Sarcoma cells secrete hypoxia-modified collagen VI to weaken the lung endothelial barrier and promote metastasis

Abstract

Intratumoral hypoxia correlates with metastasis and poor survival in sarcoma patients. Using an impedance sensing assay and a zebrafish intravital microinjection model, we demonstrated here that the hypoxia-inducible collagen-modifying enzyme lysyl hydroxylase PLOD2 and its substrate collagen type VI (COLVI) weaken the lung endothelial barrier and promote transendothelial migration. Mechanistically, hypoxia-induced PLOD2 in sarcoma cells modified COLVI, which was then secreted into the vasculature. Upon reaching the apical surface of lung endothelial cells, modified COLVI from tumor cells activated integrin β1 (ITGβ1). Furthermore, activated ITGβ1 co-localized with Kindlin2, initiating their interaction with F-actin and prompting its polymerization. Polymerized F-actin disrupted endothelial adherens junctions (AJ) and induced barrier dysfunction. Consistently, modified and secreted COLVI was required for the late stages of lung metastasis in vivo. Analysis of patient gene expression and survival data from The Cancer Genome Atlas (TCGA) revealed an association between the expression of both PLOD2 and COLVI and patient survival. Furthermore, high levels of COLVI were detected in surgically resected sarcoma metastases from patient lungs and in the blood of tumor-bearing mice. Together, this data identifies a mechanism of sarcoma lung metastasis, revealing opportunities for therapeutic intervention.

Introduction

Soft tissue sarcomas (STS) are a broad group of rare and heterogeneous tumors derived from mesenchymal tissue, accounting for approximately 1% of all cancers (1). An estimated 13,000 people were diagnosed with STS in the United States in 2022, with a mortality rate of ~40% (2). Metastasis is the leading cause of cancer deaths among STS patients, and the most frequent metastatic site for STS is the lung (3). While local sarcomas can be successfully treated with standard surgical resection and radiation, metastatic sarcomas are generally incurable. Three-year survival rates after complete resection of lung metastasis range only from 30% to 42% and chemotherapy is ineffective in extending the survival of patients with metastatic tumors (4, 5). The scarcity of molecular studies on sarcoma metastasis has proven an impediment to developing novel therapeutic strategies to curb disease progression. In addition, STS are heterogeneous in terms of clinical presentation and molecular alterations, resulting in diverse patient outcomes and a complex milieu of underlying disease mechanisms. STS encompasses over 70 histologically unique subtypes, with the most common diagnoses being undifferentiated pleomorphic sarcoma (UPS), liposarcoma, leiomyosarcoma, and synovial sarcoma (6, 7). Current data suggest that UPS, which commonly results in lethal pulmonary metastases, may not represent a distinct sarcoma subtype; instead, it may be a collection of phenotypes common to and manifested by multiple subtypes as they worsen to stages and grades wherein their tissue/cell type of origin is no longer discernible (8). Therefore, elucidating the molecular and cellular mechanisms controlling UPS metastasis may be critical to the development of effective therapeutic strategies to treat multiple types of sarcoma.

Consistent with their mesenchymal origins, STS exhibit high expression of mesenchymal genes and extensive release of extracellular matrix (ECM) components (9-11). The diverse roles of ECM components make them central mediators in the metastatic cascade. The desmoplastic reaction and mechanical pressures within the primary tumor propel migration through the interstitial ECM, foster dysfunctional neoangiogenesis, and vascular dissemination. Adhesion signaling facilitates intravasation and extravasation, and protects tumor cells from the harsh environment within the circulation. Finally, deposition and remodeling of the tumor ECM, by various collagen modifying enzymes, creates a microenvironment that promotes survival and growth of disseminated tumor cells (12).

Intratumoral hypoxia is associated with human sarcoma progression; the hypoxia master regulator HIF1A and its targets, many of which are ECM-modifying enzymes, are among the most reliable predictors of metastatic potential in sarcoma patients (13, 14). In a previous study, we showed that the HIF1A target and collagen-modifying enzyme, procollagen-lysine, 2-oxoglutarate 5-dioxygenase (PLOD2), promotes sarcoma metastasis by promoting the formation of a pro-tumor migration collagen network (10). This network acts as a supportive scaffold, facilitating tumor cell migration toward blood vessels and thereby promoting their ability to escape the primary lesion. However, the role of PLOD2 in vascular dissemination to the lung is still unknown. Further, a growing body of evidence shows that PLOD2 is highly expressed in a diverse array of tumor types wherein it promotes metastasis; however, the underlying mechanisms remain unclear (15-29). Therefore, delineating the molecular and cellular mechanisms of PLOD2-mediated metastasis may facilitate the development of novel therapeutics for a broad spectrum of cancers.

Here we investigate the impact of PLOD2 and PLOD2-modified collagens in sarcoma metastasis using human STS cells, metastasis-bearing lung specimens from human patients, and multiple UPS cell lines from genetically-engineered mouse models (GEMM) that were described previously (10). Tumors formed in these models, LSL-Kras^G12D/+; Ink4a/Arf^fl/fl (KIA) and Kras^G12D/+; Trp53^fl/fl (KP) mice, recapitulate human UPS morphologically, histologically, and transcriptionally; cells derived from these tumors can efficiently metastasize to the lung upon subcutaneous and tail vein injections (30). These cell lines, together with the cells derived from human soft tissue sarcoma patients, allow for the investigation of molecular mechanisms that control STS lung metastasis.

To study metastatic cells in vivo we use a microinjection model of zebrafish embryos that carry a fluorescent endothelial reporter. The zebrafish embryo is an invaluable animal model for studying the metastatic cascade, and has a number of unique advantages: the embryos develop externally following fertilization, making them easily accessible to manipulation and imaging; they are transparent and permit the generation of transgenic lines with cell-specific fluorophores. Utilizing the transparent feature of zebrafish embryos, we can track the real-time movement of tumor cells in a living organism, and closely interrogate the localization and patterns of migrating tumor cells at a single-cell resolution in a robust and quantitative manner, which is currently a major challenge in other models. Importantly, several studies have shown that vascular gene expression is conserved between zebrafish and humans (31-33). By one-day post fertilization (dpf), large blood vessels of the zebrafish such as the dorsal aorta, posterior cardinal vein, and common cardinal vein have developed; endothelial cell junctions have formed, and tubulogenesis has occurred to allow for blood flow (34). By 2 dpf, the zebrafish embryo has developed an intricate and complete vascular system (35). Importantly, during the first month of development, zebrafish do not have a mature adaptive immune system (36); therefore, they do not reject xenografts. Collectively, these features make the zebrafish embryo an appropriate and useful system for studying endothelial integrity and transendothelial migration of tumor cells.

Using the aforementioned tools and other in vitro and in vivo assays, we identify a microfibrillar collagen, collagen type VI (COLVI), as an essential player in promoting sarcoma lung metastasis. We propose that COLVI is a putative substrate of PLOD2 and demonstrate that PLOD2-modified COLVI weakens the endothelial barrier and promotes transendothelial migration of tumor cells. We further identify integrin β1 (ITGβ1) as the endothelial receptor engaging COLVI, and Kindlin2 as the downstream signaling molecule responsible for mediating barrier disruption in a FAK/Src independent context.

Together, our current data identifies the PLOD2-COLVI-ITGβ1-Kindlin2 axis as a novel mechanism of vascular dissemination; these findings will open up new opportunities for therapeutic intervention of sarcoma metastasis.

