Chemotherapy treatment induces pro-invasive changes in liver ECM composition

Abstract

Metastasis accounts for 90% of cancer-related deaths, yet the mechanisms by which cancer cells colonize secondary organs remain poorly understood. For breast cancer patients, metastasis to the liver is associated with poor prognosis and a median survival of 6 months. Standard of care is chemotherapy, but recurrence occurs in 30% of patients. Systemic chemotherapy has been shown to induce hepatotoxicity and fibrosis, but how chemotherapy impacts the composition of the liver extracellular matrix (ECM) remains unknown. Individual ECM proteins drive tumor cell proliferation and invasion, features that are essential for metastatic outgrowth in the liver. First, we find that the ECM of livers isolated from chemotherapy-treated MMTV-PyMT mice increases the invasion, but not proliferation, of metastatic breast cancer cells. Proteomic analysis of the liver ECM identified Collagen V to be more abundant in paclitaxel-treated livers. We show that Collagen V increases cancer cell invasion via α1β1 integrins and MAPK signaling, while also increasing the alignment of Collagen I, which has been associated with increased invasion. Treatment with obtustatin, an inhibitor specific to α1β1 integrins, inhibits tumor cell invasion in decellularized ECM from paclitaxel-treated livers. Overall, we show chemotherapy treatment alters the liver microenvironment, priming it as a pro-metastatic niche for cancer metastasis.

Introduction

Metastasis, the dissemination of cancer cells from the primary tumor, accounts for 90% of cancer related deaths [1]. Once they have reached the vasculature or lymphatics, cancer cells will exit the circulation and establish tumors in secondary organs. These metastatic lesions are more difficult to remove by surgery, visualize using traditional methods due to their small sizes, and treat because of their local tissue microenvironments. Understanding the mechanisms that promote metastatic colonization in cancer patients is essential to improve patient outcomes. Triple-negative breast cancer (TNBC) is the most aggressive subtype of breast cancer and represents up to 20% of all breast cancers worldwide. Neoadjuvant taxane chemotherapy, alone or in combination with doxorubicin and cyclophosphamide, followed by surgical resection is the standard of care [2]. However, up to 30% initially diagnosed with localized disease will suffer recurrence within 5 years after chemotherapy treatment, suggesting that tumor cells disseminate early from a histologically non-invasive primary tumor and subsequently grow in secondary organs. TNBC tumors metastasize primarily to the lungs, brain, bones and liver. Irrespective of the stage of the primary tumor at diagnosis, patients with breast cancer liver metastasis (BCLM) have worse overall prognosis with a median survival of less than 6 months [3]. When surgical resection in the liver is possible, the median survival increases to 16 months, but the projected 5-year survival is only 8.5% [4,5]. Detailed intravital imaging studies have shown that metastatic outgrowth in the liver requires both tumor cell proliferation and invasion and suppression of cell invasion reduces metastatic burden in the liver [6]. Overall, this highlights a critical need to understand the features of the local liver microenvironment that promote metastasis formation via their effects on tumor cell proliferation and invasion to ultimately identify new therapies to treat metastatic disease.

A major acellular component of both normal and pathological tissue microenvironments is the extracellular matrix (ECM) which is a complex network of proteins that provides structural support to tissues and cells. The ECM also serves as an active substrate for cancer cells to adhere to and provides biophysical and biochemical cues that can promote formation of metastases by initiating signals for invasion, survival, and proliferation [7-11]. In healthy livers, the ECM makes up a very limited amount of the total liver tissue area with collagens I, III, IV, and V being the most abundant [12]. Studies have shown that liver fibrosis, characterized by increased deposition of collagens, promotes growth of cancer cells, and is associated with increased number of liver metastases [13,14]. Cytotoxic chemotherapies, which are administered systemically, have long been known to induce liver damage and toxicity, affecting tissue function as well as structure [15]. Chemotherapy induces injury to the liver manifested as vessel damage and necrosis which can alter the composition of stromal and immune cells in the liver and cause secretion and remodeling of the ECM [16-18]. However, the effects of chemotherapy on the liver ECM composition and how these changes affect TNBC metastatic outgrowth have not yet been well characterized.

Collagen V is an ECM protein that has been implicated in liver damage and fibrosis. Studies show that collagen V abundance increases concomitantly with collagen I by immunostaining in the early stages of hepatic fibrosis and continues to do so until end stage cirrhosis [12,19]. Upregulation of collagen V in the liver has also been associated with increased liver metastasis for multiple cancer types such as pancreatic, colon, and breast cancer in matrisome characterization studies [13,20,21]. Structurally, collagen V copolymerizes with collagen I and is a key determinant in regulating collagen I fiber assembly, geometry, and strength [22-24]. Deletion of collagen V severely impacts collagen I organization in a range of tissues [25]. Collagen V is reported to signal directly through α1β1 integrin to drive cell migration [26], but most of its effects on cellular phenotypes are thought to be mediated by its ability to induce collagen I fibrillogenesis and fiber organization. It is also well established that collagen I fiber organization can drive breast cancer cell invasion and metastasis by activating β1 integrin and down-stream FAK signaling [27-29]. However, the role of collagen V in driving metastatic outgrowth remains understudied. In this study, we leverage our recently published decellularization method to investigate the effects of ECM of metastatic sites on tumor cell proliferation and invasion. We use this method to show that the ECM of livers isolated from mice treated with standard of care chemotherapy drives tumor cell invasion, an important step in the process of liver metastatic colonization. We use mass spectrometry to characterize the liver ECM of chemotherapy-treated tumor-bearing mice and identify collagen V as more abundant in the livers of mice treated with paclitaxel. Further, we show that collagen V increases TNBC cancer cell invasion via α1β1-mediated FAK and ERK1/2 signaling. Finally, treatment with obtustatin, an inhibitor specific to α1β1 integrins, decreases collagen V-driven invasion and suppresses paclitaxel-associated ECM-driven invasion of TNBC on native liver ECM.

Results

Decellularization can be used to isolate the native ECM of liver and lung

We previously developed a novel experimental pipeline that successfully isolates the native ECM of mammary tumors to study effects on cancer cell proliferation and invasion [30]. We applied this method to liver and lung tissues to allow us to study the effects of the ECM of these tissues on metastatic outgrowth. Tissues were isolated from adult mice and treated with low dose sodium dodecyl sulfate (SDS) and show successful decellularization over 96 h for liver tissues and 72 h for lung tissues (Fig. 1A-B). Decellularization with low-dose SDS removed all cell nuclei in lung and liver tissue as shown by hematoxylin and eosin (H&E) staining, while maintaining overall tissue structure and ECM abundance as shown by Picrosirius Red staining, and immunofluorescence staining for Collagen I (Fig. 1C). We have previously shown that this method does not significantly alter Collagen I fiber length and orientation [30]. We also used Western Blot to validate the removal of soluble intracellular proteins and show enrichment of fibronectin ECM protein (Fig. 1D). These data demonstrate that decellularization of liver and lung tissue with low-dose SDS removes cellular components and to generate ECM scaffolds (dECM) from native tissue.

