A Zebrafish Model Discovers a Novel Mechanism of Stromal Fibroblast-Mediated Cancer Metastasis

Tumor tissues contain diverse cell populations that relentlessly communicate with each other in the tumor microenvironment (TME; refs. 1–5). In addition to cancer cells, infiltration of other cellular components including vascular cells, inflammatory cells, fibroblasts, and other immune cells significantly contributes to tumor growth, invasion, and metastasis (1–5). CAFs often represent an irreversibly activated fibroblast population that expresses alpha-smooth muscle actin (α-SMA; refs. 6–8). Once activated, CAFs produce various molecular signaling molecules to facilitate cancer growth and invasion. In addition, CAFs communicate with other nonmalignant cellular components including vascular and inflammatory cells to execute their multifunctions in modulation of tumor growth and invasion. For example, activated CAFs produce inflammatory cytokines and angiogenic factors to indirectly modulate tumor growth and metastasis (8). Metastasis is the major cause of mortality in cancer patients and occurs often at the early stage during cancer development (9, 10). Cancer metastasis is a complex process involving intravasation, extravasation, formation of the initial metastatic niche, and regrowth of metastatic nodules (10). A clinically detectable metastatic mass often represents the endpoint of the metastatic cascade. Advanced imaging technologies for early detection of tiny metastatic masses, especially microscopic cancer nodules, are not available. While it is known that most cancers possess the intrinsic features for metastasis, the exact events of cancer cell dissemination from the primary site and intravasation into the circulation are not known. In particular, it is unknown if metastasis already occurs when primary tumors are in tiny sizes.

Owing to their transparent and immunoprivileged features, developing zebrafish offers an outstanding opportunity to study single cancer cell behaviors from human and other species origins (11). Particularly, advanced imaging technologies have made it plausible for reconstitution of the TME with the multicolor-coded diverse cell populations (11). The color-coded TME reconstitution in zebrafish would allow us directly visualize single cancer cell migration, interaction with other host cells, metastatic intravasation, and extravasation in a living fish body. Particularly, this metastatic model provides a unique opportunity to study cancer metastasis at the early stage during cancer development. In this study, we showed that CAFs were a crucial cellular component in TME for promoting cancer metastasis. In several cancer models, we provided compelling evidence to demonstrate that CAFs but not normal fibroblasts promote cancer cell metastasis when primary tumors were at microscopic sizes. Moreover, metastatic cancer cells remained tightly associated with CAFs, supporting the essential role of CAFs for subsequent extravasation and the metastatic niche formation. Unraveling mechanisms underlying CAFs in cancer metastasis has conceptual and clinical implications for cancer therapy.

All animal experiments were approved by the Northern Stockholm Experimental Animal Ethical Committee. Zebrafish embryos of the transgenic strain expressing enhanced GFP under the fli1 promoter (Fli1:EGFP; ref. 12) were raised at 28°C under standard experimental conditions. Zebrafish embryos at the age of 24 hpf were incubated in aquarium water containing 0.2 mmol/L 1-phenyl-2-thio-urea (PTU, Sigma). At 48-hpf, Fli1:EGFP zebrafish embryos were dechorionated with a pair of sharp-tip forceps and anesthetized with 0.04 mg/mL of tricaine (MS-222, Sigma). Anesthetized embryos were subjected for microinjection. Various tumor cells and fibroblasts were labeled in vitro with 2 μg/mL of 1,1′-Dioctadecyl-3,3,3′,3′- tetramethylindocarbocyanine perchlorate (DiI, Sigma) or a Vybrant DiD cell-labeling solution (Life Technologies), respectively. Approximately 500 tumor cells or a mixture consisting of 300 cancer cells and 200 fibroblasts were resuspended in DMEM (Hyclone) and 5 nL of the cell solution were injected into the perivitelline space (PVS) of each embryo by an Eppendorf microinjector (FemtoJet 5247, Eppendorf and Manipulator MM33-Right, Märzhäuser Wetziar). Non-filamentous borosilicate glass capillaries needles were used for injection and the injected zebrafish embryos were immediately transferred into PTU aquarium water. Zebrafish embryos were monitored 72 hours for investigating tumor invasion and metastasis using a fluorescent microscope (Nikon Eclipse C1).

