Omental Preadipocytes Stimulate Matrix Remodeling and IGF Signaling to Support Ovarian Cancer Metastasis

Mature adipocytes comprise the bulk of the omentum and preadipocytes differentiate to form new adipocytes as needed (Supplementary Fig. S1A; ref. 22 ); however, studies to date on the omental TME do not account for the differential contributions of these two cell types to omental metastasis. To study these differences, we cocultured ACI-23 cells, a low-grade serous ovarian cancer cell line, with either preadipocytes derived from omental tissue of nondiabetic, nonsmoking female donors or with mature lipid-storing adipocytes differentiated from the donor preadipocytes. Although ACI-23 cells are classified as low-grade, they maintain a substantial population of tumor-initiating cells (TIC), which are required to establish secondary tumors ( 23 ). Commercially available omental preadipocytes were differentiated into mature adipocytes and stained with Oil Red O to confirm adipogenic potential via lipid accumulation (Supplementary Fig. S1B) as well as evaluated for expression of differentiation-associated genes CEBPA , FASN , FABP4 , and PPARG , all of which increased by Day 14 (Supplementary Fig. S1C). Preadipocyte surface markers were assessed via flow cytometry, and preadipocytes were CD24 − , CD31 − , CD34 − , CD44 + , CD45 − , Pref-1 − , CD38 − , and CD54 + , with CD54 indicating a commitment to an adipocyte lineage (Supplementary Fig. S1D–S1E; refs. 24 – 28 ). Preadipocytes and mature adipocytes were cocultured with ovarian cancer cells, after which, the cancer cells, preadipocytes, and mature adipocytes were subjected to RNA sequencing along with monoculture controls (Supplementary Fig. S2A; Supplementary Table S3). Mature adipocytes had 19 differentially expressed genes (DEG), whereas preadipocytes had 119 DEGs after coculture with ovarian cancer cells (Supplementary Fig. S2B and S2C). Notably, there was very little overlap in the transcriptional changes in either the preadipocytes or mature adipocytes in response to ovarian cancer cells. Functional analysis of DEGs in Reactome revealed that preadipocytes were enriched for immune cell recruitment, metabolism, and matrix organization, all crucial factors in cancer progression. Considerably fewer pathways were induced in mature adipocytes using the same criterion; however, we did observe activation of receptor tyrosine kinase signaling (Supplementary Fig. S2D). Note that some heterogeneity is expected within these populations given the differentiation status of individual cells within each culture that could potentially be reflected in the DEGs. We next tested the effect of preadipocytes on tumorigenesis in vivo . CD19-labeled ACI-23 cells were subcutaneously injected in a limiting dilution with or without a constant number of preadipocytes. The resultant ratios of cancer cells to preadipocytes were 40:1, 1:1, and 1:40 ( Fig. 1A ). Mice that received the 1:40 injection had no difference in tumor initiation and a significant decrease in overall tumor growth relative to mice injected with tumor cells alone ( Fig. 1B ; top). Mice injected with the 1:1 ratio had a slight increase in the rate of tumor initiation and no difference in overall tumor growth. Notably, mice injected with the 40:1 ratio grew tumors quickly and reached clinical endpoints 2 weeks earlier than the other groups ( Fig. 1B , middle). Remarkably, mice that received the 1:40 injection all formed tumors within 18 days, whereas mice injected with the same number of cancer cells alone were unable to form tumors ( Fig. 1B ; bottom). An extreme limiting dilution analysis ( 29 ) was highly significant at the 1:40 ratio (Supplementary Fig. S2E). The 1:40 injection was repeated with either preadipocytes or adipocytes as well as preadipocyte-only controls, which demonstrated that both populations significantly enhance tumor initiation and growth although neither preadipocytes nor mature adipocytes form tumors when injected alone without cancer cells (Supplementary Fig. S2F–S2I). Importantly, this highlights preadipocytes as a comparable and significant contributor to tumor formation in the adipose TME. Omental preadipocytes enhance the tumorigenic potential of ovarian cancer cells. A, ACI-23 ovarian cancer cells were subcutaneously injected into the left flank of 8-week-old female athymic Nu/J mice in the presence or absence of human omental preadipocytes (pAdip) at three ratios (40:1, 1:1, 1:40). Ratios are expressed as cancer cells: preadipocytes. B, Tumor initiation (left) and growth (right) in subcutaneous injection model. Two-way ANOVA with Sidak post hoc test for growth ( n = 8). C, Transplantation assay (T) was performed using ACI-23 cells from the 1:40 with pAdip tumors that were isolated via FACs and reinjected using the same methods as the initial experiment (±pAdip). Tumors from the transplantation assay are shown as imaged at the time of harvest from each mouse. Mice that did not develop tumors are not shown. N = 10 mice per condition. D, Table of results for tumor initiation; cell dose column contains the number of cancer cells with the cell ratio. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. ( A, Created with BioRender.com.) We reasoned that if the increase in tumor initiation was due to an increase in TICs, as indicated by the extreme limiting dilution analysis, resultant tumors should have an increase in TICs. Tumor cells isolated via CD19 (Supplementary Fig. S3A) were used in a transplantation assay of the 1:40 dilution to test whether co-injection with preadipocytes imparted a persistent phenotypic change to the cancer cells. Mice co-injected with FACS-isolated tumor cells and fresh preadipocytes at the 1:40 dilution mirrored the results of the first study; however, cancer cells that were reinjected without preadipocytes did not retain their tumor-initiating ability ( Fig. 1C and D ). This finding suggests preadipocytes mediate a transient increase in the tumorigenic potential of ovarian cancer cells, potentially through paracrine signaling that activates growth or survival pathways. To confirm tumors were comprised primarily of tumor cells we performed flow cytometry and immunohistochemical analyses. The injected tumor cells were labeled with CD19 and tumors from both the high and middle dilutions had no significant differences in the frequency of CD19 + cells at harvest, indicating preadipocytes did not proliferate postinjection (Supplementary Fig. S3A). Moreover, ACI-23 cells are predominantly CD44 − (80%), and they differ greatly in cellular size and granularity ( 30 ). Flow cytometry on tumors from the transplantation assay confirmed that all tumor cells were CD44 − , indicating that the growth observed was representative of the cancer cells and not the preadipocytes (Supplementary Fig. S3B–S3E). