Syndecan-Mediated Ligation of ECM Proteins Triggers Proliferative Arrest of Disseminated Tumor Cells

Many patients with cancer harbor myriad, ostensibly dormant disseminated tumor cells (DTC) within their bodies (1). The vast majority of these DTCs are found as mitotically quiescent solitary cells, indicating that the inability of solitary DTCs to proliferate represents a major obstacle that precludes the eventual formation of macroscopic metastases (2). We and others previously studied the role of cancer cell:extracellular matrix (ECM) interactions, mediated by major ECM receptor integrins, in regulating the behavior of DTCs (3–5). Specifically, we characterized the behaviors of three mouse mammary carcinoma cell lines: D2.0R, D2.1, and D2A1 (6). As we found, after extravasating into the lung parenchyma of host mice, the nonaggressive D2.0R and D2.1 cells failed to assemble mature adhesion plaques containing integrin β₁ and therefore could not activate focal adhesion kinase (FAK), whose activity was critical in these cells for ERK activation and proliferation (Supplementary Fig. S1A and S1B). In contrast, the aggressive D2A1 cells did develop mature adhesion plaques, activated FAK and ERK, and ultimately proliferated rapidly (7).

Importantly, these findings did not explain why the absence of such integrin β₁-mediated adhesion signals should result, on its own, in the failure of the D2.0R and D2.1 cells to proliferate following extravasation in vivo. Thus, various growth factors present in the lung parenchyma, acting via their cognate receptors, should have provided an alternative source of mitogenic signals; however, for unknown reasons, these growth factors failed to activate Ras/ERK signaling and thereby enable the rapid proliferation of these nonaggressive cells (Supplementary Fig. S1C–S1E; refs. 8, 9).

In the present study, we identify a key signaling cascade accounting for the failure of growth factors and their receptors to drive DTC proliferation. We demonstrate that ligation of ECM proteins by syndecans triggers recruitment to the plasma membrane of the Par-3:Par-6:atypical PKC protein complex, which results in the functional inactivation of KSR scaffolding proteins, the blockade of Ras/ERK signaling, and ultimately, the proliferative quiescence of DTCs.

Lentiviral vectors that constitutively express protein-coding DNA sequences were generated using pLV-blast, pLV-empty, pLV-hygro, pLV-neo, or pLV-puro as a backbone (10). Transient expression vectors for short-guide RNA were generated using PX458 (addgene) as a backbone.

The mouse mammary carcinoma cell lines D2.0R, D2.1, D2A1 and 4TO7 were gifts from Dr. Fred R. Miller (Barbara Ann Karmanos Cancer Institute, Detroit, MI); the transformed mouse mammary epithelial cell line EpH4-Ras was a gift from Dr. Hartmut Beug (The Research Institute of Molecular Pathology, Vienna, Austria); the mouse mammary carcinoma cell line TS/A was a gift from Dr. Pier-Luigi Lollini (University of Bologna, Bologna, Italy); the human breast cancer cell line SUM149 and SUM159 were gifts from Dr. Stephen P. Ethier (Medical University of South Carolina, Charleston, SC) and maintained in accord with the sender's protocol. The mouse mammary carcinoma cell line EMT6, the mouse melanoma cell lines B16F0, B16F1, and B16F10, the mouse prostate cancer cell line TRAMP-C2 as well as the human breast cancer cell lines MCF7 and MDA-MB-231, were obtained from the ATCC; the 293FT cells were purchased from Thermo Fisher Scientific and maintained according to the provider's protocol. The Balb/c mouse origin of the murine mammary carcinoma cell lines D2.0R, D2.1, D2A1, EpH4-Ras, 4TO7, and 4T1 cells were confirmed by major histocompatibility complex (MHC) haplotype analysis. The D2.0R, D2.1, and the D2A1 cells were further authenticated by confirming their differing affinities to peanut agglutinin that were reported previously (11). All the cell lines mentioned above are routinely tested for Mycoplasma contamination every 6 months.