Cancer cells and fibroblasts used in this study included human HuH7 hepatocellular carcinoma cells (HCC; ATCC); human A549 lung adenocarcinoma epithelial cells (NSCLC; kindly provided by Dr. Tatiana Pavlova, the Karolinska Institute, Sweden); human A431 squamous carcinoma cells (kindly provided by Dr. Keiko Funa, Gothenburg University, Sweden); human MiaPaCa-2 pancreatic ductal adenocarcinoma cells (PDAC; catalog no. FS0095, ATCC); human MDA-MB-231 breast cancer cells (ATCC HTB-26); human SW480 colon cancer cells (ATCC catalog no. ab01589); mouse MC38 colon adenocarcinoma cells (CRC; kindly provided by Dr. Ruben Hernandez, University of Navarra, Pamplona, Spain); murine E0771 breast cancer cells (catalog no. 940001, CH3 BioSystems); normal human prostate fibroblasts (NF;catalog no. P10893, Innoprot); human colorectal cancer (CRC)–derived CAFs (Isolated from a human colorectal cancer, kindly provided by Dr. Johan Hartman at the Department of Oncology-Pathology, Karolinska Institutet, Stockholm, Sweden); human prostate cancer–derived CAFs (PC-CAFs, catalog no. 159366A1, Asternad); and mouse cancer-derived CAFs (isolated from a T241 tumor). T241 tumors grown in GFP-transgenic mice were cut into small pieces and digested at 37°C for 60 minutes in a solution containing 0.15% collagenases I and II. After PBS washing, cells were filtered through a 40-μm cell strainer and were centrifuged at 1,500 rpm for 10 minutes. Single cells were stained with a rat anti-mouse CD31 antibody (1:100; catalog No. 553370, BD-Pharmingen) for 45 minutes, followed by incubation on ice for 20 minutes with an Alexa Fluor 555–labeled goat anti-rat antibody. GFP⁻ and GFP⁺ cell populations were analyzed and sorted by FACS (MoFlo XTD, Beckman Coulter) and were cultured in 10% FBS-DMEM. Cells less than 10 passages were used through experiments. HuH7, A431, MiaPaCa-2, MDA-MB-231, and MC38 cancer cells were grown in 10% FBS-DMEM. E0771 cells were grown in 10% FBS-RPMI-1640 medium. A549 cells were grown in 10% FBS-IMEM. Human normal fibroblasts and CAFs were grown in a Fibroblast Medium Kit (catalog no. P60108, Innoprot), supplemented with 10% FBS. All cell lines were not authenticated after purchase or transferred from other laboratories but were routinely tested negative for mycoplasma by using the Mycoplasma Detection Kit (Lonza Inc.). HGF was purchased from the R&D Systems Inc., PDGF-BB was purchased from the PeproTech, FGF-2 was purchased from the Phamacia & UpJohn, and TGFβ-1 was purchased from the Thermo Fisher (catalog no. RP 39357).

Fresh human colon cancer tissues were provided from the Karolinska University Hospital, and all tissue studies and handling were approved by the Stockholm Regional Ethical Committee (Ethical number 2016/957-31). Patient samples were transferred to research laboratories anonymously. Fresh cancer samples were washed with serum-free DMEM, cut into small pieces, and were transferred to a 0.15% collagenase IV solution, followed by incubation at 37°C for 40 minutes. Digested cells were filtered through a 40-mm cell strainer (BD Biosciences) and centrifuged at 1,500 rpm for 10 minutes. The single-cell suspension was incubated in a Fibroblast Medium Kit (catalog No. P60108, Innoprot) for 24 hours, allowing fibroblasts to attach on culture plates. Unattached cells were removed after 24-hour incubation, and the adherent cells were further cultivated for experiments. Isolated fibroblasts from tumor tissues were defined as CAFs. This method yielded a relatively pure population of fibroblasts. Cultured CAFs less than five passages were used for our experiments.

DiI-labeled HuH7 single-cell suspension at the density of 2.5 × 10⁴ cells/mL was mixed at a 3:1 ratio with normal fibroblasts or CAFs. A same number of tumor cells were used as a control. Mixed and unmixed cells were transferred onto a 12-well plate with flat bottom (Corning, Costar 12-well cell culture plate, Sigma Aldrich, catalog No.CLS3513) and were cultured for 24 hours. After confluency of monolayers, a sterile razor blade was used to create a uniformed cell-free wound of 0.5 × 1.0 cm on each well. The plates were washed 3 times with PBS to remove floating cells. Bright-field and fluorescent pictures were taken immediately after scratch (0 hour). Cells were cultured for 48 hours with 2% FBS-DMEM containing HuH7 alone or mixed cells with 2% FBS-IMDM and 2% FBS-DMEM. Cells were photographed and migrating cancer cells were counted manually using a Photoshop CS5 software program.

