Tumor progression to metastasis is not cancer cell autonomous, but rather involves the interplay of multiple cell types within the tumor microenvironment. Here we identify asporin (ASPN) as a novel, secreted mesenchymal stromal cell (MSC) factor in the tumor microenvironment that regulates metastatic development. MSCs expressed high levels of ASPN, which decreased following lineage differentiation. ASPN loss impaired MSC self-renewal and promoted terminal cell differentiation. Mechanistically, secreted ASPN bound to BMP-4 and restricted BMP-4–induced MSC differentiation prior to lineage commitment. ASPN expression was distinctly conserved between MSC and cancer-associated fibroblasts (CAF). ASPN expression in the tumor microenvironment broadly impacted multiple cell types. Prostate tumor allografts in ASPN-null mice had a reduced number of tumor-associated MSCs, fewer cancer stem cells, decreased tumor vasculature, and an increased percentage of infiltrating CD8+ T cells. ASPN-null mice also demonstrated a significant reduction in lung metastases compared with wild-type mice. These data establish a role for ASPN as a critical MSC factor that extensively affects the tumor microenvironment and induces metastatic progression.
These findings show that asporin regulates key properties of mesenchymal stromal cells, including self-renewal and multipotency, and asporin expression by reactive stromal cells alters the tumor microenvironment and promotes metastatic progression.
Tumors do not develop in isolation and components of the tumor microenvironment demonstrably influence many cancer hallmarks including metastasis (1, 2). A growing body of evidence now indicates that cells of mesenchymal origin including mesenchymal stem cells, mesenchymal stromal cells, and fibroblasts, collectively referred to herein as MSCs, reside in the tumor microenvironment and contribute to cancer progression (2–5). MSCs are a heterogeneous population of multipotent stromal cells that give rise to physiologic cell lineages including osteoblasts, adipocytes, and chondrocytes (6). MSCs also give rise in part to pathologic cell lineages including cancer-associated fibroblasts (CAF) and fibroblasts associated with injury or inflammation (2, 3, 7, 8). Similar to MSCs, CAFs are highly heterogeneous and dynamic, which may account for the reported pleiotropic roles of MSCs and CAFs in cancer progression as either potentiators of primary and metastatic cancer development (4, 5, 9–14) or restrainers of cancer growth (15–17). The seminal properties of MSCs in the tumor microenvironment have yet to be fully established. Because of the conservation between gene expression programs in development and cancer (18, 19), key factors may be similar between MSCs in development and MSCs in the tumor microenvironment.
Asporin (ASPN), a member of the small leucine-rich proteoglycan (SLRP) family of extracellular proteins, is markedly enriched in fetal prostate mesenchyme during development (18, 20). ASPN is similarly expressed in reactive stromal cells, including CAFs, in the tumor microenvironment of prostate (20–22), breast (23), scirrhous gastric (24), and pancreatic (25) cancers. The role of ASPN in prostate MSCs and in the tumor microenvironment is poorly understood. Intriguingly, ASPN has been reported to regulate mediators of MSC differentiation including BMP-2 (26) and TGFβ1 (27). Yet, the roles of ASPN in MSC self-renewal and multipotency have not been definitively determined.
ASPN has been shown to induce migration of scirrhous gastric (24, 28) and pancreatic (25) cancer cells; nevertheless, ASPN expression in the tumor microenvironment may have a dual role in regulating cancer progression (21–25, 29), possibly due to polymorphisms in the ASPN aspartic acid (D)-repeat domain length (21). ASPN expression in prostate reactive stromal cells is associated with adjacent local cancer aggressiveness and with worse oncologic outcomes (21, 22). However, two, of the over 10 reported, germline polymorphisms in the ASPN D-repeat domain length have been shown to be differentially associated with the development of metastatic prostate cancer following surgery (21). ASPN with 14 D-repeats (ASPN D14) was associated with an increased risk while ASPN D13 was associated with a reduced risk of metastatic progression (21).
Herein, we report that ASPN functions as a novel, secreted MSC factor and a key driver of metastatic development. We establish a role for ASPN in regulating fundamental properties of MSCs including self-renewal, differentiation, and migration. We demonstrate that ASPN expression is highly enriched in MSCs, and its expression decreases during differentiation to connective tissue lineages. Our data show that ASPN regulates MSC self-renewal and restricts MSC differentiation through regulation of BMP-4 signaling. ASPN-null mice have fewer MSCs in the bone marrow and an enriched population of intermediate (or more differentiated) MSCs in the prostate. While most MSC-derived progeny have decreased ASPN expression, high ASPN expression is conserved between MSCs and CAFs in both primary and metastatic tumors. Prostate allograft tumors in ASPN null-mice have an altered tumor microenvironment with fewer tumor-associated MSCs, decreased vasculature, and an increased percentage of infiltrating CD8+ T cells. Tumors in ASPN-null mice also have a reduced number of cancer stem cells and a marked decrease in metastatic potential. These findings suggest that ASPN is an important regulator of MSC multipotency and metastatic development.
Materials and Methods
IHC (30), immunofluorescence (30), immunoblotting (21), RNA isolation and quantitative real-time PCR (30), colony-forming unit assay (31), cell proliferation (32), migration (32), cytoskeletal remodeling (32), MSC isolation and differentiation (5, 30, 32–36), the PELICAN study (37), and the CP1 E. coli model of prostate inflammation (38) have been previously described and are detailed in the Supplementary Materials and Methods.