For CRISPR/Cas9-mediated gene knockout in the D2.1 cells, cells were transfected with PX458-based plasmid vector, which co-expresses an sgRNA targeting the gene of interest, the Cas9 nuclease, and the GFP protein. Seventy-two hours after the transfection, single GFP-positive cells were sorted into 96-well plates by fluorescence-activated cell sorting (FACS). Following expansion of individual clones, the expression levels in these clones of the gene of interest were analyzed either by flow cytometry (for syndecans) or immunoblotting (for Par proteins and PKCλ/ι).

Immunoblotting and immunoprecipitation were carried out as described previously (7). Most of the immunoprecipitation and/or immunoblotting results presented are representative of multiple experiments. The numbers of replicate experiments that were conducted for each of the immunoblot panels presented in this study are as follows: three times or more: Figs. 1A, D, 2B, C, E, 3A, C, E, 4A, 5A, D, and 6C; Supplementary Figs. S3A, S3C, S3E, S5A, S5C, S5D, S6A, S7A and S12B; twice: Figs. 3B, 6A; Supplementary Figs. S1A, S2B, S2C, S3D, S5B, S6C, S8A-D, S11C, S12C, S13B, S13D, S15A, S15B and S15D; once: Supplementary Figs. S1D, S6D, S7B, S9A, S16A and S16C.

Separation of cells into cytosol, membrane/organelle, and nuclei/cytoskeleton fractions were carried out using FractionPREP Cell Fractionation Kit (BioVision) in accordance with the manufacturer's protocol.

Immunostaining of in vitro cultured cells and frozen sections of the lungs were carried out as described previously. Images were acquired using confocal microscopy, which was performed on a Zeiss LSM 700 laser scanning confocal system equipped with Axio Observer (Carl Zeiss; ref. 7).

All animal experiments were approved by the Committee on Animal Care at the Massachusetts Institute of Technology under the protocol number 1014-109-17 and conducted in compliance with their guidelines. Tail-vein injection, intracardiac injection and mammary fat pad implantation were carried out as described previously (7). Metastasis development was monitored by bioluminescence imaging, which was carried out on the IVIS Spectrum In Vivo Imaging System (PerkinElmer) following intraperitoneal injection of 10 mL per gram body weight of 15 mg/mL D-luciferin (PerkinElmer).

Statistical analyses were carried out by the Student t test, unless otherwise indicated.

We undertook to explore the functionality of growth factor-mediated signaling pathway in DTCs using the three different D2 cell types as model systems. Specifically, we examined the apparent failure of the operation of the Ras/ERK cascade—which can be depicted as Ras ⇒ Raf ⇒ MEK ⇒ ERK—observed in the nonaggressive D2.0R and D2.1 cells in vivo. We found that neither c-Raf nor MEK exhibited differing levels of activating phosphorylation among the three different D2 cell types following their extravasation into the lung parenchyma (Fig. 1A and B). In contrast, ERK displayed intense phosphorylation and thus functional activation in the extravasated, aggressive D2A1 cells but not in the other two nonaggressive D2 cell types analyzed in parallel (Fig. 1A). We also noted that none of these three kinases exhibited noticeably different levels of phosphorylation among the various D2 cell types when they were propagated in monolayer cultures in vitro (Fig. 1A). Collectively, these observations suggested that after extravasating into the lung parenchyma, signaling through the Ras/ERK cascade was blocked at a point between MEK and ERK in the nonaggressive D2.0R and D2.1 cells (Fig. 1C). Similar defects in MEK-dependent ERK activation were observed in two other nonaggressive mouse mammary carcinoma cell types, 4T07 and EpH4Ras (12, 13), again following their extravasation into the lung parenchyma (Supplementary Fig. S2A and S2B).

We proceeded to test the effect of overexpressing constitutively active oncoproteins sitting at the top of the Ras/ERK cascade, namely NeuNT, H-RasV12, and c-Raf 22W (Fig. 1A). All of these oncoproteins had minimal stimulatory effects on the phosphorylation of ERK in the nonaggressive D2.0R and D2.1 cells that had infiltrated into the lung parenchyma as well as on their subsequent ability to develop metastatic colonies (Fig. 1D and E, Supplementary Fig. S2C). In contrast, these oncoproteins successfully stimulated activation of both MEK and ERK when expressed in the D2.0R and D2.1 cells that were growing in serum-starved monolayer cultures (Supplementary Fig. S3A). These observations confirmed the defective operation of Ras/ERK signaling in the nonaggressive cancer cells that had recently extravasated into the lung parenchyma and prompted us to study further the mechanistic basis behind this defect.