Matrigel invasion assay was performed as previously described (13). Single-cell suspension of DiI-labeled HuH7 tumor cells, fibroblasts, or CAFs (2 × 10⁴ cells/mL) were prepared in 10% FBS-DMEM. Tumor cells were mixed with fibroblasts, or CAFs at a 1:1 ratio and cell mixtures were transferred to a low adhesive 96-well plate with rounded bottom (Nunc low cell binding MicroWell plate; catalog no. Z721093-8EA, Sigma), followed by 24-hour coculture. Sphere-containing medium (50 μL) was mixed with 100 μL of Matrigel in each of 96-well plate. Bright-field and fluorescent pictures were taken immediately (0 hour). Cells were cultured for 48 hours with 2% FBS-DMEM containing HuH7 alone or mixed cells with 2% FBS-IMDM and 2% FBS-DMEM. Cells were photographed and migrating cancer cells were counted manually using a Photoshop CS5 software program.

Paraffin-embedded tissue sections in 5-mm thickness were stained with hematoxylin and eosin (H&E) using our standard protocols (10, 13–27). For α-SMA and FSP1 staining, the paraffin-embedded tissue sections were incubated at 4°C overnight using a mouse anti-human α-SMA antibody (1:4; catalog no. Clone 1A4 Dako) mixed with a rabbit anti-human/mouse FSP1 antibody (1:200, catalog no. 07-2274, Millipore). A mixture of secondary antibodies comprising an Alexa Fluor 555-labeled goat anti-mouse antibody (1:200; Invitrogen) and an Alexa Fluor 488-labeled donkey anti-rabbit antibody (1:200; Invitrogen) was used. Paraffin-embedded tissue sections were incubated at 4°C overnight with a goat anti-human/mouse desmin antibody (1:200; catalog no. AF3844, R&D Systems) mixed with a rabbit anti-human/mouse PDGFRα antibody (1:200; catalog no. 3164, Cell Signaling Technology). A mixture of secondary antibodies comprising an Alexa Fluor 555-labeled goat anti-mouse antibody (1:200; Invitrogen) and an Alexa Fluor 488-labeled donkey anti-rabbit antibody (1:200; Invitrogen) was used. Positive signals were captured using a fluorescence microscope equipped with a Nikon DS-QilMC camera (Nikon Corporation).

For α-SMA and PDGFRα staining, the fixed CAFs and NFs were incubated at 4°C overnight using a mouse anti-human α-SMA antibody (1:4; catalog no. Clone 1A4, Dako) mixed with a rabbit anti-human/mouse PDGFRα antibody (1:200; catalog no. 3164, Cell Signaling Technology), followed by staining with a mixture of secondary antibodies comprising an Alexa Fluor 555-labeled goat anti-mouse antibody (1:200; Invitrogen) and an Alexa Fluor 488-labeled donkey anti-rabbit antibody (1:200; Invitrogen). For desmin and FSP1 staining, the fixed CAFs and NFs were incubated at 4°C overnight with a goat anti-human desmin antibody (1:200; catalog no AF3844, R&D Systems) mixed with a rabbit anti-human/mouse FSP1 antibody (1:200, catalog no. 07-2274, Millipore). A mixture of secondary antibodies comprising an Alexa Fluor 555-labeled goat anti-mouse antibody (1:200; Invitrogen) and an Alexa Fluor 488-labeled donkey anti-rabbit antibody (1:200; Invitrogen) was used. Positive signals were captured using a fluorescence microscope equipped with a Nikon DS-QilMC camera (Nikon Corporation).

