Mechanisms underlying the propensity of latent lung adenocarcinoma (LUAD) to relapse are poorly understood. In this study, we show how differential expression of a network of extracellular matrix (ECM) molecules and their interacting proteins contributes to risk of relapse in distinct LUAD subtypes. Overexpression of the hyaluronan receptor HMMR in primary LUAD was associated with an inflammatory molecular signature and poor prognosis. Attenuating HMMR in LUAD cells diminished their ability to initiate lung tumors and distant metastases. HMMR upregulation was not required for dissemination in vivo, but enhanced ECM-mediated signaling, LUAD cell survival, and micrometastasis expansion in hyaluronan-rich microenvironments in the lung and brain metastatic niches. Our findings reveal an important mechanism by which disseminated cancer cells can coopt the inflammatory ECM to persist, leading to brain metastatic outgrowths. Cancer Res; 77(8); 1905–17. ©2017 AACR.
Lung adenocarcinoma (LUAD) is the most frequently diagnosed subtype of lung cancer (1). The Cancer Genome Atlas (TCGA) has identified numerous mutations in primary LUADs (2). Despite this genetic heterogeneity, LUADs can be categorized into three distinctive subgroups based on their bulk gene expression profiles and morphology. These subtypes, termed terminal respiratory unit (TRU), proximal proliferative (PP), and proximal inflammatory (PI), partially overlap with other transcriptomic classifications (3–5). TRU tumors are well differentiated and have a good prognosis compared with PP and PI subtypes. High-grade PI tumors (previously referred to as squamoid LUADs) in particular, aberrantly express inflammatory genes as well as epithelial markers more typical of the proximal airways (2, 3). LUADs expressing proximal-like genes have an elevated risk of metastasis and poor prognosis even when resected at early stages (2, 4). Hence, malignant dissemination may occur early in these patients with rapid metastatic relapse ensuing. Moreover, the lung tumorigenic capacity of certain LUADs may also dictate their ability to form detectable metastasis. The mechanisms linking these processes are incompletely understood.
Malignant growth requires the cooption of a permissive tumor microenvironment (TME). Nonmalignant cell types within the TME include activated fibroblasts, endothelial cells, and various cells of the immune system. The recruitment and functions of stromal cells are coordinated by cell surface ligands, soluble growth factors, cytokines, and their respective receptors (6). In addition, the TME is composed of extracellular matrix (ECM), which consists of collagens, proteoglycans, and non-proteinaceous glycosaminoglycans (7). ECM molecules can directly activate paracrine or autocrine cell signaling, form a biomechanical scaffold for adherent cells, and remodel tissue architecture during inflammation (8, 9). One of the most rate limiting steps of cancer metastasis is the ability of disseminated tumor cells (DTC) to outgrow after seeding distant organs (10). Brain is the predominant site of metastasis for treatment refractory lung cancers (11, 12). After lung cancer cells extravasate across the blood–brain barrier (BBB), they remain bound to the local vasculature (13). Extracellular cues from this metastatic niche may prolong latency or induce DTC outgrowth, thus affecting the transition between dormancy and clinical recurrence. The ECM composition of the brain metastatic niche in particular remains underexplored, and identifying its effects on micrometastasis outgrowth may provide novel therapeutic opportunities to control relapse in patients with minimal residual disease.
In this study, we sought to identify mechanisms of DTC outgrowth and their association with different molecular subgroups of lung cancer. A comprehensive transcriptomic analysis of resected human primary tumors reveals that the expression of specific ECM molecules and their corresponding receptors are defining features of LUAD subtypes. Our data further suggest a novel mechanism that enhances tumor cell outgrowth in the lungs and brain through the cooption of inflammatory ECM components by dissemination competent LUAD cells.
Materials and Methods
Molecular subtype classification and network analysis
Affymetrix data from the National Cancer Institute Director's Challenge Consortium (NCI DCC), MSKCC2, and Samsung datasets, as well as RNA-seq data from TCGA, were processed as described in Supplementary Methods. Classification of subtypes was performed by hierarchical clustering. TRU, PP, and PI subtypes were identified in the DCC using the TCGA classifier (3) with updated nomenclature from TCGA. For the TCGA cohort, assigned classification from the integrated study (2) was used. For the alveolar-like classification, DCC and TCGA cohorts were clustered as previously described (4). The ECM-associated interactome and functional enrichment analysis was performed using ToppGene (14).