To investigate the defective operations of the Ras/ERK signaling cascade observed in the recently extravasated nonaggressive cancer cells in vivo, we decided to employ an in vitro model system. Among five different culture conditions tested, propagation in the Matrigel-based three-dimensional (3D) culture (Matrigel on-top, MoT; ref. 5) most faithfully recapitulated the behavior of the three different D2 cell types in the lung parenchyma (Supplementary Figs. S3A–S3E, S4). Using this 3D MoT culture model, we then tested the involvement of proteins that potentially modulate the level of ERK phosphorylation independent of the state of MEK phosphorylation, which included two classes of scaffolding proteins—kinase-suppressor of Ras (KSR) proteins and IQ-domain GTPase-activating proteins (IQGAP)—and MAP kinase phosphatases (MKP; Fig. 2A; refs. 14–18).

We observed by immunoblotting that none of the KSR, IQGAP and MKP isoforms tested exhibited noticeably different expression levels between the 2D monolayer and the 3D MoT cultures of the nonaggressive D2 cell types (Fig. 2B). Intriguingly, however, Phos-tag analysis (19) revealed that both KSR1 and 2, but none of the IQGAP and MKP isoforms tested, exhibited elevated phosphorylation when the nonaggressive D2.1 cells were growing in the 3D MoT culture relative to the level observed when these cells were propagated in the 2D monolayer culture (Supplementary Fig. S5A). Indeed, using a phosphorylation site-specific antibody, we found that KSR1 protein was highly phosphorylated on its S392 residue—a modification known to result in the loss of its scaffolding function (20)—in all of the nonaggressive cell types tested here, specifically when these cells were growing in the 3D MoT culture and following the extravasation of these cells into the lung parenchyma, but not when they are growing as 2D monolayers (Fig. 2B and C, Supplementary Fig. S5B).

Mechanistically, KSR1 S392 phosphorylation leads to the binding of this scaffolding protein to 14-3-3 proteins, causing its sequestration from the plasma membrane (the site of active Ras/Raf/MEK/ERK signaling) to the cytosol (Fig. 2D; ref. 20). Indeed, subcellular fractionation analysis revealed that both KSR1 and KSR2 were primarily present in the membrane/organelle fraction when the nonaggressive D2.0R and D2.1 cells were propagated in the 2D monolayer cultures, but were relocalized largely to the cytosol when these cells were growing under the 3D MoT cultures (Fig. 2E, Supplementary Fig. S5C). Hence, both KSR isoforms are functionally inactivated through phosphorylation and resulting cytosolic relocalization when the nonaggressive cancer cells were surrounded by 3D ECM both in vitro and, presumably, in vivo.

As might have been anticipated, the expression of the constitutively active, nonphosphorylatable KSR1 S392A mutant (20) efficiently restored ERK phosphorylation and proliferation in the D2.1 cells that were growing in the 3D MoT culture (Supplementary Fig. S5D–S5E). Moreover, expression of this KSR1 S392A mutant also stimulated the ability of the otherwise-indolent D2.1 and EpH4-Ras cells to colonize the lungs following tail-vein injection into mice (Fig. 2F, Supplementary Fig. S5F), reinforcing the functional significance of KSR phosphorylation in controlling the proliferation of DTCs.