Methods and Materials

Zebrafish embryo microinjection

All procedures on zebrafish (Danio rerio) were approved by the Institutional Animal Care and Use Committee of the University of Pennsylvania. Fertilized zebrafish eggs of the transgenic strain expressing EGFP under the Fli promoter Tg(Fli:EGFP) or mCherry under the Flk promoter Tg(Flk:mCherry) were incubated at 28 °C in E3 solution and raised using standard methods. Embryos were transferred to E3 solution containing 0.2 mM1-phenyl-2-thio-urea (PTU, Sigma) six hours post-fertilization to prevent pigmentation. Proteases were added to PTU-E3 solution on Day 1 to dechorionate the fish embryos. At 48 h post-fertilization, zebrafish embryos were anesthetized with 0.03% tricaine (Sigma) and then transferred to an injection plate made with 1.5% agarose gel for microinjection. Approximately 200 KIA cells expressing mCherry or EGFP (~5 nL) suspended in PBS supplemented with 0.5 mM EDTA were injected into the perivitelline space or the circulation valley of each embryo using a XenoWorks Digital Microinjector (Sutter Instrument). Pre-pulled non-filamentous borosilicate micropipettes were used for the microinjection (Tip ID 30 μm, base OD 1 mm, length 5.5 cm, Fivephoton Biochemicals). After injection, the fish embryos were immediately transferred to PTU-E3 solution, and screened for successful injections. Embryos that show approximately 200 tumor cells at the correct locations were included for further monitoring and those that do not meet these criteria were excluded. Included embryos were kept at 33 °C, and examined every day to monitor tumor migration using a fluorescent microscope (Olympus Ix81). A spinning disk confocal microscope (Olympus Ix81 with an Andor iXon3 EMCCD camera) was used for further screening for extravasation and live imaging. Images were analyzed using software FIJI.

Mouse models

All procedures on mice were approved by the Institutional Animal Care and Use Committee of the University of Pennsylvania. For transplant tumors 1×10⁶ KIA cells were injected subcutaneously into the flanks of nu/nu mice (Charles River Laboratories). Tumor growth was monitored every other day, and animals were euthanized after two weeks. For tail vein injection, 75,000 KIA cells were injected into each nu/nu mouse; animals were euthanized five weeks post injection. For assessment of pulmonary metastasis, lungs were excised, fixed in 4% paraformaldehyde, and dehydrated in an ethanol gradient; the entire lungs were then scanned by microCT. In each injection experiment, at least 6 mice per experimental group were used.

Cell Culture, treatment and lentiviral transduction

KIA cells were derived from LSL-Kras^G12D/+/Ink4a/Arf^fl/fl tumors in 2010 as described previously (10). KP cells were derived from LSL-Kras^G12D/+/Trp53^fl/fl tumors by David Kirsch M.D., Ph.D. (Duke University) in 2010. KIA and KP cells were authenticated by genotyping. STS-109 cells were derived from a human UPS tumor by Rebecca Gladdy, M.D. (University of Toronto) in 2018, and authenticated by Small Tandem Repeat (STR) analyses in our lab using the Penn DNA sequencing facility. HT1080 and HEK-293T cell lines were purchased from ATCC (CCL-121 and CRL-3216 respectively). MPLMEC cells were purchased from Cell Biologics (C57-6011) and HMVEC-L cells were purchased from Lonza (CC-2527). Cells underwent mycoplasma testing using MycoAlert, and the most recent testing was conducted in May 2022. Between thawing and use, cells were cultured for no more than 10 passages. Sarcoma cells were cultured in DMEM with 10% FBS, 1% L-glutamine, and 1% penicillin/streptomycin. MPLMEC cells were cultured in Endothelial Cell Medium (Sciencell, 1001) and in culture vessels precoated with Gelatin-Based Coating Solution (Cell Biologics, 6950); MPLMEC cells were dissociated using Accutase Cell Dissociation Solution (Sigma Aldrich, A6964). HMVEC-L cells were cultured in Microvascular Endothelial Cell Growth Medium-2 (Lonza, CC-3202) and subcultured using ReagentPack Subculture Reagents (Lonza, CC-5034). For shRNA-mediated knockdown of targets, lentiviral particles bearing pLKO.1 shRNA plasmids were generated in HEK-293T cells. HEK-293T cells were transfected overnight with pLKO.1 empty vector, nonspecific shRNA, or target-specific shRNA and viral packaging plasmids, using the Fugene 6 reagent protocol (Promega, E2693). The following shRNA pLKO.1 plasmids were employed: pLKO.1-Scrambled (RRID:Addgene_1864), pLKO.1 Hif1α shRNA (TRCN0000054448, TRCN0000054449), pLKO.1 Plod2 (TRCN0000076408, TRCN0000076411). pLKO.1 Col6a1(TRCN0000091533, TRCN0000091534), COL6A1(TRCN0000116961), Col6a2 (TRCN0000091228, TRCN0000091230). For viral packaging, the third-generation lenti-vector system (pMDLg/pRRE (RRID:Addgene_12251), pRSV-Rev (RRID:Addgene_12253), and pMD2.G/VSVG (RRID:Addgene_12259)) was used. Supernatant was harvested from cultures at 36 hrs and 60 hrs post-transfection, and concentrated using polyethylene glycol-8000 (Millipore Sigma, 25322-68-3). Integrin β1 stimulating antibody and IgG control were purchased from R&D Systems (MAB17781(RRID:AB_2129940) and MAB002 (RRID:AB_357344) respectively). Recombinant human PLOD2 protein was a kind gift from Dr. Jonathan M. Kurie’s laboratory at MD Anderson Cancer Center. Low-oxygen conditions (0.5% O₂) were achieved in a Baker-Ruskinn in vivO₂ work station or an Eppendorf hypoxia incubator.

Human Samples

All studies involving human samples were conducted in accordance with the U.S. Common Rule and approved by the Institutional Review Board at the University of Pennsylvania. Metastases-bearing lung samples were obtained from surgically resected lungs from de-identified STS patients undergoing therapeutic surgical resection at the Hospital of the University of Pennsylvania, with written informed consent from the patients. No information on sex, age, or weight.

Tissue histology, immunohistochemistry and cellular immunofluorescence

Paraformaldehyde-fixed and paraffin-embedded tissues were sectioned to 5-micron thickness. Immunohistochemistry of tissue sections with antibodies to collagen VI (Abcam, Ab6588, RRID:AB_305585) was performed on Leica Bond autostainer with bond polymer refine detection kit (Fisher Scientific, NC0318637). Collagen was stained using Masson's Trichrome Kit (Sigma Aldrich, HT15-1KT). Elastin was stained using Elastic Stain Kit (Sigma Aldrich, HT25A-1KT). Stained slides were digitally scanned by the Pathology Core Laboratory at the Children’s Hospital of Philadelphia. For immunofluorescences, cells were fixed with 4% paraformaldehyde, blocked in 10% normal goat serum-0.3% triton-x-100 in PBS, and then stained with the following antibodies: anti-VE-cadherin (Santa Cruz, sc-9989, RRID:AB_2077957), anti-Plod2 (R&D Systems, MAB4445, RRID:AB_2165768), anti-ColVI (ProteinTech, 17023-1-AP, RRID:AB_2229737), anti-Kindlin2 (Proteintech, 11453-1-AP, RRID:AB_2262660), and anti-activated integrin β1 (Millipore Sigma, MAB2079Z, RRID:AB_2233964). F-actin was stained with Phalloidin-Rhodamine (Biotium, 00027). Images were captured using an Olympus IX81 spinning disk confocal microscope, Leica TCS SP8 confocal microscope, or Leica Stellaris 5 confocal microscope. Images were analyzed using software FIJI (RRID:SCR_002285).

