Obesity-associated Adipose Stromal Cells Promote Breast Cancer Invasion Through Direct Cell Contact and ECM Remodeling

Abstract

Obesity increases the risk and worsens the prognosis for breast cancer due, in part, to altered adipose stromal cell (ASC) behavior. Whether ASCs from obese individuals increase migration of breast cancer cells relative to their lean counterparts, however, remains unclear. To test this connection, multicellular spheroids composed of MCF10A-derived tumor cell lines of varying malignant potential and lean or obese ASCs were embedded into collagen scaffolds mimicking the elastic moduli of interstitial breast adipose tissue. Confocal image analysis suggests that tumor cells alone migrate insignificantly under these conditions. However, direct cell-cell contact with either lean or obese ASCs enables them to migrate collectively, whereby obese ASCs activate tumor cell migration more effectively than their lean counterparts. Time-resolved optical coherence tomography (OCT) imaging suggests that obese ASCs facilitate tumor cell migration by mediating contraction of local collagen fibers. Matrix metalloproteinase (MMP)-dependent proteolytic activity significantly contributes to ASC-mediated tumor cell invasion and collagen deformation. However, ASC contractility is also important, as co-inhibition of both MMPs and contractility is necessary to completely abrogate ASC-mediated tumor cell migration. These findings imply that obesity-mediated changes of ASC phenotype may impact tumor cell migration and invasion with potential implications for breast cancer malignancy in obese patients.

Graphical Abstract

A collagen-embedded multicellular spheroids platform and advanced imaging approaches are used to investigate the effect of obesity associated adipose stromal cells (ASCs) on tumor cell migration. Direct cell-cell contact with ASCs enables collective tumor cell migration and obese ASCs were more effective than their lean counterparts. Both matrix metalloproteinase-dependent proteolytic activity and ASC contractility are important during this collective migration.

1. Introduction

The obesity epidemic has become a major health crisis in the United States due, in part, to the fact that obesity increases the risk and worsens the prognosis for many cancers including breast cancer. ^[1,2] The link between obesity and breast cancer is best established for postmenopausal, estrogen-receptor positive breast cancer, but obesity has also been associated with the pathogenesis of other breast cancer subtypes including triple negative breast cancer. ^[1–3] Excess body weight is typically linked with altered metabolism, chronic inflammation, and abnormal secretion of adipokines and hormones by adipocytes. ^[4–6] As a consequence, most previous work has focused on investigating the role of obesity-associated endocrine functions in breast cancer. However, microenvironmental perturbations of breast tissue may be similarly important and could stimulate the activation of otherwise non-malignant cells into overt disease. ^[4,7,8]

Fibrotic remodeling of mammary tissue can promote breast cancer development and progression by altering both the biochemical and biophysical microenvironment. ^[9,10] While most previous work was performed in the context of tumors, we and others have shown that obesity may induce similar changes. ^[7,11] In tumors, activated fibroblasts (also termed myofibroblasts or cancer-associated fibroblasts [CAFs]) deposit and remodel fibrillar extracellular matrix components (ECM) including collagen type I and fibronectin.^[12] The resulting increase in ECM density and stiffness as well as microarchitectural changes contribute to tumor development, invasion, and metastasis. ^[13–15] Interestingly, obesity leads to an enrichment of myofibroblasts in breast adipose tissue independent of cancer. Obesity-associated myofibroblasts reside in interstitial regions of adipose tissue as part of the adipose stromal cell (ASC) fraction and promote tumor cell malignancy by altering the density, stiffness, and microarchitecture of local ECM. ^[7,16] As tumor cells invade breast adipose tissue by exploring fibrotic ECM and myofibroblast-rich interstitial regions, ^[17] it will be critical to define whether or not obesity-dependent changes of ASCs and the resulting differences in ECM remodeling may promote invasion of otherwise non-invasive mammary epithelial cells.

Myofibroblasts can promote tumor invasion via both biochemical and biophysical mechanisms. In particular, myofibroblasts have been shown to generate physical guidance channels for cancer cells through matrix metalloproteinase (MMP)-dependent degradation of the ECM. ^[18] Furthermore, myofibroblasts can use contractility-based mechanisms to promote tumor cell invasion by exerting physical forces on the surrounding ECM and/or the cancer cells themselves. ^[19–22] Whether or not ASCs employ similar mechanisms to promote tumor cell invasion and if ASCs from obese versus lean patients are more effective in doing so remains unclear.

To study tumor cell migration and invasion as a function of obesity-associated changes of ASCs, multicellular spheroids can be embedded into collagen type I, the primary ECM component of interstitial adipose tissue. With a diameter of 200–500 μm, tumor spheroids exhibit typical characteristics of tumors including an outer proliferating zone, inner quiescent zone, and necrotic cores due to limited diffusion of nutrients and oxygen. ^[23,24] Furthermore, tumor spheroids can recapitulate heterotypic cell-cell interactions and cell-ECM interactions in a more physiologically relevant manner than conventional 2D culture systems. ^[23,25] However, monitoring cellular and ECM dynamics of collagen-embedded spheroids poses technical challenges, as these studies require imaging large volumes of interest at high resolutions over extended periods of time. To overcome limitations of conventional confocal microscopy, including limited imaging depth due to optical scattering, potential cell stress due to phototoxicity, and photobleaching of cellular labels, alternative imaging methods such as optical coherence tomography (OCT) can be utilized to study cell-ECM interactions. ^[26,27] OCT uses low temporal coherence light to image scattering contrast in a label-free manner and new techniques such as traction force optical coherence microscopy (TF-OCM) can separate cell movements and traction force-dependent ECM deformations in millimeter-sized volumetric fields-of-view with temporal coverage spanning minutes to days. ^[27]

Here, we leverage collagen-embedded multicellular spheroids in combination with advanced imaging approaches including confocal microscopy and OCT to investigate the role of lean versus obese ASCs in promoting tumor cell migration. Furthermore, it was our goal to delineate the mechanistic contributions of MMP- and cell contractility-dependent changes of collagen remodeling to ASC-mediated tumor cell invasion. Results from these studies provide insights into the role of ASCs in promoting tumor cell invasion into adipose tissue and contribute knowledge that leads to a better understanding of the obesity-breast cancer connection.