We sought to identify the kinase(s) catalyzing the observed phosphorylation and resulting inactivation of KSR proteins. On the basis of the previously reported role of Par-1a and Par-1b kinases as mediators of KSR1 S392 phosphorylation (20, 21), we generated, via CRISPR/Cas9-mediated gene knockout, the Par-1a-deficient (ΔPar-1a) and Par-1b–deficient (ΔPar-1b) derivatives of D2.1 cells. Subsequent experiments using these D2.1 cell derivatives revealed that D2.1-ΔPar-1b cells and, more modestly, D2.1-ΔPar-1a cells exhibited reduced KSR1 phosphorylation on S392, an elevated ERK phosphorylation, and an increased proliferation rate—all relative to the levels observed in their parental counterparts when these cells were propagated under the 3D MoT cultures (Supplementary Fig. S6A–S6B). Moreover, the wild-type (WT), but not the kinase-dead mutant (K82R), of Par-1b restored KSR1 S392 phosphorylation in the D2.1-ΔPar-1b cells growing in 3D MoT culture (Fig. 3A). We further found that the D2.1-ΔPar-1b cells failed to exhibit detectable KSR1 S392 phosphorylation following extravasation into the lung parenchyma (Supplementary Fig. S6C). Hence, Par-1b, aided possibly by Par-1a, plays a critical role in phosphorylating and thereby inactivating the KSR1, and presumably KSR2, scaffolding protein(s) (14), ultimately blocking ERK activation and proliferation in the nonaggressive D2.1 cells growing under the 3D conditions (Fig. 2D, Supplementary Fig. S6D).

We proceeded to determine whether the elevation of Par-1b–dependent KSR phosphorylation, observed in 3D conditions, could be accounted for by change(s) in the expression level and/or the phosphorylation state of the Par-1b kinase (22). As we found, the level of phosphorylated Par-1b on T593 (or the corresponding T595 residue in human cells), but not the overall level of Par-1b protein, was elevated in multiple nonaggressive cancer cell types when they were growing in the 3D MoT culture relative to the level observed in the 2D monolayer culture of the same cell type (Fig. 3B, Supplementary Fig. S7A–S7B). Moreover, both D2.0R and D2.1 cells displayed intense phosphorylation of Par-1b on T593 following their infiltration of the lung parenchyma (Fig. 3B). Importantly, the nonphosphorylatable T593A mutant of Par-1b, unlike its wild-type counterpart, failed to drive KSR1 phosphorylation (on S392) when expressed in the D2.1-ΔPar-1b cells that were growing under the 3D MoT conditions (Fig. 3A). Together, these observations suggested that the Par-1b protein itself underwent increased activating phosphorylation (on T593 [mouse]/T595 [human]) when the nonaggressive cancer cell types were growing under the 3D conditions, resulting in its increased ability to phosphorylate and thereby inactivate its KSR substrates.

Contrary with this notion, we found via an in vitro kinase assay that the nonphosphorylatable Par-1b T593A mutant was slightly more (1.7-fold increase in phospho-KSR1 [S³⁹²] band intensity) efficient than its wild-type counterpart in catalyzing KSR1 phosphorylation (Supplementary Fig. S8A). Intriguingly, however, immunoprecipitation analyses revealed that the Par-1b WT protein, but not its T593A mutant, displayed a stronger physical association with KSR1 in the 3D MoT culture than in the 2D monolayer culture when ectopically expressed in the D2.1-ΔPar-1b cells (Fig. 3C and Supplementary Fig. S8B–S8C). Hence, the phosphorylation of Par-1b on T593 contributes critically to the increased physical association between Par-1b and its KSR substrates, which was observed specifically when the D2.1 cells were growing in the 3D MoT culture, potentially accounting for the observed enhancement of Par-1b–mediated KSR phosphorylation in these cells (Fig. 3D).

The phosphorylation of human Par-1b on T595 is reported to enable the translocation of this kinase from the plasma membrane into the cytosol, doing so by enhancing Par-1b:14-3-3 binding (22). Consistently, we observed that in the D2.0R and D2.1 cells growing under the 3D MoT conditions, the highly phosphorylated Par-1b protein (on T593) was primarily localized to the cytosol, whereas in the 2D monolayer culture of the same cells, the minimally phosphorylated Par-1b protein was more abundant in the membrane/organelle fraction (Fig. 3E, Supplementary Figs. S5C–S8D). We noted that the pattern of Par-1b subcellular localization mirrored that of the KSR1/2 scaffolding proteins, which appear to shuttle back-and-forth between the plasma membrane and cytosol (23). When taken together with the increased physical association between Par-1b and KSR1/2 observed in the 3D MoT culture (Fig. 3C), this caused us to hypothesize that when the nonaggressive cancer cells are propagated in the 3D conditions, the increased phosphorylation of Par-1b on T593 liberates this kinase from the plasma membrane, allowing it to bind to and form a stable complex with the KSR proteins that are already present in the cytosol, leading to Par-1b–mediated phosphorylation and inactivation of KSR scaffolding proteins (Fig. 3D, Supplementary Fig. S8E).

