Reprogramming of Normal Fibroblasts into Ovarian Cancer-Associated Fibroblasts Via Non-Vesicular Paracrine Signaling Induces an Activated Fibroblast Phenotype

Abstract

Cancer-associated fibroblasts (CAFs) are key contributors to ovarian cancer (OC) progression and therapeutic resistance through dysregulation of the extracellular matrix (ECM). CAFs are a heterogenous population derived from different cell types through activation and reprogramming. Current studies rely on uncharacterized heterogenous primary CAFs or normal fibroblasts that fail to recapitulate CAF-like tumor behavior. Here, we present that conditioned media from ovarian cancer lines leads to an increase in the activated state of fibroblasts demonstrated by functional assays and up-regulation of known CAF-related genes and ECM pathways. Phenotypic and functional characterization demonstrated that the conditioned CAFs expressed a CAF-like phenotype, strengthened proliferation, secretory, contractility, and ECM remodeling properties when compared to resting normal fibroblasts, consistent with an activated fibroblast status. Moreover, conditioned CAFs significantly enhanced drug resistance and tumor progression. Critically, the conditioned CAFs resemble a transcriptional signature with involvement of ECM remodeling. The present study provides mechanistic and functional insights about the activation and reprogramming of CAFs in the ovarian tumor microenvironment mediated by non-vesicular paracrine signaling. Moreover, it provides a translational based approach to reprogram normal fibroblasts from both uterine and ovarian origin into CAFs using tumor-derived conditioned media. Using these resources, further development of therapeutics that possess potentiality and specificity towards CAF/ECM-mediated chemoresistance in OC are further warranted.

Graphical Abstract

INTRODUCTION

Ovarian cancer (OC) is one of the deadliest forms of cancer in women, where more than 80% of patients develop chemotherapy resistance, resulting in advanced recurrence and eventually death [1]. It is usually diagnosed at a late stage and a high tumor grade, with a very low rate of women being diagnosed at an early stage. Treatment barriers in ovarian cancer have been increasingly attributed to stromal reprogramming and activation, extracellular matrix (ECM) remodeling, angiogenesis, and drug delivery [1-3]. A known source for these barriers is stromal cell types, which are a main contributor to hallmarks of cancer like tumor recurrence, metastasis, and chemotherapy resistance [3-5]. There is evidence that stromal cells are recruited by the tumor microenvironment (TME) through tumor and stromal cell crosstalk, supplying tumor-associated stromal cells to a developing tumor, and contributing to its proliferation [4, 6]. Stromal cells are the main secretors of the ECM providing a structural framework of connective tissue throughout the body. In tumors, the most important type of stromal cells are cancer-associated fibroblasts (CAFs), a type of activated fibroblast.

CAFs may be derived from several cell types including fibroblasts, mesenchymal stem cells (MSCs), or epithelial and endothelial cells through activation and reprogramming by cancer cells, epithelial to mesenchymal transition (EMT), and endothelial to mesenchymal transition (EndMT), respectively [7]. Stromal cell types are recruited by chemokines, cytokines, and growth factors to specific cancer sites and the dysregulated ECM at these sites activates them into CAFs. These different sources of activation lead to heterogeneity of CAFs and a variety of different subtypes, many of which have not been clearly identified yet. CAFs are key contributors to tumor progression and therapeutic resistance through the remodeling of the tumor ECM composition and structure [8]. An example of this is CAFs ability to induce EMT through the upregulation of transforming growth factor beta (TGF-β) [9]. CAFs are inflammatory and are mainly known to be pro-tumorigenic, largely attributed to their production of cytokines like interleukin-6 (IL-6) and CC-chemokine ligand 2 (CCL2/MCP-1) that interfere with T cell function [10-12]. Thus, CAFs are considered to be notable players in promoting many aspects of tumor function and a promising targeted therapeutic option.

Current models studying the stromal cell influence on OC rely heavily on normal fibroblasts including NIH3T3 fibroblasts, non-ovarian tissue origin such as dermal and lung fibroblasts or immortalized ovarian fibroblasts (TRS3 and NOF151-hTERT) [13-16]. Unfortunately, the effects of CAFs and normal fibroblasts in ECM remodeling, OC progression, and drug resistance are significantly different [17-20], making normal fibroblasts a less ideal source. Alternatively, primary CAFs can be isolated from tumors and cultured in vitro, but their lifespan is finite; they have shown limited expansion capacity and the use of cells from different patients can result in non-reproducible results associated with the technologies and methodologies from each laboratory, and the lack of characterization and heterogeneity between patients [21, 22]. Therefore, there is a need for translational based approaches that allow for the reprogramming of normal fibroblasts into CAFs with a reproducible, cost-effective, and clinically relevant approach.

The overall objective of this investigation is to reprogram normal fibroblasts from both human uterine fibroblasts (HUFs) and immortalized ovarian fibroblasts NOF151-hTERT origin (NOFs) into CAFs using ovarian tumor-derived conditioned media, followed by phenotypic and functional characterization of the conditioned CAFs. Our results demonstrate the importance of activated CAFs on several OC processes including contractility, ALDH activity, tumor progression, cytokine profile, and ECM signatures, which further underscore the need for functional assessment of CAFs phenotypes. Mechanistically, we have addressed that this activation occurs via non-vesicular paracrine signaling. In the absence of mechanistic and functional characterization of the reprogramming of normal fibroblasts into CAFs, the development of effective anticancer therapies that mitigate CAF-mediated chemotherapy resistance in OC will likely remain difficult.

RESULTS

Conditioned CAFs Express CAF-Like and Stromal Markers

Normal fibroblasts were cultured in tumor-derived conditioned media from ovarian cancer cell lines (KURAMOCHI and SKOV-3) in order to reprogram them into ovarian cancer-associated fibroblasts (CAFs) (Figure 1A). First, we phenotypically assessed the reprograming of normal uterine fibroblasts into uterine CAFs (uCAFs) by evaluating their cell morphology and surface marker expression. All stromal cells including HUFs, conditioned CAFs, and primary OC-CAFs were monitored routinely for morphological changes over passages. The cells retained their fibroblast-like morphology in culture over passages (Figure 1B) and no morphological changes were found. Cell surface marker expression was assessed for all the stromal cell types, as well as the OC cell lines at early passages (P2-P4) given the limited expansion capacity of the primary CAFs. HUFs, primary OC-CAFs, both conditioned uCAFs, and the adipose MSCs expressed CAF markers FAP and CD29, which are indicators of activated fibroblasts [23, 24]. HUFs, KURA-uCAFs, OC-CAFs, and adipose MSCs showed negative PDGFRα expression, while SKOV-uCAFs exhibited low expression. HUFs and adipose MSCs were negative for α-SMA, while both CM-uCAFs and OC-CAFs display low expression (Figure 1C). HUFs, primary OC-CAFs, both conditioned uCAFs, and the adipose MSCs expressed stromal markers CD90 and CD73, which are classical MSC markers [25], and vimentin, which is highly expressed by various types of fibroblasts [26]. The corresponding relative mean fluorescence intensity (rMFI) was calculated for each marker expressed by each cell type (Figure 1D). All stromal cell types remained negative for the epithelial marker EpCAM/CD326 and the immune marker CD45 (Figure S1A). KURAMOCHI and SKOV-3 OC cell lines were validated for negative FAP expression and positive CD326 expression (Figure S1B). FAP expression was also evaluated in the conditioned uCAFs over passages and confirmed that FAP expression was retained in the later passages up to at least P9 (Figure S1C).

Figure 1. Conditioned cancer-associated fibroblasts (CM-CAFs) express CAF-like and stromal markers.

(A) Schema of the experimental design using tumor-derived conditioned media to reprogram normal fibroblast (uterine or ovarian) into cancer associated fibroblasts. (B) Representative images of 2D cultures of HUFs, KURA-uCAFs, SKOV-uCAFs, and OC-CAFs, scale bar = 400μm. (C) Representative flow cytometry histograms of expression where percent positive cells are represented for CAF-related markers (FAP, CD29, PDGFRα, α-SMA), stromal markers (CD90, CD73, Vimentin), and respective fluorescence minus one (FMO) controls in HUFs, KURA-uCAFs, SKOV-uCAFs, OC-CAFs, and adipose-derived MSCs. Cells from early passages P2-P4. (D) Representative graphs of the relative mean fluorescence intensity (rMFI) for each marker per cell type.