Cell lines and cell culture
PC-3, DU-145, WPMY-1, TRAMP-C2, and HEK293T cell lines were obtained from the ATCC. The B6MycCaP cancer cell line was a kind gift from Dr. Leigh Ellis (Roswell Park Cancer Institute, New York, NY). All cell lines were maintained in either DMEM (DU-145, WPMY-1, TRAMP-C2, B6MycCaP) or RPMI1640 (PC-3) supplemented with 10% FBS (Corning), and penicillin/streptomycin (Life Technologies). After thawing from frozen stock, cell lines were used prior to passage 7. The WPMY-1-ASPN variant–expressing cell lines were generated and cultured as described previously (21). Human MSCs were isolated from tissue and cultured in RoosterNourish-MSC (RoosterBio) as described previously (5, 39). Mouse MSCs were cultured in DMEM supplemented with 10% FBS (Corning), Glutamax (Life Technologies) and penicillin/streptomycin (Life Technologies). B6CaP organoids were generated from C57BL/6J Hi-Myc allografts and cultured using an adapted protocol from prior reports (40, 41). Briefly, B6CaP allograft tumors were finely minced with a scalpel, digested in DMEM/F12 + 10% FBS + 1:10 dilution of collagenase/hyaluronidase for 1 hour at 37°C, triturated in prewarmed 1× PBS + DNAse I, and filtered through a 40-μm cell strainer. Cells were embedded in growth factor–reduced (GFR) Matrigel, plated on ultra-low attachment plates (Corning), and cultured in advanced DMEM/F12 supplemented with 10% charcoal-stripped FBS, B-27, GlutaMAX, HEPES, and penicillin/streptomycin, recombinant mouse EGF (10 ng/mL), TGFβ inhibitor A83-01 (200 nmol/L), ROCK inhibitor Y-27632 (10 μmol/L), and DHT (100 nmol/L). For harvest and passage, Matrigel-embedded organoids were incubated in prewarmed Dispase (5 U/mL) and subsequently trypsinized for single-cell isolation. Cell lines were authenticated by short tandem repeat analysis and confirmed Mycoplasma free by PCR testing [Johns Hopkins University (JHU) Genetic Resources Core Facility].
Prostate cancer and inflammation study
Tissue from radical prostatectomies performed at Johns Hopkins School of Medicine (Baltimore, MD) from 2009 to 2011 were examined for ASPN expression in cancer-adjacent stroma and in inflammation-adjacent stroma. Four-micron–thick radical prostatectomy sections were stained for ASPN (Sigma) by IHC. Cases were scored by a urologic pathologist for ASPN expression in stroma adjacent to cancer and in distinct areas of stroma adjacent to chronic inflammation. Chronic inflammation was defined by clusters of 20 or more lymphocytes. Of the 15 cases selected, 13 cases contained both cancer and distinct areas of chronic inflammation. Using established scoring schemes (21), ASPN intensity was evaluated and assigned an incremental score of 0 (negative), 1 (weak), 2 (moderate), or 3 (strong). The extent of staining was assigned a percentage from 0% to 100%. An ASPN score was calculated by multiplying the intensity score and the extent score (H-score).
All animal procedures were performed under a JHU-approved Institutional Animal Care and Use Committee protocol. Mice were purchased from The Jackson Laboratory and maintained under standard pathogen-free conditions. B6;129S5-Aspntm1Lex/Mmucd were generated through the Mutant Mouse Regional Resource Center (MMRRC) at University of California Davis (Davis, CA), a NIH-funded strain repository. B6;129S5-Aspntm1Lex were donated to MMRRC by Genentech, Inc. Exon 2 was targeted (NCBI accession AF316825.1) by homologous recombination. The embryonic stem cell (ESC) line was derived from 129S5/SvEvBrd from Lexicon Genetics. B6;129S5-Aspntm1Lex mice were then fully backcrossed at least nine generations to the C57BL/6J background. Mice were genotyped by PCR of genomic DNA with wild-type and mutation-specific primers.
B6CaP model of metastasis
The B6CaP allograft model was derived from a C57BL/6J Hi-Myc mouse metastatic lesion. Cells from the metastatic tissue were isolated and serially passaged in C57BL/6J mice. When injected subcutaneously with B6CaP allograft cells, 60%–70% of C57BL/6J mice developed lung metastases. For this study, we injected 5.5 × 106 B6CaP cells in 100 μL PBS subcutaneously into Aspn+/+, Aspn+//−−, and Aspn−/− mice (n = 6 mice per group), resected tumors once they reached 1 cm3, waited 11 weeks postresection, euthanized the mice, and inspected the lungs for metastases. Lungs were formalin-fixed, paraffin-embedded, sectioned, and stained as described previously (42).
To assess the in vivo MSC population in adult mouse prostates or B6CaP tumors by flow cytometry, the tissue of interest was isolated, mechanically dissociated, digested in RPMI + 10% FBS + 1:10 dilution of collagenase/hyaluronidase for 1 hour at 37°C, triturated in prewarmed PBS + DNAse I, and filtered through a 40-μm cell strainer. If necessary, red blood cells were lysed by briefly incubating the cell pellet in ice-cold ACK lysis buffer (Quality Biological). Cells were Fc-blocked (Miltenyi Biotec) for 10 minutes in 4°C. Dead cells were discriminated using the LIVE/DEAD Fixable Aqua Dead Cell Stain Kit (Thermo Fisher Scientific) and samples were stained with the following antibodies from BioLegend: CD45, CD29, CD105, Sca-1, Ly6G, Ly6C, CD11b, CD11c, CCR2, F4/80, MHC II, CD3, CD4, CD8a, FoxP3, CD44, CD62L, as well as CD3 from BD Pharmingen. Stained samples were analyzed on an LSR II flow cytometer (BD Biosciences) and quantified using FlowJo.