2. Results and Discussion

2.1. ASCs promote collective migration of premalignant cancer cells via direct cell-cell contact

To study the potential effect of obesity-associated ASCs on the migration of otherwise non-invasive cells, a co-culture model was developed to mimic key aspects of how ASCs and premalignant tumor cells may interact within collagen-rich interstitial regions of adipose tissue. The design of this model was inspired by histological features of a clinical specimen from a lymph-node negative, triple breast cancer patient in which tumor cells and stromal cells are intermixed and in direct contact with each other (Figure 1a). First, monoculture spheroids composed of only premalignant MCF10AT1 cells were formed under non-adherent conditions as described previously (Figure 1b). ^[28] MCF10AT1, initially derived from MCF10A mammary epithelial cells through H-Ras mutation, ^[29] are considered premalignant but can develop into invasive cells under appropriate conditions, thus representing an appropriate model system to test the pro-invasive effects of ASCs. ^[30] Following spheroid formation, MCF10AT1 were embedded into 6 mg mL⁻¹ collagen using a custom-made device for improved confocal imaging (Figure 1c). This collagen concentration yielded hydrogels with an elastic modulus of 2214.5 +/− 218 Pa mimicking the mechanical properties of collagen-rich interstitial ECM of obese breast tissue (Figure 1d). ^[7] While MCF10AT1cells could readily invade collagen at low concentrations (1–3 mg mL⁻¹ [Figure S1a]), they were not able to invade into 6 mg mL⁻¹ collagen matrices (Figure 1c, S1a) over a culture period of up to 4 days. These results suggest that the mechanical properties of collagen present within interstitial obese adipose tissue may prevent, rather than encourage, invasion of premalignant cells.

Figure 1: Adipose stromal cells promote collective invasion of MCF10AT1 cells via direct cell-cell contact.

a. Immunohistochemical staining showing intermixing of tumor (cytokeratine-positive) and stromal (α-SMA positive) cells in a human lymph node-negative, triple negative breast cancer sample. Scale bars represent 200 μm. b. Schematic overview of spheroid formation and c. embedding of spheroids in collagen. Brightfield images showing MCF10AT1 spheroids in collagen (6 mg mL⁻¹) on day 0 and day 2. Scale bars represent 100 μm for brightfield images and 50μm for immunofluorescence (IF) images. d. DMTA testing of collagen scaffolds showing stress strain curves and the corresponding average elastic modulus, 2214.5 +/− 218 Pa (n=4). e. Brightfield and IF images visualizing the role of ASC in invasion of MCF10AT1 spheroids. i) Effect of conditioned media (CM) from ASC spheroids on tumor spheroid (green). Red arrow indicates invasion site. ii) Effect of ASCs mixed uniformly in collagen surrounding tumor spheroid (green). Red arrow indicates invasion site. iii) Effect of direct cell-cell contact by mixing a constant number of MCF10AT1 cells (5,000 cells per spheroid) with an increasing number of ASCs; i.e., 50 ASCs (1:100), 500 ASCs (1:10) or 5,000 ASCs (1:1) to form co-culture spheroids that were embedded into collagen scaffolds. f. Effect of patient-derived ASCs on invasion of MCF10AT1 as quantified by image analysis of co-culture spheroids following two days of culture. Scale bars represent 100 μm. *p< 0.05 and **p<0.01.

To test whether ASCs could promote the invasion of MCF10AT1 cells into dense collagen and if ASC-secreted soluble factors play a role in this process, wild type (WT) ASCs were isolated from the mammary fat pads of female C57BL/6J mice. As myofibroblasts secrete factors that can promote tumor cell migration (e.g. transforming growth factor beta 1 (TGF-β1) ^[31,32] and stromal-derived factor 1 [SDF-1] ^[33,34]), we first tested whether WT-ASCs promote cancer cell invasion via paracrine signaling. To this end, WT-ASC spheroids were formed and used to collect ASC-conditioned media (Figure 1e, i). However, ASC-conditioned media had little to no effect on the invasion of collagen-embedded MCF10AT1 spheroids, suggesting that ASC-secreted soluble factors alone are not sufficient to promote MCF10AT1 invasion into dense collagen. Next, we tried mixing WT-ASCs into the collagen bulk prior to encapsulating the MCF10AT1 spheroids to mimic the distribution of ASCs in interstitial ECM. While tumor cells did not invade significantly under these conditions, small invasions were observed when ASCs were located immediately adjacent to the tumor spheroid (Figure 1e, ii, arrow). As direct cell-cell contact with CAFs can promote migration and invasion ^[21], we next formed MCF10AT1-WT-ASC co-culture spheroids to increase direct cell-cell contact between both cell types. Indeed, this approach led to robust tumor cell invasion into collagen, whereby increasing the ratio of WT-ASCs to tumor cells correlated with increased invasion (Figure 1e, iii). Tumor cells invaded as groups of cells and were led by ASCs that stained positive for the myofibroblast marker α-smooth muscle actin (α-SMA). ^[15,31,35] Interestingly, co-culture spheroids in which tumor cells were intermixed with ASCs isolated from human breast adipose tissue similarly promoted invasion (Figure 1f) suggesting that our data are of potential relevance to humans. Collectively, these data indicate that ASCs promote collective invasion of premalignant tumor cells via direct cell-cell contact rather than secreted factors alone, and that myofibroblasts play a role in this process. Given the abundance of ASCs in adipose tissue stroma ^[35], all subsequent experiments using co-culture spheroids were performed with a 1:1 ratio of tumor cells: ASCs.

2.2. Obese ASCs promote tumor cell collective migration more effectively than lean ASCs

Since ASC-mediated invasion of tumor cells appeared to be related to α-SMA positive cells (Figure 1d) and because obesity-associated adipose tissue contains increased levels of α-SMA-positive myofibroblasts ^[7], we next compared invasion of lean versus obese ASCs, and the ability of both cell types to promote collective invasion of MCF10AT1 cells. ASCs were isolated from age-matched WT and leptin-deficient B6.Cg-Lepob/J (ob/ob) mice, a commonly used genetic mouse model of obesity that has comparable interstitial fibrosis as diet-induced mouse models of obesity (Figure 2a). ^[7,36] Initially, we characterized the invasion pattern of WT and ob/ob ASC mono-culture spheroids (Figure S1b). When embedded into dense (6 mg mL⁻¹) collagen gels, both WT- and ob/ob-ASC spheroids invaded readily and featured α-SMA positive cells at the leading edge of invasions (Figure 2b, i). Yet ob/ob-ASCs migrated over longer distances as compared to WT-ASCs. Because α-SMA-positive myofibroblasts have been described to invade using both ECM degradation and contractility-dependent fiber alignment ^{[12,19,20,37]}, we next tested whether obesity impacts how ASCs preferentially use these migration patterns. Confocal reflectance image analysis of ASC invasion associated with fiber contraction (Type I) versus cell-sized collagen degradation pores (Type II) revealed that obese ASCs explore cell contractility-dependent Type I invasion more frequently than lean ASCs (Figure 2b, ii). Correspondingly, lean ASCs had an increased tendency to form large degradation pores relative to obese ASCs (Figure S2a). These results suggest that obese ASCs are more invasive relative to lean ASCs due, in part, to altered migration mechanisms.