Conditioned Uterine CAFs Possess Enhanced Expansion Capacity, a CAF-Like Cytokine Profile, and Deposit ECM as Activated Fibroblasts

To demonstrate the activation of the reprogrammed conditioned uCAFs, we performed experiments that measured cell proliferation, cytokine secretion, and ECM deposition in HUFs, both conditioned uCAFs, and primary OC-CAFs. While HUFs and OC-CAFs showed limited expansion of less than 2-fold and for limited passages (less than P4), conditioned uCAFs showed significantly enhanced expansion capacity (2-4 fold) compared to primary HUFs and OC-CAFs, as well as expanded proliferation over longer passages (P7) (Figure 2A). When OC-CAFs were exposed to the tumor-derived CM, there was not increased proliferation, highlighting the role of CM in activation and reprogramming of normal fibroblasts. In order to account for the potential influence of ovarian media as the source of activation, we exposed HUFs to ovarian media and it did not enhance or prolong their expansion capacity compared to HUFs in their original media (Figure 2A). Cytokine secretion profile was assessed in all stromal lines, as well as their culture media. IL-6, IL-8, MCP-1, IGFBP-2, osteprotegerin (OPG), VEGF, TIMP-1, MMP-2, and MMP-3 were the most highly secreted chemokines/cytokines by the 4 cell types (Figure 2Bi). Differences in cytokine profile secretion were identified among both uCAF lines, where IL-6, VEGF, and MMP-3 were significantly higher for SKOV-uCAFs than for KURA-uCAFs, or where IGFBP-2 was significantly more secreted by KURA-uCAFs than the other cell types. Importantly, 7 of the 12 analytes studied clearly showed that the uCAFs had significantly higher secretory profile than HUFs revealing a clear reprogramming into CAF-like phenotype. Primary OC-CAFs secrete the same cytokines as the uCAFs, just at lower normalized expression, potentially indicating their prolonged activated status. The same analytes were measured in the corresponding media and although levels of IL-6, IL-8, VEGF, MCP-1, IGFBP-2, and TIMP-1 were present in the conditioned media, those levels were about 3-5-fold lower, ensuring the secretion profiles identified were truly derived from the cells and not their corresponding media (Figure 2Bii). OC-CAFs were the only cell type showing high expression of LIF, but it should be noted that part of it is coming from their different culture media. Collagen type I deposition was assessed by immunohistochemistry to demonstrate that the conditioned uCAFs deposited ECM-like activated fibroblasts. Figure 2C shows that there is higher intensity average of collagen expression in both conditioned uCAFs than in HUFs, with OC-CAFs having the highest collagen intensity average yet comparable to uCAFs, confirming their fibroblast activation.

Figure 2. Conditioned uterine CAFs possess enhanced expansion capacity, CAF-like cytokine profile, and deposit ECM as activated fibroblasts.

(A) Proliferation of HUFs, KURA-uCAFs, SKOV-uCAFs, and OC-CAFs in culture over passages presented as the expansion fold after 7 days for each passage; HUFs were also cultured in ovarian (OV) media and OC-CAFs were also cultured in CM. (B) Normalized expression of cytokines expressed in media and in conditioned media from 4 days cultures of HUFs, KURA-uCAFs, SKOV-uCAFs, and OC-CAFs, respectively, where i) represents cytokine expression in cell conditioned media and ii) cytokine expression in culture media. (C) Representative IHC images of HUFs, KURA-uCAFs, SKOV-uCAFs, and OC-CAFs for Collagen type I expression on day 7 with quantified intensity average $(\text{lavg})(\text{Iwp}+\text{Ip}+\text{Isp})∕(\text{Nwp}+\text{Np}+\text{Nsp}))$, scale bar = 100μm. Results are shown as mean ± standard deviation and were analyzed using ANOVA where *p>0.05, **p>0.001, ***p>0.0001, ****p>0.00001.

Conditioned Uterine CAFs Are Implicated in Contractility, Tumor Promotion, and Drug Resistance

Collagen gel contractility assays, tumor progression and drug resistance in 3D co-cultures, ALDH expression, and spheroid formation were assessed to functionally characterize the conditioned uCAFs. HUFs, both conditioned uCAFs, and primary OC-CAFs were examined for collagen contractility capabilities by measuring the percent change of the area of collagen 3D matrices over 24 hours as an indication of fibroblast activation. While HUFs produced a very slight contraction of collagen gels of about 20%, primary OC-CAFs and conditioned uCAFs contracted significantly more compared to HUFs with about 50% and 60-70% contraction, respectively (Figure 3A and B). CAFs have shown clear roles in tumor promotion and induction of drug resistance and in order to assess these mechanisms, physiologically relevant 3D models allowing cell-cell and cell-ECM interactions were used (Figure 3C). Co-culture of cancer and conditioned uCAFs or OC-CAFs enhanced cancer growth significantly about 2.5-3-fold compared to both cancer alone and cancer in 3D co-culture with HUFs for both cancer lines (KURAMOCHI and SKOV-3) (Figure 3D). Similarly, 3D co-cultures of cancer and corresponding conditioned-uCAFs significantly reduced cell killing by carboplatin treatment when compared to cancer alone or cancer co-cultured with HUFs for both lines and have no significant differences with primary OC-CAFs (Figure 3E). Preliminary assessment confirmed that both lines are sensitive to carboplatin in our 3D cultures and 30μM is a concentration close to half maximal effective concentrations for both cell types (Figures S2A and S2B). Spheroid formation, integrity, and composition were also assessed in co-culture conditions. Conditioned uCAFs and OC-CAFs produced spheroids with significantly greater area, length, and integrity than when cancer cells were co-cultured with HUFs (Figure 3F, 3G and Figure S3A. After digestion, spheroids showed similar epithelial/stromal distribution when compared to the co-cultures with the different stromal types, although there is a slight trend of higher epithelial composition with conditioned uCAFs and OC-CAFs compared to co-cultures with HUFs for both OC lines (Figure S3Band S3C). ALDH activity was assessed by aldefluor expression and conditioned uCAFs and OC-CAFs had significantly higher ALDH+ cells than HUFs (Figure 3H and 3I). Finally, we assessed whether the reprograming of HUFs with conditioned cancer media could affect the conditioned uCAFs tumorigenic potential. Neither HUFs, conditioned uCAFs, nor OC-CAFs induce tumors in vivo when implanted in immunodeficient mice. Mice had a normal weight gain over 6 weeks and no tumors were palpable (Figure S3D).

Figure 3. Conditioned uterine CAFs are implicated in contractility, tumor promotion and drug resistance.

(A) Representative images of collagen contractility assay for HUFs, KURA-uCAFs, SKOV-uCAFs, and OC-CAFs at 0 hours and 24 hours. (B) Collagen contraction quantification denoted as percentage of area change after 24 hours. (C) Schema of co-culture experimental design in 3D cultures as 3D matrices and spheroids. (D) Effect of stromal cells (HUF, CM-uCAF, OC-CAF) on cancer growth (KURAMOCHI, SKOV-3) where each cancer cell line was cultured in a 3D matrix for 7 days either alone or in co-culture with HUF, CM-uCAF, or OC-CAF, respectively; shown as the fold of cancer alone (n = 3). (E) Effect of carboplatin (30uM) on cancer survival (KURAMOCHI, SKOV-3) cultured in 3D matrix either alone or in co-culture with stromal cells (HUF, CM-uCAF, OC-CAF) for 7 days, normalized to untreated (PBS control) (n = 3). (F) Effect of stromal cells (HUF, CM-uCAF, OC-CAF) on cancer (KURAMOCHI, SKOV-3) spheroid formation in culture where cancer was co-cultured with respective stromal cells and grown for 21 days by comparing spheroid length and area measurements. (G) Representative Cytation3 images of day 21 KURAMOCHI (top) and SKOV (bottom) cancer spheroids co-cultured with HUFs, KURA-uCAFs, and OC-CAFs, respectively. Scale bar = 300μm. (H) Representative flow cytometry images showing gating of ALDH positive cells by comparing aldefluor stained cells to cells stained with the DEAB control. (I) ALDH activity shown as percentage of ALDH positive cells in HUFs, KURA-uCAFs, SKOV-uCAFs, and OC-CAFs. Results are shown as mean ± standard deviation and were analyzed using ANOVA where *p>0.05, **p>0.001, ***p>0.0001, ****p>0.00001.

Conditioned Uterine CAFs Reveal a Transcriptional Signature with Involvement of the ECM/Matrisome

To understand how tumor conditioned media alters the transcriptome of the CM-uCAFs, we compared the transcriptional profiles of HUFs, CM-uCAFs, and primary OC-CAFs by bulk RNA sequencing. We found 748, 790, and 1543 differentially expressed genes (DEG) when KURA-uCAFS, SKOV-uCAFs, and primary OC-CAFs were compared to HUFs respectively (Figure 4A). To identify pathways enriched by the DEGs in conditioned uCAFs and OC-CAFs, gene ontology analysis revealed ECM-related pathways (extracellular matrix structural constituent, extracellular matrix binding, laminin binding, integrin binding, and growth factor activity) to be similarly enriched in both conditioned uCAFs and OC-CAFs when compared to HUFs, pathways relevant to provide structural support to the ECM [27, 28] (Figure 4B). When the DEGs were compared we found a 29%, 28%, and 14% overlap of the genes for KURA-uCAFs, SKOV-uCAFs, and OC-CAFs respectively (Figure 4Ci). Significantly, out of the total DEGs produced, 84, 92, and 186 belonged to the matrisome, an ensemble of genes encoding for ECM-related genes, clearly highlighting the implication of CM-uCAFs and OC-CAFs in the matrisome/ECM. Importantly, mesenchymal subtypes characterized by CAF presence have been linked to a worse outcome in OC [29]. When tumors of mesenchymal subtype were compared to HUFs and the generated DEGs were compared to the DEGs from CM-uCAFs to HUFs, we found that a total of 371 overlapped genes accounted for 60% of the total DEGs for the CM-uCAFs, clearly highlighting the relevance of the CM-CAFs-like transcriptional profile in patient outcomes (Figure 4Cii). Moreover, the stromal cells were ranked according to three distinct CAF scores [30-32] and robustly showed that both CM-uCAFs and OC-CAFs significantly ranked higher for CAF scores than HUFs (Figure 4D). Similarly, considering the clear implication of CAFs in ECM remodeling, three ECM scores were produced [33, 34], and again both CM-uCAFs and OC-CAFs significantly ranked higher for ECM scores than HUFs (Figure 4E). Comparably, some of the DEGs that were significantly expressed in the reprogrammed CAFs compared to HUFs are matrisome/ECM related including IGFBP3, ANGLPTL4, SPON1, SLPI, VTN, and VWA1 (Figure 4F) [35-39]. Conditioned media induced profound changes in the transcriptome of the HUFs, particularly in genes significantly enriched as part of pathways, such as growth factors like TGF-β, collagens, the ECM/matrisome, and immunomodulatory cytokines, and as anticipated we see upregulation of these genes in the CM-uCAFs and downregulation in the HUFs (Figure 4G).