HEK293T cells were transfected with a murine ASPN-Myc-FLAG tagged expression vector (Origene) or vehicle control using Lipofectamine 3000 (Thermo Fisher Scientific) according to the manufacturer's protocol. Two days posttransfection, cells were washed with PBS and incubated in serum-free DMEM overnight. HEK293T cells were then treated with 0.3 μg/mL recombinant murine BMP-4 protein (R&D Systems) or vehicle and incubated in 5% CO2 at 37°C for 30 minutes. After BMP-4 incubation, conditioned media was filtered through a 0.45-μm filtration unit (Millipore) and one third was set aside for the “input” sample. The remaining two thirds was treated with a protease inhibitor cocktail (Sigma), precleared with mouse IgG agarose (Sigma) for 1 hour rotating at 4°C, and then incubated with anti-FLAG M2 magnetic beads (Sigma) overnight rotating at 4°C. Beads were washed three times with ice-cold TBS and proteins were eluted from the beads with 3XFLAG peptide (Sigma). Input conditioned media was concentrated using an Amicon Ultra-4 centrifugal filter unit with a 10 kDa NMWL (Millipore) according to the manufacturer's protocol.
Gene expression analysis
Gene expression profiling was performed for Aspn+/+ and Aspn−/− fetal MSCs expanded to passage 3 and cultured under normal growth conditions to 70%–80% confluency prior to RNA isolation. Three biological replicates were analyzed per experimental group and Affymetrix mouse Clariom D arrays were run. Bioinformatics analyses were performed in R. Datasets were normalized at a transcript level for core transcripts with the R/Bioconductor package oligo and annotated with the NetAffx Biological Annotations. Artifacts from technical batches were removed with ComBat (43) implemented in the R/Bioconductor package SVA (44). Differential expression analyses were performed on ComBat adjusted data using the R/Bioconductor package LIMMA (45). Genes with FDR-adjusted P values below 0.05 and absolute log fold change greater than or equal to two were called statistically significant. Gene set analysis was performed on the Kyoto Encyclopedia of Genes and Genomes (KEGG) cell-cycle pathway and the GeneMAPP TGFβ family signaling pathway obtained from the NetAffx Biological Annotations to confirm hypothesized functional mechanisms with a one-sided Wilcoxon gene set test implemented in LIMMA.
Quantification and statistical analysis
Unless otherwise indicated, statistical comparisons between two groups were performed using a two-tailed unpaired Student t test. Statistical comparisons between multiple groups were performed using one-way ANOVA with Tukey or Newman–Keuls multiple comparisons, as indicated in the figure legends. Statistical significance was defined as a P < 0.05, and P values were indicated with asterisks in the figures as follows: *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.0001. All statistical comparisons were performed using GraphPad Prism software (v5.0).
ASPN is highly expressed in MSCs
ASPN expression was examined in primary human MSCs isolated from fetal prostate, benign prostate, prostate cancer, and bone marrow. Primary heterogeneous cultures were expanded and shown to be enriched for MSCs by flow cytometry expression profile (CD73+, CD90+, CD105+, CD14−, CD20−, CD34−, CD45−, and HLA-DR−) and by multi-lineage (osteoblast, adipocyte, and chondrocyte) differentiation potential (5). Compared with human primary benign prostate epithelial cells (PreC) and a human benign prostate epithelial cell line (RWPE-1), MSCs from all sites had significantly elevated ASPN expression (Fig. 1A). Similar to human MSCs (20), Aspn mRNA and protein were distinctly expressed in urogenital sinus mesenchyme (UGM) compared with urogenital sinus epithelium (UGE) of the developing mouse prostate (Fig. 1B and C). MSCs were isolated from both fetal and adult mouse prostate and adult mouse bone marrow, expanded in culture, and then examined for Aspn expression. Mouse MSCs showed increased Aspn expression compared with UGE (Fig. 1B). In contrast to MSCs and fetal prostate mesenchyme, ASPN expression was markedly reduced in adult mouse benign prostate stromal cells as determined by IHC (Fig. 1C). These data suggest that ASPN is highly expressed in both bone marrow–derived and tissue-resident human and mouse MSCs. However, the role of ASPN in regulating key properties of MSCs such as self-renewal, differentiation, and migration has not been fully established.
ASPN regulates MSC self-renewal
To determine the role of ASPN in MSCs, we generated Aspn−/− mice (B6J-Aspntm1Lex/Mmucd), which were then fully backcrossed at least nine generations to the C57BL/6J background. The first coding exon of Aspn, exon 2, was targeted for deletion by homologous recombination (ref. 46; Supplementary Fig. S1A and S1B), which resulted in the loss of full-length Aspn transcript and protein (Supplementary Fig. S1C and S1D). Aspn−/− mice were viable, fertile, had comparable prostate size and histology, and had similar expression of other SLRP family members including Decorin (Dcn), Biglycan (Bgn), and Extracellular matrix protein 2 (Ecm2) as Aspn+/+ mice (Supplementary Fig. S1E–S1K).
To determine how ASPN regulates MSCs in vivo, we compared bone marrow–derived MSCs from Aspn+/+ and Aspn−/− male mice at 24 weeks of age by flow cytometry for MSC markers (CD45−, CD105+, CD29+, and Sca-1+). By these markers, both Aspn+/+ and Aspn−/− mice had a single population of bone marrow–derived MSCs; however, Aspn−/− bone marrow had significantly fewer MSCs than bone marrow from Aspn+/+ mice (Fig. 1D; Supplementary Fig. S2A and S2B). We next compared prostate MSCs from Aspn+/+ and Aspn−/− male mice at 24 weeks of age. In contrast to the single MSC population detected in the bone marrow, both Aspn+/+ and Aspn−/− prostates had two distinct populations of MSCs that were delineated by the level of Sca-1 expression (Sca-1hi vs. Sca-1med; Supplementary Fig. S2C and S2D). Prior studies have shown that MSC stem progenitors with the capacity for both self-renewal and differentiation are Sca-1hi, while more differentiated intermediate MSC progenitors with proliferative and differentiation capacity, but with reduced self-renewal ability, are Sca-1med (47). Interestingly, Aspn−/− prostates were slightly enriched for Sca-1med–intermediate MSCs compared with Aspn+/+ prostates (Fig. 1E; Supplementary Fig. S2C and S2D), suggesting that in the absence of ASPN, prostate-derived MSCs are more likely to differentiate and may thereby have a decreased potential for self-renewal.