Fig. 1.

Decellularization of lung and liver generates native ECM enriched scaffold. (A) Schematic diagram of decellularization and isolation of lung and liver ECM scaffold. (B) Representative images of lung and liver during decellularization with SDS up to 96 h, which leads to the isolation of a translucent ECM scaffold. (C) Hematoxylin and Eosin staining, Picrosirius Red, and immunostaining for Collagen I before and after decellularization of liver and lung tissue with low dose SDS (scale bar, 200 μm) (D) Western blot of liver and lung ECM before and after decellularization and stained for ECM and cellular components.

Liver dECM isolated from paclitaxel-treated tumor-bearing mice promotes TNBC cell invasion

We next set out to investigate how tissue-level changes in liver and lung ECM composition after systemic treatment with chemotherapy impacted tumor cell proliferation and invasion, two features essential for metastatic outgrowth [6]. We used the transgenic PyMT-MMTV mouse model which closely mimics clinical progression of human breast cancer and treated the mice with paclitaxel or doxorubicin, drugs that are standard of care for TNBC patients. At about 12 weeks old or when overall tumor burden reached at least 500 mm³, mice were treated with 4 cycles of doxorubicin, paclitaxel, or vehicle control every 5 days (Fig. 2A) [31]. Mice were sacrificed 3 days after the last treatment cycle. Treatment with paclitaxel and doxorubicin significantly slowed the growth of primary mammary tumors compared to vehicle over the course of 4 treatment cycles (Fig. 2B). However, there were no differences in the number of liver metastatic foci relative to total tissue area at sacrifice between vehicle and chemotherapy-treated mice (Fig. 2C) We also found no difference in the total area of metastatic foci normalized to the total area of liver tissue and average size of individual metastatic lesions between vehicle and chemotherapy-treated mice (Fig. S1A-C). In the lung, we also saw no difference in the number of lung metastatic foci between vehicle and chemotherapy-treated mice (Fig. S1D). Together, these data show that chemotherapy effects on primary tumor growth do not correlate with effects on metastatic burden. This leads us to hypothesize that chemotherapy treatment may be inducing changes in the local environment of these organs that continue to support metastatic outgrowth, which requires both tumor cell proliferation and invasion, even during chemotherapy treatment.

Fig. 2.

dECM from livers from paclitaxel-treated mice increases tumor cell invasion. A) Schematic diagram of study design. PyMT-MMTV mice at 12 weeks old were treated with vehicle, paclitaxel (10 mg/kg), or doxorubicin (5 mg/kg) every 5 days for 4 cycles. Liver and lung tissues were excised and decellularized to obtain native ECM scaffolds. Decellularized ECM scaffolds were reseeded with PyMT-GFP cells to determine effects on invasive speed. (B) Fold change in mammary tumor burden at each treatment day for vehicle-, paclitaxel-, and doxorubicin-treated mice groups. (C) Number of metastatic lesions in the liver per μm² quantified from H&E-stained sections from vehicle, paclitaxel-, and doxorubicin-treated mice. Data from 11 mice for vehicle, 17 mice for paclitaxel, and 7 mice for doxorubicin groups pooled from 3 independent experiments. (D) Representative rose plots of individual cell paths seeded on liver dECM from vehicle-, paclitaxel-, and doxorubicin- treated PyMT-MMTV mice. The dotted lines represent 100 μm in each direction (E) Cell invasion speed and (F) percent of cells undergoing cell proliferation of PyMT-GFP cells seeded onto liver dECM from vehicle-, paclitaxel-, and doxorubicin-treated PyMT-MMTV mice. Experiments from 10 mice for vehicle, 17 mice for paclitaxel, and 7 mice for doxorubicin groups pooled from at least 3 independent experiments with 5 fields of view imaged per experiment and 4–8 cells tracked per field of view with n = 213 cells for vehicle, n = 1012 cells for paclitaxel, and n = 320 cells for doxorubicin. Data show mean with SEM, significance determined by t-test with *p < 0.05, **p = 0.004.

The local microenvironments of the lung and liver are composed of resident tissue specific cells, stromal cells, immune cells, vasculature and the ECM. To start to dissect the effects of chemotherapy on these microenvironments, we focused on investigating the role of the ECM in supporting metastatic outgrowth in the liver and lungs. To investigate the functional effects of chemotherapy treatment on liver and lung ECM, we isolated and decellularized tissues as previously described (Fig 1). Treatment naïve PyMT-MMTV derived cells (PyMT-GFP) labeled with green fluorescent protein (GFP) were then seeded onto the decellularized ECM (dECM) scaffolds overnight to determine effects of lung and liver ECM isolated from chemotherapy-treated, tumor-bearing mice on tumor cell proliferation and invasion, properties that support metastatic outgrowth (Movie S1). PyMT-GFP cells invaded significantly faster on liver dECM isolated from paclitaxel-treated mice, but not doxorubicin-treated mice, compared to vehicle (Fig. 2D-E). The dECM from paclitaxel- and doxorubicin-treated liver had no effect on tumor cell proliferation (Fig. 2F). There was also no significant difference in invasion of TNBC cells reseeded on lung dECM isolated from paclitaxel- relative to vehicle-treated mice (Fig. S1E-F). Based on these findings, we chose to focus solely on characterizing the liver in subsequent experiments. We next investigated the role of the primary tumor. We seeded PyMT-GFP cells onto liver dECM isolated from wild type FVB/NJ, non-tumor bearing mice treated with the same chemotherapy regimen. There was no difference in cell invasion speed on liver dECM from paclitaxel- or doxorubicin- treated mice, compared to vehicle (Fig. S2A-B). These data demonstrate that chemotherapy-treatment of tumor-bearing mice induces changes in the liver ECM, but not the lung ECM, that promote tumor cell invasion, a phenotype essential to metastatic outgrowth, but not proliferation.

Paclitaxel treatment increases abundance of type V collagen in the liver

We then wanted to determine the compositional changes in the liver ECM resulting from chemotherapy treatment to identify the ECM proteins that could be contributing to our observation of increased cancer cell invasion. We utilized tandem mass tag (TMT) quantitative proteomics using previously published and validated protocols [21,30] to identify protein-level changes to the liver ECM. The resulting dataset (Fig. S3A) was analyzed with principal component analysis (PCA) and shows that each drug can have distinct effects on ECM composition (Fig. S3B,C). We identify several proteins that were more abundant in paclitaxel- and doxorubicin-treated liver when compared to vehicle ((log2(FC) > 0.3, Fig. 3-B, Supplementary File 1). A total of 33 proteins and 73 proteins were differentially abundant in paclitaxel- and doxorubicin- treated liver ECM, respectively compared to dECM from the vehicle group and categorized as matrisome (collagens, proteoglycans, glycoproteins) and matrisome-associated (ECM regulators, ECM-affiliated proteins, and secreted factors) (Fig. 3B) (Supplementary File 1). Because only the liver dECM from paclitaxel-treated mice induced tumor cell invasion (Fig 2), we chose to focus on ECM proteins more abundant in paclitaxel- vs. doxo-rubicin-treated ECM. COL5A1 and A3 were identified to be more abundant in paclitaxel-treated livers, relative to vehicle-treated livers, with no significant change in doxorubicin-treated livers. To validate our findings from mass spectrometry, we used immunofluorescence staining in an independent group of chemotherapy-treated PyMT-MMTV mice. Immunostaining for collagen V confirmed a significant increase in collagen V abundance in paclitaxel-treated livers (Figs. 3C-D, S3D, S3E). We also quantified collagen V in liver dECM in chemotherapy-treated, non-tumor bearing mice and find no significant difference (Fig. S3F,G). Together, these data represent the first protein-level characterization of chemotherapy-treated liver matrisome and demonstrate that paclitaxel treatment leads to an increase in collagen V levels in the livers of tumor-bearing mice.