Quantitative PCR (qPCR) was performed according to a standard protocol (28). In brief, total RNA samples were prepared by a GeneJet RNA Purification Kit (catalog no. K0731, Thermo Scientific) and a cDNA Synthesis Kit (catalog no. K1632, Thermo Scientific). A Power SYBR Green Master Mix Kit (catalog no. 4367659, Applied Biosystems) and a Step One Plus real-time PCR system (Applied Biosystems) were applied. The following paired primers were used: human Gapdh forward: 5′-CCAGCAAGGACACTGAGCAA-3′ and human Gapdh reverse: 5′-GGGATGGAAATTGTGAGGGA-3′; human α-Sma forward: 5′-AGGGGGTGATGGTGGGAATG-3′ and human α-Sma reverse: 5′-GCCCATCAGGCAACTCGTAAC-3′; human Fap forward: 5′-GCTAACTTTCAAAAACATCTGGAAAAATG-3′ and human Fap reverse: 5′-GTAATATGTTGCTGTGTAAGAGTATCTCC-3′; mouse Ccl5 forward 5′- GCTGCTTTGCCTACCTCTCC-3′ and mouse Ccl5 reverse 5′-TCGAGTGACAAACACGACTGC-3′; mouse Cxcl12 forward 5′-TGCATCAGTGACGGTAAACCA-3′ and mouse Cxcl12 reverse 5′-TTCTTCAGCCGTGCAACAATC-3′; mouse Il33 forward 5′-ATGGGAAGAAGCTCATGCTG-3′ and mouse Il33 reverse 5′-CCGACGACTTTTTCTGAAGG-3′; mouse Il1rl1 forward 5′-ATTCAGGGGACCATCAAGTG-3′ and mouse Il1rl1 reverse 5′-CGTCTTGGAGGCTCTTTCTG-3′; mouse Gapdh forward 5′-CCAGCAAGGACACTGAGCAA-3′ and mouse Gapdh reverse 5′- GGGATGGAAATTGTGAGGGA-3′.

FGF-2 (10 ng/mL), HGF (10 ng/mL), PDGF-BB (10 ng/mL), and TGFβ-1 (10 ng/mL) were used to stimulate the fibroblasts for 24 hours. The growth factor–stimulated fibroblasts together with HCC or NSCLC cells were coinjected into the perivitelline space of each zebrafish embryo using the same method described above.

Mouse stromal cells were treated with or without 100 ng/mL PDGF-BB for 72 hours and total RNA was extracted with an RNA extraction kit (Thermo Fisher Scientific). Samples (triplicates per group) were used for the gene expression-profiling analyses. Normalization and analysis for differentially expressed genes were performed using robust multi-array analysis and significance analysis of microarrays (SAM) through the R statistical software packages, oligo and samr. Heatmaps were presented for up- and downregulation of gene expression using the Multiple Experiment Viewer system (version 4.7). Gene array data were deposited in the Gene Expression Omnibus, accession number GSE33717.

Statistical analysis was performed using the standard Student t test for pair comparisons and ANOVA analysis for multiple factors. Data were presented as means of determinants (±SEM) and P values of *, P < 0.05; **, P < 0.01; ***, P < 0.001 were deemed as statistically significant, very significant, and extremely significant, respectively.

To study the role of CAFs in facilitating cancer invasion at the initial steps of the metastatic cascade, we developed a zebrafish cancer model in which different cell populations in the TME can be visualized by color codes. Solid tumors contain high proportions of a fibrotic component that is often associated with an invasive and metastatic phenotype. To visualize the fibrotic component in human tumors, we defined CAFs in a freshly dissected human colorectal cancer (CRC) using several specific markers. H&E staining showed a high proportion of the fibrotic component in this invasive and metastatic tumor (Supplementary Fig. S1A). Immunohistochemical staining with fibroblast specific markers (15), including alpha-smooth muscle actin (α-SMA), fibroblast specific protein-1 (FSP-1), desmin, and PDGFRα, validated the existence of the high CAF content in this invasive CRC tumor (Supplementary Fig. S1A).

We next isolated and defined these CAFs and human normal fibroblasts for expression of FSP-1, desmin, and PDGFRα (Supplementary Fig. S1B). Interestingly, human normal fibroblasts only expressed a very low level of α-SMA (Supplemental Fig. 1B and C). These findings demonstrate that isolated CAFs were α-SMA⁺ myofibroblasts, which is known to be associated with cancer invasion. In contrast, human SW480 CRC cancer cells with an epithelial origin lacked detectable levels of fibrotic markers (Supplementary Fig. S1C and S1D). Together, these data show that primary CAFs isolated from a human CRC possess distinct marker features compared with healthy human fibroblasts.

We next investigated the functional properties of CAFs in cancer invasion using in vitro cancer cell migration and invasion assays. In a cell motility assay, a mixture of cancer cells and CAFs was cocultured at a 3:1 ratio for 48 hours, followed by creating a uniform wound. After 24 hours, migrating cancer cells in the wound area of each sample were counted. CAFs isolated from a human prostate cancer and isolated from a human CRC markedly promoted tumor cell migration (Fig. 1A and B). Normal human fibroblasts lacked the ability to facilitate cancer cell migration. In contrast, normal human fibroblasts significantly inhibited cancer cell migration in this in vitro assay (Fig. 1A and B).