Figure 2: Obese ASCs promote collective tumor cell migration more effectively than lean ASCs.

a. Schematic showing isolation of ASCs from WT and ob/ob mice. b. Confocal image analysis of WT- and ob/ob-ASC mono-culture spheroid migration into collagen. (i) Maximum projection of IF images (left) and corresponding single slice (right) showing ASCs migrating out of spheroids into collagen (right). Comparison of WT- and ob/ob-ASC migration distances. Scale bars represent 200 μm for the maximum projections and 100 μm for the single slice images. (ii) Zoomed-in images (20 μm scale bar) of individual cells associated with contracted collagen (Type I migration) and adjacent to degradation pores (Type II migration) and corresponding quantification. c. IF images showing invasion of co-culture spheroids into collagen. Zoomed-in images of α-SMA+ ASCs leading the invasion of GFP+ MCF10AT1 cell clusters. Quantification shows the sizes of protruding invasion strands and the number of ASCs leading each strand. 200 μm scale bar for overview and 100μm scale bar for zoomed-in images. d. Quantification of ASC-led invasions across the MCF10A series, and with MDA-MB231 and PDX-derived organoids (X indicate missing data points due to lack of spheroid formation as shown in Figure S3a) as measured via brightfield image analysis. Student t-tests were used for comparisons between two groups and two-way ANOVAs were used for comparisons between three or more groups. *p < 0.05, ## or **p <0.01, ###p < 0.001 and ****p<0.0001.

To test whether these differences translate to ASC-mediated changes of tumor cell invasion, co-culture spheroids composed of MCF10AT1 and either lean or obese ASCs were formed (Figure S1b) and embedded into collagen. Interestingly, ob/ob-ASCs promoted the collective invasion of tumor cells more effectively than WT-ASCs as ob/ob-ASC-guided invasions were associated with fewer ASCs and were larger than WT-ASC-guided invasions (Figure 2c). Indeed, live-cell imaging using spinning disk confocal microscopy confirmed that ASCs were the driver of collective invasion as they migrated out of the spheroids first, while GFP-MCF10AT1 cells only followed subsequently (Figure S2b, Supplemental video 1, 2). This pro-invasive effect of ob/ob-ASCs was not limited to the MCF10AT1 cell line as isogenic MCF10DCIS.com and MCF10CA1a of the MCF10A progression series responded similarly (Figure 2d, S3). Moreover, co-culture spheroids of triple negative MDA-MB231 cells or PDX-derived cells also migrated more when co-embedded with ob/ob-ASCs versus WT-ASCs although both cell types were unable to form spheroids in mono-culture (Figure S3). These results suggest that obese ASCs can stimulate the invasion of premalignant and malignant tumor cells into interstitial, dense collagen more effectively than their lean counterparts.

2.3. ASC-mediated tumor cell invasion correlates with MMP-dependent collagen remodeling

The formation of physical guidance cues due to MMP-mediated degradation of collagen contributes to CAF-mediated migration of tumor cells. ^[38,39] Accordingly, we have found that both lean and obese ASCs can form degradation pores when migrating into collagen (Figure 2b, ii), and we have previously shown that similar pores promote invasion of other cells. ^[40] To more directly investigate the importance of ASC-dependent MMP activity to tumor cell migration, we added the broadband MMP inhibitor Batimastat ^[41] to both mono- and co-culture spheroids embedded in collagen. Batimastat decreased growth-mediated expansion of MCF10AT1 mono-culture spheroids, but had no effect on invasion (Figure 3a). In contrast, MMP inhibition significantly reduced, but did not fully abrogate invasion of either WT- or ob/ob-ASC mono-cultures. Similarly, addition of Batimastat to tumor cell co-cultures with WT- or ob/ob-ASCs reduced, but did not completely block invasion. Interestingly, MMP-inhibited co-cultures containing ob/ob-ASCs invaded more than cultures containing WT-ASCs (Figure 3a, b) suggesting again that alternative mechanisms may contribute to ob/ob-ASC-mediated invasion.

Figure 3: ASC-mediated collective invasion of tumor cells depends on MMPs.

a. Brightfield images of mono-culture (MCF10AT1, WT ASCs and ob/ob ASCs) invasion patterns with or without Batimastat over a period of four days. Image analysis of bright-field images showing differences relative to control groups (untreated) across all conditions. Scale bars represent 100 μm. b. Brightfield images of WT-ASC and ob/ob-ASC co-culture spheroids. Scale bars represent 100 μm. Quantification of invasion areas from bright-field images with Batimastat treatment. c. Representative single slice images for WT-ASC co-culture and ob/ob-ASC co-culture spheroids with or without Batimastat treatment to visualize different invasion patterns. d. Representative single slice image and zoomed-in views of non-invasive (top) and invasive regions (bottom) of a Batimastat-treated ob/ob-ASC co-culture spheroid. Collagen intensity distribution along the corresponding dashed blue line are shown on the right. Scale bar represents 100 μm. Two-way ANOVA with multiple comparison used for all graphs.

As CAFs can mediate tumor cell invasion via contractility-dependent ECM remodeling ^[20] and because obesity increases ASC differentiation into contractile myofibroblasts ^[7], we hypothesized that obesity-dependent changes in ASC contractility may explain MMP-independent differences in cell migration between ob/ob- and WT-ASCs co-cultures. To address this possibility, we first visualized the microarchitecture of collagen in Batimastat-treated ASC-MCF10AT1 co-cultures using confocal reflectance microscopy. While invading control spheroids exhibited significant remodeling of collagen networks, the respective MMP-inhibited cultures formed a dense ring of collagen around the spheroid edges consistent with growth-induced collagen compression (Figure 3c). ^[42] In clinical samples, collagen densification and alignment parallel to the tumor border has been classified as tumor-associated collagen signatures-2 (TACS-2) and correlates with less invasive tumors in patients. ^[14] Accordingly, dense collagen at the spheroid boundary prevented cell migration in our studies, while regions lacking this dense layer of collagen were readily invaded by cells (Figure 3d). Interestingly, the continuity of the spheroid-associated collagen ring appeared more compromised in Batimastat-treated ob/ob- versus WT-ASC co-cultures (Figure 3c, right) suggesting that the invasion of MMP-inhibited spheroids may be related to differences in ECM remodeling between lean and obesity-associated ASCs and possibly their intrinsic variations of cell contractility.