Fig. 3.

Characterization of the matrisome of chemotherapy-treated livers. (A) Volcano plots comparing paclitaxel- and doxorubicin- treated liver dECM relative to vehicle control. Arrow shows COL5a1 and COL5a3. (B) Pie chart of paclitaxel- and doxorubicin-treated liver ECM showing number of differentially expressed proteins by category relative to vehicle control by Mann-Whitney test with Benjamini-Hochberg correction. (C) Immunostaining of native liver dECM sections from vehicle-, paclitaxel-, and doxorubicin-treated mice for collagen V (scale bar, 100 μm) and (D) Mean fluorescent intensity of Collagen V signal within tissue area. Mean fluorescent intensity of collagen V signal within tissue area. Experiments from 3 mice for vehicle, 6 mice for Paclitaxel and 6 mice for doxorubicin, with 4 fields of view per tissue. Data represent mean with SEM, significance determined by t-test with **p < 0.05.

Collagen V promotes TNBC cell invasion

Next, we sought to determine if collagen V contributes to the increased invasion of cancer cells observed when reseeded on chemotherapy-treated liver dECM. We assessed the effect of collagen V on cancer cell invasion using a tunable and physiologically relevant 3D spheroid model, as previously described [30,32,33]. In this assay, tumor cells are embedded in a Collagen I matrix, which is the most abundant ECM protein in the liver, in addition to Collagen V. Spheroid outgrowth is dependent on tumor cell proliferation and invasion, mimicking the two features required for metastatic colonization [34]. We used the mouse PyMT cell line and a human TNBC cell line (MDAMB231). For both cell lines, spheroids grew significantly larger in the presence of collagen V when compared to collagen I only (Fig. 4A-B). To determine whether increased spheroid growth by Collagen V is due to an effect on proliferation, we assessed the effect of individual collagen I, collagen V and combination of collagen I and V on cell viability. We find that there is no significant difference in cell number when cells are plated in 2D on Collagen V or Collagen I (Fig. S4A). We then evaluated the effect of collagen V on 3D single cell invasion. We find that TNBC cells invade significantly faster when seeded in a combination of collagen I and collagen V gels than collagen I gel alone (Fig. 4C-D, MovieS2). We also quantified the proliferation events in the 3D single cell invasion assay and find no significant difference in cell proliferation when cells were seeded in 3D with or without collagen V (Fig. S4B). To determine effects of collagen V in isolation from collagen I, we quantified 2D cancer cell speed of TNBC cells seeded on collagen I- or collagen V-coated plates. We find that both collagen I and collagen V significantly increase tumor cell migration speed when compared to control (Fig. S4C). Together, this data show that in Collagen V alone can promote cell motility in 2D, and single cell and spheroid invasion in 3D in Collagen I gels.

Fig. 4.

Collagen V is a driver of TNBC cell invasion. (A) Representative images of TNBC seeded 3D spheroid invasion assay with collagen I alone or with combined collagen I and V (scale bar, 500 μm) and (B) quantified fold change in area after 5 days of growth. Data show mean with SEM from 3 independent experiments, with at least 3 independent spheroids per experiment (C) Representative rose plots of MDAMB231 and PyMT TNBC cell invasion paths in a 3D single cell invasion assay with collagen I or combined collagen I and V. The dotted lines represent 300 μm and 200 μm, respectively, in each direction(D) quantified cell invasion speed. Data show mean with SEM from 3 independent experiments with each replicate represented as a different opacity, *p < 0.01, ***p < 0.001, ****p < 0.0001 by t-test.

Collagen V alters collagen I fiber organization

Collagen V is known to be a key regulator of collagen I fiber assembly by forming heterotypic fibrils with collagen I [23]. Changes in collagen V abundance has been shown to alter collagen assembly and modifies cell behavior [35]. Therefore, we investigated whether the increased invasive seen in the presence of Collagen V is associated with a change in collagen fiber alignment. Structure of collagen I fibers was visualized by second harmonic generation (SHG) in our 3D spheroid assay and analyzed by CT-FIRE and OrientationJ as previously described [36,37]. ECM is determined to be aligned if greater than 55% of fibers are oriented within (+/−) 15° of the mode [36]. We find that in MDAMB231 and PyMT spheroids, the presence of collagen V leads to a significant increase in the percentage of fibers aligned, which reaches around 60% in the presence of collagen V (Fig. 5A-C). Additionally, analysis with CT-FIRE show that collagen V increases fiber straightness and length, but decreases fiber width MDAMB231 spheroids (Fig. 5D-F). In PyMT spheroids, collagen V increases fiber length and decreases fiber width, but has no effect on fiber straightness (Fig. 5D-F). To determine concentration dependent effects of collagen V, we seeded TNBC cells in our spheroid invasion assay with collagen I alone or in combined collagen I and V in increasing concentrations with 5 μg/mL, 10 μg/mL, and 20 μg/mL of collagen V. At 20 μg/mL of collagen V, fold change of spheroid growth was significantly increased compared to collagen I alone. At 5 μg/mL of collagen V we also show increased spheroid growth when compared to collagen I, but not at 10 μg/mL of collagen V (Fig. S5A-B). We also perform SHG on the spheroids at increasing collagen V concentrations and find that at 10 μg/mL and 20 μg/mL of collagen V, the percentage of collagen I fibers aligned is significantly increased (Fig. S5C-E). Finally, we perform SHG on liver dECM isolated from vehicle-, paclitaxel-, and doxorubicin-treated mice and show that collagen I alignment is increased in vivo in liver dECM isolated from paclitaxel-treated mice (Fig. 5G-I). These experiments demonstrate that collagen V-driven invasion of TNBC cells is associated with increased alignment, straightness and length of collagen I fibers, features which have been associated with increased tumor cell invasion.

Fig. 5.