To further study the functional features of CAFs in cancer invasion, we developed a Matrigel assay in which cancer cells and CAFs were mixed at a 1:1 ratio. Again, CAFs isolated from two independent tumor types, but not normal fibroblasts, markedly promoted cancer cell invasion (Fig. 1C and D). It appeared that CRC-CAFs exhibited more potent effects than PC-CAFs in promoting cancer invasion, perhaps reflecting the intrinsic invasive property of various cancers. These data demonstrate that CAFs promote cancer cell invasion in these in vitro assays.

We recently developed a zebrafish model to study cancer metastasis, particularly to decipher the initial steps of the metastatic cascade, which remains a challenging issue in other animal models such as in rodent cancer models. The transparent and immune-privileged nature of zebrafish embryos allowed us to monitor human or mouse tumor cell migration and metastasis in the living fish body even when primary tumors were at microscopic sizes. Approximately 500 DiI-red–labeled human HCC cells were injected into the perivitelline space of each zebrafish embryo. Tumor cell invasion and metastasis in the zebrafish body were monitored at 72 hours postinjection. Injection of HCC cells alone resulted in dissemination of tumor cells from the original primary tumor sites (Fig. 2A). However, only a limited number of cancer cells were detectable at the surrounding regions of primary tumors and metastatic tumor cells were rarely found in distal regions such as the trunk region of the zebrafish embryo (Fig. 2A and D).

To study behaviors of CAFs in the zebrafish body, CAFs, shown in green, were labeled with DiD dye and 500 CAF cells were implanted into the perivitelline space of each zebrafish embryo. Surprisingly, CRC-CAFs and PC-CAFs were highly “metastatic” and substantial numbers of CAFs were disseminated to distal regions of the zebrafish body (Fig. 2B–E). These findings show that CAFs were highly motile and “metastatic” in the zebrafish body.

To study the role of CAFs in facilitating cancer metastasis in TME, we developed a zebrafish model in which different color-coded CAFs and cancer cells were coimplanted into the zebrafish body (Fig. 3A; ref. 11). Implantation of HCC tumor cells alone resulted in a modest effect of cancer metastasis (Fig. 3B, D, and E). Coimplantation of HCC cells with normal human fibroblasts significantly reduced the number of metastatic cancer cells (Fig. 3C–E), suggesting that the healthy fibroblasts might inhibit cancer cell motility and metastasis. However, coimplantation of CRC-CAFs and PC-CAFs with cancer cells markedly increased the number of metastatic cells in different parts of the zebrafish body (Fig. 3C–E). The disseminated tumor cells were visible in the head and trunk regions of the zebrafish body. Quantification analysis showed that the number of distal metastatic tumor cells was markedly higher in CRC-CAF-HCC- and PC-CAF-HCC-coimplanted zebrafish (Fig. 3F). Surprisingly, most distal metastatic cancer cells were associated with CAFs that showed double color positivity (Fig. 3C). Approximately, 80% of the distal metastatic cancer cells were associated with CAFs (Fig. 3F and G). These findings provide compelling evidence that CAFs facilitate cancer cell metastasis and the intimate interactions between cancer cells and CAFs remain even after the intravasation event.

To further validate these findings, we performed similar experiments with a human lung cancer cell line (A549 NSCLC). The metastasis-promoting effect of CAFs was further validated in this cancer model (Fig. 4). Again, a majority of metastatic cancer cells remained association with CAFs. Similar data were also obtained from other tumor models including squamous carcinoma, pancreatic ductal adenocarcinoma (PDAC), and breast cancer, showing that CFAs markedly promoted cancer metastasis (Fig. 5). We also provided compelling evidence showing that mouse CAFs isolated from a fibrosarcoma also markedly promoted mouse breast cancer and CRC cancer metastasis (Supplementary Figs. S2 and S3). A video clip showed dissemination of CAF-cancer cell complex from the original tumor site (Video 1). Together, several independent cancer models demonstrate that CAFs display marked effects in promoting metastasis and remained in close association with metastatic cancer cells in distal tissues.