Figure 4. Conditioned uterine CAFs reveal a CAF-like transcriptional signature with involvement of the ECM/matrisome.

(A) Volcano plots of significant upregulated and downregulated genes from bulk RNA-seq results comparing OC-CAFs, SKOV-uCAFs, and KURA-uCAFs to HUFs, respectively. Red = p-value and Log2 fold change, Blue = p-value, Green = Log2 fold change, and Gray = Not significant. (B) Dot plots of significantly enriched GO:MF terms in differentially expressed genes from bulk RNA-seq comparing OC-CAFs, SKOV-uCAFs, and KURA-uCAFs to HUFs, respectively. (C) Venn diagrams comparing the overlap of differentially expressed genes (DEGs) between i) OC-CAFs versus HUFs, SKOV-uCAFs versus HUFs, and KURA-uCAFs versus HUFs and between ii) DEGs from patients with mesenchymal subtype versus HUFs and CM-uCAFs versus HUFs; number in parentheses denotes number of matrisome ECM-related DEGs. (D) CAF score of HUFs, KURA-uCAFs, SKOV-uCAFs, and OC-CAFs when scored by GSVA using three different CAF-related gene sets. E) ECM score of HUFs, KURA-uCAFs, SKOV-uCAFs, and OC-CAFs when scored by GSVA using three different sets of matrisome ECM-related genes, respectively. (F) Heatmap of differentially expressed genes regulated by reprogramming of normal fibroblasts into CAFs. (G) Heatmaps of differentially expressed genes of KURA-uCAF and SKOV-uCAF (CM-uCAFs) versus HUFs related to collagens, matrisome, TGF-beta (TGFB) signaling, anti-inflammatory activity, and pro-inflammatory activity. Results are shown as the mean and were analyzed using ANOVA where *p>0.05, **p>0.001, ***p>0.0001, ****p>0.00001. All analyses used an FC cutoff of 1.5.

Conditioned Uterine CAFs Preserve CAF Phenotype and Functionality After Removal of Tumor-Derived Conditioned Media

To demonstrate that our conditioned CAFs have been transformed/reprogrammed into CAFs and that they remain CAF-like when removed from tumor-derived conditioned media, we conducted retrieval experiments, comparing uCAFs kept in the conditioned media to uCAFs switched to ovarian media. Conditioned CAFs at post-retrieval passages P1, P3, and P5 (R-P1, R-P3, and R-P5) showed the same contractility potential as conditioned uCAFs (Figure 5A). Moreover, the retrieval uCAFs promoted growth the same as the corresponding conditioned uCAFs with no changes to tumor promotion in both CM-uCAFs lines (Figure 5B). Finally, we also confirmed that the classical CAF marker (FAP) was retained (Figure 5C), as well as stromal CD73 expression (Figure 5D), while epithelial CD326 expression remained negative (Figure 5E) over the 5 post-retrieval passages for both lines.

Figure 5. Conditioned uterine CAFs preserve CAF phenotype after removal of tumor-derived conditioned media.

(A) Representative images of collagen contractility assays at 0 hours and 24 hours for KURA-uCAFs and SKOV-uCAFs at post-retrieval passages P1, P3, and P5. Collagen contraction is quantified as percentage of area change after 24 hours. (B) Effect of KURA-uCAFs and SKOV-uCAFs at post-retrieval passages P1, P3, and P5 on cancer growth (KURAMOCHI, SKOV3) where cancer was cultured in a 3D matrix in co-culture with either the conditioned KURA-uCAFs and SKOV-uCAFs or the retrieval passages of KURA-uCAF and SKOV-uCAF, respectively, for 7 days; shown as the fold of corresponding conditioned-media CAFs (CM-uCAF) (n = 3). (C) Representative flow cytometry histograms of FAP expression in conditioned KURA-uCAFs and SKOV-uCAFs at post-retrieval passages P1, P3, and P5 and corresponding non-retrieved CM-uCAFs. (D) Representative flow cytometry histograms of stromal marker (CD73) expression in conditioned KURA-uCAFs and SKOV-uCAFs at post-retrieval passages P1, P3, and P5 and corresponding non-retrieved CM-uCAFs. (E) Representative flow cytometry histograms of epithelial marker (CD326) expression in conditioned KURA-uCAFs and SKOV-uCAFs at post-retrieval passages P1, P3, and P5 and corresponding non-retrieved CM-uCAFs.

Non-Vesicular Paracrine Signaling is Responsible for Fibroblast Reprogramming into CAFs

In order to determine if the reprogramming of CAFs via tumor-conditioned media is vesicular dependent or not, the conditioned media was subjected to ultracentrifugation where either EVs (both exosomes (Exo) and microvesicles (MV)) were isolated and used as EVs CM or removed and used as EVs free CM (Figure 6A). Initially, validation of removal of EVs was achieved by Nanosight particle analyzer. While EVs CM showed a clear peak at around 180nm for exosomes and a tail over 250nm, the EVs-free CM have zero presence of particles of a relevant vesicle size for both KURA-CM and SKOV-CM (Figure 6B). Both KURA and SKOV EVs free CM and EVs CM cells maintained their fibroblast-like morphology in culture (Figure S4A). Then, both non-vesicular and vesicular CM were compared to full CM from both cell lines in some of the key functional assays demonstrating CAF activation. First, uCAFs were tested for their collagen contractility potential and we identified that for both KURA and SKOV-uCAFs the EVs free CM had a similar contraction to the whole CM (around 60%), and very limited contraction was achieved by the uCAFs in the EVs CM (20%), clearly indicating that the non-vesicular CM induced contractility potential to the uCAFs (Figure 6C and 6D). Second, ALDH activity revealed that EVs-free CM induced higher ALDH+ cells than EVs CM for KURA-uCAFs in line with the previous results (Figure 6E). When uCAFs exposed to either vesicular or non-vesicular CM were co-cultured with corresponding cancer lines in 3D cultures, cancer growth was significantly more promoted in EVs free CM than by EVs CM and similar to the levels of growth seen in the whole CM (Figure 6F). Considering these last results, the same 12 analytes were screened by Luminex in the EVs free and EVs CM, as well as the secretion profile of the uCAFs exposed to the non-vesicular and vesicular CM. As previously seen, the cancer CM cytokine content is much lower than the uCAFs secretion, and it was identified that the majority of the paracrine signaling it’s contained in the EVs free CM (marked in orange) when compared to the EVs CM (marked in purple). When the uCAFs CMs were profiled, most of the cytokine total content was contained in the EVs free CM (marked in red) when compared to the EVs CM (teal) and the same profile previously identified in Figure 2B was replicated. IL-6, IL-8, MCP-1, IGFBP-2, OPG, VEGF, TIMP-1 and MMP-3 were the most highly expressed cytokines from both KURA and SKOV-uCAFs (Figure 6G). Overall, these results indicate that non-vesicular paracrine signaling is responsible for HUFs reprogramming into uCAFs using tumor-derived CM.

Figure 6. Non-vesicular paracrine signaling is responsible for fibroblast reprogramming into CAFs.