Collagen V increases collagen I fiber alignment. (A) Representative images of second harmonic generation of 3D MDAMB231 and PyMT spheroids with collagen I alone or with combined collagen I and V and fiber overlay generated by CT-FIRE analysis (scale bar, 50 μm). (B) Distribution of averaged occurrence of fibers from 3 independent experiments aligned along the normalized mode angle and (C) quantified percentage of fibers aligned within 15° of the mode angle from averaged occurrence by OrientationJ plugin with ImageJ. (D) Fiber straightness (E) fiber length, and (F) fiber width as quantified by CT-FIRE. Data show mean with SEM from 3 independent experiments. (G) Representative images of second harmonic generation of liver dECM isolated from vehicle, paclitaxel, and doxorubicin mice (scale bar, 50 μm) with (H) distribution of averaged occurrence of fibers from 5 independent experiments aligned along the normalized mode angle and (I) quantified percentage of fibers aligned within 15° of the mode angle from averaged occurrence by OrientationJ plugin with ImageJ. *p < 0.01, ***p < 0.001, ****p < 0.0001 by t-test.

Collagen V activates FAK and ERK signaling in TNBC cells

We next dissected the signaling pathways mediating collagen V-driven invasion. Collagen V has been reported to signal through α1β1 integrin receptor, which can also be activated by collagen I, and activates down-stream RAS/MAPK and FAK signaling pathways [20,38,39]. We seeded MDAMB231 and PyMT cells on full-length collagen I only, collagen V only, and combined collagen I and V coated wells for 24 h and collected lysates for Western blotting. We probed for phosphorylation status of ERK1/2 (Thr202/Tyr204) and FAK (Tyr397) signaling in both cell lines (Fig. 6A and D). We show a significant increase in phosphorylation of ERK1/2 in MDAMB231 (Fig. 6B) and PyMT (Fig. 6E) cells when seeded on combined collagen I and V ECM compared to no ECM in PyMTs and compared to no ECM and collagen I alone in MDAMB231. In the PyMT cells, we see significant increase in phosphorylation of FAK (Fig. 6F) when seeded on combined collagen I and V compared to no ECM or collagen I alone. However, there was no significant increase in phosphorylated FAK in MDAMB-231 (Fig. 6C). These data demonstrate that collagen V activates of both FAK and ERK1/2 signaling in PyMT cells and ERK1/2 signaling in MDAMB231 cells.

Fig. 6.

Collagen I and V activates FAK and ERK signaling in mouse and human TNBC cells. (A) Representative western blot images of MDAMB231 lysates seeded on No ECM, collagen I, collagen V, and collagen I and V and immunoblotted for the indicated proteins and phosphoproteins. (B) Quantified expression p-ERK1/2 relative to total ERK1/2 and (C) p-FAK relative to total FAK normalized to loading control. (D) Representative western blot images of PyMT-2900 lysates seeded on No ECM, collagen I, collagen V, and collagen I and V and immunoblotted for indicated proteins and phosphoproteins. (E) Quantified expression of p-ERK1/2 relative to total ERK1/2 and (F) p-FAK relative to total FAK normalized to loading control. Data show mean with SEM, *p < 0.01, **p by ANOVA. Data from at least 4 independent experiments.

FAK signaling and α1β1-driven ERK1/2 signaling mediate collagen V-driven invasion

To determine which signaling pathways mediate Collagen V increases invasion, we used existing small molecule inhibitors targeting ERK1/2 (ulixertinib), FAK (defactinib), and α1β1 integrins (obtustatin) in vitro cell invasion assays. FAK and ERK1/2 signaling pathways are known to be key regulators of cell motility and play a critical role in cell proliferation, differentiation, and survival [40,41]. MDAMB231 and PyMT cells were seeded in a collagen I gel and with or without collagen V as described above and treated with ulixertinib, defactinib, or obtustatin. The doses were chosen based on previous literature and published IC50 values, to minimize effects on cell viability [42-47]. In the PyMT cell line, inhibition of α1β1 integrin with obtustatin significantly reduced the growth of spheroids in the presence of collagen V but not in the collagen I only condition (Fig. 7A-B). In both cell lines, inhibition of ERK1/2 with ulixertinib and FAK with defactinib reduced the growth of spheroids in both collagen I only conditions and collagen I combined with collagen V (Fig. S6A-B). We show similar results in our 3D single cell invasion assay. Inhibition of α1β1 integrin receptors did not reduce cell invasion in collagen I only conditions, but inhibition of α1β1 integrin receptors significantly reduced 3D invasion in the presence of collagen V. Inhibition of FAK with defactinib and ERK1/2 with ulixertinib significantly reduced 3D invasion in both collagen I only and combined collagen I and V conditions for both cell lines (Figs. 7C-D, S6C-D). We then set out to determine whether integrin α1β1 mediates collagen V-driven invasion via ERK1/2 phosphorylation, since we found that Collagen V led to consistent pERK activation in both cell lines (Fig. 6). We show that ERK1/2 phosphorylation is increased in combined collagen I and V and that obtustatin treatment reduces the amount of phosphorylated ERK1/2 in PyMT cells (Fig. 7E-F). In MDAMB231 cells, obtustatin had no difference in phosphorylation of ERK1/2 when plated on Collagen I+V (Fig. S6E-F). Additionally, we analyze cell speed of TNBC cells seeded on collagen I or collagen V coated plates treated with Obtustatin. We show that in the 2D context, inhibition of α1β1 integrins with Obtustatin decreased 2D cell speed in cells seeded on collagen V matrix, but not in collagen I (Fig. S6G). This demonstrates that collagen V signals primarily through α1β1 integrins. Together, our data show that Collagen V can activate pERK signaling, drive tumor cell migration in 2D, an effect abrogated by inhibition of α1β1 by obtustatin. In 3D, Collagen V enhances tumor cell invasion, can drive changes in Collagen I organization, and in combination with Collagen I, can activate ERK1/2 signaling in α1β1 integrin-dependent manner. We therefore conclude that Collagen V contributes to increased invasion directly by activating integrin signaling and indirectly by increasing Collagen I organization.

Fig. 7.

Inhibition of α1β1 integrins and ERK/ FAK signaling abrogates collagen V-driven invasion in 3D. (A) Representative images of PyMT cells seeded 3D spheroid invasion assay with collagen I alone or with combined collagen I and V (scale bar, 500 μm) with added 2 nM obtustatin (α1β1 integrin inhibitor), 1 μM defactinib (FAK inhibitor), and 0.3 μM ulixertinib (ERK inhibitor) and (B) quantified fold change in area after 5 days of growth (C) Representative rose plots of PyMT cell invasion paths in a 3D single cell invasion assay with collagen I or combined collagen I and V and with added 2 nM obtustatin (α1β1 integrin inhibitor), 1 μM defactinib (FAK inhibitor), and 0.3 μM ulixertinib (ERK inhibitor). The dotted lines represent 100 μm in each direction (D) quantified cell invasion speed. Data show mean with SEM from 3 independent experiments, *p < 0.01, ***p < 0.001 by t-test relative to vehicle control. (E) Representative western blot images of PyMT cells plated on 2D plates with No ECM, collagen I, collagen V, or combined collagen I and V with and without 2 nM obtustatin(α1β1 integrin inhibitor) and immunoblotted for the indicated proteins and phosphoproteins. (F) Quantified expression of p-ERK1/2 relative to total ERK1/2 normalized to loading control. Data show mean with SEM from 4 independent replicates, *p < 0.05, **p < 0.01 by t-test relative to vehicle control.