Cell lines were cultured as recommended by the ATCC. Doxycycline treatment was performed for 3 days at 1 μg/mL. Cell lines listed in Supplementary Methods were purchased directly from the ATCC or identification was performed on established lines using STR DNA profiling (Genewiz and the Yale Cell Line Authentication Service). Lines were tested for Mycoplasma every 6 months using the ATCC Universal Mycoplasma Detection Kit. Hyaluronan-specific hyaluronidase (HSE; H1136, Sigma) was reconstituted in 0.01% BSA in PBS. For organoid experiments, following 3 days of doxycycline treatment, cells were pretreated with 5 U/mL HSE or vehicle for 6 hours. A total of 2×103 tumor cells per well in RPMI, doxycycline (0.5 μg/mL), 2% FBS, 5% growth factor reduced Matrigel (BD Biosciences), and vehicle or HSE were cultured in 24-well ultra-low attachment plates (Corning) for 3 to 14 days with media replenished every 3 days. In cocultures, 2×103 tumor cells were mixed with 1.0×104 astrocytes. Organoid outgrowth was quantified on the basis of tumor cell luciferase activity measured by the Luciferase Assay System (Promega).
Athymic nu/nu aged-matched male mice (5–7 weeks) were purchased commercially and doxycycline was administered through their diet. For orthotopic lung tumors, 1 × 105 cells were injected intra-tracheally. For metastatic tumors, 5×104 cells were injected intra-arterially. Tumor incidence and growth were quantified by bioluminescence. All work was done in accordance with the Yale Institutional Animal Care and Use Committee.
Western blot, fluorescent image acquisition, and quantification
For Western blots, total protein lysate was harvested from cells using RIPA lysis buffer. For immunofluorescent staining, tissue was fixed overnight in 4% PFA at 4°C, embedded into OCT, and lung (8 μm) or brain (8 and 30 μm) was sectioned. Slides were blocked in 3% BSA/TBST for 30 minutes at room temperature. Primary antibodies were incubated overnight at 4°C in 0.3% BSA/TBST+0.5% Triton-X100. Confocal images were analyzed with Volocity (PerkinElmer) and Image J (NIH). All images are max projections of z-stacks. HA, GFAP, Cl-Casp3, Ki67, and GFP positive cells/clusters were imaged by immunofluorescent microscopy and quantified using ImageJ software and manually counted for the number of GFP-positive, DAPI-positive cells. All primary antibodies and concentrations are listed in Supplementary Methods.
mRNA/miRNA qRT-PCR and bisulfite-sequencing PCR
Total RNA was extracted using RNeasy mini kit for mRNA and miRNeasy mini kit for miRNA (QIAGEN). Five-hundred ng (mRNA) or 50 ng (miRNA) was reverse transcribed into cDNA using the iScript cDNA Synthesis Kit (Bio-Rad) for mRNA or the TaqMan microRNA Reverse Transcription Kit (Applied Biosystems) for miRNA. For mRNA, cDNA PCR was performed using SYBR Green master mix (Applied Biosystems). For miRNA, cDNA PCR was done using TaqMan Universal Master Mix, no AmpErase UNG (Applied Biosystems) and primers (20X) from the TaqMan MicroRNA Assay. Reactions were run in quadruplicates. Data were normalized to housekeeping genes GAPDH or GUSB for mRNA and RNU48 for miRNA and represented as mean ± SD. For mir-34A methylation analysis, genomic DNA was extracted using DNeasy Blood & Tissue kit (Qiagen). Bisulfite conversion was performed with EpiTect kit (Qiagen) and CpG island 157 (hg19) was amplified by PCR. The PCR products were gel-purified and subcloned into pGEM-T Easy vector (Promega) followed by transformation into competent cells for blue-white screening. Eight to 10 positive colonies were sequenced with T7 primers. All relevant primer sequences are listed in Supplementary Methods.