(A) Schema of the experimental design using EVs and EVs free conditioned media. Conditioned media was centrifuged to separate the non-vesicular and vesicular fractions containing microvesicles (MV) and exosomes (Exo) of the conditioned media. (B) Concentration of vesicles in particles/mL based on particle size comparing EVs free CM and EVs CM for both KURAMOCHI and SKOV-3. (C) Representative images of collagen contractility assay for KURA-uCAFs, EVs free KURA-uCAFs, EVs CM KURA-uCAFs, SKOV-uCAFs, EVs free SKOV-uCAFs, and EVs CM SKOV-uCAFs at 0 hours and 24 hours. (D) Collagen contraction quantification denoted as percentage of area change after 24 hours. (E) ALDH activity shown as percentage of ALDH positive cells in EVs free KURA-uCAFs, EVs CM KURA-uCAFs, EVs free SKOV-uCAFs, and EVs CM SKOV-uCAFs. (F) Effect of stromal cells (CM-uCAF, EVs free CM-uCAF, and EVs CM-uCAF) on cancer growth (KURAMOCHI, SKOV-3) where cancer was cultured in a 3D matrix for 7 days either alone or in co-culture with CM-uCAF, EVs free CM-uCAF, and EVs CM-uCAF, respectively; shown as the fold of cancer alone (n = 3). (G) Normalized expression of cytokines expressed in KURA EVs free CM and SKOV EVs free CM (in purple) and KURA EVs CM and SKOV EVs CM (in orange) compared to cytokines expressed in the conditioned media from the KURA-uCAF EVs free CM and SKOV-uCAF EVs free CM cells (in red) and KURA-uCAF EVs CM and SKOV-uCAF EVs CM cells (in teal). Results are shown as mean ± standard deviation and were analyzed using ANOVA where *p>0.05, **p>0.001, ***p>0.0001, ****p>0.00001.

Normal Ovarian Fibroblasts Were Reprogrammed into oCAFs and Validated Phenotypically and Functionally as Activated CAFs

To expand on the clinical relevance of our established protocol reprogramming normal uterine fibroblasts into CAFs, the same approach was validated using a non-commercially available immortalized normal ovarian fibroblast line (NOF-151-hTERT). NOF were reprogrammed with the same conditioned media (KURA and SKOV) and were phenotypically and functionally characterized as ovarian-derived conditioned CAFs (oCAFs). NOFs and oCAFs maintained their fibroblast-like morphology in culture (Figure S5A). Cell surface marker expression revealed similar patterns for NOF when compared to HUFs, where they were FAP+, CD29+, PDGFRα-, α-SMA- for CAF-like markers, CD73+, CD90+, Vimentin+ for stromal markers, and negative for CD45 and CD326 immune and epithelial markers, respectively (Figure 7A and Figure S5B). oCAFs retained FAP and CD73 positive expression indicating their activated stromal-like status in culture. While NOFs showed about a 2-fold expansion over passages, conditioned oCAFs had a slight increase in proliferation capacity, which is not surprising for an immortalized line (Figure 7B). When oCAFs and NOFs were functionally characterized, NOFs were found to barely induce collagen contractility when compared to reprogrammed oCAFs showing a 50% collagen contraction (Figure 7C). ALDH activity was significantly enhanced in both oCAFs with about 26% for KURA and 23% for SKOV ALDH positive cells when compared to 9% positive cells for NOFs (Figure 7D). Promotion of cancer growth in 3D cultures revealed that while cancer co-culture with NOFs does not significantly increase tumor growth, oCAFs significantly promoted tumor growth about 2-fold (Figure 7E). Finally, cytokine secretion profile from NOF and oCAFs, as well as NOF media cytokine profile, were evaluated. Similarly to the HUFs and uCAFs cytokine profiles, IL-6, IL-8, MCP-1, IGFBP-2, VEGF, TIMP-1, and MMP-3 were the highest cytokines secreted by oCAFs and significantly more abundant than in NOFs (Figure 7F). Differences between KURA-oCAFs and SKOV-oCAFs were identified as well with some cytokines including IL-6, IL-8, VEGF, MMP-3, and more predominantly for IGFBP-2. In addition, NOFs media did not contain any of the relevant analytes studied. These results suggest that normal ovarian fibroblasts are reprogrammable to functional oCAFs using tumor-derived conditioned media, with a similar pattern seen in the uterine fibroblasts.

Figure 7. Normal ovarian fibroblasts were reprogrammed into oCAFs and validated phenotypically and functionally as activated CAFs.

(A) Representative flow cytometry histograms of expression of CAF-related marker FAP and stromal marker CD73 with respective fluorescence minus one (FMO) controls in NOFs, KURA-oCAFs, and SKOV-oCAFs. (B) Proliferation of NOFs, SKOV-oCAFs, and KURA-oCAFs in culture over passages presented as the expansion fold after 7 days for each passage (mean ± SD, n = 4). (C) Representative images of collagen contractility assay for NOFs, SKOV-oCAFs, and KURA-oCAFs at 0 hours and 24 hours. The graph represents collagen contraction quantification denoted as percentage of area change after 24 hours. (D) ALDH activity shown as percentage of ALDH positive cells in NOFs, SKOV-oCAFs, and KURA-oCAFs. (E) Effect of stromal cells (NOF, SKOV-oCAF, KURA-oCAF) on cancer growth (KURAMOCHI, SKOV-3) where cancer was cultured in a 3D matrix for 7 days either alone or in co-culture with NOF, SKOV-oCAF, or KURA-oCAF, respectively; shown as the fold of cancer alone (n = 3). (F) Normalized expression of cytokines expressed in plain NOF media and the conditioned media from NOF CM, KURA-oCAF CM, SKOV-oCAF CM, and OC-CAF CM cells, respectively. Results are shown as mean ± standard deviation and were analyzed using ANOVA where *p>0.05, **p>0.001, ***p>0.0001, ****p>0.00001.

Non-Vesicular Paracrine Signaling Drives oCAFs Activation

To further validate the results identified in the HUFs reprogramming, vesicular and non-vesicular conditioned media were evaluated in oCAFs activation. oCAFs collagen contractility potential was evaluated for both KURA and SKOV-oCAFs. The EVs free CM had a similar contraction to the full CM (around 50%) and very limited contraction was achieved by the oCAFs in the EVs CM (around 20%), clearly indicating that the non-vesicular CM induced contractility potential to the oCAFs (Figure 8A and 8B), as it was shown with the uCAFs. ALDH activity revealed that EVs-free CM induced higher ALDH+ cells than EVs CM for both oCAFs in line with the previous results (Figure 8C). Cancer growth in 3D co-cultures was significantly higher in EVs free CM than by EVs CM and comparable to the full CM (Figure 8D). When the oCAFs CM were profiled, most of the cytokine total content was identified in the EVs free CM (marked in red) when compared to the EVs CM (teal) and the same profile previously identified in Figures 2B and 6G was replicated. IL-6, IL-8, MCP-1, IGFBP-2, OPG, VEGF, MMP-2, and TIMP-1 were the most highly expressed cytokines from both KURA and SKOV-oCAFs (Figure 8E). Overall, these results indicate that non-vesicular paracrine signaling is responsible for NOFs reprogramming into oCAFs using tumor-derived CM, reinforcing the previous results shown for HUFs.

Figure 8. Non-vesicular paracrine signaling drives oCAFs activation.

(A) Representative images of collagen contractility assay for NOFs, KURA-oCAFs and SKOV-oCAFs, EVs free KURA-oCAFs and SKOV-oCAFs, and EVs CM KURA-oCAFs and SKOV-oCAFs at 0 hours and 24 hours. (B) Collagen contraction quantification denoted as percentage of area change after 24 hours. (C) ALDH activity shown as percentage of ALDH positive cells in EVs free KURA-oCAFs and SKOV-oCAFs and EVs CM KURA-oCAFs and SKOV-oCAFs. (D) Effect of stromal cells (SKOV-oCAF and KURA-oCAF, EVs free KURA-oCAFs and SKOV-oCAFs, and EVs CM KURA-oCAFs and SKOV-oCAFs) on cancer growth (KURAMOCHI, SKOV-3) where cancer was cultured in a 3D matrix for 7 days either alone or in co-culture with SKOV-oCAF, KURA-oCAF, EVs free KURA-oCAFs and SKOV-oCAFs, or EVs CM KURA-oCAFs and SKOV-oCAFs, respectively; shown as the fold of cancer alone (n = 3). (E) Normalized expression of cytokines expressed in the conditioned media from EVs free KURA-oCAFs CM, EVs free SKOV-oCAFs CM, EVs KURA-oCAFs CM, and EVs SKOV-oCAFs CM cells, respectively. Results are shown as mean ± standard deviation and were analyzed using ANOVA where *p>0.05, **p>0.001, ***p>0.0001.

DISCUSSION

In this study, we were able to demonstrate an accessible method to reprogram normal uterine and ovarian fibroblasts into cancer associated fibroblasts (CAFs) using ovarian tumor-derived conditioned media. Using an optimized translatable-based approach the reprogrammed CAFs were validated phenotypically and functionally. This approach relies on paracrine secretion transformation, avoiding complicated co-culture techniques where cells are in physical interaction. This study demonstrates that ovarian tumor-derived conditioned media, specifically non-vesicular paracrine signaling, is sufficient to induce fibroblast reprogramming into CAFs, which is in line with other reports using secreted factors from conditioned media to reprogram normal cells into cancer-associated ones [40-44].