Inhibition of α1β1 decreases TNBC invasion on liver dECM from paclitaxel-treated mice

Finally, we wanted to determine whether α1β1-driven signaling, which is activated by collagen V, is responsible for the increased cell invasion induced by paclitaxel treatment on liver dECM scaffolds, as seen in Fig. 2D-E. PyMT-GFP cells were seeded onto liver dECM isolated from paclitaxel- and vehicle-treated mice and treated with the α1β1 inhibitor obtustatin. With an independent set of mice, we successfully replicated the previous phenotype of increased TNBC cell invasion on paclitaxel-treated liver dECM compared to vehicle-treated liver dECM. Treatment with obtustatin to inhibit α1β1 integrins on vehicle-treated liver dECM did not reduce cell invasion speed. However, on paclitaxel-treated liver dECM, obtustatin significantly suppressed paclitaxel-associated invasion of TNBC cells (Fig. 8A-B). This demonstrates that α1β1-driven signaling, which is activated by collagen V contributes to ECM-mediated breast cancer cell invasion in the liver induced by paclitaxel treatment.

Fig. 8.

Inhibition of α1β1 integrins decreases TNBC cell invasion on liver dECM from paclitaxel-treated mice. (A) Representative rose plots of PyMT cell invasion paths reseeded on vehicle- or paclitaxel-treated liver dECM with or without 2 nM obtustatin treatment. The dotted lines represent 100 μm in each direction (B) quantified cell invasion speed. Experiments from 4 mice per group, data show mean with SEM, significance determined by t-test with *p < 0.05, ****p < 0.001.

Discussion

In the present study, we identify a mechanism by which systemic chemotherapy treatment of tumor-bearing mice may promote a pro-metastatic environment for metastatic outgrowth in the liver through changes in the liver ECM composition. Chemotherapy remains the most widely used and effective treatment for patients with cancer. However, development of chemoresistance and disease recurrence continues to be an ongoing problem. For TNBC patients treated with neoadjuvant chemotherapy, over 30% of patients experience metastatic recurrence [3]. Previous studies show increasing evidence that chemotherapy, in addition to killing proliferating tumor cells, can induce changes that promote metastasis and priming of metastatic sites [48-52]. Understanding the systemic effects of chemotherapy, particularly on secondary sites of metastases, is critical to prevent and target recurrence. Here, we focus on the ECM, whose composition and organization is regulated by tumor and host stromal cell types, and is known to regulate a range of tumor phenotypes. We demonstrate feasibility of decellularization to isolate native tissue ECM scaffolds of metastatic organs to study the role of local ECM in driving phenotypes that support metastatic colonization. We perform the first proteomic characterization of liver ECM isolated from breast tumor-bearing mice treated with chemotherapy and show that several ECM proteins are dysregulated in the liver after chemotherapy treatment. We identify collagen V to be more abundant in liver ECM of paclitaxel-treated mice. We find that collagen V drives increased tumor cell invasion, a feature essential for metastatic outgrowth, by regulating collagen I organization, and signaling via α1β1 integrins and down-stream ERK1/2 and FAK activation. Importantly, inhibition of α1β1 integrins reduced tumor cell invasion in dECM from chemotherapy-treated mice. Together, these data illustrate the first example of chemotherapy-induced pro-metastatic niche formation involving the ECM.

First, we describe a novel pipeline to study the effect of whole tissue ECM of metastatic sites on tumor cell behaviors using decellularization. We are able to reseed fluorescently labeled cells on the intact dECM scaffolds from lung and liver and perform live imaging to evaluate effects on tumor cell proliferation and invasion. While this method of isolation had been used in the past, previous studies have consistently pulverized and lyophilized the frozen decellularized organs to then use as a substrate for cancer cells [53]. As a consequence, the physical and structural properties of the tissue are completely lost. Our method of decellularization directly from tissues allows us to isolate the whole ECM without disruption of the organ-specific tissue architecture and composition. Another method has been to grow three-dimensional cell culture models with liver cells in vitro to produce cell-derived ECM matrices [54]. However, these methods fail to capture the complexity and heterogeneity of the in vivo ECM, which is secreted and remodeled by a range of cell types. Our method allows us to study cell responses to organ ECM, irrespective of the cell that secreted it. Finally, recapitulation of the in vivo ECM microenvironment can also be achieved by coating exogenous ECM substrates onto synthetic polymer scaffolds [55]. Overall, our data show that liver dECM isolated from paclitaxel-treated, tumor bearing mice, but not lung dECM, increases tumor cell invasion, but not proliferation. Liver dECM isolated from doxorubicin-treated mice do not have the same effect, suggesting a drug-specific effect. Our method provides a straightforward strategy to determine the effect of whole-tissue ECM on tumor cell phenotypes and investigate the functional impact of systemic insults on tissue ECM.

Our proteomic characterization of the liver ECM shows that both paclitaxel and doxorubicin induce significant changes in the ECM composition. Paclitaxel-treated mice have increased abundance of collagen V, which has been previously implicated in liver damage and has been described as a biomarker of a fibrotic liver microenvironment [12,56,57]. Colorectal cancer liver metastases also have upregulated expression of collagen V in liver ECM [21,58].

Previous studies have shown that paclitaxel treatment can have a protective effect on liver fibrosis induced by different physical and chemical injuries, and characterized by Collagen I staining, hydroxy-proline content or alpha-SMA staining [59,60]. The effect on ECM composition was not evaluated in this experimental paradigm. Further, in these studies, paclitaxel treatment is given after the development of liver fibrosis and at a relatively low dose (0.3 mg/kg, administered 3 times weekly), a dose 30 times lower than the therapeutic dose used in our study. Paclitaxel treatment alone had no effect. Therefore, it is possible that sub-therapeutic doses have different effects than the high doses required to treat tumor models. Based on these studies, it would be interesting to investigate the effect of different doses of paclitaxel on liver ECM composition and how therapeutic doses of paclitaxel impact ECM composition in mice that have already have fibrosis.

The mechanisms by which paclitaxel stimulates ECM production in the liver and the cell types responsible for this process remains unclear. The main ECM-producing cells in the liver are the hepatic stellate cells (HSCs). HSCs are well-studied in the context of fibrosis and shown to deposit a multitude of ECM proteins which include collagens, laminin, fibronectin, tenascin-C, and vitronectin [61,62]. Previous studies have established that hepatic stellate cells (HSCs) in the liver are responsible for mediating the process of liver repair and regeneration [63]. Therefore, it is possible that chemotherapy activates liver damage and repair pathways that initiate the production of collagen V by HSCs to promote a pro-metastatic liver microenvironment. Conversely, paclitaxel has been shown to have protective effects against cell models of liver fibrosis by inhibition of TGFβ in HSCs [59,64]. However, the interplay of HSCs, and liver stromal cells such as hepatocytes, as well as the influence of cancer-dependent factors have not been studied. Drugs are primarily metabolized by hepatocytes, which are also impacted by the cytotoxic effects of chemotherapies that inherently target DNA, RNA, or protein synthesis [17]. Hepatocytes themselves have been shown to produce ECM proteins like collagens and work in concert with HSCs to repair and regenerate liver tissues [56,65]. Macrophages in the liver have also been shown to increase fibronectin synthesis in response to issue damage [66]. Finally, while we have focused on the role of the ECM in this study, it is well known that liver damage also activates systemic host-mediated tissue repair and regeneration signals by releasing cytokines and chemokines, activating pro-inflammatory pathways, as well as liver hypertrophy and expansion [67-69]. Therefore, it will be important for future studies to dissect how these changes in the liver microenvironment also impact metastatic outgrowth.