3′UTR reporter assay
The wild type (WT) or mutant 3′- untranslated regions (UTR) of HMMR (784 bp) were cloned via PCR at the 3′ end of a Renilla luciferase gene in psiCHECK2 (Promega). H2030-BrM3 cells (3×104) were plated (24-well plates) in quadruplicates and transfected with 100 nmol/L of Control or miR-34a mimic and 100 ng of psiCHECK2-HMMR-3′UTR using Lipofectamine 2000 (Invitrogen). A vector that constitutively expresses Renilla luciferase (CMV-Renilla) was also transfected separately for normalization. Forty-eight hours posttransfection, Renilla activity was measured using the Dual-Luciferase Reporter Assay System (Promega).
Experimental data are presented as mean ± SEM. P values for in vitro data and tumor volume are calculated by the two-tailed Student t test. Survival curves and metastatic burden were analyzed by log-rank test and Mann–Whitney test, respectively, using specific time points or area under the curve.
Molecular subtypes of LUAD express distinct ECM-associated genes
We hypothesized that aberrant levels of ECM molecules would be associated with the heterogeneity and metastatic competence of primary LUADs. We therefore examined the expression of several categories of defined lung ECM associated proteins (15–17) across reproducible molecular classes of LUADs, including the TRU, PP, and PI subtypes. The expression of collagens, proteoglycans, and specifically hyaluronic acid–binding proteins (HABP) was significantly increased in LUADs of the poor prognosis PI subtype (Fig. 1A; Supplementary Fig. S1A). We further reasoned that the enrichment of functional ECM components could be accompanied by the overexpression of their receptors, which relay signals between tumor cells and the TME. We identified genes encoding cell surface proteins that were differentially expressed in the PI subtype and performed a molecular interactome/functional enrichment analysis between these proteins and lung ECM molecules (Fig. 1B). Several cell surface receptors that bind directly or indirectly to ECM molecules were overexpressed in the PI tumors (Fig. 1C; Supplementary Table S1). For example, ITGB1 interacts with both collagens and proteoglycans and mediates lung cancer cell adhesion (18).
Amongst other ECM receptors identified in our analysis, hyaluronan mediated motility receptor (HMMR) is notable based on the following observations. First, HMMR is a receptor for hyaluronic acid (HA), a nonsulfonated glycosaminoglycan that is overproduced during lung inflammation, fibrosis, and in the stroma of certain cancers (19, 20). Second, HMMR is increased in multiple malignancies (19, 21). HMMR was overexpressed in LUADs relative to normal lung tissue, and we also found that several HMMR isoforms were further increased in the PP and PI subtypes (Fig. 1D and E; Supplementary Fig. S1B). Another HA receptor is CD44, which can interact with HMMR (22). Average and isoform-specific CD44 expression did not correlate with the poor prognosis LUAD subtypes (Supplementary Fig. S1C and data not shown). Expression of major HA synthesizing (HAS1-3) or degrading (HYAL1-2) enzymes trended lower in tumors but was not clearly associated with subtypes (Supplementary Fig. S1C). However, versican (VCAN), a secreted proteoglycan that binds to HA and HA receptors (23), was significantly elevated in the bulk of PI tumors (Fig. 1D and E). Finally, we previously identified a partially overlapping subgroup of metastatic LUADs classified by alveolar gene expression (4). HMMR and VCAN were also overexpressed in these aggressive LUADs [alveolar-low or distal airway stem-like (DASC-like); Supplementary Fig. S1D and S1E]. Thus, the expression of specific HA-binding proteins is significantly increased in molecularly defined subsets of LUADs.
Regulation of the glycosaminoglycan receptor HMMR in metastatic LUAD cells
HMMR is a major determinant of extracellular HA's effects on cells, but the regulation of its expression and in vivo function in LUAD are poorly understood. When compared with stromal cells known to express HMMR, HMMR protein was high in 13 human LUAD cell lines (Supplementary Fig. S2A). HMMR was localized both at the cell surface/cytoplasm and in the nucleus of LUAD cells (Supplementary Fig. S2B). Across a repository of cell lines from late stage LUAD, HMMR copy-number variations (CNV) correlated with its expression (Supplementary Fig. S2C). However, HMMR CNVs or mutations were not frequently detected in primary LUADs (data not shown). Alternatively, using independent methods for identifying miRNA–mRNA interactions in LUADs (Supplementary Table S2), we found a partial inverse correlation between the expression of mir-34 family members, particularly mir-34a, and its predicted target HMMR (Fig. 2A; Supplementary Fig. S2D). Moreover, methylation of MIR34A CpG clusters inversely correlated with mir-34a expression and, conversely, positively correlated with HMMR expression in resected LUADs (Supplementary Table S3). These results are consistent with MIR34A-C being tumor-suppressor genes that can be silenced by DNA methylation (24–26).