One predicament in defining CAFs is the lack of a pan-specific marker, or a consensus of markers to identify CAFs, adding to the difficulty in differentiating CAFs from other cell types. Consequently, phenotypic and functional characterizations are needed in order to fully characterize CAFs. To date, CAFs are defined as cells that express different levels of mesenchymal biomarkers such as FAP, α-SMA, platelet-derived growth factor receptor alpha (PDGFRα), and vimentin, and lack the expression of markers for epithelial or hematopoietic cells [26, 45]. Our data supports that HUFs and NOFs also expressed FAP, making it an unreliable marker for CAF differentiation. Exposure to serum in culture media has been described to induce a wound healing response in normal fibroblasts, which induces similar genes expressed by CAFs and tumor cells [46]. The preservation of FAP expression in the CM-CAFs over passages should be considered carefully as it could be just an indicator of activation through culture and exposure to serum. With this in consideration, our studies relied upon functional assays to more accurately confirm that the conditioned CAFs display a greater activation status when compared to normal fibroblasts. The results of the functional assays show that although some of the cell surface markers are similar between the normal fibroblasts and both conditioned CAFs (uCAFs and oCAFs), the CM-CAFs were functionally activated while normal fibroblasts were not in terms of collagen contractility, ALDH activity, induction of tumor progression, promotion of drug resistance, and up-regulation of CAF-related genes and ECM pathways. Importantly, functional CAF subsets sustain a unique cytokine expression profile shaping the tumor microenvironment. Significant advances have been made in pancreatic cancer CAF subtyping and characterization of their different functionality [47-51], yet investigations in OC are limited. Recent OC studies have yielded four CAF subtypes, described as CAF S1–S4 based on the differential expression of five fibroblast markers (FAP, CD29, α-SMA, FSP1, and PDGFRβ) [32]. Critically, a CAF-S1 subtype expressing FAP^High SMA^Med-High FSP1^Med-High CD29^Med-High PDGFRβ^Med was associated with worse survival. Similarly in a recent study, OC-CAFs identified as FAP^High SMA^Low subpopulation were related with worse overall survival [52]. The characterized ovarian conditioned uCAFs express FAP^High SMA^Low CD29^High PDGFRα^Neg and FAP^High SMA^Low CD29^High PDGFRα^Low for KURA-uCAFs and SKOV-uCAFs, respectively (Figure 1). Likewise, Tothill et al. [29] reported four distinct subtypes in high-grade serous ovarian carcinoma, with a poor prognosis subtype displaying strong stromal response, correlating with extensive desmoplasia. When the transcriptomic profile of this mesenchymal subtype was compared to HUFs and those transcriptomic changes of the conditioned CAFs against HUFs, we found a significant 60% overlap of DEGs. Additionally, a 2020 study [53] differentiated CAF states between FAP^High and FAP^Low and showed that FAP^High CAFs were more aggressive at promoting chemotherapy, tumor invasion, and proliferation in high-grade serous ovarian cancer patient samples. This is consistent with our FAP^High CAFs, which were shown to enhance tumor growth and drug resistance in 3D co-cultures with OC cells. Another study identified two subpopulations of CAFs in epithelial OC by their expression of seven CAF-related markers in TCGA datasets, associating a subtype cluster B (PDGFRα, PDGFRβ, ACTA2, THY1, PDPN, FAP, and COL1A1) with worse prognosis and advanced stage, while subtype cluster A corresponded to better prognosis and an earlier stage [30]. The conditioned uCAFs showed a significantly higher CAF score than HUFs when scored for these seven genes (Figure 4). Mishra et al. [31] determined that by exposing human bone marrow-derived stem cells (hMSCs) to tumor conditioned media, the MSCs differentiate into a CAF-like phenotype expressing CAF-related genes. Similarly, when we used the top genes upregulated by the conditioned exposed hMSCs to assign a CAF score, KURA-uCAFs and OC-CAFs expressed higher scores against HUFs (Figure 4), supporting the use of tumor conditioned media to reprogram stromal cells into CAFs. These outcomes together suggest that the generated conditioned uCAFs resemble a CAF stromal-like subtype, which was associated with worse outcomes in other clinical studies.

In terms of functional characterization of the CAFs, several avenues were addressed to provide a deep and thorough investigation. An accepted marker of fibroblast activation is collagen contractility [54, 55]. The contractile forces of activated fibroblasts are increased by signals from the ECM and α-SMA [56] and can be determined by a collagen gel contraction assay. Through this assay, our conditioned uCAFs and oCAFs displayed a significant contraction of collagen type 1, which confirmed fibroblast activation. Furthermore, a higher ALDH expression in CAFs has been linked to increased stemness, where cancer-associated MSCs from human epithelial ovarian cancer compared to normal ovary-derived MSCs had a 2-fold increase in the percentage of ALDH+ cells [57], which was consistent with our findings in uCAFS and oCAFs. Another widely known trait of CAFs is their ability to induce tumor promotion and drug resistance in the TME [4, 58]. We confirmed these characteristics in our conditioned CAFs by performing co-culture experiments with cancer cells in 3D gel matrices and spheroids. Various studies have shown that 3D scaffold models work well to mimic in vivo-like drug resistance and tend to cause higher drug resistance in cell lines, thus in order to see clinical drug efficacies in our 3D experiments we determined growth parameters and the concentrations of various chemotherapeutics on individual cell lines in our 3D drug response assays, as previously described [59, 60]. Both SKOV-3 and KURAMOCHI cell lines were found to be similarly sensitive to the drug carboplatin in 3D culture and 30μM is a concentration close to half maximal effective concentrations for both cell types, which is why this concentration was chosen for our 3D co-culture experiments. In both experiments, our results show that when conditioned uCAFs are co-cultured with OC, there is a significant increase in tumor growth and an increased potential to promote chemoresistance when compared to cancer alone and cancer co-cultured with normal fibroblasts (HUFs). Similarly, oCAFs significantly increased tumor growth in 3D cultures. Because of CAFs tumor-promoting role in the TME, co-culturing CAFs with OC likely improved cancer cell attachment, growth, cell-cell interactions, and cell-ECM interactions [61, 62]. These results confirm that both characterized conditioned uCAFs and oCAFs are activated fibroblasts recapitulating key functional CAF subtypes.

The ECM is significantly gaining attention as a key contributor to OC tumor progression and recurrence [63, 64]. To better study the ECM, the matrisome was established as an ensemble of over 1000 ECM-related genes that encode for two groups of ECM proteins, either matrisome-associated or core matrisome. ECM-associated proteins include ECM-affiliated proteins, ECM regulators, and secreted factors known or suspected to bind core ECM proteins, and the core matrisome includes glycoproteins, collagens, and proteoglycans. Stromal cells are the main contributor to ECM secretion and to matrisome expression. We identified that the reprogramming of HUFs into uCAFs induces severe transcriptomic changes in the matrisome over both matrisome-associated and core matrisome categories. Our results indicate that many matrisome-related genes and collagens are upregulated by the CM-uCAFs and downregulated by the HUFs, which supports the role conditioned CAFs play in ECM remodeling [65, 66]. Moreover, our data clearly implicated the conditioned uCAFs with TGF-beta signaling, which is associated with tumor promotion, cancer cell invasion and migration, metastasis, and poor overall survival [15, 67, 68]. Conditioned CAFs were associated with the upregulation of related TGF-beta signaling pathways SMAD and STAT [69-71]. The immunomodulatory role of the conditioned uCAFs was confirmed by the upregulation of pro-inflammatory genes and downregulation of anti-inflammatory genes when compared to HUFs. This coincides with previous research highlighting CAFs role in tumor-promoting inflammation [11, 72]. Of note, the anti-mullerian hormone (AMH) gene is also upregulated by the conditioned uCAFs, which has been recently linked to the modulation of cancer-associated mesothelial cells [41]. These transcriptomic findings confirm a CAF-like transcriptional signature expressed by the OC conditioned CAFs and their potential role in ECM remodeling within the OC microenvironment.

Importantly, the OC conditioned CAFs possess a CAF-like phenotype, strengthened proliferative, secretory, contractility, and ECM remodeling properties when compared to resting normal fibroblasts, as it has been previously documented [17, 73, 74], and their activated fibroblast status was demonstrated in Figures 1 to 4. It is understood that CAFs have pro-tumorigenic functions by secreting tumor-promoting signals including IL-6, IL-8, and MCP-1 [68, 75], and they contribute to the development of resistant cancer phenotypes following chemotherapy by paracrine signaling via cytokines and exosomes [73, 76-78]. The cytokines/chemokines IL-6, IL-8, and MCP-1 are known to function as tumor promotors while also supporting the migration of OC cells [27-29]. Specifically, TIMP-1 is an important tissue inhibitor of matrix metalloproteinases (MMPs) governing ECM remodeling [31] and MMP-3 has been shown to have a regulatory role in ECM remodeling [79]. A 2017 study showed that in prostate cancer progression, CAFs expressed lower levels of MMP-3 than normal fibroblasts where it was upregulated in prostate cancer cells, but downregulated in CAFs, highlighting the expression switching that occurs within the TME [80]. Dkk-1 is known to be both upregulated or downregulated in various cancer types and has been known to function as tumor suppressive [81-83]. As a member of the TNF receptor superfamily, OPG has been found to inhibit tumor cell apoptosis that is induced by TRAIL leading to cancer progression, and it has also been found to be an effective biomarker for distinguishing between healthy tissues and ovarian cancer tissues [84, 85]. Increased expression of VEGF has been associated with worse outcomes in ovarian cancer patients due to its role in tumor angiogenesis [86, 87]. IGFBP-2 overexpression was identified as increasing ovarian cancer cell invasiveness [88] and thus has been cited as a potential biomarker for ovarian cancer [89]. Similarly, in our transcriptomic analysis we found that Dkk-1 and MMP-3 were upregulated by HUFs, while VEGFA, IGFBP-3, and TIMP-3 were upregulated by CAFs, highlighting the expression of cytokines that are crucial to OC progression.