The role of the primary tumor in driving changes in ECM composition in the liver will need to be examined in further details. We find that tumor cells seeded on the decellularized liver ECM isolated from chemotherapy treated, non-tumor bearing mice show no differences in cell invasion speed, suggesting an important role for tumor secreted factors in the changes in ECM composition we see in the liver. Tumor-secreted exosomes have been shown to travel to secondary organs to prime the pre-metastatic niche. For example, pancreatic cancer derived exosomes induced changes to the liver microenvironment to form a pre-metastatic niche and demonstrating the potential of cancer cell derived intercellular communication [70]. Further, chemotherapy can also induce secretion of exosomes carrying heparanase, an enzyme capable of remodeling the primary tumor ECM and promoted survival of cancer cells [70,71]. Therefore, it is possible that the tumor cells already present in the liver could also directly secrete cues locally that impact ECM composition in the liver. Studies to elucidate the mechanisms by which chemotherapy stimulates ECM remodeling cues and deconvolute the crosstalk between cancer cell- or stromal cell- driven ECM remodeling post-chemotherapy treatment will be critical to develop better strategies to reduce liver metastasis.

Finally, we show that collagen V enhances tumor cell invasion, can drive changes in Collagen I organization and activate ERK1/2 signaling in α1β1 integrin-dependent manner. Upregulation of collagen V in the liver has been associated with increased liver metastasis in pancreatic, colon, and breast cancer [13,20,21], although the detailed mechanism by which collagen V drives this process remains poorly understood. Although only representing 10–16% of total collagen of the liver, the role of collagen V as a key regulator of collagen I assembly and structure has been well established [19, 72]. Knockout of Col5 in mouse embryos resulted in fewer collagen fibrils assembled, altered structures and organization, and disrupted architecture across multiple tissues including the liver [24,25]. We show that collagen V increases collagen alignment and fiber length, while decreasing fiber width in the presence of breast cancer cells and that livers from paclitaxel-treated tumor-bearing mice which have increased collagen V abundance also have increased collagen I alignment. These are consistent with previous in vitro study characterizing mixed solutions of increasing soluble collagen V with collagen I [73]. Clinical data show that patients with increased collagen fiber alignment in mammary tumors indicates poor patient prognosis due to stiffer mechanical properties which enhance cancer cell invasion [29,74,75], however whether increased Collagen I fiber alignment in the liver leads to more metastatic outgrowth, either by increasing invasion or proliferation, is not well known. Liver fibrosis, marked by increased collagen deposition such as collagen I, has been associated with increased liver metastases [14]. It has separately been shown that liver fibrosis also increases deposition of collagen V [19,58]. Our studies suggest that increased Collagen I alignment in the liver may be associated with metastatic outgrowth, although further studies are needed to dissect this. Collagen V has also been shown to promote adhesion, migration and metastasis of pancreatic cancer by directly activating β1 integrins [26]. We show that collagen V on its own can directly activate pERK signaling and drive 2D cell migration in human TNBC cell lines, an effect dependent on α1β1 integrins. Further, we show that collagen V, in the presence of collagen I, leads to significant increased activation of ERK1/2 in both cell lines studied, and FAK signaling in only one cell line. Inhibition of α1β1 integrin with obtustatin decreased collagen V-driven invasion in the presence of collagen I and significantly reduced ERK1/2 activation. At the dose of obtustatin used, inhibition of α1β1 had no effect on tumor cell invasion in collagen I alone, but it did in gels with both collagen I and collagen V, which not only have increased ERK1/2 signaling, but also have a more aligned collagen I network and decreased fiber width. Whether the reorganization of collagen I by collagen V further increases the activation of α1β1 in this context to promote cell invasion and make the cells more sensitive to α1β1 inhibition by obtustatin is unclear, although it has been shown in other contexts that highly dense or bundled collagen can impact cell adhesion dynamics and integrin activation [76-78]. Based on our data, we conclude that collagen V contributes to increased invasion directly by activating integrin signaling and indirectly by increasing collagen I organization. Future work and experimental in vitro models with engineered ECM molecules will be needed to uncouple the effects of collagen I fiber organization and width and α1β1-driven signaling in this context.

With over 650,000 patients receiving chemotherapy treatment in the US each year, it is important to understand the clinical consequences of cytotoxic chemotherapy in the progression of cancer [51]. Managing the burden of recurrent metastatic cancer continues to be a complex, multifaceted problem. Though this research is specific to the effects of chemotherapy on the liver ECM within a model of breast cancer, the findings and methods can be applied to other primary tumors and secondary organs to determine tumor-specific and organ-specific effects [79]. By further establishing the ECM as a key regulator in the metastatic cascade and microenvironment, this work can lead to the development of novel ECM-targeting drugs and ECM-based biomarkers to aid in tracking cancer progression and drug resistance.

Experimental procedures

Study design

This study was designed to investigate ECM-driven phenotypes associated with chemotherapy treatment in commonly metastasized organs from TNBC. Specifically, we sought to use our novel dECM experimental pipeline to characterize how chemotherapy changes ECM composition in organs colonized by breast cancer metastases, and how those changes affect cancer cell behavior. We used the immunocompetent MMTV-PyMT genetic mouse model of TNBC to ensure detection of chemotherapy-induced ECM remodeling. Upon reaching a tumor burden of 500 mm³, mice were randomized into paclitaxel, doxorubicin, or vehicle groups. Paclitaxel and doxorubicin chemotherapies were chosen because they are employed in the most common chemotherapy regimen used in clinical settings. Discovery and validation of collagen V as a chemotherapy-induced ECM protein was conducted using tissues from separate mice to reduce the chance that this observation was an artifact of a single experiment. Researchers were blinded during imaging and quantification of collagen V-stained MMTV-PyMT liver tissue sections. In vitro experiments in this study were completed with a minimum of three replicates to ensure reproducibility. We excluded data from reseeding assays only when it was apparent that cells did not adhere to dECM, such as when the majority of cells in the field of view died during the experiment. All other data that was produced in the discussed experiments was presented.

