nm23- transfectants into mice has shown a statistically significant reduction in tumor metastatic potential by nm23 re-expression, with no effect on primary tumor size (3–11). This effect stands as the current operational definition of a metastasis suppressor. The metastasis suppressor field is exploding. The advent of microarray profiling of gene expression, using both rodent model systems and human tumors, has produced a wealth of candidate genes. Chromosome transfections have also identified functional genomic regions. Confirmed metastasis suppressors include Nm23, Differentiation related gene (Drg-1) (12), Src-suppressed C Kinase substrate (SSeCKs) (13), Vitamin D3 up-regulated protein 1 (VDUP)(14), the CRSP3 transcriptional coactivator (14), Mitogen activated protein kinase kinase 4 (15), Raf kinase inhibitor Rkip (16), RhoGDI2 (17), Brms1 (18), Kiss-1 (19), Claudin-4 (20), and Kai1 (21). Most of these genes consistently have no effect on proliferation in vitro or primary tumor size in vivo. Most of the metastasis suppressor genes would not have been identified a priori as being involved in the metastatic process. In particular, they do not appear to be involved in the well-studied processes of invasion or angiogenesis. As is also the case for many tumor suppressors, an extensive but incomplete hunt for the biochemical activity or pathway responsible for the suppression of metastasis is underway. Three conclusions have emerged from these studies: 1) Many metastasis suppressors act at the final stage of tumor cell colonization at the metastatic site. For both the Mkk4 and Kiss1 suppressors, transfectants reach the lungs at frequencies comparable to control transfectants but fail to either grow or survive (22,23). This observation provides strong evidence that re-expression of metastasis suppressor expression by micrometastatic tumor cells may have therapeutic effects on cancer progression. Metastasis suppressors are usually transcriptionally silenced, rather than mutated, in aggressive tumors. Thus, if well-tolerated agents could be identified that elevate metastasis suppressor expression and therefore function in micrometastatic tumor cells, this strategy might constitute a new therapeutic endeavor. 2) New activities continue to be linked to the metastatic process. Brms regulates gap junctional signaling between cells (24). Mkk4 may control the apoptotic response to stresses that occur in metastatic sites (25). Kiss-1 is thought to encode a precursor to a migration inhibitory cytokine, which binds to a G protein–coupled receptor (26–28). SSeCKS is a scaffold protein for the protein kinase A (PKA) and PKC pathways (13). 3) Metastasis suppressors affect multiple points of well-known signal transduction pathways. The Map kinase pathway is influenced by Mkk4, Rkip, and Nm23 interaction with Ksr (29), a scaffold for the Erk pathway. Brms1 interacts with histone deacetylase complexes and is postulated to influence transcription (30), and CRSP3 is a transcriptional coactivator. Invasion-related signaling pathways are influenced by RhoGDI2. These data highlight the important intersection of metastasis suppressor research and classical drug discovery, validation, and preclinical testing: Most drugs are now tested in vivo on short-term xenografts of primary tumors, the majority of which do not have metastatic activity. However, many drugs impact the signaling pathways listed. It therefore stands to reason that in patients with metastatic disease the readout of these pathways may be different than in patients with localized disease caused by loss of metastasis suppressor expression. A drug with efficacy on a primary tumor may not show similar effects on a metastatic lesion, in which signaling pathways are quantitatively altered. Drugs must therefore be tested using metastatically competent preclinical models (31). Despite 15 years of work, more remains unknown about Nm23 than is known. However, research on this gene holds promise for both basic research and translational advances. Nm23 is known to be a family of eight proteins occurring in all cellular compartments (32). We know that this gene possesses metastasis suppressive activity in a variety of tumor cell types. In vitro correlates of suppression include reduced invasion, motility and soft agar colonization, and induction of differentiation. Differentiation remains one of the key correlates of altered Nm23 expression in multiple model systems. Both in vitro and in vivo studies support a role for this gene in breast differentiation. Control- and nm23-H1 transfectants of the human MDA-MB-435 breast carcinoma cell line were cultured three-dimensionally in Matrigel in the laboratory of Dr. Mina Bissell. The nm23-H1 transfectants formed ascinar structures, secreted the basement membrane proteins laminin and type IV collagen to the basal side of the ascinus, and