Paracrine signaling is the key communication system that cells use to activate surrounding cells. Extracellular vesicles (EVs) are key carriers of these signaling molecules, but not all paracrine signaling is vesicular dependent [90]. Non-vesicular paracrine signaling occurs through other secretion methods like secreted proteins, cytokines, and soluble factors [91]. By studying CAF-related paracrine signaling, Wessoly et al. [92] concluded that CAF activity is linked to chemotherapy failure in HGSOC and ovarian cancer in general. We speculated that OC conditioned media reprograms normal fibroblasts into CAFs through epigenetic changes driven by cytokines, exosomes (miRNAs), and/or cell-free DNA (cfDNA) present in the conditioned media [93]. To investigate this, we performed experiments to examine the effect of vesicular and non-vesicular paracrine signaling in CAF reprogramming by growing the HUFs and the NOFs in both vesicle conditioned media and vesicle free conditioned media from SKOV-3 and KURAMOCHI, as well as analyzed the CM cytokine content. We found that fibroblasts exposed to vesicle free CM showed CAF-like functionality similar to regular CM, while the vesicle media did not display CAF-like functionality. Our results suggest that non-vesicle paracrine signaling is the main driver of CAF reprogramming in our experimental approach. In ovarian cancer, other studies found that the downregulation of miR-214 and miR-31 partnered with the upregulation of miR-155 was responsible for the activation of normal fibroblasts into CAFs [94]. Also, it has recently been demonstrated that three different exosomal miRNAs derived from triple-negative breast cancer activate stromal fibroblasts [95]. Albrengues et al. observed an epigenetic switch initiated by the cytokine LIF in three cancer types that sustained the pro-invasive phenotype of CAFs [96]. Moreover, Filatova et al. found that cfDNA originating from tumors can penetrate recipient cells and increase oncogenicity [97]. These previous studies suggest epigenetic changes and paracrine signaling as mediators of CAF reprogramming. Further studies will need to elucidate other epigenetic involvement in the reprogramming of conditioned CAFs, but our studies indicate non-vesicular paracrine signaling drives reprogramming of HUFs and NOFs into CAFs using tumor-derived conditioned media.

About 90% of ovarian tumors are epithelial tumors and the most common epithelial ovarian tumors are high-grade serous ovarian tumors, accounting for roughly 70% of all epithelial ovarian carcinomas [98, 99]. Clear cell epithelial cancer is the second most common subtype, accounting for 5-11% of epithelial ovarian carcinomas [100].Treatment is similar for both subtypes, although clear cell leads to a worse prognosis than high-grade due to clear cell chemoresistance to platinum treatment [101]. Our study incorporates the cell line KURAMOCHI, which resembles true high-grade subtype with high confidence, as well as SKOV-3, a frequently used cell line as high-grade but with low suitability score and ARID1A mutations which strongly associate with clear cell [102, 103]. Both these cell lines have successfully reprogrammed HUFs and NOFs into functional activated CAFs; of note their cytokine profile and gene expression profiles in ECM remodeling are similar but with unique changes to each cell type, which should be further investigated. We want to acknowledge that with the four functionally characterized CM-CAFs generated, we cannot recapitulate the high heterogeneity of patients, but we have provided a clinically relevant protocol to reprogram normal fibroblasts into CAFs using tumor-derived conditioned media, and more cell lines should be further investigated and validated using this tool. Moreover, our studies robustly validated the reprogramming of HUFs and NOFs. It has been shown that CAFs originate from local fibroblasts in OC both in animal models and in vitro [104, 105], and we further confirmed that uterine and ovarian fibroblasts can be reprogrammed into activated ovarian CAFs. HUFs were selected as they are commercially available by a trusted provider such as the American Type Culture Collection (ATCC) and due to their gynecological origin. In order to reflect the ovarian location where ovarian tumors spread and further grow, we repeated the reprogramming experiments and key functionality assays and confirmed that conditioned oCAFs generated for normal ovarian fibroblasts resemble the same functional activated status as that when using uterine fibroblasts. Significantly, our results showed that functionally the CM-CAFs are like the activated primary ovarian CAFs. The approach presented in this study provides an excellent opportunity for investigations into ovarian cancer progression and chemoresistance with well-characterized ovarian CAFs, offering an alternative to the use of normal fibroblasts, immortalized lines, or heterogenous primary sources.

To address the innovation of our findings, our results provide four well-characterized reprogrammed CAF lines that were not available before in the scientific community, which successfully proliferate in cell culture, while functionally resembling activated ovarian CAFs with strengthened secretory, contractility, and ECM remodeling properties when compared to resting normal fibroblasts. To our knowledge, this is the first publication that provides a comprehensive functional characterization of reprogrammed CAFs using tumor-derived conditioned media and that mechanistically links the reprogramming to non-vesicular paracrine signaling. Our publication additionally delivers an easy to follow and reproducible translational based protocol to reprogram and characterize CAFs in culture.

Taken together, our study demonstrates that the reprogramming of normal fibroblasts into ovarian CAFs with tumor conditioned media induces an activated state of the fibroblasts that was demonstrated in functional assays and in the up-regulation of known CAF-related genes and pathways involved in the ECM/matrisome. The significance of this study is three-fold as it not only provides a clinically relevant and easy protocol to reprogram normal fibroblasts into CAFs, but it also provides four functionally well-characterized CM-CAFs to the community, which currently do not exist, and mechanistically addresses that non-vesicular paracrine signaling drives this reprogramming and activation. Our results are expected to have an important positive impact because they will provide strong evidence for further development of therapeutics that possess potentiality and specificity towards CAF/ECM-mediated chemotherapy resistance in OC.

LIMITATIONS OF THE STUDY

Future investigations will need to validate other cell-type origins besides uterine fibroblasts and an immortalized ovarian fibroblast line. Unfortunately, immortalized fibroblast lines undergo genetic alterations that can create unique gene patterns that might express attributes or functions distinct to the original cells [84]. While primary CAFs derived from tumors would be ideal in terms of the origin and activated fibroblast status, lack of characterization considering that ovarian cancer patients present wide heterogeneity, makes it challenging to broadly adopt their use. The development of characterized CAFs lines could limit some of those variables. Importantly, retrieval experiments were performed to determine the stable reprogramming of normal uterine fibroblasts into CAFs. CAFs removed from tumor conditioned media remained CAF-like phenotypically and functionally. CAFs can be derived from many different cell types [10], thus new studies will need to investigate reprogramming with tumor conditioned media of these other cell types, like MSCs, epithelial, and endothelial cells. Additionally, the believed origin for ovarian cancer is the fallopian tube [106, 107], hence further studies should investigate fallopian tube fibroblasts instead of uterine fibroblasts and ovarian fibroblasts and characterize the functional changes and differences of the conditioned CAFs deriving from different cells of origins. Future studies will be needed to more accurately investigate the conditioned CAFs tumorigenic properties in vivo. Our experiments showed that the CM-uCAFs alone in vivo do not have a tumorigenic effect, but a co-culture of the cancer cells and CM-uCAFs in vivo should be explored to reflect the pro-tumoral role CAFs have in OC that was confirmed in our 3D co-culture experiments.

METHODS

Reagents

Type I collagenase, thrombin, trans-4-(aminomethyl)cyclohexanecarboxylic acid (AMCHA), 10% neutral buffered formalin, and calcium chloride (CaCl₂) were purchased from Sigma-Aldrich (Saint Louis, MO, USA). Lipophilic tracer 3,3′-dioctadecyloxacarbocyanine perchlorate (DiO, excitation, 488 nm; emission 525/50 nm) was purchased from Invitrogen (Carlsbad, CA, USA). Counting beads for flow cytometry experiments were purchased from Biolegend (424902, San Diego, CA, USA). The drug carboplatin (CARBO) used for drug resistance studies was purchased from MedKoo (Morrisville, NC, USA). For the contraction assay, rat tail collagen type I was purchased from R&D Systems (Minneapolis, MN, USA) and glacial acetic acid was purchased from Fisher Scientific (Hampton, NH, USA).