To examine MSC self-renewal in vitro, total MSCs were isolated from Aspn+/+ and Aspn−/− bone marrow, compact bone, and prostates at 24 weeks of age and then plated at equal numbers in vitro for colony-forming unit (CFU) assays. Compared with Aspn+/+, Aspn−/− bone marrow-, compact bone-, and prostate-derived MSCs formed significantly fewer CFUs (Fig. 1F–K; Supplementary Fig. S3A and S3B). To determine further the role of ASPN in self-renewal, fetal MSCs were isolated from Aspn+/+ and Aspn−/− mice and examined at early (P2) and late (P7) passage for CFU ability. CFU number was comparable between early passage Aspn−/− and Aspn+/+ fetal MSCs; however, late passage Aspn−/− fetal MSCs formed significantly fewer colonies compared with late passage Aspn+/+ fetal MSCs (Fig. 1L–N; Supplementary Fig. S3C and S3D). When added to late passage Aspn−/− fetal MSCs, recombinant mouse ASPN did not increase CFU number. Collectively, these data suggest that ASPN is necessary to sustain a subpopulation of MSCs with self-renewal capacity, but exogenous ASPN is not able to rescue self-renewal once MSCs have lost this capacity.
ASPN restricts MSC proliferation and differentiation
The loss of bone marrow MSCs in Aspn−/− mice suggests that ASPN may regulate MSC differentiation. Consistent with a more quiescent phenotype, Aspn+/+ fetal MSCs had a significantly decreased proliferation rate compared with Aspn−/− fetal MSCs (Fig. 2A). Microarray analyses of gene expression changes between Aspn+/+ and Aspn−/− fetal MSCs demonstrated that genes annotated to the cell cycle by KEGG were significantly downregulated in Aspn+/+ compared with Aspn−/− fetal MSCs (Fig. 2B). Consistent with a more proliferative phenotype, Aspn−/− fetal MSCs had elevated expression of several DNA synthesis–regulatory genes including Cdc45, Mcm2, Mcm3, Mcm5, and Mcm6. In contrast, Aspn+/+ fetal MSCs showed increased expression of genes that restrict cell-cycle progression including Cdkn2a. Interestingly, two genes with reported germline SNPs associated with aggressive prostate cancer, Cdkn1a and Cdkn1b (48), were differentially elevated between Aspn+/+ and Aspn−/− MSCs.
To more clearly define the role of ASPN in regulating MSC differentiation, early passage (P3) fetal MSCs from Aspn+/+ and Aspn−/− mice were cultured in osteogenic, adipogenic, and chondrogenic differentiation–inducing media and then analyzed for in vitro lineage differentiation. Differentiation to all three lineages (osteoblast, adipocyte, and chondrocyte) was significantly greater in Aspn−/− fetal MSCs compared with Aspn+/+ fetal MSCs as measured by Alizarin Red, Oil Red O, and Alcian Blue staining, respectively (Fig. 2C–H). In addition, lineage-specific differentiation markers such as Runx2, Adipoq, and Col10a1 were significantly higher in Aspn−/− fetal MSCs compared with Aspn+/+ fetal MSCs when cultured in osteogenic, adipogenic, and chondrogenic differentiation-inducing media, respectively (Fig. 2I–K). Compared with control media, Aspn expression in Aspn+/+ fetal MSCs significantly decreased in differentiation-inducing media, indicating that Aspn is downregulated during differentiation (Fig. 2L–N). Consistent with these findings, recombinant mouse ASPN decreased osteogenic differentiation of Aspn−/− fetal MSCs (Supplementary Fig. S3E and S3F). Interestingly, late passage (P8) Aspn−/− fetal MSCs showed increased differentiation in the absence of differentiation-inducing media (Supplementary Fig. S3G–S3K). Collectively, these data suggest that ASPN restricts MSC proliferation and inhibits early MSC differentiation prior to lineage specification.
ASPN binds to BMP-4 and restricts BMP-4–induced signaling and differentiation
How ASPN regulates MSC differentiation is not fully known. ASPN has been reported to regulate many extracellular signaling molecules (27, 28, 49–52), including multiple TGFβ family members (26, 27, 49, 51–53). Most TGFβ family members have pleiotropic and lineage-specific roles during differentiation; however, BMP-4 distinctly induces progenitor differentiation to osteoblast, adipocyte, and chondrocyte lineages (54). Microarray analysis of gene expression changes between Aspn+/+ and Aspn−/− fetal mouse MSCs demonstrated that TGFβ family signaling pathway genes annotated in GeneMAPP, including genes restricted to BMP signaling, were significantly downregulated in Aspn+/+ compared with Aspn−/− fetal MSCs (Fig. 3A). Consistent with this, BMP-4–induced SMAD1/5/9 phosphorylation was significantly decreased in Aspn+/+ compared with Aspn−/− fetal mouse MSCs (Fig. 3B and C). Similarly, recombinant mouse ASPN also restricted BMP-4–induced signaling in Aspn−/− fetal mouse MSCs (Fig. 3D and E).
Prior studies have shown that polymorphisms in the aspartic acid (D)-repeat length of human ASPN differentially regulate signaling (27) and that germline D-repeat variants are differentially associated with prostate oncologic outcomes (21). The germline variant of human ASPN containing 14-D repeats (ASPN D14) was shown to be significantly associated with metastatic prostate cancer recurrence after surgery while homozygosity for the ASPN D13 variant was significantly associated with a reduced risk of metastatic recurrence (21). To determine whether ASPN D14 and ASPN D13 differentially regulate BMP-4 signaling in human prostate stromal cells, we generated independent clones (A and B) of the prostate fibroblast cell line, WPMY-1, overexpressing either ASPN D13 or ASPN D14. Independent clones of WPMY-1 expressing the neomycin empty vector alone (Neo) were also generated as control lines. WPMY-1, which are germline ASPN D13/13, did not express ASPN under standard conditions (Fig. 3F; ref. 21). Similar to mouse ASPN, ASPN D14 significantly restricted BMP-4–induced signaling in WPMY-1 cells compared with WPMY-1-Neo and WPMY-1-ASPN D13 (Fig. 3F and G). BMP-4–induced signaling was slightly decreased in WPMY-1-ASPN D13 compared with WPMY-1-Neo, but was not significant (Fig. 3F and G).