We proceeded to identify the kinase(s) that catalyze(s) Par-1b T593 phosphorylation. Others had previously reported that this phosphorylation is mediated by either of the two atypical protein kinase C (aPKC) isoforms—PKCλ/ι and PKCζ (22). Indeed, the D2.1 cells that had been rendered deficient for PKCλ/ι—the major aPKC isoform expressed in these cells (Supplementary Fig. S9A)—failed to display detectable phosphorylation of Par-1b on T593 even under the 3D MoT conditions (Fig. 4A).

In fact, this aPKC-dependent Par-1 phosphorylation comprises a critical component of the molecular program governing cell polarity, where mutual antagonism between Par-1 and Par-3:Par-6:aPKC complex (i.e., the aPKC complex) for residence on the plasma membrane results ultimately in the mutually exclusive localization patterns of the Par-1 and aPKC complex on this membrane, providing a mechanistic foundation for multiple distinct types of cell polarity (Supplementary Fig. S9B; refs. 24, 25). This prompted us to examine the subcellular localization of PKCλ/ι and other regulators of cell polarity. As we found, all three components of the aPKC complex were localized predominantly in the membrane/organelle fraction when the nonaggressive D2.0R and D2.1 cells were growing in the 3D MoT, but not in the 2D monolayer, cultures (Fig. 3E, Supplementary Fig. S5C). This contrasted starkly with the precisely opposite behavior of the Par-1b protein (Fig. 3E).

These observations were reinforced by our subsequent immunofluorescence analysis, which revealed that Par-3 is primarily distributed evenly along the plasma membrane of the nonaggressive D2.1, EpH4-Ras, and 4TO7 cells, whereas Par-1b mainly exhibits diffuse cytosolic localization, when these cells were growing in the 3D MoT cultures (Fig. 4B–D, Supplementary Fig. S10A–S10B). We also found that fluorescent fusion proteins containing the PKCι (clover-PKCι) were largely spread uniformly over the plasma membrane, whereas the fusion protein containing Par-1b (clover-Par-1b) was mostly dispersed throughout the cytosol (Fig. 4E, Supplementary Fig. S10C–S10D), this being observed in both the D2.1 and EpH4Ras cells residing in the lung parenchyma. To summarize, when these nonaggressive cancer cell types were growing under the 3D conditions, the inner surface of the plasma membrane appeared to be occupied by the components of the aPKC complex; the Par-1 kinase behaved in the opposite fashion, being excluded from the plasma membrane and liberated into the cytosol.

We also discovered that ECM proteins surrounding individual cancer cells in the 3D conditions appeared to contribute functionally to the above-mentioned localization patterns of cell polarity-regulating proteins. Thus, Par-3 failed to accumulate at the plasma membrane in the D2.1 cells that were growing in suspension culture, in which these cells were not exposed to abundant ECM proteins (Supplementary Fig. S10E). This finding caused us to speculate that the ECM proteins surrounding the cancer cells induce the accumulation of the aPKC complex to the plasma membrane, triggering in turn the eviction of Par-1 to the cytosol (Supplementary Fig. S11A).

To test this model, we examined the effects in the nonaggressive D2.1 cells of eliminating the expression of Par-3, the protein critical to the stable attachment of the aPKC complex to the plasma membrane (25). These experiments revealed that these Par-3–depleted D2.1 cells, unlike their Par-3-intact counterparts, failed to exhibit (i) eviction of Par-1b from the membrane fraction into the cytosol, (ii) elevated phosphorylation of Par-1b and KSR1 on their respective T593 and S392 residues, and (iii) impaired ERK activation, even when propagated under the 3D MoT conditions of culture (Fig. 5A and B, Supplementary Fig. S11B–S11C). Moreover, the Par-3–depleted D2.1 cells proliferated more rapidly than did the control D2.1 cells in 3D MoT culture (Supplementary Fig. S11D–S11E), and most importantly, these Par-3–depleted cells, unlike their parental counterparts, successfully colonized the lung tissue following tail-vein injection into mice (Fig. 5C; refs. 26, 27). Together, these observations supported the notion that ECM-dependent recruitment of the aPKC complex to the plasma membrane contributes critically to the signaling pathway that is responsible for the shutdown of Ras/ERK cascade in the nonaggressive cancer cells growing under the 3D conditions.