To validate the correlation between LUAD aggressiveness, miR-34a repression, and HMMR upregulation, we analyzed gene expression in a well-characterized poorly metastatic LUAD cell line, H2030, in comparison with its highly metastatic cell subpopulation, H2030-BrM3, which colonizes distant organs when transplanted in mice (27). HMMR mRNA and protein were increased in metastatic H2030-BrM3 cells relative to their matched parental cells (Fig. 2B and 2C). Conversely, miR-34a and miR-34b were repressed in the BrM3 cells relative to its parental line, and miR-34c was undetectable in both cell populations (Fig. 2D). Moreover, CpG sites closest to the MIR34A promoter (hg19 chr1: 9242205-9242815) are more frequently methylated in the BrM3 cells (Fig. 2E; yellow region). Ectopic expression of a miR-34a mimic in the metastatic H2030-BrM3 cells inhibited a HMMR-3′UTR reporter, and this effect was abrogated when the miR-34a seed sequence was mutated (Fig. 2F; Supplementary Fig. S2E and S2F), indicating that miR-34a can directly target HMMR. Finally, increasing miR-34a reduced HMMR in four independent LUAD cell lines (Fig. 2G and H; Supplementary Fig. S2G). Our data suggest that HMMR expression in LUAD cells correlates with metastatic competence and is epigenetically regulated in part by miR-34a.
HMMR mediates tumor initiation in the lungs and distant organ metastasis
HMMR expression correlates with survival (HR, 1.34; P = 0.0018), but this association is not significant when adjusting for the TCGA classification (TRU, PP, PI; HR, 0.91; P = 0.7931), consistent with the observation that HMMR is a marker of PI tumors. HMMR levels associated with clinical progression as well as distant metastasis (Supplementary Fig. S2H–S2J), suggesting that HMMR may enhance both locoregional and distant relapse. To test the in vivo function(s) of HMMR, we decreased its expression in metastatic cell lines, using two independent doxycycline-inducible short hairpin RNAs (Fig. 3A; Supplementary Fig. S3A and S3B). Reduction of HMMR in H2030-BrM3 cells did not affect subcutaneous tumor formation (Fig. 3B). To model the inflammatory TME of LUADs at risk for relapse, we injected H2030-BrM3 cells into the lungs of mice after injuring their airways, which significantly increases lung tumorigenesis (data not shown; ref. 28). In this setting, knockdown of HMMR impaired the tumorigenic capacity of H2030-BrM3 cells as early as 14 days following tumor cell injection (Fig. 3C and D; Supplementary Fig. S3C). Delaying shRNA induction until 7 days after LUAD cell transplantation (Supplementary Fig. S3D) to reduce HMMR in established nodules did not affect tumor progression (Fig. 3E and F). Using an established LUAD allograft model (29), knockdown of Hmmr in murine LUAD cells had a similar effect on early lung tumor outgrowth in the absence of exogenous injury (Supplementary Fig. S3E–S3G).
We next examined whether HMMR was also required for metastasis formation in distant sites. Although DTCs are detected from BrM3 derived orthotopic lung tumors (27), quantitative analysis of metastasis in this setting is limited by high lung tumor burden and its associated morbidity (data not shown). To directly quantify metastatic colonization, we injected H2030-BrM3 cells into the arterial circulation of mice. Knockdown of HMMR decreased the incidence of metastasis (Fig. 4A). We obtained similar results using PC9-BrM3 cells, an independent metastatic LUAD cell subpopulation of the PC9 line, which expresses constitutively high levels of HMMR (Fig. 4B; Supplementary Fig. S4A and S4B). Knockdown of HMMR primarily reduced the incidence of brain metastasis but also attenuated metastasis in other organs such as limb bone, lung, liver, and adrenal glands (Fig. 4C and D; Supplementary Fig. S4C and S4D). Tumor burden was decreased 20 days after DTC entry into circulation using independent models (Fig. 4E and F). We conclude that the ECM receptor HMMR regulates a selective step at early stages in the formation of lung tumors and metastasis in a TME-dependent manner.