To determine whether ASPN directly binds to BMP-4, mouse ASPN was overexpressed in HEK293T cells and then incubated with recombinant mouse BMP-4. Immunoprecipitation of secreted ASPN from the media demonstrated binding to BMP-4 (Fig. 3H). To determine whether BMP-induced signaling is necessary for differentiation in the absence of ASPN, we interrogated the ability of a BMP inhibitor to rescue differentiation in Aspn−/− MSCs. To do this, we compared osteogenic-induced differentiation in Aspn+/+ and Aspn−/− fetal MSCs in the absence of BMP-induced signaling by treating MSCs with the BMP inhibitor LDN-193189 or vehicle. Inhibition of BMP-induced signaling significantly restricted osteogenic-induced differentiation in Aspn−/− fetal MSCs as assayed by Alizarin Red staining and by expression of osteogenic-induced genes (Fig. 3I and J). To determine whether BMP-4 could rescue differentiation in Aspn+/+ fetal MSCs, we compared osteogenic-induced differentiation in Aspn+/+ and Aspn−/− fetal MSCs in the presence of exogenous recombinant mouse BMP-4. BMP-4 strongly induced differentiation in Aspn+/+ fetal MSCs (Fig. 3I and J). Taken together, these findings support a model by which mouse ASPN and human ASPN D14 restrict BMP-4–induced signaling and differentiation of MSCs.
ASPN expression pattern is distinctly conserved between MSCs and reactive stromal cells in the tumor microenvironment
MSCs also give rise, in part, to pathologic cell lineages including CAFs and injury or inflammation-associated fibroblasts (2, 3, 7, 8, 39, 47, 55, 56). We and others have previously reported that ASPN is highly expressed by reactive stromal cells in the microenvironment of primary tumors, with minimal to undetectable expression by primary adenocarcinoma cells or adenocarcinoma-derived cell lines (refs. 21–25; Fig. 4A and B; Supplementary Fig. S4A). In fact, ASPN expression in cancer cell lines negatively correlated with the epithelial marker EPCAM (Pearson r: −0.2478; P < 0.0001) and positively correlated with MSC markers including THY1 (Pearson r: 0.4064; P < 0.0001), ENG (Pearson r: 0.3607; P < 0.0001), VCAM (Pearson r: 0.1237; P < 0.0001), and ALCAM (Pearson r: 0.0952; P < 0.0019; Supplementary Fig. S4A–S4G; ref. 57). Elevated ASPN in primary prostate cancer from the TCGA was significantly associated with a worse disease/progression-free survival (Fig. 4C; refs. 58, 59). Similar to human prostate cancer, both Aspn mRNA and protein expression were elevated in the tumor microenvironment of mouse models of prostate adenocarcinoma compared with benign adjacent prostate stroma and stroma associated with mouse prostatic intraepithelial neoplasia (mPIN; Fig. 4D–G; Supplementary Fig. S4H).
To determine whether ASPN is also expressed in the metastatic tumor microenvironment, prostate cancer metastases from a Johns Hopkins School of Medicine rapid autopsy cohort (n = 15 patients; 60 metastases; ref. 37) were analyzed for ASPN expression by IHC and quantified by H-score. Elevated ASPN expression was detected in reactive stromal cells at various metastatic sites including lymph node, bone, and soft tissue such as lung and liver (Fig. 4H; Supplementary Fig. S4I). ASPN expression in metastatic prostate cancer adjacent stroma was elevated compared with benign prostate, lymph node, lung, and liver stroma, and it was comparable with ASPN expression in primary prostate cancer (Gleason sum ≥6) adjacent stroma (Fig. 4I). Thus, we demonstrate that ASPN is highly expressed in MSCs isolated from multiple sites and in reactive stromal cells associated with both primary and metastatic prostate cancers.
To determine whether ASPN expression is selective to reactive stroma in the tumor microenvironment or is generally conserved among reactive stroma, we next examined ASPN expression in human and mouse reactive stroma associated with prostate inflammation. Human prostate sections obtained after radical prostatectomy for prostate cancer that contained regions of both benign chronic inflammation and prostate cancer, but in distinct locations (≥2mm apart), were analyzed for ASPN expression by IHC. Compared with reactive stroma associated with human prostate cancer, stroma associated with distinct areas of chronic prostate inflammation expressed significantly lower levels of ASPN that were comparable with benign adjacent stroma (Fig. 4J and K). Similar to human prostate inflammation, CP1 E. coli–induced prostate inflammation, which is characterized by a marked vimentin-positive stromal response (38), was largely negative for ASPN expression (Fig. 4L–N). ASPN expression in inflammation adjacent stroma was comparable with benign prostate stroma and was only detected in perivascular regions. Collectively, these data suggest that ASPN expression is selectively conserved between MSCs and cancer-associated reactive stromal cells.
ASPN induces MSC and cancer cell migration
A fundamental property of MSCs is their ability to migrate to both local and distant areas of tissue injury or damage. Migration was significantly elevated in Aspn+/+ mouse fetal MSCs compared with Aspn−/− mouse fetal MSCs (Fig. 5A and B). Consistent with elevated motility, Aspn+/+ fetal MSCs cells had increased cytoskeletal remodeling compared with Aspn−/− MSCs as assayed by spontaneous motions of Arg-Gly-Asp (RGD)-coated microbeads that bind to the cytoskeletal network through cell surface integrin receptors (Fig. 5C and D). Similar dynamics of increased migration and cytoskeletal remodeling were also observed in WPMY-1-ASPN D14–overexpressing cells compared with WPMY-1-ASPN D13–overexpressing cells and WPMY-1-Neo control cells (Fig. 5E–H).