The above scheme motivated us to study further the mechanism(s) responsible for the ECM-dependent recruitment of the aPKC complex to the plasma membrane. Specifically, we postulated the involvement of cell-surface ECM receptors whose cytoplasmic tails interact directly with component(s) of the aPKC complex. To identify such receptors, we conducted IP with an anti-FLAG antibody from the lysate of D2.1 cells that ectopically express FLAG-tagged versions of either Par-3 or Par-6γ.

This revealed that syndecan-4 coprecipitated with both FLAG-Par-3 and FLAG-Par-6γ, whereas syndecan-1 coprecipitated only with FLAG-Par-3. In contrast, none of the other known ECM receptors tested, including integrin β₁, β₃ and β₄, CD44, and β-dystroglycan, was found to coprecipitate with either FLAG-Par-3 or FLAG-Par-6γ (Fig. 5D). Syndecans are transmembrane glycoproteins that have multiple heparan sulfate (HS) side-chains covalently attached to their ectodomains (28). Acting through the physical associations mediated by these HS chains, syndecans serve as receptors for a number of extracellular proteins, including a diverse array of ECM constituents (29, 30). Upon ligand binding, the syndecan core proteins are assembled into higher-order oligomers, which results in the recruitment via their cytosplasmic tails of various cytosolic proteins (Supplementary Fig. S12A; ref. 28). On the basis of this, we speculated that syndecan-mediated ligation of ECM proteins contributes to the relocalization of the aPKC complex from cytosol to the plasma membrane, ultimately inducing proliferative quiescence in the nonaggressive cancer cells growing under the 3D conditions.

Consistent with this notion, we noted that enhanced shedding of the ectodomain of syndecan-1 has been associated with metastasis development and poor clinical outcome in multiple cancer types (31, 32). Even more intriguing, the four distinct syndecan isoforms all share a tetrapeptide sequence E-F-Y-A (or “EFYA motif”) in their cytoplasmic tails, which binds to the PDZ domain—a protein interaction module carried by both Par-3 and Par-6 (24, 28). This prompted us to determine whether syndecans recruit Par-3 and Par-6 via EFYA:PDZ interactions.

In fact, recombinant GST-tagged full-length syndecan-1 and syndecan-4 (GST-Sdc1 and GST-Sdc4), but not their respective mutants lacking the C-terminal EFYA tetrapeptide motif (GST-Sdc1-ΔEFYA and GST-Sdc4-ΔEFYA), successfully captured FLAG-Par-3 and FLAG-Par-6γ from extracts of 293FT cells overexpressing these FLAG-tagged proteins (Supplementary Fig. S12B). Moreover, the mutant of Par-3 that lacks its second PDZ domain and the mutant of Par-6γ that lacks its only PDZ domain both failed to interact with GST-Sdc1 and GST-Sdc4 (Supplementary Fig. S12C). Hence, syndecan-1 and -4 are able to bind directly to the PDZ domains of Par-3 and Par-6γ via their EFYA motifs.

We proceeded to test the role of syndecans in governing the behaviors of nonaggressive cancer cells, doing so by knocking out the genes encoding all three of the syndecan isoforms (syndecan-1, -3, and -4) expressed in the D2.1 cells (Supplementary Fig. S12D–S12E). As we found, these syndecan triple-knockout cells (D2.1-ΔSdc cells) failed to exhibit accumulation of the components of the aPKC complex to the plasma membrane in the 3D MoT culture (Fig. 6A and B, Supplementary Fig. S13A), which was accompanied by (i) reduced phosphorylation of Par-1b and KSR1 on their T593 and S392 residues, respectively (Fig. 6C); (ii) impaired physical interactions between Par-1b and KSR1 (Supplementary Fig. S13B); (iii) restoration of ERK phosphorylation and rapid proliferation (Fig. 6C, Supplementary Fig. S13C). Similarly, the concomitant expression of dominant-negative deletion mutants of syndecan-1, -3, and -4 (Sdc-1-ΔEFYA, Sdc-3-ΔEFYA, Sdc-4-ΔEFYA, respectively) also resulted in impaired phosphorylation of Par-1b and KSR1, enhanced phosphorylation of ERK, and increased proliferation rate when the D2.1 cells were growing under the 3D MoT culture condition (Supplementary Fig. S13D and E).