Zebrafish embryo live imaging preparation

Fish embryos were anesthetized in E3 fish water containing 0.01% tricaine and transferred to 1.5% low melting point agarose (Promega, V2111) in E3 water, prewarmed to 42°C and maintained at that temperature. Drops of agarose, each containing a fish, were transferred to imaging dishes. Using a zebrafish metal probe, the fish were positioned flat on the dish bottom under the microscope. Once the agarose containing the fish solidified, a sufficient amount of tricaine-E3 solution was added to fully submerge the agarose drops, keeping them moist during imaging. The fish were imaged, and then carefully released from the agarose under the microscope using the metal probe.

Western Blotting and qRT-PCR

Whole cell lysates were prepared in RIPA buffer. Proteins were electrophoresed and separated by SDS-PAGE and transferred to nitrocellulose membranes and probed with the following antibodies: rabbit anti-Hif1α (Cayman Chemical, 10006421, RRID:AB_409037), rabbit anti-Gapdh (Cell Signaling Technology, 2118S, RRID:AB_561053), rabbit anti-Plod2 (ProteinTech, 21214-1-AP, RRID:AB_10733347), rabbit anti-Fak (Cell Signaling Technology, 13009, RRID:AB_2798086), rabbit anti-pFak (Tyr397) (Cell Signaling Technology, 8556, RRID:AB_10891442), rabbit anti-Src (Cell Signaling Technology, 2109T, RRID:AB_2106059), rabbit anti-pSrc (Tyr416) (Cell Signaling Technology, 6943T, RRID:AB_10013641), and rabbit anti-ColVI (ProteinTech, 17023-1-AP, RRID:AB_2229737). Band densities of Western blots were quantified with software FIJI and normalized to the control sample. Total RNA was isolated from cells using the RNAeasy minikit (Qiagen, 74104). mRNA was reverse transcribed using the High-Capacity RNA-to-cDNA kit (Applied Biosystems, 4387406). Transcript expression was determined by quantitative PCR of synthesized cDNA using the Applied Biosystems ViiA7 real-time PCR system. Target cDNA amplification was measured using TaqMan primer/probe sets (Applied Biosystems) for all targets and housekeeping control Hprt1.

Co-Immunoprecipitation (Co-IP)

Confluent KP cells were cultured under hypoxia for 18 hours, and then incubated in 1% paraformaldehyde containing culture media for 15 min to crosslink protein complexes. After three PBS washes, cells were lysed in Co-IP buffer (20 mM Tris HCl pH 8, 137 mM NaCl, 1% NP-40, and 2 mM EDTA) containing protease/phosphatase inhibitors (Sigma-Aldrich, PPC1010-5ML). 500 μg of cell lysates were incubated with 12.5 μg of Plod2 mAb (R&D, MAB4445, RRID:AB_2165768) at 4°C overnight. The lysate-antibody mixture was then incubated with 2.5 mg of Dynabeads M-280 Sheep Anti-Mouse IgG (Thermo Fisher Scientific,11202D) at 4°C overnight. Unbound supernatant was collected, and the beads with bound proteins were washed with PBS three times, then boiled in Laemmli buffer for 15 min. Total cell lysates, unbound post-IP fractions, and IP fractions were then examined by Western Blot.

ECIS assay

ECIS arrays (Applied Biophysics, 8W10E+ PC) were activated with 10 mM cysteine solution (Applied Biophysics, 001-cysteine) for 15 min at 37°C, coated with gelatin-based coating solution (Cell Biologics, 6950) for 10 min, and rinsed with PBS followed by endothelial growth medium. Endothelial cells were seeded at confluent density on the array in their growth medium. A 100% confluent monolayer with matured barrier function was achieved overnight, and then treatment was added. Cells were then monitored for 12-16 hours. All resistance/capacitance values post treatment were normalized to values prior to the treatment.

Conditioned media treatment in endothelial cells

Cell culture vessels, including dishes for downstream Western Blotting, chamber slides (Thermo Scientific, 154526) for downstream immunofluorescent staining, and cysteine pre-treated ECIS arrays for downstream barrier function assessment, were coated with a gelatin-based solution (Cell Biologics, 6950). Endothelial cells were then seeded at a confluent density into these pre-coated vessels and cultured overnight using their respective growth medium. For the generation of conditioned media, sarcoma cells were seeded at a confluent density in endothelial growth medium and cultured in either a regular tissue culture incubator (21% O₂) or a hypoxia workstation/incubator (0.5% O₂) for 16-18 hours. Following this incubation period, conditioned media was collected, centrifuged, and applied to endothelial cells that had formed a confluent monolayer.

Phospho Explorer Antibody Array

Mouse primary lung microvascular endothelial cells, treated with conditioned media from normoxic or hypoxic KP cells expressing either shScr or shCol6a1, were rinsed five times with ice-cold PBS. Subsequently, protein extraction, biotinylation, conjugation, and detection steps were performed following the product protocols for the Phospho Explorer Antibody Array (Full Moon Biosystems, PEX100) and Antibody Array Assay Kit (Full Moon Biosystems, KAS02). Post preparation, the arrays were briefly dried using compressed nitrogen and sent to Full Moon Biosystems for scanning and analysis.

Generation and staining of decellularized ECMs

4-well chamber slides (Thermo Scientific, 154526) or chambered coverglass (Thermo Scientific, 155382) were coated with gelatin-based coating solution for 1h at 37°C, and then treated with 2.5% glutaraldehyde at room temperature for 30 minutes. The slides were washed twice with PBS and 50,000 KP cells per well were seeded. Cells were transferred to a hypoxia station upon becoming confluent, and maintained for 5 days at 0.5% oxygen in growth media supplemented with 50 ug/mL L-ascorbic acid (Sigma a7506); media was replenished daily. For decellularization, cells were washed twice with wash buffer 1 (100mM Na₂HPO₄, 2mM MgCl₂, 2mM EGTA (pH 9.6)), and lyzed with lysis buffer (8mM Na2HPO4, 1% NP-40 (pH 9.6)). Cell debris were removed with wash buffer 2 (10mM Na2HPO4, 300mM KCl (pH 7.5)) and then ddH₂O four times. dECMs were then stored in PBS. For immunofluorescent staining, dECMs were fixed in 4% paraformaldehyde, blocked in 10% normal goat serum-0.3% triton-x-100 in PBS, and then stained with rabbit anti-collagen VI (ProteinTech, 17023-1-AP, RRID:AB_2229737).

Statistics

Data are presented as mean ± SD, unless otherwise indicated. Two-tailed t test or One-way ANOVA test was performed for most of the studies to evaluate the differences between the control and experimental groups. P ≤ 0.05 was considered statistically significant. Significance is indicated by the presence of asterisks “* (≤ 0.05), ** (≤ 0.01), and *** (≤ 0.001)”. GraphPad Prism (RRID:SCR_002798) was used to conduct all statistical analyses.

Data availability

All raw data generated in this study are available upon request from the corresponding author. The patient gene expression data (“TCGA-SARC”) analyzed in this study were obtained from the Cancer Genomics Hub (CGHub) at https://cghub.ucsc.edu/ in 2015, and the survival dataset (“TCGA, Cell 2017”) was downloaded from cBioPortal at http://www.cbioportal.org/study/summary?id=sarc_tcga_pub in 2019; only UPS cases included in the TCGA-SARC publication were used in our analysis.