As a secreted protein in the tumor microenvironment, ASPN may also induce cancer cell migration. We examined cancer cell migration in response to conditioned media from Aspn+/+ and Aspn−/− fetal MSCs. Conditioned media from Aspn+/+ fetal MSCs increased B6MycCaP (60) and TRAMP-C2 mouse prostate cancer cell migration compared with Aspn−/− fetal MSC-conditioned media (Fig. 5I and J; Supplementary Fig. S5A and S5B). Conditioned media from WPMY-1-ASPN D14 cells also increased PC-3 and DU-145 migration compared with conditioned media from WPMY-1-ASPN D13 cells and conditioned media from WPMY-1- Neo cells (Fig. 5K and L; Supplementary Fig. S5C and S5D). Interestingly, a 1:1 mix of conditioned media from WPMY-1-ASPN D14 and WPMY-1-ASPN D13 cells also increased PC-3 migration similar to conditioned media from WPMY-1-ASPN D14 alone, thereby suggesting a dominant function of ASPN D14 relative to ASPN D13 (Fig. 5K and L).
ASPN-induced migration is calcium-dependent
How ASPN regulates prostate stromal and cancer cell migration is not fully understood. Migration in Aspn−/− fetal MSCs was neither rescued by the BMP inhibitor LDN-193189 nor further restricted by BMP-4, suggesting that ASPN regulates migration by an alternative mechanism (Supplementary Fig. S5E). To determine whether ASPN is able to directly enhance migration as opposed to functioning indirectly through the expression of other secreted factors, migration was examined in Aspn−/− cells in the presence of exogenous recombinant mouse ASPN. Elevated migration was observed in Aspn−/− cells when cultured with exogenous ASPN (Fig. 6A), suggesting that ASPN can, at least in part, directly enhance migration. To better determine how ASPN regulates migration, Aspn+/+ MSCs, Aspn−/− MSCs, WPMY-1-ASPN D14, WPMY-1-ASPN D13, and WPMY-1-ASPN Neo were compared by microarray for gene expression. Differentially expressed genes in Aspn+/+ MSCs compared with Aspn−/− MSCs as well as in WPMY-1-ASPN D14 compared with WPMY-1-ASPN Neo and in WPMY-1-ASPN D13 compared with WPMY-1-ASPN Neo were examined for KEGG Pathway enrichment. Genes annotated to cell adhesion, actin cytoskeleton, and cytokine–cytokine receptor interaction by KEGG were altered in cells expressing ASPN (Aspn+/+ MSCs, WPMY-1-ASPN D14, WPMY-1-ASPN D13) compared with cells deficient for ASPN (Aspn−/− MSCs and WPMY-1-ASPN Neo; Supplementary Fig. S5F). Because mouse ASPN and human ASPN D14 enhanced migration compared with ASPN-null and ASPN D13 cells, KEGG pathways selective to Aspn+/+ MSCs and WPMY-1-ASPN D14 were highlighted as a means to determine pathways potentially involved in ASPN-mediated migration. Interestingly, genes annotated to calcium signaling, Hedgehog signaling, chemokine signaling, WNT signaling, and TGFβ signaling by KEGG were altered in Aspn+/+ MSCs and WPMY-1-ASPN D14 when compared with cells deficient for ASPN (Fig. 6B and C). ASPN has been reported to bind calcium, likely through its poly-aspartate domain (61). Consistent with this, hierarchical clustering showed differential expression of several Gene Ontology (GO) calcium-related genes between Aspn+/+ and Aspn−/− fetal MSCs (Fig. 6D). Similarly, differential expression of calcium-related genes was also observed in WPMY-1-ASPN D14 compared with WPMY-1-ASPN D13 and WPMY-1-Neo (Fig. 6E). To determine whether ASPN-mediated migration depends on calcium, Aspn−/− MSCs were examined for migration in the presence of exogenous recombinant mouse ASPN and an extracellular calcium chelator, BAPTA. BAPTA restricted ASPN-induced migration, suggesting that extracellular calcium is necessary for ASPN's promigratory function (Fig. 6F and G).
ASPN alters the primary tumor microenvironment
To determine the role of ASPN in the tumor microenvironment, we utilized a novel allograft generated from a C57BL/6J Hi-Myc mouse metastatic lesion (B6CaP). This allograft is not to be confused with the B6MycCaP cell line used for migration studies that was generated on a C57BL/6N background (60). B6CaP allografts consisted of both cancer and tumor-infiltrating host cells including reactive stromal cells. Because of this stromal infiltration, B6CaP allografts highly expressed Aspn (Supplementary Fig. S6A). B6CaP organoids, which were enriched for cancer cells, had undetectable Aspn expression further establishing that Aspn expression was restricted to infiltrating stromal cells (Supplementary Fig. S6A). Subcutaneous growth of B6CaP allografts was similar in Aspn+/+, Aspn+/−, and Aspn−/− mice as average tumor volume at resection and average time to resection were not significantly different between mice (Fig. 7A and B). Despite comparable Aspn expression at tumor inoculation, B6CaP allografts in Aspn−/− mice had significantly reduced levels of Aspn at resection (Supplementary Fig. S6B).