Animal studies

All animal studies were reviewed and approved by the Tufts University Institutional Animal Care and Use Committee. Transgenic mice bearing the polyomavirus middle T antigen under control of the mouse mammary tumor virus promoter (MMTV-PyMT) were obtained from The Jackson Laboratory (Bar Harbor, ME). Female MMTV-PyMT mice were grown until overall tumor burden reached 500 mm³ at about 12 weeks of age before being randomized into chemotherapy treatment groups. Paclitaxel (Taxol) was resuspended in 5% dimethyl sulfoxide, 40% polyethylene glycol 3000, and 5% Tween 80 in dH2O and administered intraperitoneally at 10 mg/kg. Doxorubicin (Adriamycin) was dissolved in phosphate-buffered saline and administered intravenously at 5 mg/kg. Appropriate vehicle controls for each drug were included in the control group. All drugs were administered for 4 cycles given every 5 days. Tumor burden was monitored using digital calipers at each treatment. Three days after the conclusion of treatment, mice were euthanized by CO₂ asphyxiation and lungs and livers were excised for further study. Histologic analysis of hematoxylin and eosin (H&E) staining of these lungs and livers were quantified for metastases. Metastatic lesions were calculated by counting the number of metastases per left lateral lobe relative to the area of tissue.

Antibodies, inhibitors, and ECM substrates

Primary antibodies used in this study include: α-fibronectin (ab2413, Abcam, Cambridge, MA), α-tubulin (DM1A, Sigma-Aldrich, St. Louis, MO), α-GAPDH (14C10, Cell Signaling Technology, Danvers, MA), α -Histone H3 (ab1791; Abcam, Cambridge, MA), α-collagen I(ab21286), α-collagen V (ab7046, Abcam), α -phospho-p44/42 ERK (Thr202/Tyr204) (4370; Cell Signaling Technology, Danvers, MA), α-p44/42 ERK1/2 (9107; Cell Signaling Technology, Danvers, MA), α -pFAK397 (3283; Cell Signaling Technology, Danvers, MA). α-FAK (D5O7U, Cell Signaling Technology).

All antibodies were used at a concentration of 1:1000, except for α -pFAK which was used at a concentration of 1:500. Pharmacological inhibitors include: obtustatin (4664, Tocris Bioscience), ulixertinib (HY-15,816, MedChemExpress), and defactinib (S7654, Selleck Chemicals). ECM substrates used were: Collagen I (CB-40,236; Fisher Scientific, Hampton, NH) and Collagen V (ab7537; Abcam, Cambridge, MA)

Cell culture

MDAMB-231-GFP cells were obtained from American Type Cell Collection (Mannassas, VA) and cultured in Dulbecco’s modified Eagle’s medium (DMEM) with 10% fetal bovine serum and 1% penicillin-streptomycin-glutamine. PyMT2900-GFP cells were a gift Professor Richard Hynes’ lab at MIT and were cultured in a 1:1 mix of DMEM and Ham’s F-12 Nutrient Mixture containing 2% fetal bovine serum, 1% bovine serum albumin (A2153-50G; Sigma-Aldrich, St. Louis, MO), EGF (10 ng/ml; PHG0313; Fisher Scientific, Hampton, NH), insulin (10 μg/ml; 12,585,014; Fisher Scientific, Hampton, NH), and 1% penicillin-streptomycin-glutamine. Cells were routinely monitored for mycoplasma contamination by polymerase chain reaction (PCR) using the Universal Mycoplasma Detection Kit (30-1012K; ATCC, Manassas, VA). All cells used in this study were mycoplasma negative.

Isolation of dECM scaffolds and protein isolation

Liver and lung-derived dECM scaffolds were produced as described previously (28). Briefly, livers and lungs dissected from MMTV-PyMT mice were submerged in 0.1% SDS in PBS (w/v) solution with rotation for several days, replacing the solution daily. Upon successful decellularization, tissues were moved to 0.05% Triton X-100 in PBS solution for 3 h, followed by washing in dH₂O for 24 h to remove residual detergents. To assess decellularization, dECM scaffolds were paraffin embedded, sectioned and stained with H&E, Picrosirius Red, and immunohistochemistry. Additional assessment of decellularization was confirmed with Western Blot. Decellularized ECM or intact tissues were lysed in 25 mM Tris, 150 mM NaCl, 10% glycerol, 1% NP 40 and 0.5 M EDTA with 1x protease Mini-complete protease inhibitor (04,693,124,001; Roche, Indianapolis, IN) and 1x phosphatase inhibitor cocktail (4,906,845,001; Roche, Indianapolis, IN) at 4 °C. In addition, for decellularized ECM, tissue homogenization was performed using a BeadBug^™ 3 Place Microtube Homogenizer (D1030; Benchmark Scientific, Sayreville, NJ). Decellularized ECM or cell homogenate was centrifuged at 21,000 g for 10 min at 4 °C and supernatant stored at −20 °C until used for Western Blot.

dECM reseeding experiments

Reseeding experiments were performed as described previously (28). Decellularized liver and lung tissue were cut into 5 mm pieces and conditioned in complete cell culture media for 24 h. On day of the experiment, tissues were reseeded with 500,000 PyMT-GFP cells and allowed to adhere for 6 h in 37C. Reseeded tissue was then moved to fresh media, stabilized, and imaged for 16 h overnight, capturing images every 10 min in an environmentally controlled chamber within the Keyence BZ-X710 microscope (Keyence, Elmwood park, NJ). For inhibitor experiments, vehicle controls or small-molecule inhibitors were added to the media 1 h before imaging began. Cells were then tracked using VW-9000 Video Editing/Analysis Software (Keyence, Elmwood Park, NJ), and invasive speed was calculated using a custom MATLAB script vR2019a (MathWorks, Natick, MA). For quantification of proliferation events, each event that a cell visibly proliferates is a single proliferation event normalized to the total number of cells in the field of view. Data presented are the result of at least three independent experiments with five fields of view imaged per experiment and 4–8 cells tracked per field of view. Distance on roseplot represent distance from certain of plot to end of grey-dotted axis line.

Sample preparation for mass spectrometry

Decellularized ECM homogenate (from ±5 mm piece of tissue) was denatured in 8 M urea and 10 mM dithiothreitol, alkylated with 25 mM iodoacetamide, and deglycosylated with peptide N-glycosidase F (P0704S; New England Biolabs, Ipswich, MA). Samples were then digested sequentially, first with endoproteinase LysC (125–05,061; Wako Chemicals USA, Richmond, VA), then trypsin (PR-V5113; Promega, Madison, WI). Samples were acidified with 50% trifluoroacetic acid and pH tested to <2.0 and centrifuged to collect supernatant. Peptide labeling with TMT10plex (90,110; Thermo Fisher Scientific, Waltham, MA) was performed according to the manufacturer’s instructions.