Cell Lines

Primary human uterine fibroblasts (HUFs) were purchased from ATCC (Manassas, VA, USA). HUFs were grown in DMEM media containing 1ng/mL fibroblast growth factor (FGF), 10% fetal bovine serum (FBS, Gibco, Life Technologies, Grand Island, NY, USA), 5ug/mL insulin, 1% penicillin/streptomycin (Corning CellGro, Mediatech, Manassas, VA, USA). Human serous ovarian cancer cell line SKOV-3 and high-grade serous ovarian cancer cell line KURAMOCHI were gifts from Dr. Paola Vermeer (Sanford Research, Sioux Falls, SD, USA). OC cell lines were grown in ovarian media, a 1:1 ratio of DMEM and Ham’s F12 with the addition of 1% penicillin/streptomycin and 10% FBS. The primary OC-CAF cell line was obtained from Vitro Biopharma (Denver, CO, USA) and was grown in MSC-GRO Vitroplus media with 1% penicillin/streptomycin. Adipose-derived MSCs were purchased from Fisher Scientific and grown in MesenPRO media growth supplement with 2% MesenPRO growth supplement, 1% L-glutamine, and 1% penicillin/streptomycin. The immortalized normal human ovarian fibroblast cell line, NOF-151-hTERT (NOFs), was a gift from Dr. Jinsong Liu and was grown in a 1:1 ratio of MDCB and Medium 199 with the addition of 10% FBS and 10ug/mL epithelial growth factor (EGF), and 1% penicillin/streptomycin. All cells were cultured at 37 °C, 5% CO₂, and 21% O₂; primary OC-CAF cells were also cultured in a hypoxic environment at 37 °C, 5% CO₂ and 1.5% O₂.

Human Samples

Blood from healthy donors was collected at Sanford Biobank in Sioux Falls, SD. Informed consent was obtained from all subjects with approval from the Sanford Health Institutional Review Board and in accordance with the Declaration of Helsinki. Peripheral blood from healthy subjects was obtained through venipuncture and collected in whole blood collection tubes (BD Lavender K2-EDTA Vacutainer). Plasma was separated by centrifugation of the blood samples at 800 x g for 10 min, then a secondary spin at 400 × g for 10 min with plasma immediately aliquoted and stored at −80 °C.

HUFs Reprogrammed into CAFs

For the reprogramming of HUFs into CAFs, tumor-derived conditioned media from both OC cell lines was collected, centrifuged, and filtered with syringes using a 0.2 um filter. Conditioned media was isolated from confluent OC cultures and kept at 4°C for no more than 1 week. Conditioned media was supplemented with 10% FBS at the time of culture. Two types of conditioned uterine-derived CAFs were generated, SKOV-3 conditioned uterine-derived CAFs (SKOV-uCAFS), and KURAMOCHI conditioned uterine-derived CAFs (KURA-uCAFs). To obtain conditioned CAFs, HUFs between passage 2 and passage 5 were cultured in conditioned media for at least one passage before being collected for analysis. Conditioned uCAFs were assessed until passage 12, including retrieval, and at least 3 different lots of HUF cells were tested throughout experiments.

NOFs Reprogrammed into CAFs

For the reprogramming of NOFs into CAFs, tumor-derived conditioned media from both OC cell lines was collected, centrifuged, and filtered with syringes using a 0.2 um filter. Conditioned media was isolated from confluent OC cultures and kept at 4°C for no more than 1 week. Conditioned media was supplemented with 10% FBS at the time of culture. Two types of conditioned ovarian-derived CAFs were generated, SKOV-3 conditioned ovarian-derived CAFs (SKOV-oCAFS), and KURAMOCHI conditioned ovarian-derived CAFs (KURA-oCAFs). To obtain conditioned oCAFs, immortalized NOFs were cultured in conditioned media for at least one passage before being collected for analysis. Conditioned oCAFs were assessed until passage 6.

EVs and EVs Free Conditioned Media

Tumor-derived conditioned media from both OC cell lines was collected when cells were confluent (after 4-5 days cultured in media), then centrifuged two times at 2500 x g for 10 minutes to ensure removal of cells. The conditioned media was then ultracentrifuged at 13,000 x g for 30 minutes using the Sorvall RC6 centrifuge to isolate microvesicles (MV). Then, the media was transferred to new tubes and filtered with syringes using a 0.2 um filter into 15 mL Sorvall tubes, saving the microvesicle pellet. The microvesicle pellet was resuspended in ovarian media without serum. The Sorvall tubes were added to WX80 tube holders and centrifuged at 117,000 x g for 2 hours using the Sorvall WX80 ultracentrifuge to isolate exosomes (Exo). Immediately, the media was collected in a new tube and saved as EVs free conditioned media. The exosome pellets were resuspended in ovarian media without serum and then combined with the resuspended microvesicles and saved as EVs conditioned media. Both the EVs CM and EVs free CM were stored at 4°C for no more than 1 week. Normal HUFs and NOFs were split into separate flasks and cultured in SKOV-3 EVs conditioned media, SKOV-3 EVs free conditioned media, KURAMOCHI EVs conditioned media, and KURAMOCHI EVs free conditioned media, respectively, and grown until confluent. SKOV-3 EVs conditioned media, KURAMOCHI EVs conditioned media, SKOV-3 EVs free conditioned media, and KURAMOCHI EVs free conditioned media were analyzed using the NanoSight particle analyzer to assess concentration of vesicles in the medias to confirm vesicle removal, following the manufacturer’s protocol [108].

Cell Expansion Analysis

Expansion of HUFs, conditioned uCAFs and oCAFs, NOF-151, and primary OC-CAFs was assessed by plating 1.0 x 10^5 cells/mL in triplicates in 96-well flat bottom plates (Fisher Scientific) and growing for 7 days between cell passages before trypsinization, and finally counting cells by Countess II (Fisher Scientific). Proliferation of each cell type is reported as the expansion fold after 7 days for each passage.

Cell Surface Marker Characterization

HUFs, conditioned uCAFs, primary OC-CAFs, SKOV-3, KURAMOCHI, NOF-151, and adipose MSCs were phenotypically characterized by analyzing CAF markers (FAP, CD29, PDGFRα, α-SMA), stromal markers (CD90, CD73, vimentin), epithelial markers (EpCAM), and immune marker (CD45) (Table S1). 0.1 x 10⁶ cell/mL cells were stained with the appropriate antibodies and analyzed by flow cytometry (BD FACS Fortessa, BD Biosciences, San Jose, CA, USA). Fluorescence minus one (FMOs) samples were used as controls. Percent positive cells are marked in the histograms and the relative MFI (rMFI) was calculated for each marker expressed by each cell type as the MFI of the specific marker divided by the MFI of the FMO control.

Tumor Progression and Drug Resistance in Physiologically Relevant 3D Cultures

HUFs, conditioned uCAFs and oCAFs, NOF-151, and primary OC-CAFs were embedded in a physiologically relevant three-dimensional (3D) culture system that recapitulates the complex biology of the TME and allows better cell-cell and cell-ECM interactions [60, 108-110]. We performed a co-culture of cancer and stromal cells to evaluate tumor promotion by the different stromal cell types. SKOV-3 and KURAMOCHI OC cells (0.3 × 10⁵ cell/mL) were prelabeled with DiO (10ug/mL) and incorporated into a 3D matrix either alone, or in a co-culture with HUFs, their respective conditioned CAFs, (SKOV-CAFs or KURA-CAFs), and primary OC-CAFs. The 3D cultures were established by mixing plasma, cells in ovarian media, a crosslinker, and a stabilizer in a 4:4:1:1 ratio, as previously described [60, 108-110]. After stabilization, ovarian media was added on top of the 3D cultures and renewed every 3 days. After 7 days of incubation, type I collagenase (20 mg/mL for 2 h at 37 °C) was used to enzymatically digest the 3D matrices. The isolated cells were stained for cell viability using a Live/Dead Blue cell stain (L34962, Thermo Fisher Scientific, Waltham, MA, USA) and blocked with 4% bovine serum albumin (BSA). Counting beads were added to each sample and a minimum 10,000 events were acquired by BD FACS Fortessa (BD Biosciences, San Jose, CA, USA) as previously described [21-24]. The data was analyzed with FlowJo program v10 (BD Biosciences, San Jose, CA, USA). Cancer cells were identified by gating high DiO signal. The effect of stromal cell types on drug resistance was also studied and carboplatin was given in order to recapitulate the standard of care treatment in OC. The same 3D matrix cultures were exposed to the drug carboplatin (30μM) against an untreated control for 7 days, and the same steps were followed as described above. Carboplatin at 30μM was given after careful preliminary characterization of our 3D cultures by dose response curves.

Collagen Gel Contraction Assay

HUFs, conditioned uCAFs and oCAFs, NOF-151, and primary OC-CAFs were incorporated into a 3D matrix made of rat tail type 1 collagen (R&D Systems) as previously described [111]. A 3 mg/mL concentration of collagen was made by diluting rat tail type 1 collagen with 0.1% acetic acid. This collagen solution was then combined with cells in media suspension to form a collagen solution with a concentration of 1mg/mL. A predetermined volume of 1 M NaOH was quickly added and then the mixture was immediately plated in a 24-well flat bottom plate (Fisher Scientific) and allowed to solidify at room temperature before adding media and dissociating the gels from the sides of the well. The 3D matrices were incubated at 37°C for 24 hours while they were observed for contraction. Images were taken with the ChemiDoc MP Imaging System (Bio Rad, Hercules, CA, USA) at time intervals of 0 h and 24 h. ImageJ software (National Institutes of Health, Bethesda, MD, USA) was used to record the diameter and area change of the matrix images over a 24-hour period.