B6CaP allografts were examined for MSCs and CAF-mediated properties including cancer stem cells, immune infiltration, and vascular development. Consistent with our in vivo and in vitro findings pertaining to the roles of ASPN in MSC self-renewal, differentiation, and migration, flow cytometry demonstrated that B6CaP tumors in Aspn−/− mice had significantly fewer MSCs (CD45−, CD105+, CD29+, and Sca-1+) compared with B6CaP tumors in Aspn+/+ mice (Fig. 7C). B6CaP tumors in Aspn−/− mice also had significantly fewer CD44+ cancer stem cells (CD45−, CD105−, CD31−, CD29+, Sca-1+, CD44+) compared with B6CaP tumors in Aspn+/+ mice (Fig. 7D). Flow cytometry for myeloid and lymphoid lineage cells demonstrated that Aspn−/− tumors had an increased percentage of tumor-infiltrating CD8+ T cells compared with Aspn+/+ cells (Fig. 7E; Supplementary Fig. S6C and S6D). IHC for tumor-associated vasculature showed that Aspn+/+ tumors were enriched for larger blood vessels compared with Aspn−/− tumors (Fig. 7F and G). These findings suggest that ASPN pleiotropically alters both the tumor and the tumor microenvironment.
ASPN promotes metastatic progression
MSCs, cancer stem cells, angiogenesis, and immune infiltration have all been shown to regulate metastatic development. Prior overexpression studies indicate that ASPN (62) and specifically ASPN D14 (21) promotes invasion and metastasis; however, this has not been determined in a genetic, immunocompetent mouse model. Thus, we examined metastatic progression in Aspn+/+ and Aspn−/− mice with B6CaP allografts. When grown subcutaneously and then resected, B6CaP allografts form lung metastases in 60%–70% of C57BL/6J mice. While primary tumor size was similar between Aspn+/+ and Aspn−/− mice, the development of lung metastases was strikingly different. B6CaP subcutaneous allografts formed lung metastases in 67% (4/6) of Aspn+/+ mice and in 50% (3/6) of Aspn+/− mice, but not in any of the Aspn−/− (0/6) mice (Fig. 7H–J). Collectively, these findings indicate that ASPN in the tumor microenvironment broadly impacts multiple cell types, and ultimately influences metastatic progression (Fig. 7K).
MSCs and CAFs reside in the tumor microenvironment and impact cancer growth and progression (4, 5, 9–13). These studies establish ASPN as a novel MSC factor expressed in the tumor microenvironment that regulates metastatic development. ASPN restricts early MSC differentiation, shown here by inhibiting BMP-4 signaling. While ASPN expression is reduced during differentiation to most MSC lineages, ASPN is highly expressed in CAFs and broadly impacts multiple cell types in the tumor microenvironment. Collectively, these findings underpin the role of ASPN as a key mediator of MSC multipotency and as a critical regulator of metastasis in the tumor microenvironment.
MSCs are a heterogeneous population of cells that include cells of mesenchymal origin including mesenchymal stem cells, mesenchymal stromal cells, and fibroblasts. We demonstrate that ASPN is expressed in MSCs isolated from distinct sites including bone marrow, adult prostate, and fetal prostate in humans and mice. MSCs give rise to physiologic connective tissue lineages, and differentiation is regulated by both conserved and lineage-specific factors (6). ASPN expression is diminished during MSC differentiation to multiple lineages, and our findings indicate that ASPN functions prior to MSC lineage commitment to suppress the differentiation of a stem population of multipotent MSCs capable of self-renewal. In the absence of ASPN, fewer MSCs reside in the bone marrow and an intermediate population of multipotent MSCs with diminished self-renewal potential predominates in the prostate. This is evidenced by our flow cytometry data demonstrating that in contrast to Aspn+/+ MSCs in the prostate, MSCs from Aspn−/− prostates were enriched for intermediate MSCs (Sca-1med) as opposed to stem MSCs (Sca-1hi). Consistent with these in vivo findings, Aspn−/− bone marrow–derived, compact bone-derived, prostate-derived, and late passage fetal MSCs had a decreased capacity for self-renewal in vitro compared with Aspn+/+ MSCs. Studies using ASPN-deficient cells demonstrated that ASPN restricts MSC differentiation in vitro along multiple lineages. Taken together, these data indicate that ASPN functions as a novel MSC factor that maintains MSC multipotency.
MSC lineage commitment and progression through differentiation is regulated by many TGFβ family members (54). ASPN is a secreted extracellular protein that has been shown to antagonize several members of the TGFβ family including BMP-2 (49), BMP-4 (51), and TGFβ1 (27). Most TGFβ family members have pleiotropic roles in regulating MSC differentiation to connective tissue lineages (54). In contrast to other TGFβ family members, BMP-4 distinctly induces early MSC differentiation along several lineages including osteoblasts, adipocytes, and chondrocytes (54). Our data demonstrate that ASPN binds to BMP-4 and restricts BMP-4–induced signaling in MSCs and support a model by which MSC-secreted ASPN locally restricts BMP-4–induced MSC differentiation.
Similar to MSCs, reactive stroma cells in multiple cancer types, including prostate (20–22), breast (23), scirrhous gastric (24), and pancreatic (25), also highly expresses ASPN while benign-associated stroma is largely negative for ASPN expression. In addition to localized cancer, we show that ASPN is widely expressed in prostate cancer metastases. In contrast to the tumor microenvironment, ASPN was not expressed in reactive stroma in the inflammatory microenvironment. This suggests that ASPN expression is not conserved across all types of reactive stroma, but has some specificity to reactive stroma in the tumor microenvironment. Tumor-associated reactive stroma is a heterogeneous population of cells that includes MSCs and CAFs. Studies support that CAFs are, in part, derived from MSC precursors (47, 55, 56). Accordingly, CAFs may be more similar to MSCs than inflammation-associated reactive stromal cells based on the data presented herein. Collectively, these findings indicate that ASPN expression is highly conserved between MSCs and the reactive stroma in both primary and metastatic tumor microenvironments.