2.4. ASC-mediated tumor cell invasion correlates with MMP-independent collagen contraction

To more directly test the potential link between ob/ob-ASCs and contractile ECM remodeling, we performed label-free time-lapse imaging of collagen fiber displacement and cell migration using a custom-built spectral domain optical coherence tomography (SD-OCT) system. ^[27] 3D OCT images were reconstructed using temporal speckle contrast to separate cells from surrounding collagen fibers. This approach allowed us to simultaneously monitor changes of collagen expansion/contraction and cell migration over extended periods of time (Figure 4a). Consistent with a lack of migrating cells, MCF10AT1 mono-culture spheroids did not yield significant collagen fiber displacement. A trend towards slightly increased values of collagen displacement at later time points can be explained by spheroid growth and the corresponding compression of the surrounding matrix (Figure 4b). In contrast, both WT- and ob/ob-ASC mono-culture spheroids contracted the surrounding collagen, whereby ob/ob- were slightly more effective than WT-ASCs. (Figure 4b, Supplemental video 3) When combined with tumor cells, ob/ob-ASCs contracted collagen more than co-culture spheroids with WT-ASCs or ob/ob-ASC mono-cultures. No difference was noted between WT-ASC mono- and co-cultures (Figure 4c, Supplemental video 4). Consistent with previous reports, these findings suggest that tumor cells can promote ASC myofibroblast differentiation via paracrine signals ^[43] and that ob/ob-ASCs may respond more to these factors than WT-ASCs. Indeed, a co-culture experiment in which ob/ob ASCs and MCF10AT1 mono-culture spheroids were embedded into collagen separately, but could communicate via soluble morphogens (Figure S4), confirmed that ob/ob-ASCs invade more effectively in the presence of tumor cell-secreted factors. ASC-secreted factors had no effect on tumor cell migration, recapitulating our findings using ASC-conditioned media (Figure 1d). Collectively, these data suggest that increased invasion of MCF10AT1 co-cultures with ob/ob- versus WT-ASCs is associated with local contractile deformation of collagen fibers.

Figure 4: TF-OCM imaging revealed a correlation between ASC-induced collagen fiber deformations and tumor cell invasion.

a. Schematic outlining OCT quantification method. The line plots in b-d were computed from the median radial displacement of collagen fibers, as measured within the region indicated by gray surface, located 150 μm from the initial surface of each spheroid, and spanning +/− 15 degrees above/below the spheroid equator. Positive radial displacements indicate expansion with respect to the spheroid core, while negative radial displacements indicate contraction. Collagen displacements were measured via the tracking of collagen fibers. As an example, the right-hand panel depicts collagen fibers at time t=0 hours (green) and time t=24 hours (purple). b. Representative images of local collagen displacement in the vicinity of mono-culture spheroids at time t=24 hours. Median radial displacements (as measured from the region illustrated in a) are plotted over time on the right, and depict the full range (minimum to maximum value) of the data obtained across all spheroids of the given type. Both WT ASCs and ob/ob-ASCs induced greater deformations than MCF10AT1 spheroids. c. Representative images of local collagen displacement in the vicinity of co-culture spheroids at time t=24 hours, with plots of median radial displacement on the right. d. (Top) Representative images of local collagen displacement in the vicinity of ob/ob-ASC co-culture spheroids at time t=24 hours, under control and Batimastat conditions, with plots of median radial displacement on the right. (Bottom) Corresponding morphology of the spheroids shown above. Color encodes depth over a range of +/–100 μm with respect to the spheroid center. e. Expanded view of the inset regions shown in d. The green and red insets (top) depict regions with greater local invasion than the regions corresponding to the blue and yellow insets (bottom). Median radial displacements within these regions over time are shown at the right. All scale bars represent 200 μm.

Next, we tested whether MMP-inhibited co-cultures of MCF10AT1 with ob/ob-ASCs also cause local collagen contraction by analyzing collagen displacement in the presence and absence of Batimastat and correlating these results with cell invasion. While Batimastat globally reduced collagen deformation, some regions of the surrounding matrix exhibited significant residual collagen contraction (Figure 4d, Supplemental video 5). Interestingly, in both control and MMP-inhibited cultures, regions with high collagen deformation correlated with more cell invasion, suggesting a functional link between both. To determine the relationship between collagen contraction and invasion in more detail, we quantified collagen displacement in two representative regions of low and high invasion for each control and Batimastat-inhibited culture (Figure 4d, e). Again, more invasive regions (Figure 4d, e: green versus blue and red versus yellow boxes) were associated with more collagen deformation (Figure 4e: green versus blue and red versus yellow graphs) regardless of condition. The non-invasive region of Batimastat-treated spheroids exhibited collagen expansion over time, while the invasive region contracted in a manner that was comparable in magnitude to invasive regions of the control spheroids (Figure 4e). These results indicate that an MMP-independent, collagen contraction-based mechanism may play a role in ob/ob-ASC-mediated tumor invasion.

2.5. MMP activity and contractility are co-requisites for invasion of ob/ob-ASC mono-cultures

Given that ob/ob-ASC-mediated tumor cell invasion appeared dependent on ECM contraction, we next determined which effect ob/ob-ASC contractility has on the invasion of the ob/ob-ASCs themselves. To this end, we pharmacologically inhibited Rho-associated protein kinase (ROCK), a key regulator of myosin light chain (MLC) phosphorylation and actomyosin contractility^[19,44], in ob/ob-ASC mono-cultures either in the presence or absence of Batimastat. A collagen contraction assay revealed that ob/ob-ASCs contract dense collagen (6 mg mL⁻¹) disks to less than 50% of their initial size over a period of 4 days. When added individually, Batimastat or Y27632, a specific inhibitor of p160ROCK, significantly reduced contraction; a slight but not significant advantage of co-inhibiting MMPs and ROCK was noted. We also tested whether Y27632 and/or Batimastat affected contractility of individual ob/ob-ASCs using a previously established microwell platform in which cells can remodel collagen fiber networks under conditions that prevent fiber detachment (Figure S5a). ^[40] Control ob/ob-ASCs expressed α-SMA, formed robust stress fibers, and were associated with local collagen fiber contraction as defined by increased matrix density. In contrast, Y27632- or Batimastat-treated ASCs expressed less α-SMA, formed fewer stress fibers, and lacked local collagen contraction. Combined treatment with Y27632 and Batimastat was slightly more effective, but not to a significant extent. These data suggest that both contractility and MMP activity are important for ob/ob-ASC interactions with collagen. However, when analyzing ob/ob-ASC invasion from collagen-embedded spheroids, only MMP but not ROCK inhibition appeared to reduce invasion (Figure 5b). Nevertheless, confocal imaging revealed that ROCK inhibition led to distinct changes of cell morphology (Figure 5c, i). Whereas both control and Y27632 groups invaded robustly, Y27632 reduced F-actin and α-SMA levels compared to controls (Figure 5c, ii, S5b). Furthermore, Y27632 reduced contractility-dependent migration (Type I), while increasing degradation-dependent migration (Type II) relative to control conditions (Figure 5c, iii). Consistent with these findings, Y27632 treatment led to the formation of larger degradation pores compared to controls (Figure 5c, iv). These data suggest that contractility contributes to the invasion of ob/ob-ASCs and may explain the residual migration of Batimastat-treated cultures previously described. Indeed, co-inhibition of both MMP and contractility significantly affected filopodia formation and reduced the number of invading ob/ob-ASCs (Figure 5c, i, v).