Effect of Stromal Cell Types on Spheroid Formation

HUFs, conditioned uCAFs, and primary OC-CAFs (0.42 x 10⁵ cell/mL) were co-cultured with either SKOV-3 or KURAMOCHI cells (0.14 x 10⁵ cell/mL), then plated on ultralow attachment 96-well microplates with clear bottoms (Corning, Corning, NY, USA). The spheroids were grown for 21 days and imaged every 7 days with Cytation 3 Imaging reader (Biotek, Winooski, VT, USA). ImageJ software was used to measure the lengths and areas of the spheroids to compare the formed spheroids over time. After 21 days, the spheroids were enzymatically digested with type 1 collagenase and stained with Live/Dead Blue cell stain, stromal marker CD90, and the ovarian cancer epithelial marker CD276 ((Table S1); then, counting beads were added, and the cells were analyzed with flow cytometry (BD FACS Fortessa). Results were analyzed in FlowJo program v10 to determine the composition of the different spheroid types.

Luminex Assay

Plain media was collected and saved in −20°C. Plain medias collected: OC-CAF, Ovarian, HUF, NOF-151, SKOV-3 EVs free, SKOV-3 EVs, KURAMOCHI EVs free, KURAMOCHI EVs. Conditioned media was cultured with corresponding cell lines for 5 days in a 96-well plate then collected and saved in −20°C. Conditioned medias (CM) collected: OC-CAF CM, SKOV-3 CM, KURAMOCHI CM, HUF-derived CM, HUF-derived SKOV-3 CM, HUF-derived SKOV-3 EVs free CM, HUF-derived SKOV-3 EVs CM, HUF-derived KURAMOCHI CM, HUF-derived KURAMOCHI EVs free CM, HUF-derived KURAMOCHI EVs CM, NOF-151 CM, NOF-derived SKOV-3 CM, NOF-derived KURAMOCHI CM, NOF-derived SKOV-3 EVs free CM, NOF-derived SKOV-3 EVs CM, NOF-derived KURAMOCHI EVs free CM, NOF-derived KURAMOCHI EVs CM. At least 3 replicates of each media type were prepared using the Human Magnetic custom Luminex assay kit (R&D Systems) which included the analytes CCL2/MCP-1, LIF, Dkk-1, MMP-2, HGF, MMP-3, IGFBP-2, Osteoprotegerin (OPG), IL-6, TIMP-1, IL-8/CXCL8, and VEGF. The plate was analyzed with the Luminex 100/200 analyzer (Luminex Corporation) following the manufacturer’s protocol.

Collagen Deposition by Immunohistochemistry (IHC)

3D matrices containing monocultures of HUFs, conditioned uCAFs, and OC-CAFs were fixed in 10% neutral buffered formalin and processed on a Leica 300 ASP tissue processor. 3D matrix slides were stained by IHC with α-collagen 1 diluted 1:50 (Table S1), with DAB as the chromogen and the slides were counterstained with hematoxylin. Omission of the primary antibody served as the negative control. The Aperio VERSA Bright Field Fluorescence & FISH Digital Pathology Scanner (Leica Biosystems, Buffalo Grove, IL, USA) was used to image the slides, which were then quantified for intensity average (Iavg) where $\text{Iavg}=(\text{Iwp}+\text{Ip}+\text{Isp})∕(\text{Nwp}+\text{Np}+\text{Nsp})$ to account for potential differences in collagen deposition in Aperio Imagescope (Leica Biosystems, Buffalo Grove, IL, USA). Iwp = intensity for weak-positive pixels; Ip = intensity for positive pixels; Isp = intensity for strong-positive pixels; Nwp = Number of Weak-Positive pixels; Np = Number of Positive pixels; and Nsp = Number of Strong-Positive pixels.

ALDEFLUOR Assay

ALDH activity was assessed in HUFs, conditioned uCAFs and oCAFs, NOF-151, and primary OC-CAFs using the ALDEFLUOR assay kit (STEMCELL Technologies, Cambridge, MA, USA). The assay was performed per manufacturer’s protocol. 0.1 x 10⁶ cell/mL cells were resuspended in the Aldefluor assay buffer containing ALDH substrate (BODIPY-aminoacetaldehyde). The ALDH inhibitor diethylamino benzaldehyde (DEAB) was added to half the cells to provide a negative control. The cell suspensions were incubated at 37°C for 45 min, suspended in PBS 1x and counting beads, and analyzed to measure fluorescence by BD FACS Fortessa. The percentage of ALDEFLUOR-positive cells for each sample was gated separately and compared to the negative controls marked by the inhibitor DEAB in FlowJo program v10.

Retrieval Experiments

Retrieval of tumor-derived conditioned media was also investigated, where conditioned uCAFs were kept in culture with the corresponding tumor-derived conditioned media or switched to ovarian media, depriving them from the conditioned media exposure. The conditioned uCAFs and retrieval uCAFs were reassessed phenotypically and functionally through marker expression, collagen contractility, and tumor growth as described above for passages 1, 3, and 5 post-conditioned media retrieval.

RNA-sequencing Analysis

RNA from the HUFs, conditioned uCAFs, and primary OC-CAFs was sequenced. Cell pellets were flash frozen in liquid nitrogen and then RNA was extracted using QIAgen RNeasy mini columns according to manufacturer’s protocol [112]. RNA was quantified using a NanoDrop spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA) and the Agilent 2100 Bioanalyzer (Santa Clara, CA, USA) was used to verify RNA quality under the support of the Functional Genomics and Bioinformatics Core at Sanford Research. Sequencing was performed by Novogene. FASTQ was used to check the read quality of the raw sequencing reads and STAR was used for mapping the reads to genome. In RStudio, the aligned read counts by gene were converted into a count matrix, which was then analyzed by DESeq2 (version 1.40.2) to identify differential expression by normalizing counts and producing differentially expressed genes (DEGs). DEGs were functionally analyzed using the clusterProfiler R package for gene ontology (GO) analysis. All volcano plots were built using the EnhancedVolcano R package and all heatmaps were built using the ComplexHeatmap R package. Using the RNA-sequencing results for HUFs, CM-uCAFs, and OC-CAFs with the GSVA R package, an ECM score was produced using three gene sets from the GSEA website for the core, associated, and complete Naba matrisome and a CAF score was produced from gene sets used in previous studies: 24 carcinoma-associated fibroblast genes upregulated by MSCs grown in tumor conditioned media [31]; 7 CAF markers used to identify two CAF subtypes [30]; and 5 CAF markers used to identify four CAF subtypes [32]. A dataset of OC patients from Tothill et al. [29] was obtained through accession GSE9899. The affy R package was used to read .CEL files containing affy IDs and export an expression data file where affy IDs were converted to Entrez gene symbols with the website DAVID. A mesenchymal score was produced with GSVA using all genes that were upregulated in the mesenchymal (C5) subtype that the paper identified, where we used only the top 10% of mesenchymal^high scoring samples (29 total). This data from the 29 mesenchymal patients was then merged and normalized in R with the RNA-sequencing results for normal HUFs, and then analyzed with DESeq2 to produce DEGs. DEG lists were annotated using the Matrisome AnalyzeR tool. The venn diagram R package was utilized to compare the overlapping DEGs among the datasets. All analyses used an FC cutoff of 1.5.

Tumorigenic Potential of Conditioned CAFs in vivo

All in vivo experiments were conducted according to protocols approved by the Institutional Animal Care and Use Committee (IACUC) at Sanford Research. 0.3 x 10⁶ cell/mL of HUFs, conditioned uCAFs, and primary OC-CAFs was injected into female, 6–10-week-old immunodeficient Swiss Webster mice intramuscularly in the flank, and tumor progression was evaluated by caliper for 6 weeks. Twice a week the mice were weighed, and tumor size was measured. The formula $\mathsf{V}(\text{mm}3)=1∕2\mathsf{x}\text{length}(\text{mm})\mathsf{x}\text{width}{(}^{\ast }2)(\text{mm})$ was used to determine tumor growth.

Data Availability Statement

Relevant RNA sequencing datasets will be made available in a data repository before publication and linked to the article. Any other data can be available upon reasonable request to the corresponding author.

Statistical Analysis

All in vitro experiments were repeated independently at least three times with triplicate samples per experiment unless otherwise stated. GraphPad Prism 5 (GraphPad Software Inc., La Jolla, CA, USA) was used for graphing results. We have presented the results as a mean ± a standard deviation value. Statistical significance was analyzed using Student’s t-test, one-way ANOVA, or two-way ANOVA, where a p value less than 0.05 was considered significant and an interquartile range (IQR) was used to remove outliers [60]. Unless stated otherwise, *p>0.05, **p>0.01, ***p>0.001, and ****p>0.0001.

Highlights.

• Reprogramming of normal fibroblast into cancer-associated fibroblast (CAFs) is mediated via non-vesicular paracrine signaling.

• CAFs transformed from uterine and ovarian fibroblasts resemble a CAF-like phenotype, which was phenotypically and functionally characterized.

• Activated CAFs possess strengthened proliferation, secretory profile, contractility, ALDH activity, tumor growth in 3D cultures, and ECM remodeling properties when compared to resting normal fibroblasts.

• Translatable-based approach to activate normal fibroblast into CAFs using tumor-derived conditioned media.