Results

PLOD2 facilitates transendothelial migration

In a previous study, we found that hypoxia-inducible Plod2 promotes lung colonization following tail vein injection of sarcoma cells in mice (10). Therefore, we hypothesized that Plod2 may be required for transendothelial migration during metastasis. To interrogate this hypothesis, we employed multiple in vivo and in vitro models beginning with a novel zebrafish microinjection system using transgenic zebrafish expressing green or red fluorescent protein in their endothelial cells (Danio Reno, Tg(Fli:EGFP) or Tg(Flk:mCherry)). While zebrafish possess a gill rather than lungs, their vasculature shares similarities with mammalian endothelial cells in various organs, including the lung. As in mammals, zebrafish endothelial cells form cell-cell junctions, including tight junctions (TJ) and adherens junctions (AJ) (31-35), creating a barrier that needs to be disrupted to facilitate tumor cell migration and metastatic extravasation.

We first asked if intratumoral hypoxia promotes distant migration of tumor cells in the zebrafish system as seen in mouse models in our previous study (10). We injected fluorescently labeled KIA cells, pre-exposed to normoxic (21% O₂) or hypoxic (0.5% O₂) conditions for 18 hours, into the perivitelline space of 2 dpf-embryos (Fig. 1A). By 18 hours post-injection, we observed a significantly higher incidence of tumor cell migration from the injection site to distant locations (dashed circle) in the hypoxia group relative to the normoxia group (Fig. 1B, C). The migrating tumor cells predominantly lodged at the caudal vein plexus (CVP), a region of small blood vessels that are topographically disordered. This localization is consistent with other reports in the literature and aligns with the current dogma stating that extravasation occurs after cancer cells become physically trapped or arrested in small capillaries (37-39).

Fig. 1. Plod2 promotes tumor cell transendothelial migration.

(A) Schematic of zebrafish embryo microinjection at perivitelline space. (B) Representative images of live fish (Tg(Fli:EGFP)) displaying distant migration of UPS cells (KIA:mCherry) at 18-hour post injection. Scale bar: 100 μm. (C) Metastasis incidence (%) in zebrafish receiving KIA:mCherry cells (n ≥ 102 embryos in each group). (D) Schematic of zebrafish embryo intravital microinjection at circulation valley. (E) Representative images of live fish (Tg(Flk:mCherry)) bearing circulating UPS cells (KIA:CopGFP) post injections. Embryos showing UPS cells at any extravascular spots at this time point were excluded and euthanized. Scale bar: 200 μm. (F) Representative confocal images of live fish displaying extravasation of UPS cells at 18-hour post injection. Scale bar: 100 μm. (G) Extravasation incidence (%) in zebrafish receiving tumor cells (n ≥ 14 embryos in each group). Two-tailed unpaired t-test. (H) Immunoblots (left) and quantification of band densities (right) showing the expression of indicated proteins in KIA and KP cells subjected to normoxia, hypoxia, or hypoxia followed by normoxia. Normalized densities are presented as mean ± SD of two independent experiments. Two-tailed unpaired t-test.

To confirm the viability of sarcoma cells following extravasation in this model, we tracked these cells for an additional four days. The extravasated cells exhibited a highly aggregated and elongated morphology (Supplementary Fig. S1A), characteristic of malignant cell migration and invasion (40). By 120 hours post-injection, the sarcoma cells had colonized the tail region surrounding the CVP, resulting in significant bending of the fish tail and noticeable impairment in fish movement. As a result, we euthanized the fish at this point for humane reasons. During the four-day time course, sarcoma cells proliferated, as determined by the observed increase in the area positive for fluorescent cells (Supplementary Fig. S1B).

To specifically interrogate the potential role of Plod2 in the observed transendothelial migration, we introduced hypoxia-pretreated KIA or KP cells expressing a control shRNA (shScr) or shPlod2 directly into the fish vasculature by injecting cells into the circulation valley (Fig. 1D). Immediately post-injection, we observed circulation of the tumor cells within the blood vessels (Fig. 1E). 18 hours post-injection, we determined the transendothelial migration of tumor cells in live embryos using fluorescent microscopy (Fig. 1F). Loss of Plod2 reduced the incidence of sarcoma cell extravasation (Fig. 1G), Importantly, sarcoma cells are returned to normoxic conditions, in this assay, upon injection into the zebrafish. This approach recapitulates dissemination of tumor cells from a mammalian hypoxic primary site into the more oxygen-rich circulatory system. We observed that Plod2 expression remains upregulated in sarcoma cells for at least 18 hours following the return to normal oxygen levels, permitting collagen modification and extravasation at distant sites (Fig. 1H).

PLOD2 weakens the endothelial barrier integrity

We hypothesized that Plod2 may facilitate transendothelial migration by disrupting lung endothelial barrier integrity. To test this idea, we adapted the Electric Cell-substrate Impedance Sensing (ECIS) assay, which is a real-time, label-free, impedance-based method that can monitor the barrier function of a cell monolayer. The measured impedance is made up of two components: capacitance and resistance. At AC frequencies > 32,000 Hz, most of the current capacitively couples through the insulating cell membranes, and the capacitance reflects electrode coverage and therefore cell confluence levels. At frequencies ≤ 4,000 Hz, most of the current flows under and between adjacent cells, and the resistance responds to changes in the spaces either between or under the cells allowing for the quantification of barrier function and cell adherence, respectively. We seeded mouse primary lung microvascular endothelial cells (MPLMEC) on electrode-coated and gelatin solution-treated 8W10E+ arrays, and let the cells form a 100% confluent monolayer with mature barrier resistance (Supplementary Fig. S2A-B). We then treated the monolayer with conditioned media (CM) from KIA and KP cells that were exposed to normoxic or hypoxic conditions, and monitored the capacitance and resistance at both low and high AC frequencies. As shown in Fig. 2A-F and Supplementary Fig. S2C-D (additional shRNAs), endothelial cells treated with hypoxic CM showed reduced resistance relative to the normoxic condition (red data points). In contrast, endothelial cells treated with media conditioned by sarcoma cells lacking Hif1α (black data points) or Plod2 (blue data points) did not show a hypoxia-induced decrease in resistance (Fig. 2A-F, Supplementary Fig. S2C-D). The capacitance, in contrast, was unaltered between conditions (Supplementary Fig. S2E), indicating that CM from sarcoma cells controls barrier function or cell adherence but has no effect on the growth or death of endothelial cells resulting in unchanged cellular confluency. Consistently, we saw no difference in the total number of endothelial cells remaining after CM treatment (Supplementary Fig. S2F). Next, we assessed the role of sarcoma secreted factors on endothelial cell-substrate adhesion at the molecular level by examining the phosphorylation of focal adhesion kinase (Fak) and its effector Src in MPLMEC cells treated with CM from KP and KIA cells. Fak is a non-receptor tyrosine kinase localized to focal adhesions — large, dynamic protein complexes organized at the basal surface of cells. Focal adhesions physically connect the extracellular matrix to the cytoskeleton and regulate cell adhesion and cell migration (41-43). We observed no difference in phospho-Fak or phospho-Src between the normoxic and hypoxic conditions (Supplementary Fig. S2G-H). Therefore, we conclude that hypoxia-induced resistance decrease in the ECIS assays reflects a disruption in barrier integrity but not in cell adherence. Collectively, these data suggest that tumor cell-intrinsic Hif1α and Plod2 are important modulators of distant endothelial barrier integrity.