Moreover, when introduced via tail-vein injection into mouse hosts, the D2.1-ΔSdc triple-knockout cells colonized the lung tissue far more efficiently than did the parental D2.1 cells (18-fold increase in the abundance of colonies containing ≥7 cells; Fig. 7A). Similarly, the knockout of all four syndecan isoforms (syndecan-1 to -4) in the nonaggressive mouse mammary carcinoma cell lines EpH4-Ras and 4TO7, as well as the concomitant knockdown of three of the major syndecan isoforms (syndecan-1, -2 and -4) expressed in the human breast cancer cell lines SUM159 and MDA-MB-231 (yielding EpH4-Ras-ΔSdc, 4TO7-ΔSdc, SUM159-ΔSdc, and MDA-MB-231ΔSdc cells, respectively), all enabled these cells to develop large metastases in the lungs more efficiently than did their parental counterparts (2.5- to 4.7-fold increase in the abundance of colonies containing ≥7 cells; Fig. 7B, Supplementary Figs. S12D, S14A–S14C). Collectively, these observations supported the notion that syndecan-mediated ligation of ECM proteins contributes critically to the quiescence of nonaggressive cancer cells following their extravasation into the lung parenchyma.

In addition, upon injection through the left ventricle of the mice, 4TO7-ΔSdc cells colonized multiple tissues, including lungs and bone marrow, more efficiently than their unmodified parental counterparts (Fig. 7C and D). Moreover, following orthotopic implantation into the mammary fat pads of host mice, both the parental 4TO7 cells and 4TO7-ΔSdc cells formed primary tumors of comparable sizes, whereas the tumors generated by the latter cells spawned metastatic progenies in multiple tissues more efficiently than did the tumors of the parental cells (Fig. 7E). Hence, syndecan-mediated mechanism of blocking post-extravasation proliferation is likely to account critically for the quiescent behaviors of nonaggressive DTCs in multiple target organs.

Intriguingly, the above-described mechanism of silencing proliferation triggered by syndecan:ECM ligation appear to be active in some intrinsically aggressive cancer cells (Supplementary Fig. S14D). Thus, the aggressive D2A1 cells, like their nonaggressive counterparts, exhibited elevated phosphorylation of Par-1b and KSR1 on T593 and S392 residues, respectively, both in the 3D MoT culture and following extravasation into the lung parenchyma (Figs. 2B and C, 3B). In addition, Par-1b and KSR1 proteins are both enriched in the cytosol of these aggressive cells specifically when they were growing in the 3D MoT culture (Supplementary Fig. S15A). However, these changes ultimately failed to determine the proliferative ability of these aggressive cancer cells whose proliferation was driven through a second, alternative signaling channel: As we demonstrated previously (7), engagement of ECM proteins by integrins and the resulting activation of FAK signaling—which is known to activate ERK by employing alternative (i.e., non-KSR) scaffolding proteins such as paxillin (14)—enable ERK activation and rapid proliferation of these cells under these 3D conditions (Supplementary Figs. S14D, S15B–S15D).

In contrast, another aggressive cancer cell line, the B16F10 melanoma cells, exhibited lower expression levels of two components of the syndecan-mediated antiproliferative signaling pathway, Par-3 and syndecan-1, than did their nonaggressive counterparts—B16F0 and B16F1 cells (Supplementary Fig. S16A–S16B). Consistently, the B16F10 cells failed to display noticeable phosphorylation of Par-1b and KSR1 even under the 3D MoT culture (Supplementary Fig. S16C). The restoration of Par-3 expression in the B16F10 cells impaired the abilities of these cells to proliferate rapidly in the 3D MoT culture in vitro as well as to develop large metastases in the lungs following intravenous injection into mice in vivo; these abilities were diminished further by the additional expression of syndecan-1. In contrast, the ectopic expression of neither Par-3, nor syndecan-1, nor both together, noticeably affected the proliferation rate of B16F10 cells in 2D monolayer culture (Supplementary Fig. S16D–S16E). These observations revealed the functional significance of reduced Par-3 and syndecan-1 expression, and the resulting defect in syndecan-mediated antiproliferative signaling machinery, in enabling the aggressive behaviors of the B16F10 cells.