LC-MS/MS analysis

Peptides were separated by reversed-phased HPLC (Thermo Easy nLC1000, Thermo Fisher Scientific, Waltham, MA) using a precolumn (made in house, 6 cm of 10-μm C18) and a self-pack 5-μm tip analytical column (12 cm of 5-μm C18; New Objective, Woburn, MA) over a 140-min gradient before nanoelectrospray using a Q Exactive HF-X mass spectrometer (Thermo Fisher Scientific, Waltham, MA). Solvent A was 0.1% formic acid, and solvent B was 80% ACN/0.1% formic acid. The gradient conditions were 2-10% B (0-3 min), 10-30% B (3-107 min), 30-40% B (107-121 min), 40-60% B (121-126 min), 60-100% B (126-127 min), 100% B (127-137 min), 100-0% B (137-138 min), and 0% B (138-140 min), and the mass spectrometer was operated in a data-dependent mode. The parameters for the full-scan MS were resolution of 60,000 across 350 to 2000 m/z (mass/charge ratio), automatic gain control 3 × 106, and maximum ion injection time of 50 ms. The full-scan MS was followed by tandem mass spectrometry (MS/MS) for the top 15 precursor ions in each cycle with a normalized collision energy of 34 and dynamic exclusion of 30 s.

Western blotting

Protein lysates were separated using SDS-polyacrylamide gel electrophoresis on a 12% polyacrylamide gel. Proteins were transferred to a nitrocellulose membrane using a TransBlot Turbo Transfer system (Bio-Rad; Hercules, CA) then were blocked in 5% nonfat dry milk in tris-buffered saline with 0.05% tween-20 and incubated in primary anti-body overnight at 4 °C with rocking. Proteins were detected using horseradish peroxidase-conjugated secondary antibodies (Jackson ImmunoResearch, West Grove, PA). Imaging was performed using a ChemiDoc MP imaging system (12,003,154; BioRad, Hercules, CA).

Immunohistochemistry

Tissue fixation, processing and sectioning was performed as previously described (28). The left lateral lobe of liver and left lobe of lung tissue dissected from MMTV-PyMT mice were fixed in fresh 4% paraformaldehyde in PBS, embedded in paraffin, and sectioned into 10 μm sections. For hematoxylin and eosin staining, sections were deparaffinized, hydrated and stained using standard procedures with hematoxylin (GHS280; Sigma-Aldrich, St. Louis, MO), and counterstained with eosin (HT110180; Sigma-Aldrich, St. Louis, MO). Stained sections were mounted with toluene (SP15–500, Fisher Scientific, Hampton, NH). For immunofluorescence, tissue sections were deparaffinized and antigen retrieval was performed in Citra Plus solution (HK057; Biogenex, Fremont, CA). Sections were then blocked in PBS with 0.5% tween 20 and 10% donkey serum and incubated with primary antibodies overnight at 4C. The following day, sections were incubated with fluorophore-conjugated secondary antibodies and 4′6-Diamidino-2-phenylindole (DAPI; D1306; Thermo Fisher Scientific, Waltham, MA) to stain cell nuclei. Sections were mounted in Fluoromount mounting medium (00-4958-02; Thermo Fisher Scientific, Waltham, MA). All slides were imaged using a Keyence BZ-X710 microscope (Keyence, Elmwood Park, NJ). Quantification of ECM signal was performed using ImageJ (National Institutes of Health, Bethesda, MD).

Spheroid invasion assay

1000 231-GFP or PyMT-GFP cells were seeded with full media in round-bottom, low-attachment 96-well plates and centrifuged at 3000 g for 3 min to induce spheroid formation. Spheroids were incubated in 37C for three days addition of an ECM mixture of 1 mg/mL collagen I (354,236; Corning, Corning, NY), 10 mM NaOH, 7.5% 10X DMEM, and 50% DMEM with or without 20 μg/mL native human collagen V (ab7537; Abcam, Cambridge, MA). Once ECM gelated, 50uL of culture media to maintain moisture and humidity in the well. In inhibitor studies, this media included DMSO as a vehicle control or small molecule inhibitors. Spheroids were imaged in 3D on the day of ECM addition and after 4 days of growth using a Keyence BZ-X710 microscope (Keyence, Elmwood Park, NJ). Spheroid invasion was quantified by measuring the area of the initial spheroid at Day 0 and at Day 4 then calculating the fold change as (Day 4 Area-Day 0 Area)/(Day 0 Area). Data presented are the result of three independent experiments with three technical replicates per experiment.

3D Single cell invasion assay

20,000 231-GFP or PyMT-GFP cells were suspended with full media and ECM mixture of 1 mg/mL collagen I (354,236; Corning, Corning, NY), 10 mM NaOH, 7.5% 10X DMEM, and 50% DMEM with or without 20 μg/mL native human collagen V (ab7537; Abcam, Cambridge, MA) in 96 well plates and incubated at 37C for 2 h prior to imaging. Once ECM gelated, 50uL of culture media was added to maintain moisture and humidity in the well. In inhibitor studies, this media included DMSO as a vehicle control or small molecule inhibitors. Z-stack images were of cells were captured every 10 min overnight for 16 h using Keyence BZ-X710 microscope (Keyence, Elmwood Park, NJ). Cells were then tracked using VW-9000 Video Editing/Analysis Software (Keyence, Elmwood Park, NJ), and invasive speed was calculated using a custom MATLAB script vR2019a (MathWorks, Natick, MA). Data presented are the result of three independent experiments with three technical replicates per experiment. Distance on roseplot represent distance from certain of plot to end of gray-dotted axis line. For quantification of proliferation events, each event that a cell visibly proliferates is a single proliferation event normalized to the total number of cells in the field of view.

Cell viability assay

Cells were plated on plates coated with 20 μg/ml ECM protein and allowed to adhere for 24 and 48 h. PrestoBlue^™ Cell Viability Reagent (A13261; Invitrogen, Carlsbad, CA) was added to each well according to the manufacturer’s instructions and incubated for 30 min at 37 °C. Fluorescence was then read on a plate reader at 562 nm. Background was corrected to control wells containing only cell culture media without cells. Data are the result of 3 independent experiments with 6 technical replicates per experiment. Fold change of fluorescence was calculated from 24 to 48 h.

2D Cell migration speed assay

20,000 231-GFP or PyMT-GFP cells were plated on plates coated with 20 μg/ml ECM protein and allowed to adhere for 2 h in 37C prior to imaging. In inhibitor studies, 2 nM Obutstatin was added to the media. Cells were imaged every 10 min overnight for 16 h using Keyence BZ-X710 microscope (Keyence, Elmwood Park, NJ). Cells were then tracked using VW-9000 Video Editing/Analysis Software (Keyence, Elmwood Park, NJ), and invasive speed was calculated using a custom MATLAB script vR2019a (MathWorks, Natick, MA). Data presented are the result of three independent experiments with three technical replicates per experiment.

Statistical analysis

GraphPad Prism v9.1.0 was used for generation of graphs and statistical analysis. For comparison between two groups, an unpaired two-tailed Student’s t-test was used and a p-value of ≤ 0.05 considered significant. For comparison between more than two groups, a one-way ANOVA with Bonferroni multiple testing correction. Data represent mean ± SEM.

Movie S1. PyMT cells seeded onto dECM scaffolds isolated from the livers of vehicle-treated mice. Epifluorescence images were captured every 20 min for 16 h and are displayed at 12 frames/s

Abbreviations:

TNBC

Triple Negative Breast Cancer

BCLM

Breast Cancer Liver Metastasis

ECM

Extracellular matrix