Knowing that the normal healthy fibroblasts lacked the ability for promoting cancer metastasis, we next stimulated these cells with growth factors targeting fibroblasts. It is known that fibroblasts are target cells for fibroblast growth factor-2 (FGF-2), hepatocyte growth factor (HGF), platelet-derived growth factor–BB (PDGF-BB), and transforming growth factor-beta 1 (TGFβ-1; ref. 6). Therefore, we stimulated the normal healthy fibroblasts with FGF-2, HGF, PDGF-BB, and TGFβ-1 and the stimulated fibroblasts were subsequently coimplanted into the perivitelline space of each zebrafish. Interestingly, FGF-2–stimulated fibroblasts markedly promoted HCC metastasis (Fig. 6). Similarly, HGF-, PDGF-BB-, TGF β-1−stimulated normal fibroblasts also significantly promoted HCC cancer metastasis (Fig. 6). We provide independent evidence demonstrating that HGF, PDGF-BB, TGFβ-1 stimulated−normal fibroblasts considerably promoted NSCLC cancer metastasis (Supplementary Fig. S4). These findings provide new mechanistic insights on fibroblast-stimulating factors in promoting cancer metastasis.

To further gain mechanistic insights on fibroblast-mediated metastasis, we analyzed molecular signaling components that are known to be involved metastasis. Knowing PDGF-BB–stimulated fibroblasts were able to promote metastasis, we have focused our efforts on studying molecular changes of the PDGF-BB–stimulated fibroblasts. Metastasis-related genes including CCLs, CCRs, CXCLs, CXCRs, ILs, ILRs, and MMPs were particularly investigated. CCLs, including CCL5, CCL2, and CCL7, were upregulated, whereas CCL9, CXCL12, and CXCL14 were downregulated (Fig. 6G and H). Interestingly, IL-33 and its receptor ST2 (Il1rl1) were markedly upregulated (Fig. 6 and Supplementary Fig. S5). These data suggest that complex molecular mechanisms may be involved in mediating fibroblast-stimulated cancer metastasis.

In this work, we provide novel mechanistic insights on CAFs in facilitating cancer metastasis. In the presence of CAFs, dissemination of cancer cells from the primary site to distal organs occurs at the early stage during cancer development. If these findings are successfully translated to clinical situations in cancer patients, they imply that tumors at microscopic size containing only a few hundred cells are already metastatic. At a microscopic size, the existence of a tiny tumor nodule in a given tissue or organ may not even be dependent on angiogenesis and cooption of the preexisting host vasculatures is sufficient for maintaining cancer cell survival (29, 30). Infiltration and activation of fibroblasts in these microscopic tumors are sufficient to promote cancer metastasis. If so, this may imply that cancer metastasis can occur in dormant tumors in which the primary tumor mass is not detectable and would not grow. Yet, these microscopic dormant tumors can cause serious troubles in cancer patients if fibroblasts become activated in tumors. Metastasis may also occur in microscopic avascular tumors in which tumor cells invade the preexisting surrounding blood vessels. Our zebrafish model provides compelling evidence to support this important concept. First, in the zebrafish model the implanted primary tumors remain similar sizes throughout the entire experimental duration. Thus, the growth-related tumor size is irrelevant to metastatic potentials. Second, the tiny microscopic tumor mass is not vascularized and dissemination of cancer cells is mediated through the preexisting surrounding host vasculatures. As various cellular components are color-coded, interactions of cancer cells and fibroblasts can be directly visualized at a single-cell level.

Our findings show that a majority of metastatic cancer cells in the circulation and distal tissues remain tightly associated with CAFs. These data imply that CAFs facilitate cancer metastasis at several distinct steps of the metastatic cascade. First, CAFs are crucially important for the initial intravasation of cancer cells into the circulation. In support of this notion, we show that CAFs are able to promote broad dissemination of cancer cells into the circulation. Secondly, the CAF–tumor cell complex, although lack of evidence, may be essential for extravasation to distal organs. Indeed, a recent study shows that fibroblasts and mesenchymal cells regulate cancer cell extravasation (31). This view was supported by several independent studies showing that metastatic cancer cells carry their fibrotic soil with them to distal organs (32–36). Thirdly, disseminated CAFs may promote the formation of initial metastatic niches and subsequent regrowth in distal tissues and organs. Thus, CAFs play dynamic roles in promoting cancer metastasis at different locations. One of the important issues underlying advantages of the CAF–tumor cell complex for dissemination is that the fibrotic cell type is the engine for promoting dissemination. Implantation of CAFs alone without cancer cells causes broad dissemination from the primary site supports the fact that CAFs enforce cancer cells moving away from their primary sites. Thus, our findings propose that the fibrotic soil carry the seeds, but not the seeds carrying soil, to distal tissues and organs. An extended view to clinical implications would be that targeting the fibrotic component in solid tumors is crucial for treatment of metastatic disease. Also, targeting fibroblasts at the early stage of cancer development is crucial for preventing metastasis. These important clinically related issues warrant further investigation.