Figure 5: Collagen invasion of ob/ob ASC mono-cultures depends on both MMP activity and contractility.

a. Representative images of collagen contraction by ob/ob-ASC mono-cultures with different treatments: Y27632 (30μM), Batimastat (20 μM), or a combination of both denoted as co-inhibition. b. Effects of inhibitors on ob/ob-ASC invasion from mono-culture spheroids as visualized and quantified from brightfield images between day 0 and day 4. Scale bars represent 100 μm. c. (i) 3D renderings and maximum intensity projections visualizing qualitative differences in invasion patterns and filopodia formation between conditions. Scale bars represent 100 μm. (ii) Quantification of F-actin and α-SMA fluorescent intensity in the presence and absence of Y27632. (iii) Image analysis of confocal images showing a Y27632-dependent decrease of collagen contraction-associated invasion (type I) while increasing degradation-associated invasion (type II). (iv) Quantification of degradation pore area for collagen-embedded ob/ob-ASC monoculture spheroids with and without Y27632. (v) Quantification cell invasion as determined by the number of nuclei outside the spheroid cores. Two-way ANOVA with multiple comparisons used for all graphs.

Given these results, similar inhibition studies were performed on ob/ob-ASC co-culture spheroids. While Y27632 alone had only small effects on ob/ob-ASC mono-culture invasion, it significantly inhibited invasion of ob/ob-ASC co-cultures (Figure 6a, S6). This difference may be explained by our findings that tumor cells secrete factors that promote ASC contractility ^[43] which may consequentially promote invasion (Figure S4c). Confocal image analysis further suggested that Y27632 treatment significantly reduced the size of individual invasion strands and impaired the localization of ASCs to the leading edges of these invasions (Figure 6b, c). Similar to previous reports describing CAFs as leader cells that guide collective cancer cell migration through force-dependent heterotypic interactions with tumor cells ^[21], α-SMA positive ob/ob-ASCs led individual tumor cell invasions in ob/ob-ASCs co-cultures (Figure 6b). In contrast, fewer ASCs led collective invasions in Y27632-treated co-cultures and were instead migrating individually or adjacent to tumor cells (Figure 6b, d, S6a). These data suggest that Y27632 treatment may stimulate compensatory mechanisms to enable tumor cell migration in co-cultures. For example, our data above (Figure 5 iii–iv) suggest that ROCK inhibition may promote ASC-dependent matrix degradation and thus, the formation of physical guidance cues. Alternatively, it is possible that Y27632 could promote tumor cell migration independently of ASCs as it can increase cell deformability and thus ameboid migration mechanisms. ^[45] However, Y27632 had no effect on the invasion of MCF10AT1 mono-cultures excluding this possibility (Figure S6b). Collectively, these results imply that ob/ob-ASC contractility contributes to collagen invasion of tumor cells. To test the relevance of this mechanism in the context of MMP inhibition, Y27632 and Batimastat were added simultaneously. Indeed, MMP and ROCK co-inhibition led to reduced invasion of ASCs from the spheroids relative to MMP-inhibited conditions (Figure 6b, S6c). Given that ASCs may promote tumor cell invasion via contractility-based mechanisms, these findings suggest that ob/ob-ASCs may promote invasion of tumor cells via both degradation- and contraction-based ECM remodeling.

Figure 6: Collective invasion of ob/ob ASC co-cultures depends on both MMP activity and contractility.

a. Brightfield images of ob/ob-ASC co-culture invasion with different treatments and corresponding quantification. Scale bars represent 100 μm. b. 3D renderings of confocal images visualizing differences in collective migration patterns between conditions. Maximum projection (scale bar: 200 μm) and zoomed-in images (scale bar: 50 μm) illustrating the morphology of individual invasion strands. c. Volume quantification of individual invasion strands in Y27632-treated and control conditions. d. Quantification of ASC migrating individually, adjacent to or at the leading edge (leading) of invasion strands for the different conditions. e. Schematic of proposed mechanisms contributing to collective tumor cell invasion in the presence of obesity-associated ASCs.

3. Conclusions

Obesity is associated with increased interstitial fibrosis mediated by myofibroblastic ASCs, but which role these cells play in tumor invasion remains unclear. Here, we combined a 3D in vitro model recapitulating aspects of ASC-tumor cell interactions and advanced imaging tools. Our results suggest that ASCs promote collective tumor cell invasion in a cell contact-dependent manner and that obese ASCs are more effective in doing so than their lean counterparts. Obese ASC-mediated differences in invasion were associated with MMP-dependent ECM degradation, but local contraction of collagen also contributed. (Figure 6e) Disrupting Rho-ROCK based contractility and MMPs simultaneously reduced migration more effectively than MMP-inhibition alone.

Different cell types explore varied migration mechanisms depending on their respective microenvironmental composition (i.e., cells, ECM, and biochemical context). Pharmacological interference with MMPs and contractility can impact all of these parameters and, hence, the migration mechanisms tumor cells explore. This should be considered when interpreting our results. Furthermore, the model systems employed here are missing important cellular and molecular component of the obese adipose tissue microenvironment, and future studies will need to evaluate our results under conditions of increased complexity. For example, cancer associated fibroblasts and thus, possibly obesity-associated ASCs can promote tumor invasion via altered fibronectin deposition ^[12] and secretion of cytokines (including SDF-1 and TGF-beta). ^[31,34] Accordingly, our previous work suggests that obesity stimulates fibronectin deposition by ASCs ^[7] and that activated ASCs increase the secretion of factors contributing to elevated tumor malignancy. ^[40,43] Although secreted factors alone were not sufficient to promote extensive invasion in our experiments, cytokines can be sequestered within the ECM (e.g., due to differential binding to fibronectin). Therefore, it is possible that the combination of altered ECM deposition and cytokine release contributed to our findings, which will need to be tested in the future.