Fig. 2. Plod2 induced by hypoxia weakens endothelial barrier integrity.

(A) Representative time course curves of resistance in endothelial cells (MPLMEC) treated with conditioned media (CM) from KIA cells in indicated groups. Resistance values were normalized to values prior to treatment. (B) Quantitation of the normalized resistance at the 16hr time point in (A). Normalized resistance in all other groups were divided by that in the shScr-normoxia group to calculate the fold changes. Fold changes are presented as mean ± SD of three independent experiments. One-way ANOVA test with multiple comparisons. (C) Immunoblot (left) and quantification of band densities (right) showing the expression of indicated proteins in KIA cells as confirmation of target knockdown and hypoxia induction. Normalized densities are presented as mean ± SD of three independent experiments. One-way ANOVA test with multiple comparisons. (D) Representative time course curves of resistance in endothelial cells (MPLMEC) treated with conditioned media (CM) from KP cells in indicated groups. Resistance values were normalized to values prior to treatment. (E) Quantitation of the normalized resistance at the 16hr time point in (D). Normalized resistance in all other groups were divided by that in the shScr-normoxia group to calculate the fold changes. Fold changes are presented as mean ± SD of three independent experiments. One-way ANOVA test with multiple comparisons. (F) Immunoblot (left) and band density quantification (right) showing the expression of indicated proteins in KP cells as confirmation of target knockdown and hypoxia induction. Normalized densities are presented as mean ± SD of two independent experiments. One-way ANOVA test with multiple comparisons. (G) Representative time course curves of resistance in endothelial cells (MPLMEC) treated with conditioned media (CM) from KIA cells pre-treated with Minoxidil or DMSO. Resistance values were normalized to values prior to treatment. (H) Quantitation of the normalized resistance at the 16hr time point in (G). Normalized resistance in all other groups were divided by that in the DMSO-normoxia group to calculate the fold changes. Fold changes are presented as mean ± SD of three independent experiments. One-way ANOVA test with multiple comparisons. (I) Immunoblot (left) and band density quantification (right) showing the expression of indicated proteins in KIA cells as confirmation of Plod2 inhibition and hypoxia induction. Normalized densities are presented as mean ± SD of three independent experiments. One-way ANOVA test with multiple comparisons.

To confirm that Plod2 mediates downstream endothelial barrier weakening, we assessed KIA cells pretreated with a pharmacological modulator that reduces Plod2 expression, Minoxidil(44). CM from KIA cells pre-treated with Minoxidil (0.5 mM) for 48 hours significantly diminished hypoxia-induced barrier weakening (Fig. 2G-H), concomitant with reduced Plod2 protein levels as shown by western blot analysis of KIA cells (Fig. 2I). These data are consistent with our previous finding that Minoxidil significantly suppresses sarcoma pulmonary metastases (10), and demonstrates the potential of targeting Plod2 as a therapeutic strategy for preventing the transendothelial migration of sarcoma cells. As PLOD2 can also be secreted by tumor cells into the extracellular space (45), we asked if the addition of PLOD2 protein can act on secreted endothelial collagens directly and exert a barrier-weakening effect on endothelial cells. We treated endothelial cells with recombinant human PLOD2 proteins generated in and characterized by Dr. Jonathan M. Kurie’s lab at MD. Anderson. These recombinant proteins have been shown to be enzymatically active (wild-type; WT) or inactive (PLOD2 D689A)(45). However, we observed that lung endothelial cells are insensitive to PLOD2 protein alone (Supplementary Fig. S3A, B), suggesting that PLOD2 doesn't weaken barrier function by directly acting on endothelial cells or by modifying their associated collagens. Thus, we began to investigate the role of PLOD2-modified tumor-derived collagens in endothelial barrier dysfunction.

Collagen VI mediates endothelial barrier weakening

We have previously demonstrated that PLOD2-dependent collagen modification is important for sarcoma lung metastasis; a high level of collagen/vessel association indicates the ability of the collagen network to deliver primary tumor cells to the local vasculature for dissemination to distant organs (10). Here we ask which PLOD2-modified collagen/s is responsible for disrupting the endothelial barrier and if this collagen is important for the transendothelial migration of tumor cells. We first analyzed metastasis-containing-lung sections from human STS patients, and stained them with Elastic tissue fibers - Verhoeff's Elastin Van Gieson (EVG) stain. EVG labels elastin membranes black, collagens pink, and all other tissues brown. As shown in Fig. 3A, tumor-adjacent normal tissue displays an organized tissue architecture of the cross-section of blood vessels. Mason’s Trichrome, a pan collagen (blue) stain, also shows a similar pattern. In contrast, cross-sections of vessels containing intravascular lesions display aberrant collagen deposition, inside and outside of the vessel lumen, as reflected by the EVG and Mason’s Trichrome staining. Approximately 40% of patient lung samples displayed these collagen-rich intravascular lesions (7 out of 18). Given the buildup of collagens in pulmonary vessels, we hypothesized that intravascular collagens may be responsible for PLOD2-mediated endothelial barrier weakening. We first asked if collagen type I, (ColI) a well-established PLOD2 substrate, mediates this phenotype; however, we observed that ColI is not relevant to PLOD2-mediated barrier disruption (Supplementary Fig. S4A-D). PLOD2 and the microfibrillar collagen type VI (COLVI) have been reported to be present in the tumor cell secretome and collectively play a pro-metastatic role (21). Importantly, COLVI genes are highly expressed in multiple types of human sarcoma (bioRxiv 2022.03.31.486627), therefore, we assessed whether COLVI is among the collagens present in the intravascular lesions. COLVI is a microfibrillar collagen encoded by multiple genes, with the three most predominant members being COL6A1, COL6A2, and COL6A3 (46). IHC of COLVI on human lung sections containing metastatic sarcoma and mouse lung sections containing metastasis from tail vein-injected KIA cells, revealed abundant COLVI deposition inside tumor-filled blood vessels (Fig. 3B, Supplementary Fig. S5A, B).

Fig. 3. Collagen VI mediates PLOD2-dependent endothelial barrier disruption.

(A) Representative images of staining with the indicated dyes on lung tissues bearing metastatic intravascular lesions and adjacent normal tissues from human STS patients (n = 18). Intravascular lesions that contain collagen deposition were detected in approximately 40% of patient lung samples. Out: area outside of the vessels; in: area inside the vessels. Scale bar: 500 μm. (B) Representative images of immunohistochemistry of COLVI and CD31 in cross sections of intravascular metastatic lesions in lung sections from human patients; Scale bar: 300 μm. (C) Col6a1 protein levels in plasma of non-tumor bearing mice (n=4) or tumor-bearing mice (n=7) detected by ELISA; tumors were introduced by subcutaneous injection of KIA cells. Two-tailed unpaired t-test. (D) Representative dot blots with indicated antibodies or ponceau staining on CM from KP cells expressing shScr or shCol6a1. CM dots were separated by a pencil line drawn on the membrane before sample loading. (E) Representative time course curves of resistance in mouse endothelial cells (MPLMEC) treated with CM from KP cells in indicated groups. Resistance values were normalized to values prior to treatment. (F) Quantitation of the normalized resistance at the 16hr time point in (E). Normalized resistance in all other groups were divided by that in the shScr-normoxia group to calculate the fold changes. Fold changes are presented as mean ± SD of two independent experiments. One-way ANOVA test with multiple comparisons. (G) Representative time course curves of resistance in mouse endothelial cells (MPLMEC) treated with CM from KIA cells in indicated groups. Resistance values were normalized to values prior to treatment. (H) Quantitation of the normalized resistance at the 16hr time point in (G). Normalized resistance in all other groups were divided by that in the shScr-normoxia group to calculate the fold changes. Fold changes are presented as mean ± SD of three independent experiments. One-way ANOVA test with multiple comparisons. (I, J) Disease-free survival of UPS patients in TCGA-SARC dataset stratified by tumor expression levels of PLOD2 and COL6A1, respectively. Each tertile (low, medium, high) represents one-third of patients. Log-rank (Mantel-Cox) test.

Consistently, we observed that tumor-bearing mice possessed higher levels of circulating ColVI than non-tumor-bearing animals (Fig. 3C). Furthermore, immuno dot blot of CM from sarcoma cells demonstrated substantial secreted ColVI (Fig. 3D). Next, we determined if ColVI is required for hypoxia-induced endothelial barrier disruption. CM collected from cultures of KP and KIA cells expressing Col6a1-specific shRNAs did not cause the hypoxia-induced barrier integrity loss, relative to control, in MPLMEC, phenocopying what we observed in CM from shHif1α and shPlod2 conditions; CM from Col6a2-lacking KP cells also showed a similar phenotype (Fig. 3E-H, Supplementary Fig. S6A, B). To interrogate this phenotype in human sarcoma, we used a cell line derived from a human STS patient (STS109) as the CM donor and human lung microvascular endothelial cells (HMVEC-L) as the CM recipient. Very similar to the murine system, we observed a COLVI-dependent barrier-weakening phenotype (Supplementary Fig. S6C-E). Interestingly, loss of COL6A1 in STS109 cells increased endothelial resistance relative to control under both oxygen conditions (Supplementary Fig. S6C, D), suggesting that individual cell lines have unique barrier disruption profiles and varying metastasis potential. Consistent with this conclusion, hypoxic CM from HT1080 human fibrosarcoma cells weakened the endothelial barrier, as seen in other cell lines (Supplementary Fig. S6F, G). However, COL6A1 depletion reduced HT-1080 cell adherence, interfering with monolayer formation and invalidating the ECIS and CM approach for this specific line.

Based upon these data, we hypothesized that PLOD2 modifies COLVI, which is then secreted into the vascular circulation where it disrupts endothelial barrier integrity. Importantly, the gene expression levels of both PLOD2 and COLVI show a trending inverse association with disease-free survival in a cohort of 44 human UPS patients (Fig. 3I, J, Supplementary Fig. S6H), indicating the potential therapeutic utility of targeting these two molecules or their interaction in UPS.

Collagen VI is a putative substrate of PLOD2

To determine the structural effect of Plod2 on ColVI, we first generated cell-derived matrixes (CDM) from KP cells expressing shScr or shPlod2 and performed immunofluorescent staining for ColVI. In control KP-derived CDM, ColVI presents as a lattice decorated with longer and straighter bundle-like structures (Fig. 4A, arrows), whereas the length and structure of ColVI bundles are significantly altered in CDM from Plod2- deficient KP cells (Fig. 4A-C). These data suggest that ColVI modification by Plod2 is critical for the higher-order assemblies of this collagen upon secretion.

Fig.4. Collagen VI is a putative substrate of Plod2.

(A) Representative confocal images of immunofluorescent staining of ColVI on decellularized matrix deposited by KP cells expressing shScr or shPlod2.) Scale bar: 100 μm. (B, C) Quantification of ColVI length (B) and degree of straightness in ColVI longer than 16 micron (C) for images represented by (A). Software CTFire was used for quantification; six fields including over 8000 and 3000 ColVI filaments were assessed for each condition for length and straightness, respectively. (D) Representative confocal images of immunofluorescent staining of Plod2 and ColVI in KP cells expressing shScr or shPlod2 and cultured under normoxia or hypoxia. Scale bar: 100 μm. (E) ColVI-Plod2 colocalized pixel maps and scatter plots for representative images of normoxic or hypoxic KP cells expressing shScr shown in (D); colocalized pixels are shown in grey scale, whereas un-colocalized pixels are shown in blue and red. (F) Quantification of Pearson’s Colocalization Value for the colocalization between ColVI and Plod2 in normoxic or hypoxic KP cells expressing shScr. Two-tailed unpaired t-test. (G) Immunoblot representing four independent experiments showing the co-immunoprecipitation of ColVI in KP cells using a monoclonal Plod2 antibody as the bait. Ratio of sample loading amount for TCL: IP: Post-IP was 1:10:1. TCL: total cell lysate; IP: immunoprecipitation; Post-IP: lysate post-immunoprecipitation.

Next, to further characterize the interaction between Plod2 and ColVI, we performed immunofluorescent co-staining of Plod2 and intracellular ColVI to assess their localization in KP cells. Under hypoxic conditions, Plod2 is significantly upregulated; intriguingly, ColVI appears aggregated in bands in response to hypoxia treatment, which occurs independent of Plod2 expression (Fig. 4D). Importantly, Plod2 and ColVI are colocalized, which is consistent with their roles as a putative enzyme and substrate pair (Fig. 4E-F). To substantiate the physical interaction between Plod2 and ColVI, we conducted endogenous co-immunoprecipitation (IP) employing a monoclonal Plod2 antibody as the bait. Through a partial IP of Plod2, we identified ColVI as a PLOD2 binding protein, thus confirming the direct physical association between these two molecules (Fig. 4G).

Consistent with the role of Plod2 as a collagen-modifying enzyme, we determined that neither Plod2 (nor Hif1α) controls the production of ColVI at the mRNA level (Supplementary Fig. S7A), indicating that Plod2/Hif1α-mediated collagen modification specifically regulate the structure/organization of ColVI, rather than its expression. Consistent with the hypoxia-induced reduction in mRNA, we also observed a decrease in ColVI protein levels, both intracellularly and in the conditioned media (Supplementary Fig. S7B, C). Despite this decrease, ColVI still induced barrier disruption and promoted metastasis under hypoxic conditions, suggesting that these effects are governed by ColVI’s structural modification by Plod2, rather than its quantity.

Collagen VI weakens endothelial adherens junctions through ITGβ1

COLVI functions both as a structural protein and a signaling molecule. It is able to bind to different components of the ECM, linking cells to the surrounding connective tissue and organizing the three-dimensional tissue architecture of skeletal muscles, tendons, bone, and cartilage (46). It also interacts with several membrane receptors and triggers intracellular signaling pathways (46). To identify the potential endothelial surface receptor that interacts with tumor-secreted ColVI, we carried out an ELISA-based phospho-protein screen on mouse endothelial cells (MPLMEC) treated with CM from KP cells cultured under normoxia or hypoxia and expressing shScr or shCol6a1 (Fig. 5A, Supplementary Table S1). The antibody arrays in the screen contain 1318 site-specific and phospho-specific antibodies covering over 30 signaling pathways. We narrowed down the top hits from the screen by comparing the fold changes between oxygen conditions (Hx vs. Nx) and between the two genotypes (shScr and shCol6a1) and identified ITGβ1 as a promising candidate molecule potentially mediating tumoral ColVI-induced signaling. Integrins, existing as heterodimers of α- and β-subunits, are a family of cell adhesion receptors essential for cell-matrix and cell-cell communications. Integrins are present in open and closed conformations that have different affinities for their extracellular ligands. In the closed conformation, integrins are inactive, and have low affinity for ligands; whereas in the open conformation, integrins are active, engage ligands with increased affinity, and can propagate intracellular signaling (47, 48). ITGβ1 has an established role in vessel integrity: during embryonic development, it regulates vascular sprouting and lumenization (49, 50); during postnatal life, it stabilizes the developing adherens junctions of newly formed vessels (51); in mature vessels, ITGβ1signaling promotes vascular leakage (52). Interestingly, ITGβ1 expression is inversely associated with survival in a cohort of 44 human UPS patients, suggesting that targeting this molecule may also be therapeutically useful (Fig. 5B).

Fig. 5. Endothelial integrin β1 is responsible for ColVI-mediated signaling and barrier weakening.

(A) Schematic of the general design of the Phospho Explorer Antibody Array screening on CM-treated mouse endothelial cells. (B) Kaplan-Meier disease-free survival curves of UPS patients in TCGA-SARC dataset (n = 44), stratified by gene expression levels of ITGβ1; each tertile (low, medium, high), represents 1/3 of patients. Log-rank test. (C) Representative confocal images of immunostaining of ITGβ1 (activated conformation) and f-actin in in human endothelial cells (HMVEC-L) treated with normoxic CM containing IgG or an ITGβ1-stimulating antibody (0.2 μg/1x10⁵ cells) (left panel) and colocalized pixel maps generated using software FIJI (right panel); colocalized pixels are shown in grey scale, whereas un-colocalized pixels are shown in blue and red. Scale bar: 50μm. (D) Integrated density of activated ITGβ1 in images represented in (C). Two-tailed unpaired t-test. (E) Pearson’s Colocalization Values (PCV) for the colocalization between activated ITGβ1 and f-actin in images represented in (C). Two-tailed unpaired t-test. (F) Representative time course curves of resistance in human endothelial cells (HMVEC-L) treated with normoxic CM containing IgG or an ITGβ1-stimulating antibody. Resistance values were normalized to values prior to treatment. (G) Quantitation of the normalized resistance at the 16hr time point in (F). Normalized resistance in the ITGβ1-stimulating antibody-treated group was divided by that in the IgG group to calculate the fold changes. Fold changes are presented as mean ± SD of three independent experiments. One-way ANOVA test with multiple comparisons. (H) Representative confocal images of immunostaining of ITGβ1 (activated conformation) and f-actin in human endothelial cells (HMVEC-L) treated with media conditioned by hypoxic KP cells expressing shScr or shCol6a1 (left panel) and colocalized pixel maps generated using software FIJI (right panel). Scale bar: 50μm. (I) Integrated density of activated ITGβ1 in images represented in (H). Two-tailed unpaired t-test. (J) Pearson’s Colocalization Values (PCV) for the colocalization between activated ITGβ1 and f-actin in images represented in (H). Two-tailed unpaired t-test.

To assess the role of ITGβ1 in endothelial barrier function, we first treated human lung microvascular endothelial cells (HMVEC-L) with an ITGβ1-stimulating antibody. We observed contraction of F-actin that partially co-localized with activated ITGβ1 (under normoxic conditions) (Fig. 5C-E), which is consistent with the established link between integrin activation and the downstream actin organization (53). Next, we performed ECIS assays using this stimulating antibody and observed a similar barrier-weakening effect as seen with hypoxic CM (Fig. 5F, G). Furthermore, we determined the activation status of ITGβ1 and its potential co-localization with F-actin in endothelial cells treated with hypoxic CM from KP cells expressing shScr or shCol6a1. Endothelial cells treated with hypoxic CM from shScr expressing KP cells phenocopied ITGβ1-stimulating antibody-treated cells, and this phenotype was dependent on the presence of tumor-derived ColVI and Plod2 (Fig. 5H-J, Supplementary Fig. S8).

Plod2 and collagen VI are important for hypoxia-induced adherens junction (AJ) disruption

Endothelial cell adhesion depends not only on adhesion molecules, such as VE-cadherin, but also on the actin filaments that support intercellular adhesions. To characterize the role of activated ITGβ1 in endothelial adhesion, we treated an endothelial monolayer with normoxic CM and the ITGβ1stimulating antibody and performed immunofluorescent staining for VE-cadherin and F-actin. We observed a higher degree of VE-cadherin disruption and internalization in endothelial cells treated with ITGβ1-stimulating antibody (Fig. 6A, B). These findings are consistent with the hypothesis that COLVI activates ITGβ1, promoting barrier disruption, transendothelial migration, and metastasis. Consistently, under hypoxic CM treatment, endothelial cells exhibited more VE-cadherin disruption and F-actin polymerization relative to endothelial cells treated with hypoxic CM from COLVI and PLOD2-deficient sarcoma cells (Fig. 6C-E, Supplementary Fig. S9). Notably, we observed that, Kindlin-2, an ITGβ1 downstream effector and a F-actin adaptor protein, exhibited more colocalization with VE-cadherin in endothelial cells treated with CM from hypoxic shPlod2/shCol6a1- expressing cells compared to the shScr- expressing cells (Fig. 6C, F). This result is consistent with previous reports showing that Kindlin2 is a component of the AJ, where its direct interaction with VE-cadherin maintains AJ integrity (4). In hypoxic control CM-treated endothelial cells, relocalization of Kindlin2 out of AJ regions occurs concomitant with its colocalization with F-actin and ITGβ1 (Fig. 6C, G, H). These data suggest that ColVI activates ITGβ1 on the apical surface of endothelial cells, triggering the assembly of a complex involving ITGβ1, Kindlin2, and F-actin. This interaction may lead to the dissociation of Kindlin2 from VE-cadherin, resulting in the disruption of VE-cadherin and subsequent weakening of adherens junctions independent of focal adhesions. This observation, supports earlier reports of a Fak/Src-independent AJ disruption mechanism (54).

Fig. 6. ColVI and Plod2 are important for hypoxia-induced adherens junction disruption.

(A) Representative confocal images of VE-cadherin and F-actin immunostaining in human endothelial cells (HMVEC-L) treated with normoxic CM containing IgG or an ITGβ1-stimulating antibody. Scale bar: 50μm. (B) Quantification of intercellular gap area represented in (A) as an indicator of VE-cadherin disruption. (C) Representative confocal images of immunostaining of VE-cadherin, Kindlin2 and F-actin in in human endothelial cells (HMVEC-L) treated with hypoxic CM from KP cells expressing shScr, shPlod2, or shCol6a1. Scale bar: 25μm. (D-F) Quantification of intercellular gap area (D), normalized integrated density of individual F-actin objects (E), and integrated density of Kindlin2 within the VE-Cadherin positive region (F) in images represented in (C). One-way ANOVA with multiple comparisons. (G, H) Representative confocal images of immunostaining of Kindlin2 and F-actin (G) and Kindlin and activated ITGβ1 (H) in human endothelial cells (HMVEC-L) treated with CM from hypoxic KP cells. CPM: colocalized pixel map. Scale bar: 25μm.

Tumor-secreted collagen VI promotes metastasis in vivo

We first determined that ColVI has no effect on the proliferation of KIA cells in vitro (Fig. 7A). In vivo, ColVI has minimal effect on the growth of primary tumors developed from subcutaneous injection of KIA cells (Fig. 7B). To determine if ColVI promotes the late phases of metastasis, such as vascular dissemination, we introduced control and Col6a1-deficient KIA cells to nude mice through tail vein injection. Five weeks post-injection, mice that received Col6a1-depleted KIA cells displayed a striking reduction in lung metastases relative to control as assessed by microCT imaging of whole lungs dissected from the animals (Fig. 7C); this observation was also confirmed by H&E staining of the lung sections (Fig. 7D), indicating that ColVI is essential for pulmonary metastasis and that loss of ColVI phenocopies Hif1α or Plod2 depletion (10). Next, we performed an intravital microinjection of control and Col6a1-deficient KIA/CopGFP or KP/CopGFP cells in zebrafish (Tg(Flk:mCherry), wherein endothelial expression of ITGβ1 and VE-cadherin has been reported (35, 55, 56). The Col6a1-deficient group had a significant reduction in the extravasation incidence (Fig. 7E-G), confirming the importance of ColVI in tumor cell transendothelial migration.