HMMR is not required for LUAD cell seeding but enhances metastatic outgrowth
We analyzed in greater detail the requirement for HMMR during rate limiting steps of tumor outgrowth in the lungs and brains. In the airways, decreasing HMMR had no effect on the seeding or proliferation of tumor cells by day 2 of transplantation (Fig. 5A–C; Supplementary Fig. S5A). However, by day 14, HMMR knockdown significantly decreased the number of LUAD cells and tumor clusters in the lungs (Fig. 5D and E; Supplementary Fig. S5B). Once in circulation, DTCs can extravasate across the BBB within 5 to 7 days, after which they die, arrest, or invade and expand as perivascular metastases (Fig. 5F; refs. 13, 30). In vitro, HMMR knockdown decreased the migration of LUAD cells, but did not affect their adhesion to an endothelial monolayer (Supplementary Fig. S5C and S5D). However, in vivo, HMMR reduction did not alter the extravasation of tumor cells in the brain (Fig. 5G; Supplementary Fig. S5E). Conversely, after 15 days, HMMR knockdown significantly reduced the number of perivascular LUAD cells and expanding micrometastases, which we define as clusters of 1 to 20 cells (Fig. 5H-J). Altogether, our in vivo data confirm that HMMR is not required for malignant seeding/extravasation, but enhances the outgrowth of micrometastases in the brain parenchyma.
HA deposition in the lung TME and brain metastatic niche
HA is detected in the stroma of high-grade human LUADs (31). We therefore characterized HA deposition over time in our in vivo model. Within 2 days of intratracheal injection, HA was abundant in the airway stroma surrounding GFP positive LUAD cells (Fig. 6A). By 14 days, HA accumulated within larger tumor nodules with some levels still present in the surrounding TME (Fig. 6A; Supplementary Fig. S6A). VCAN is also detected in the stroma of human lung cancers (32) and was expressed predominantly in the lung microenvironment (Supplementary Fig. S6B). During fibrosis, HA is produced and catabolized by fibroblasts and endothelial cells respectively (33, 34). At all timepoints, LUAD cells were surrounded by endomucin-positive endothelial cells (Supplementary Fig. S6C) and alpha smooth muscle actin (αSMA)-positive cells (Fig. 6B), which include activated fibroblasts (35). Stromal HA more significantly overlapped with αSMA-positive cells, especially at early time points (Fig. 6B). LUAD cells can produce some HA, but pulmonary fibroblasts synthesize higher amounts of HA and express more VCAN in culture (Supplementary Fig. S6D and S6E).
Because HMMR expression significantly impacts metastasis to the central nervous system (CNS) and this is the predominant site of relapse in lung cancer patients, we next focused on analyzing HA in brain metastases. HA staining is normally diffuse in the adult murine cortex (36). Strikingly however, by day 15, 90% (238/264) of brain micrometastases were surrounded by a pericellular halo of HA, which increased over time (Fig. 6C and D; Supplementary Fig. S6F). In addition to tumor cells, potential sources of HA in the brain are endothelial cells and reactive astrocytes, which associate with metastases and reactive glia synthesize HA during cerebral scarring (36). Although all micrometastases were bound to collagen IV positive blood vessels (Supplementary Fig. S6G), HA accumulation correlated closely with GFAP-positive astrocytes in contact with tumor cells (Fig. 6E; Supplementary Fig. S6H). In addition, proliferating astrocytes produced elevated amounts of HA in culture (Supplementary Fig. S6I). Finally, at later time points (day 60), HA deposition remained concentrated at the invasive edge compared with the inner core of large macrometastases (Fig. 6F and G). Altogether our results indicate that LUAD cells express HMMR and produce some HA, but additional HA can accumulate from the stromal interface, particularly during micrometastasis expansion.