Our work was motivated in the context of obesity-associated differences in ASC phenotype, but it has direct relevance to how cell-mediated differences in biomaterials properties (e.g. degradation and mechanical properties including 3D microarchitecture and fiber structure) may promote cancer progression. In particular, ASCs and biomaterials including collagen are commonly used for breast reconstruction therapy following mastectomy. ^[46] However, it remains poorly understood whether or not this approach could contribute to activation of residual tumor cells and thus, increased risk for recurrence. Our results suggest that implanted or resident ASCs mediate structural changes of collagen that correlate with the ability of ASC to stimulate invasion of premalignant tumor cells. Indeed, cellular contractility can lead to the alignment of fibers in collagen matrices ^[22,47] and other materials. ^[48] These structural changes stimulate cellular mechanosignaling, which is key to both ASC myofibroblast differentiation as well as tumor cell aggression. ^[7,10,15] As such our work is of direct relevance to the field of regenerative medicine and will contribute to the design of safe biomaterials for reconstructive therapies in breast cancer patients.

In summary, our studies increase understanding of the functional connections between microenvironmental changes in obesity and breast cancer progression.

4. Experimental Section

Animal use:

All experiments using cells isolated from animals were approved by the Cornell University Institutional Animal Care and Use Committee (IACUC) under protocol number 2009–0117.

Patient-derived tissue and cells:

The patient sample was accessed through the University of Virginia Biorepository and Tissue Research Facility. This sample was selected from a patient with a definitive diagnosis of node-negative, triple negative breast cancer who received no treatment prior to tumor resection. Samples were de-identified before use. All procedures involving human tissue were performed in accordance with the ethical standards of the Institutional Review Board at the University of Virginia under protocol IRB-HSR 13310. To isolate human ASCs, mammary adipose tissue was resected from patients with normal body weight who underwent bi-lateral breast reduction, and human adipose stromal cells were isolated as described below. All procedures were performed according to a protocol (IRB 1510016712) approved by the Institutional Review Board at Weill Cornell Medicine and Cornell University.

Immunohistochemistry of patient sample:

Formalin-fixed, paraffin-embedded sections were deparaffinized in xylene and rehydrated in ethanol. Citrate-based antigen retrieval (Vector Labs) was performed before samples were permeabilized in 0.1% Triton X-100 and then blocked in 4% goat serum. Samples were incubated with a pan-cytokeratin antibody (Thermo) and α-SMA antibody (eBioscience) overnight at 4°C. Samples were incubated with secondary antibodies for 1 hour at RT. Finally, samples were stained with DAPI (Sigma-Aldrich), mounted with Fluoromount-G (SouthernBiotech), and sealed. All antibodies were used at dilutions recommended by the manufacturer for paraffin-embedded tissues. Images were processed using ImageJ and Photoshop.

Cell culture:

Murine adipose stromal cells (ASCs) were isolated from inguinal fat of 10 week-old B6.Cg-Lepob/J (ob/ob) mice and their age-matched C57BL/6J wild-type (WT) controls (Jackson Laboratories) according to previously published protocols. ^[7] Human ASCs were isolated from resected breast adipose tissue of female lean (body mass index [BMI] between 18–25) patients using the same protocol. ^[7] MCF10AT1, MCF10DCIS.com and MCF10CA1a cell lines were obtained from the Barbara Ann Karmanos Cancer Institute. MCF10A and MDA-MB231 cell lines were purchased from ATCC. MCF10AT1 cells were transfected with a commercially available turbo-green fluorescent protein (GFP) vector (Thermo). Successfully transfected GFP+ cells were sorted on a BD FACS Aria cytometer. Organoids established from triple negative patient-derived xenografts (PDX) were received from the Englander Institute for Precision Medicine at Weill Cornell Medicine as 3D cultures suspended in Matrigel and were dissociated into single cells prior to spheroid formation. ^[49]

ASCs were cultured in DMEM/F12 media supplemented with 100 U mL⁻¹ Penicillin-Streptomycin and 10% fetal bovine serum. MCF10A, MCF10AT1, MCF10DCIS.com and MCF10CA1a cells were cultured in enriched DMEM/F12 media supplemented with 5% horse serum, 10 μg mL⁻¹ insulin, 0.5 μg mL⁻¹ hydrocortisone, 100 ng mL⁻¹ cholera toxin, 20 ng mL⁻¹ EGF, and 100 U mL⁻¹ Penicillin-Streptomycin. MDA-MB231 cells were cultured in DMEM media supplemented with 100 U mL⁻¹ Penicillin-Streptomycin and 10% fetal bovine serum. PDX cells were cultured in enriched DMEM/F12 media according to previously published protocols. ^[49] All cell lines were maintained at 37°C and 5% CO₂, with media changes every two days. 10μM Batimastat (Tocris) and 30μM Y27632 (R&D Systems), used for inhibition studies, was supplemented to fresh media every two days. For co-culture experiments, a 1:1 ratio of the respective media for each cell type was used unless otherwise noted.

3D spheroid formation and invasion experiments:

96-well tissue culture plates were coated with 50 μL of 1.5% agarose diluted in DMEM/F12 per well that solidified to form a non-adherent coating. Tumor cells were seeded alone or in co-culture with ASCs in each well of the agarose-coated plate and placed on a rotating shaker (60rpm) overnight to form multicellular spheroids. For invasion analysis, polydimethylsiloxane (PDMS) silicone elastomer (Dow Corning) rings (8mm inner diameter, 10mm outer diameter) were covalently bonded to 18mm glass coverslips following plasma treatment (Harrick Plasma). The inner glass surface of each coverslip was then treated with 1% [v/v] poly-ethylenimine (Sigma-Aldrich) and 0.1% [v/v] glutaraldehyde (ThermoFischer Scientific) to promote collagen attachment. High concentration rat tail collagen I (Corning) was reconstituted and neutralized with sodium hydroxide (NaOH) and 10X DMEM/F12 to a final concentration of 1, 3, or 6 mg mL⁻¹ before the addition of cells or multicellular spheroids. The collagen gels were placed at 4°C for 15 min, 20°C for 15 min, and at 37°C for 15 min to promote the formation of thick collagen fibers. Each gel was then immersed with media and cultured at 37°C for 2–4 days unless otherwise note. Brightfield images were taken every day prior to fixation.