Although in-depth molecular mechanisms underlying fibroblast-mediated cancer metastasis need to be elucidated, we can reasonably speculate several possibilities based on our findings. Of great interest, PDGF-BB–stimulated fibroblasts express high levels of IL-33 and its receptor ST2, which has been recently shown to promote cancer metastasis (13). Expression of both ligands and receptors in fibroblasts indicates the existence of an autocrine loop for fibroblast activation. Thus, the IL-33–ST2 pathway might be important for PDGF-BB–induced metastasis though activation of CAFs. Another interesting possibility is that PDGF-BB is able to promote myofibroblast formation, which is also crucial for induction of an invasive phenotype (15). The third possibility is that PDGF-BB induces chemokines particularly CCL5 that promotes cancer cell invasion and metastasis. It is known that CCL5 mediates breast cancer metastasis through a tumor stromal mechanism (37). Thus, sophisticated molecular mechanisms might be involved in fibroblast-mediated cancer metastasis. Perhaps, each fibroblast-stimulating factor, including members in the FGF, PDGF, TGF-β, and HGF families, uses different but yet overlapping mechanisms for promoting cancer metastasis. This interesting view warrants future studies.

Clinically, a frequently encountered situation is that the first sign of cancer disease is metastasis without overt primary tumors (38). In such a situation, the primary tumor may remain dormant and avascular and is unable to grow into a detectable mass. However, tumor cells can be disseminated to the circulation and subsequently spread to various tissues and organs. In distal tissues and organs, owing to environmental changes, metastatic tumor cells can be vascularized and grow to a clinically detectable size. Perhaps, infiltration of CAFs in these microscopic tumors plays a determinant role in cancer metastasis. Indeed, metastasis occurs at the early stage of pancreatic ductal adenocarcinoma (PDAC) development and PDAC is one of the most fibrotic cancer types of all cancers. Also, the amount of the fibrotic component is reversely correlated with survival. In our experimental settings, a mixture of cancer cells and CAFs were cocultured at a 3:1 ratio, which is clinically relevant in some most common cancer types such as lung cancer and PDAC. Taken together, our present work provides a novel mechanistic insight into fibroblast-mediated cancer metastasis. Targeting the fibrotic compartment would be an important approach for cancer therapy.

No potential conflicts of interest were disclosed.

Conception and design: T. Alkasalias, Y. Cao

Development of methodology: Y. Yang, T. Pavlova, Y. Cao

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Lim, T. Pavlova, T. Alkasalias, J. Hartman, Y. Cao, C. Liu, Y. Zhang

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Lim, Y. Yang, Y. Cao, C. Liu, Y. Zhang, K. Hosaka, X. Wang, Y. Lu, G. Nie

Writing, review, and/or revision of the manuscript: S. Lim, T. Alkasalias, J. Hartman, Y. Cao, C. Liu, Y. Zhang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Lim, L. Jensen, X. Xing, Y. Cao

Study supervision: L. Jensen, Y. Cao

Y. Cao's laboratory is supported through research grants from the European Research Council (ERC) advanced grant ANGIOFAT (Project no 250021), the Swedish Research Council (K2012-67X-21130-04-5), the Swedish Cancer Foundation (140659), the Karolinska Institute Foundation (C166658073), the Karolinska Institute distinguished professor award (C166646352), the Torsten Soderbergs Foundation (M144/12), the Maud and Birger Gustavsson Foundation (C166655073), the NOVO Nordisk Foundation (BtN/AIR 2014), and the Knut and Alice Wallenbergs Foundation (C166646483). Tatiana Pavlova and Twana Alkasalias were supported by grants from the Swedish Research Council (2015-02410), the Swedish Cancer Society (15 0591) and Cancer Research Institute, New York (July 2016-June 2017). Guohui Nie is supported by Fund for High Level University Construction of Medical Discipline (2016031638), China.

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.