HMMR enhances LUAD cell survival in the presence of ECM and stromal cells
To elucidate the mechanism(s) by which HMMR regulates metastatic outgrowth, we modeled in vivo interactions through a series of in vitro experiments. HMMR was not required for the viability of H2030-BrM3 cells cultured as monolayers at high cell density (Fig. 7A) nor under serum starved conditions (Supplementary Fig. S7A). The contextual requirement for HMMR in vivo, and the distinct accumulation of HA around metastases indicate that the effects of HMMR may depend on the presence of other ECM proteins and stromal cells. Previous studies demonstrated that metastatic cells can coopt vascular basement membrane (BM) proteins for outgrowth (37). In addition, we noted that PI LUADs overexpress genes associated with the BM (Supplementary Fig. S7B) suggesting that the BM ECM contributes to the biology of PI LUADs. Matrigel consists primarily of collagen IV and laminin, two ECM proteins abundant in the BM of CNS blood vessels (38). When cultured with Matrigel, H2030-BrM3 cells formed tumor organoids. HMMR knockdown decreased the outgrowth of these organoids (Fig. 7B; Supplementary Fig. S7C). Similarly, organoid outgrowth was reduced when treated with recombinant hyaluronidase (HSE; Fig. 7B), which specifically degrades extracellular HA (Supplementary Fig. S7D). Coculture of tumor cells with HA producing astrocytes increased the invasive outgrowth of LUAD cells, a phenotype that was then restricted by HMMR knockdown in LUAD cells or by HSE treatment of the cocultures (Fig. 7C and D; Supplementary Fig. S7E and S7F). Combining HSE treatment with HMMR knockdown had minimal additive effects (Fig. 7D; right), indicating that HMMR and extracellular HA have some redundant functions. HMMR can activate various intracellular pathways in a context-dependent manner (39). Reduction of HMMR decreased activation of Mitogen-activated protein kinase/ERK and AKT in LUAD cells grown as ECM embedded organoids (Fig. 7E) but not when these cells were cultured on adherent plates in the absence of exogenous ECM (Supplementary Fig. S7G). Finally, coculture of LUAD organoids with astrocytes decreased cleaved-caspase-3–positive apoptotic tumor cells (Fig. 7F), whereas HMMR knockdown increased LUAD cell apoptosis (Fig. 7G) without significantly altering proliferation (Supplementary Fig. S7H). Thus, HMMR enhances metastatic cell survival by potentiating ECM-mediated signaling.
In summary, our findings ascribe a novel function for HMMR and HA in mediating the survival of dissemination competent LUAD cells and enhancing the outgrowth of micrometastasis.
Lung ECM-associated proteins define aggressive subsets of early-stage LUADs
The mechanisms underlying the heterogeneity of LUADs and their variable propensity for local and distant relapse are incompletely understood. We noted that the differential expression of ECM-associated genes is a defining property of known molecular subtypes of primary LUADs. The ECM is a major constituent of the TME and modulates molecular and biophysical interactions between malignant cells and stromal cells. Excessive deposition of particular ECM components is a hallmark of pulmonary inflammation and fibrosis. Likewise, LUADs of the PI subtype display a gene expression pattern akin to fibrosis. The concomitant overexpression of ECM interacting proteins by LUAD cells may endow them with a selective growth advantage in this inflammatory microenvironment as well as increase their likelihood of forming metastases.
HMMR links lung tumor outgrowth to metastatic relapse
A notable ECM-binding protein whose expression was enriched in the PI subtype was HMMR (a.k.a RHAMM/CD168). HMMR is a receptor for HA that accumulates during pulmonary inflammation. HA and HMMR upregulation has been correlated with the aggressiveness of other types of cancers. The role of HMMR in tumor cells has been mainly attributed to cell motility through HA (39, 40). In addition, an intracellular pool of HMMR can bind to the mitotic spindle and may have HA-independent functions (41). HA is preferentially detected in the stroma of high-grade lung cancer (31) and HMMR expression can correlate with lung cancer patient survival (21). However, the mechanism by which HMMR is activated, its relationship to inflammatory tumors, and its function during lung cancer relapse in vivo was to our knowledge unknown.
We demonstrated that HMMR expression in LUAD cells is regulated by miR-34a, although other mechanisms such as HMMR gene amplification may also occur. Loss of miR-34a can cooperate with mutant Kras and p53 haplo-insufficiency to accelerate lung cancer progression (24). HMMR overexpression enhances Ras-mediated transformation of fibroblasts in a manner that is dependent on the HA-binding domain of HMMR (42). These findings suggest a mechanism by which known oncogenic pathways can coopt the fibrotic ECM. Upon studying the requirement for HMMR in metastatic LUAD cells, we did not observe additional defects in the proliferation of adherent cells after HMMR knockdown. As previously suggested, reducing HMMR decreased the motility of LUAD cells in vitro. However, this did not correlate with a defect in their ability to seed the lung or extravasate across the BBB in vivo. Rather, HMMR was required for the outgrowth of dissemination competent LUAD cells. Given the reported pleiotropic functions of HMMR, its overexpression in LUAD cells may modulate other stages of cancer progression beyond those we characterized. Altogether, our data illustrate how lung tumor initiation and micrometastasis outgrowth can be mechanistically linked through HMMR activation.