We note that inactivation of this syndecan-mediated anti-proliferative signaling machinery also accompanies malignant progression of human tumors. Thus, in breast cancer, genetic amplification of KSR1 or KSR2 (encoding KSR1 and KSR2 scaffolding proteins, respectively), homozygous or heterozygous deletion of PARD3 (encoding Par-3), as well as homozygous or heterozygous deletion of any of the four genes encoding various syndecan isoforms (SDC1 to SDC4), have all been associated closely with shorter survival periods of the patients (Supplementary Fig. S17A–S17C; ref. 33). Hence, intrinsically aggressive cells can overcome the antiproliferative effect of the syndecan-mediated signaling machinery either by shutting down the function of this machinery or activating a second, alternative mechanism of stimulating ERK activation.

The successful development of metastases involves the disruption of previously established interactions between cancer cells and their native microenvironment within primary tumors, followed by the adaptation of these cells to novel tissue microenvironments after successful dissemination and extravasation. Recent efforts along this line have identified specific host tissue components that hinder proliferation of DTCs, which include modulators of growth factor signaling and ECM proteins (34, 35). Our research has revealed yet another dimension of post-extravasation control of DTC proliferation. Specifically, we and others have noted that certain cancer cell types that lack aggressive post-extravasation proliferation in vivo also halt proliferation when cultured sparsely within various types of 3D gels of ECM proteins in vitro. This indicated that the interaction of individual cancer cells with a variety of ECM constituents can impede the proliferation of these nonaggressive cells (4, 5, 36, 37).

Importantly, the initial induction of quiescence in the recently extravasated, solitary DTCs may not result in an irreversible exit of these cells from the mitotic cycle. Thus, some of these initially quiescent or dormant DTCs can resume proliferating under certain conditions via a process that has been termed “awakening” (38). Once awakened, previously dormant DTCs may proceed to colonize vital organs and progress to terminal disease, revealing a critical demand for a better therapeutic management of DTCs. However, the management of DTCs has been hampered by the lack of information on the physiologic state of these cells. We propose that the present study has uncovered a critical component of the signaling machinery that governs the inability of these cells to undertake rapid proliferation after initially encountering the ECM proteins present in the microenvironment of recently invaded host tissues (Supplementary Fig. S18).

The mechanisms described herein are likely to be operative in a wide variety of tissue sites of dissemination. However, we suspect that there also are subsequent barriers to successful colonization that arise as DTCs confront unfamiliar tissue-specific microenvironments to which they must adapt if they succeed in forming metastatic colonies. An improved understanding of the mechanism behind DTC dormancy and awakening may, in turn, pave the way to developing strategies for the effective therapeutic management of these potentially harmful cells.

No potential conflicts of interest were disclosed.

Conception and design: T. Shibue, R.A. Weinberg

Development of methodology: T. Shibue

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T. Shibue

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T. Shibue

Writing, review, and/or revision of the manuscript: T. Shibue, R.A. Weinberg

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): F. Reinhardt

Study supervision: R.A. Weinberg

We are grateful to A.W. Lambert for critical reading of the article; M. Brooks, T. Chavarria, Keck Imaging and FACS Facilities of the Whitehead Institute, and Animal Imaging and Preclinical Testing and Histology Core Facilities of the Koch Institute for assistance; H. Beug, F.R. Miller, and V.M. Weaver for reagents. R.A. Weinberg is an American Cancer Society research professor and a Daniel K. Ludwig Foundation cancer research professor. This work was funded by grants from National Institutes of Health (R01 CA0784561 and P01 CA080111), Samuel Waxman Cancer Research Foundation, Breast Cancer Research Foundation, and Ludwig Fund for Cancer Research.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.