Fig. 7. Collagen VI promotes sarcoma lung metastasis in vivo.

(A) Growth curves of KIA cells expressing shScr or shCol6a1 in vitro. (B) Volumes of tumors developed from subcutaneously injected KIA cells expressing shScr or shCol6a1. 1x10⁶ KIA cells were injected to each flank of a nude mice. In each group, at least 6 mice were used. (C) Representative mouse whole lung microCT scan showing metastasis developed from tail vein injection of KIA cells expressing shScr (left), and % of lungs showing metastasis in the shScr and shCol6a1 groups (right); at least 7 mice were used in each group. (D) Representative H&E staining and ColVI immunohistochemistry on lung sections in the indicated groups. Sections were prepared from the lung samples used in (C) post-microCT scan. Scale bar: 2 mm. (E) Representative confocal images of live zebrafish embryos (Tg(Flk:mCherry) displaying extravasation of tumor cells infected with shCol6a1 at 18-hour post intravital injection. Scale bar: 50 μm. (F, G) Extravasation incidence (%) in fish receiving KIA:CopGFP (F) or KP:CopGFP cells (G) expressing shScr or shCol6a1. At least 19 embryos in total were used in each group. Two-tailed unpaired t-test.

Discussion

Consistent with their mesenchymal origins, sarcoma cells produce and deposit large quantities of ECM into the tumor microenvironment (9). ECM proteins regulate many hallmarks of cancer and are major drivers of metastasis (12). However, the molecular interactions of ECM components with their receptors, resulting downstream signaling events, and corresponding phenotypic effects on STS progression are poorly understood. Moreover, most studies on collagens and cancer assess the role of collagens as a whole, whereas the diverse contributions of individual collagens in tumors have been frequently overlooked. Therefore, there is a clear need for strategies to identify and study individual collagens and the signaling pathways they activate. Previously, we demonstrated that PLOD2 gene expression levels are associated with metastasis in UPS patients; the HIF1A/PLOD2 pathway regulates sarcoma cell migration in a collagen-dependent manner (10). In this study, we discovered that COLVI specifically is a putative substrate of PLOD2, and we observed that both molecules regulate lung endothelial barrier integrity and the subsequent transendothelial migration of tumor cells. Deletion of ColVI in tumor cells reduced extravasation in zebrafish and lung colonization in mice. These new findings identify COLVI as an essential driver of hypoxia-mediated metastasis in sarcomas.

COLVI is expressed in a wide range of tissues and plays important roles in cartilage, bone, and skeletal muscle function, nervous system regeneration and myelination, and immune cell recruitment and polarization (46). It also regulates the adhesion and migration of a variety of mesenchymal cells (57, 58). Several studies showed that COLVI promotes tumor metastasis in ovarian, breast, and pancreatic cancer; however, the underlying molecular mechanisms remain elusive (21, 59-62). Here, we demonstrate that COLVI weakens lung endothelial barrier integrity by activating endothelial ITGβ1. Dysregulation of integrin functions has been extensively reported in cancer and is associated with disease progression (63-67). The discovery of ITGβ1 as a critical regulator of transendothelial migration will provide additional rationale for the ongoing efforts of targeting this molecule as a cancer therapeutic strategy (NCT04608812)(68).

The functional link between integrins and adherens junctions (AJs) is well established. Integrins, when activated at cell-ECM focal adhesion complexes at the basolateral surface of endothelial cells, by various stimuli, can change cytoskeleton dynamics and intracellular tension. These alterations can, in turn, affect the localization, stability, and function of VE-cadherin (69-72). However, recent reports have demonstrated the presence and functional roles of apical integrins, though the underlying signaling mechanisms remain elusive (73, 74). In this study, we discovered a novel integrin-AJ pathway initiated at the apical surface of endothelial cells. In this non-canonical paradigm, tumor derived- and hypoxia modified-COLVI activates ITGβ1 at the apical surface. Activated ITGβ1 then causes relocalization of Kindlin2 away from AJ, triggering F-actin polymerization and contraction. This signaling cascade finally results in VE-cadherin internalization and AJ disruption (Fig. 8).

Fig. 8. Graphical model of PLOD2- and COLVI-mediated endothelial barrier weakening and lung metastasis.

Briefly, in a sarcoma, hypoxia-induced PLOD2 modifies COLVI, which is then secreted into the vasculature. The mechanism and kinetics of the COLVI vascular circulation remains to be studied. Upon reaching the apical surface of the lung endothelium, COLVI activates apical integrin β1 on the endothelial cells. Activated ITGβ1 then recruits Kindlin2 from the adherens junction, triggering F-actin polymerization and contraction. This signaling cascade finally results in VE-cadherin internalization and adherens junction disruption. Weakened adherens junctions then allow tumor cells to transmigrate, leading to lung metastasis.

This signaling pathway bypasses the focal adhesions, likely due to the spatial and functional nature of apical ITGβ1 and its interaction partners F-actin and Kindlin2. Apical integrins may be interacting with unique pools of actin, such as the membrane-proximal pool of F-actin (MPA) (73, 75), rather than focal adhesions-associated actin. Similarly, Kindlin-2 resides not only in focal adhesions, its canonical location, but also in AJ (76), as evidenced in our study. Therefore, the AJ-localized Kindlin2 and the potentially specific pool of F-actin may serve a functionally distinct role, upon engaging with the apical ITGβ1. Interacting together, the ColVI- ITGβ1- Kindlin2- F-actin signaling axis may directly regulate the stability of AJ without affecting focal adhesions.

In the future, several questions remain to be studied. It’s important to further characterize the kinetics of Plod2-mediated ColVI modification. Additionally, investigating whether the secreted ColVI could circulate systematically without associating with the tumor cells or whether it remains associated with the tumor cells post-secretion is warranted. Moreover, future studies should also address whether ColVI can create a long-term weakening effect on the pulmonary endothelium in a 'niche-creating' fashion, or temporarily open endothelial junctions to facilitate the transmigration of the tumor cells to which it adheres.

Together, our current findings suggest a new mechanism in tumor-induced endothelial barrier weakening and support the translation of novel collagen-/collagen enzyme-based tools to target STS metastasis. We have used the PLOD2 modulator, Minoxidil, to address the importance of PLOD2; however, since Minoxidil lacks specificity to this protein (44), the discovery of specific and potent PLOD2 inhibitors is clearly needed. Based on our findings, it seems prudent to pursue clinical agents that would inhibit metastasis by compromising COLVI modification or signaling in conjunction with the current standard of therapy. Lastly, as hypoxia and PLOD2 expression are important prognostic indicators in many solid tumors, conclusions drawn from the current studies may be applicable in multiple tumor contexts.

Significance.

Collagen type VI modified by hypoxia-induced PLOD2 is secreted by sarcoma cells and binds to integrin β1 on endothelial cells to induce barrier dysfunction, which promotes sarcoma vascular dissemination and metastasis.