HMMR mediates DTC survival and expansion in an HA rich niche
HA was abundant in tumor-bearing lungs and correlated with the appearance of activated fibroblasts, which are known to accelerate cancer progression. These data are consistent with myofibroblasts being a pervasive source of HA during inflammation and fibrosis (33), although other inflammatory cell types and the tumor cells themselves can also produce HA. Interestingly, HA in the brain was more concentrated around micrometastatic lesions. During cerebral ischemia, reactive astrocytes produce HA (36). Analogously, we found the appearance of reactive astrocytes to correlate with increased HA deposition around DTCs in vivo, and astrocytes can promote LUAD cell survival and organoid expansion in vitro. In addition to inducing an inflammatory response from glia, micrometastasis expand and invade along existing blood vessels in the CNS (13, 30, 37). When stabilized, this perivascular niche can induce tumor cell dormancy (43), whereas micrometastatic outgrowth requires remodeling of the vasculature (30), which also sustains the expansion of normal stem cells in the distal airways (44) and CNS (45). Because HA can regulate angiogenesis and vascular permeability (46) it may act as a driver of perivascular remodeling. DTCs overexpressing HMMR may have a competitive advantage within this niche.
HMMR enhanced the expansion of LUAD cells cultured as organoids in contact with ECM. Hyaluronidase treatment recapitulated the effects of HMMR knockdown, consistent with HMMR being a cellular mediator of endogenously produced HA function. Reduction of HMMR impaired survival signals in LUAD cells. Because HA deposition can aggregate other ECM molecules and HMMR lacks a signaling domain, it may enhance DTC survival as a consequence of mechanosensing from HA-ECM complexes, which can be derived both from tumor cells as well as stromal cells. This provides a pathway by which DTCs can partially sustain ECM mediated cell survival on their own, yet further coopt ECM components produced by the inflammatory stroma to amplify metastatic outgrowth, particularly in the brain. HA can also bind to other receptors, some of which are expressed on immune cells (47). Moreover, HA can be modified into chains of different lengths, which have distinct properties (48). Finally, incorporation of HA into the ECM affects vascular perfusion (49). HA may thus have additional effects on the brain TME, which could be modulated for therapeutic treatment.
Several molecular determinants of cancer cell migration and invasion have been described. However, at the time of diagnosis, a significant proportion of LUAD patients are already likely to harbor DTCs and distant organ metastasis. Consequently, the identification of HMMR upregulation as a novel mechanism by which LUAD cells can persist in the primary and metastatic sites is of specific clinical interest. Inhibiting HMMR- and HA-mediated interactions with the ECM may be explored as a therapeutic strategy to constrain minimal residual disease and the expansion of micrometastasis in LUAD patients.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: L.E. Stevens, R.S. Herbst, D.X. Nguyen
Development of methodology: L.E. Stevens, K. Brewer
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L.E. Stevens, W.K.C. Cheung, S.J. Adua, A. Arnal-Estapé, M. Zhao
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L.E. Stevens, W.K.C. Cheung, S.J. Adua, A. Arnal-Estapé, M. Zhao, Z. Liu, R.S. Herbst, D.X. Nguyen
Writing, review, and/or revision of the manuscript: L.E. Stevens, S.J. Adua, A. Arnal-Estapé, K. Brewer, R.S. Herbst, D.X. Nguyen
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L.E. Stevens, M. Zhao, D.X. Nguyen
Study supervision: D.X. Nguyen
This study was funded by grants from the Free to Breathe Foundation and National Cancer Institute (R01CA166376 and R01CA191489 to D.X. Nguyen), and from the Yale SPORE in Lung Cancer (P50CA196530 to R.S. Herbst),
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.
We thank Joan Massagué and Xiang F. Zhang for comments on the article, and Monte Winslow for murine cell lines.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
- Received July 22, 2016.
- Revision received December 28, 2016.
- Accepted January 19, 2017.
- ©2017 American Association for Cancer Research.