Conditioned media (CM) from mono-culture ASC spheroids (5,000 cells per spheroid) was collected after 24 hours of culture and added to mono-culture MCF10AT1 spheroids as depicted in Figure 1D (i). For Figure 1D (ii), 5,000 ASCs were mixed homogeneously into collagen before the addition of a single mono-culture MCF10AT1 spheroid. Co-culture spheroids were formed of ASCs mixed with MCF10AT1 cells at different ratios (1:1, 1:10, or 1:100). All co-culture spheroids were formed at a 1:1 ratio unless otherwise noted. The duration of experiments was 4 days.

Histology:

Spheroids were fixed with 4% PFA (paraformaldehyde) for 20 minutes and washed in 1X PBS three times. 10–15 spheroids per condition were added to a centrifuge tube and stained with Eosin for identification during later sectioning. Spheroids were then embedded in HistoGel (Thermo), placed in cassettes, and stored in 70% ethanol. Samples were sent to the Cornell College of Veterinary Medicine Animal Health Diagnostic Center for paraffin embedding, sectioning, and staining with hematoxylin and eosin (H&E)

ASC microwell experiments:

Collagen microwells were prepared as previously described.^[40] In brief, PDMS was cast into silicon wafer molds with SU-8 micro-patterning and then demolded. Individual molds were plasma treated and bonded to 24-well plates. The previously described collagen gels were seeded with 2 million ASCs per mL and cast into the microwell molds prior to crosslinking. Inhibitors were added as described and media was changed every two days until samples were fixed for imaging.

Immunofluorescence staining of samples:

Samples were fixed with 4% PFA for 20 minutes and washed in 1X PBS three times. Samples were then permeabilized with 0.05% Triton X-100 (Alfa Aesar) in 1X PBS and blocked with 1% BSA (Thermos Fisher Scientific) in 1X PBS before incubation with a primary antibody against α-SMA (Abcam) overnight at 4°C. The following day, samples were incubated with AlexaFluor 488 or 568 secondary antibodies, Alexa Fluor 647 Phalloidin, and 4’,6-diamidino-2-phenylindole (DAPI) for one hour at room temperature. After staining, samples were washed three times with 1X PBS. Samples were stored at 4°C in the dark prior to imaging.

Confocal microscopy and image analysis:

Whole field of view stacks of spheroids or representative zoomed-in images (n=3 per sample) were acquired on Zeiss LSM 710, upright 880, and inverted 880 microscopes. C-Apochromat 10x/0.45 W, W Plan-Apochromat 20x/1.0 Korr DIC M27 75mm, LD LCI Plan-Apochromat 25x/0.8 Imm Korr DIC M27, C-Apochromat 40x/1.20 W Korr M27, and C-Achroplan 32x/0.85 W Korr M27 objectives were used for imaging various experimental set ups. Collagen fibers were visualized with either reflectance imaging or second harmonic generation imaging. 512 pixel × 512 pixel images were acquired individually or 2 by 2 tile images were stitched together to capture an overview image for each sample. Fluorescence channels were set up using smart set up options in Zeiss Zen software and customized to add either a reflectance or second harmonic generation (SHG) channel. Images were averaged 4 times during acquisition to reduce background noise. Time-lapse images of co-culture spheroids were acquired with a 4x/0.16 objective on an Andor Revolution Spinning Disk Confocal with one channel for brightfield and one channel for GFP signals. Z stack images (10 μm steps) were acquired at 10 min intervals. The brightfield channel is presented as a single focus plane and the fluorescence channel is presented as a 2D maximum intensity projection of each stack. The videos were processed using the IDL image visualization language (L3Harris Geospatial). At least three spheroids were imaged per condition and four representative images were acquired per sample. Experiments were repeated three times. All images were processed in ImageJ, MATLAB, and Imaris. Detailed methods for image contrast enhancement are described in the image presentation section. Microsoft Excel and GraphPad Prism were used for subsequent data analysis and plotting purposes.

Measurement of collagen pore size and intensity distribution:

3D z-stack images of nuclei, F-actin, and collagen were acquired for pore size analyses. Degradation pores were defined as empty collagen regions that co-localize with cell bodies. If the pore size was bigger than the nuclei of a given cell, it was defined as type II migration. If there was no visible degradation of collagen near the cell, it was defined as type I migration. Each experimental group contained at least three samples with representative 3D z-stack images for quantification. For analysis of collagen intensity distribution, a line was drawn along the diagonal direction of representative reflectance images. The distribution of collagen intensity was calculated using the “plot profile” function in ImageJ and re-plotted in GraphPad.

Image analysis of invasion:

Brightfield images were hand traced for boundaries in ImageJ to determine the cross-sectional area of spheroids. The percent change in cross sectional area after invasion was calculated as $\left(\frac{\mathit{\text{final\hspace{0.17em}area}}}{\mathit{\text{initial\hspace{0.17em}area}}}-1\right)*100\%$. Brightfield images of spheroids were taken daily to monitor changes in invasion area over time. To determine the size of individual invasion strands, 2D maximum intensity projections in the XY plane were generated from 3D confocal z-stack images as depicted in Figure 2c. Strands of invading cancer cells were manually traced to determine the average projected invasion area per strand. For Figure 6c, the 3D object counter function in ImageJ was used to determine the 3D volume of invasion strands. Volumes greater than 3,000 μm³ were included for analysis.

To quantify ASC location in co-culture spheroids, 2D maximum intensity projections in the XY plane were generated from 3D confocal z-stack images. ASCs were identified as GFP- cells and their relative locations to GFP+ cancer cells within invasion strands were manually documented. “Leading cells” are defined as ASCs located at the front end of invasion strands; “adjacent cells” are defined as ASCs located along the invasion strands, “individual cells” are defined as ASCs migrating outside the spheroids with no direct contact to invasion strands. Cell counts were pooled and averaged over three spheroids per condition. The relative percentage of each cell type was measured as the number of ASCs of a given type over the total number of ASCs quantified.