produced sialomucin (33). Recently, a team from France reported a knockout mouse for nm23-M1(34). Nm23-M1-/- mice exhibited growth retardation and pronounced mammary defects. In virgin mice, ductal elongation and branching was poor and the mammary gland failed to fill the fat pad. These morphological differences were overcome in pregnancy, but a functional defect persisted in feeding pups. In cells of neural origin, including PC12 cells, C6 gliomas, and neuroblastomas, increased Nm23 expression has been associated with differentiation (35–39). It is likely that the fundamental cellular role of Nm23 is in differentiation and that this same role contributes to a lower metastatic threshold in cells that express this gene. The Nm23 pattern of expression in tumor cohorts has been somewhat clear in breast cancer but has been subject to an unending debate in other cancer types. Nm23 is not an independent prognostic or predictive factor in any cancer cell type. The breast cancer data support the conclusion that altered Nm23 expression levels may be of functional significance in human, as opposed to mouse or cell line, cancer. Of the many possible reasons for variability in immunohistochemical studies, two stand out. The specificity of various anti-Nm23 antibodies for the eight family members varies, as do their nonspecific binding patterns. Second, quantitation systems for proteins whose reduced expression signals poor outcome are conceptually difficult. What should constitute the poor-prognosis subset? A computer-generated cutpoint? Any low-expressing cells? Mostly low-expressing cells? Is the percentage of cells staining or rather the intensity of staining important? Is the comparison of tumor to surrounding normal cells important? The contribution of different grading systems to the disparity of conclusions is likely important. We rationalized that it takes only a few metastatic cells to kill a patient and classified tumors with a nest of low–Nm23 expressing cells as a putative poor-prognosis subset. This grading system generated a statistically significant correlation in our laboratory (40) as well as in others’ (41–43). Cutpoints where mixed populations of Nm23+ and Nm23- cells fell into the poor-prognosis category also generated statistically significant correlations (44,45). Nm23 expression patterns in cohort studies remain more than an academic question, however. If we identify compounds to elevate Nm23 expression in micrometastatic tumor cells, patients will be enrolled in a clinical trial based on low Nm23 expression in their primary tumor. It’s time to get it “right.” I remain intrigued by studies in which Nm23 expression is consistently increased concomitantly with aggressiveness, as in neuroblastoma (46,47). The mechanism behind these stimulatory effects is incompletely known and may involve mutation (48), alternate forms of the protein (49), or involvement in distinct signaling pathways. Determining the biochemical mechanism(s) by which Nm23 suppresses metastasis is of vital importance. Papers describing a new binding partner or activity for this protein appear almost weekly, and five biannual meetings have been held on this topic. At least four classes of Nm23 biochemical activities have been described that may mediate its biological effects, including protein–protein interactions (39,50–59), regulation of GTP-binding protein function (60–65), DNA-associated activities (66–69), and histidine-dependent protein phosphotransferase activity (70–72). Site-directed mutagenesis of nm23-H1, followed by transfection of the wild-type or mutated cDNAs into a human breast carcinoma cell line and readout of in vitro motility, have linked the histidine protein kinase activity of Nm23-H1 with motility suppression (72–74), and we continue to search for physiologic substrates of this unusual activity. One substrate is the Kinase suppressor of ras (Ksr), a scaffold protein for the Erk Map kinase pathway (29). We are working on the hypothesis that Nm23-H1 binding and phosphorylation of Ksr may inactivate this scaffold protein, reducing Erk activation and metastatic colonization. The next frontier in metastasis suppressor research lies in its incorporation into the therapy of cancer patients. Recent papers have documented that standard chemotherapeutic agents may work in part through modulation of metastasis suppressor expression and function (75). For Nm23, several compounds have been identified that can elevate its expression in metastatically competent cell lines in vitro(76,77). The field now awaits evidence of in vivo elevation of metastasis suppressor expression and concomitant reduction in metastatic burden in model systems, as well as improvements in the specificity of the agents. Such agents could contribute to the search for molecularly defined therapies that can render cancer a chronic, treatable disease.