These studies indicate that ASPN broadly impacts both the tumor and the tumor microenvironment. In the absence of ASPN, B6CaP allografts had fewer tumor-associated MSCs supporting that ASPN regulates the number of multipotent MSCs in the tumor microenvironment. In addition, the immune infiltration and vasculature was significantly altered in Aspn−/− mice. Effects on other cell types in the tumor microenvironment and on cancer stem cells may be directly mediated by ASPN or indirectly regulated through MSC deficiencies. In addition to impacting the tumor microenvironment, ASPN may promote tumor progression by directly enhancing migration of both stromal and cancer cells. ASPN has been implicated in the activation of multiple migratory pathways including FGF-2 (50), IGFR (51), RAC1 (24), CD44 (24, 25), and EGFR (28, 62). Our data support that exogenous ASPN is able to enhance migration, likely, by a mechanism independent of its ability to restrict BMP4-induced signaling. Our findings support that ASPN-mediated migration is dependent on extracellular calcium and that genes annotated to the KEGG calcium signaling pathway are differentially expressed in ASPN-expressing cells compared with ASPN-null and ASPN D13 cells. Future studies will be needed to determine how ASPN mechanistically regulates MSC and prostate cancer cell migration.
Mouse and human ASPN are highly homologous, including the D-repeat domain. While C57BL/6 mice have eight D-repeats interrupted by a single asparagine (N), the D-repeat length in humans is variable. Recent reports suggest that the D-repeat domain length may regulate cancer progression (21, 29). Patients harboring germline allele(s) of ASPN D14 were more likely to develop metastatic prostate cancer, while homozygous germline ASPN D13 was associated with a reduced risk of metastatic prostate cancer (21). Animal models have shown that ASPN D14 promotes metastatic prostate cancer progression (21), while ASPN D13 restricts breast (63) and prostate (21) cancer progression. Our data show that ASPN D14 increased stromal and cancer cell migration compared with ASPN D13. Intriguingly, prostate cancer patient–based data suggest that ASPN D14 may be dominant over ASPN D13. Consistent with this, mixed conditioned media from ASPN D13- and ASPN D14–expressing cells increased cancer cell migration similar to conditioned media from ASPN D14–expressing cells alone while conditioned media from ASPN D13 did not increase cancer cell migration. Collectively, our findings reveal that ASPN D14 and ASPN D13 may differentially regulate molecular and cellular pathways that are fundamental to metastasis.
In conclusion, our data identify ASPN as a novel MSC factor that is conserved in the tumor microenvironment and promotes metastatic progression. Further studies are warranted to determine the efficacy and feasibility of therapeutically targeting ASPN in prostate cancer, and potentially other solid tumor types.
Disclosure of Potential Conflicts of Interest
T.L. Lotan reports receiving other commercial research support from GenomeDx, Ventana/Roche, Myriad Genomics, and is a consultant/advisory board member for Janssen. E. Davicioni has ownership interest (including stock, patents, etc.) in GenomeDx. D.L.J. Thorek has ownership interest (including stock, patents, etc.) in Diaprost AB and is a consultant/advisory board member for Diaprost AB. B.H. Park reports receiving a commercial research grant from Abbvie, reports receiving other commercial research support from Foundation Medicine, has ownership interest (including stock, patents, etc.) in Loxo Oncology, and is a consultant/advisory board member for Loxo Oncology. No potential conflicts of interest were disclosed by the other authors.
Conception and design: R.M. Hughes, B.H. Park, P.J. Hurley
Development of methodology: R.M. Hughes, W.N. Brennen, P.J. Hurley
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R.M. Hughes, B.W. Simons, H. Khan, V. Kugler, S. Torquato, D. Theodros, M.C. Haffner, T.L. Lotan, J. Huang, E. Davicioni, S.S. An, R.C. Riddle, D.L.J. Thorek, I.P. Garraway, J.T. Isaacs, W.N. Brennen, P.J. Hurley
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R.M. Hughes, B.W. Simons, H. Khan, V. Kugler, S. Torquato, E. Davicioni, S.S. An, R.C. Riddle, D.L.J. Thorek, E.J. Fertig, W.N. Brennen, B.H. Park, P.J. Hurley
Writing, review, and/or revision of the manuscript: R.M. Hughes, H. Khan, S. Torquato, D. Theodros, E. Davicioni, S.S. An, D.L.J. Thorek, I.P. Garraway, E.J. Fertig, W.N. Brennen, B.H. Park, P.J. Hurley
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R.M. Hughes, H. Khan, R. Miller, S. Torquato, E. Davicioni, P.J. Hurley
Study supervision: P.J. Hurley
We thank Edward M. Schaeffer, Ashley E. Ross, Daniel Ardeljan, Leigh Ellis, Daniele M. Gilkes, Inês Godet, Timothy E. Krueger, and Karen Cravero for thoughtful discussions and for sharing resources and experimental protocols. We thank the Johns Hopkins Sidney Kimmel Comprehensive Cancer Center Flow Cytometry and Cell Imaging Core Facilities supported by the Cancer Center Core Grant P30 CA006973. We also thank the Prostate Cancer Biorepository Network (PCBN), supported by the Department of Defense Prostate Cancer Research Program, Department of Defense Award No. W81XWH-10-2-0056 and W81XWH-10-2-0046. We acknowledge the use of tissues procured by the National Disease Research Interchange (NDRI) with support from NIH grant 2 U42 OD011158. This work was supported by Department of Defense Prostate Cancer Research Program W81XWH-11-PCRP-IDA (to P.J. Hurley), The Patrick C. Walsh Prostate Cancer Fund (to P.J. Hurley), The Willowcroft Foundation (to P.J. Hurley), The Johns Hopkins Catalyst Award (to P.J. Hurley), The American Cancer Society 131356-RSG-17-160-01-CSM (to P.J. Hurley), The National Cancer Institute/NIH RO1CA211695-01A1 (to P.J. Hurley, B.W. Simons, T.L. Lotan, W.N. Brennen, R.M. Hughes), R01CA194024 (to B.H. Park), and R01CA201035 (to D.L.J. Thorek), and the Commonwealth Foundation (to B.H. Park). W.N. Brennen acknowledges support from the Prostate Cancer Foundation as the Clay and Lynn Hamlin Young Investigator. B.H. Park acknowledges support from Susan G. Komen as a Komen Scholar.
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.