Positive pixels and intensity measurements:

2D maximum intensity projections in the XY plane were generated from 3D z-stack images. Core regions of the spheroids were automatically segmented using the Otsu thresholding method in ImageJ for the Batimastat and co-inhibition treatment groups and manually traced in control and Y27632 treatment groups. Masks of the total spheroid projections were then generated using the Triangle thresholding method in ImageJ. The total spheroid and core region masks were imported into MATLAB and used to find the distance from the spheroid perimeter to the core using the bwdist function in the Image Processing Toolbox. A histogram distribution plot was used for data presentation. For intensity comparisons, 2D sum projections in the XY plane were generated from 3D z-stack images. Projections were thresholded with the Otsu method in ImageJ to identify a region of interest (ROI) mask for each channel. ROI masks were then processed with the smooth function and overlaid with original sum projections and the overlapping areas were measured in ImageJ. The number of nuclei per spheroid was processed in the same manner, with a sum projection followed by Otsu thresholding in ImageJ. The total number of nuclei was quantified with the analyze particles function in ImageJ within the same field of view. Three representative images per sample were acquired for analysis.

Maximum projection image presentation and 3D rendering:

3D z-stack images were projected as 2D images in the XY plane using the maximum intensity projection function in ImageJ. 3D z-stack images were also visualized with maximum intensity projections and 3D renderings in Imaris. Images from Figure 1d, 2c, 3c, 5c, and 6b were processed in MATLAB to normalize the contrast of each channel by its 99.99th percentile, followed by a maximum intensity projection. This enhanced contrast was used to aid representation of spheroid morphology. Representative overview images were projected using maximum intensity projection and split into individual channels with the same spectrum of brightness contrast for all samples within each signal in Figure S4b and S5b.

TF-OCM set-up and quantification of deformation:

Time-lapse images of spheroids were acquired on a previously described custom-built spectral domain optical coherence tomography (SD-OCT) microscope^[27] equipped with an Olympus LCPlan N 20X/0.45 IR air immersion objective and upgraded with a motorized stage for parallel imaging of multiple spheroids. For each spheroid, images were acquired every 40 minutes for a total duration of 48 hours, beginning approximately 1 hour after embedding. All spheroids (2–3) for a given experimental condition (mono-culture versus co-culture and control versus MMP inhibition) were imaged in a single session.

3D OCT images were reconstructed using an image formation protocol ^[27] customized for enhanced image stability. Since OCT microscopes record optical scattering contrast, the resulting OCT images simultaneously recorded time-varying spheroid/cell morphology, collagen degradation, and traction force-mediated collagen deformation. The images were separated into computed/synthetic “cell” and “collagen” channels by leveraging the burst acquisition and speckle reduction protocols detailed in a previously published paper. ^[27] The resulting “cell” images were used to generate renderings of spheroid invasions as depicted in Figure 4d.

Time-varying displacements of the collagen matrix were measured with respect to the first image (acquired at t=0 hours) of the “collagen” channel via elastic image registration with the built-in MATLAB function imregdemons. The output displacement field u was given in Cartesian coordinates (i.e., u(r) = 〈u_x (x,y,z), u_y (x,y,z), u_z(x,y,z)〉). Radial components of the measured displacement field were computed from these data via the relation: u_r(r;r₀) = u(r) · (r – r₀)/|(r – r₀)|, where r₀ = 〈x₀, y₀, z₀〉 is defined as the center of the spheroid body at the initial time point. z₀ was determined by the axial depth (z) for which the initial spheroid body displayed the greatest lateral area (i.e., z₀ = arg max A_sph(z)). x₀ and y₀ were obtained from the lateral center of mass of the spheroid body.

The line plots in Figure 4 b–d depict the median radial displacement of collagen fibers as measured across a fixed set of positions, located 150 μm from the surface of the initial spheroid body and lying within +/-15° of its ‘equator’. A precise description of this operation is as follows: Assume the initial spheroid body is approximately spherical with center r₀ and radius R₀ (computed as $R_0=\sqrt{A_{\mathit{\text{sph}}}\left(z_0\right)/\pi }$). The median radial displacement was then obtained via:

$${\widehat{u}}_{r,\Omega }=\text{median} \left[\left\{u_r\left(\mathbf{\text{r}};{\mathbf{\text{r}}}_0\right):\mathbf{\text{r}} \in \Omega \right\}\right],$$

where

$$\Omega =\left\{\mathbf{\text{r}}\in {ℝ}^3:|\mathbf{\text{r}}-{\mathbf{\text{r}}}_0|=R_0+150\mu \text{m}\right\}\cap \left\{\mathbf{\text{r}}\in {ℝ}^3:\frac{|\left(\mathbf{\text{r}}-{\mathbf{\text{r}}}_0\right)\cdot \langle x-x_0,y-y_0,0\rangle |}{|\left(\mathbf{\text{r}}-{\mathbf{\text{r}}}_0\right)|\times |\langle x-x_0,y-y_0,0\rangle |}\ge \text{cos}\left(15°\right)\right\}$$

Mechanical testing of collagen:

TA Instruments DMA Q800 was used to measure the compressive moduli of collagen scaffolds via dynamic mechanical thermal analysis (DMTA). In brief, collagen scaffolds (1 mm thickness and 12 mm diameter) were placed in between two plates and compressed from 0.001 N until 0.1 N at a rate of 0.01 N/min for four tests. Force was calculated from the applied force and strain was calculated from the thickness change of the collagen scaffolds. Data from each DMTA test/sample (n = 4) was fit using local polynomial regression fitting in R. For the range of strain values that was measured for all tests/samples, the predicted stress values based on curve fitting were used to calculate mean and standard deviation. The summarized data are plotted as mean +/− standard deviation as smooth fit and overlaid with individual test measurements from four tests. Elastic moduli were calculated from the slope of the stress-strain curves (between 0– 6% strain) and averaged over four samples.

Collagen gel disk contraction assay:

6 mg mL⁻¹ collagen gels were prepared as described above. Individual ASCs were mixed with collagen before crosslinking at 2 million cells per mL. 100 μl of this solution was added to individual wells of 96-well plates and allowed to as described for the 3D spheroids invasion assays. Collagen gels were then gently removed from the wells and transferred to a new 24-well plates for four days of culture with or without treatment. Contraction was monitored with photos taken over the plates every day until fixation. Percent contraction was measured by normalizing day 4 surface area to day 0 surface areas across all conditions.

Graphics:

Schematic illustrations were created using BioRender.

Statistical analysis:

All experiments were performed with at least 2–3 independent biological replicates and each of these experiments was performed with at least 3 technical replicates unless otherwise stated. Student t-tests, two-way ANOVAs, and multiple comparison tests with Tukey post-hoc analysis were used to compare differences between samples as indicated in the figure legends. P-values less than 0.05 were considered significant. All data were plotted as mean +/− standard deviation (SD). Confocal microscopy and OCT images as well as the corresponding image analysis are representative of one experiment.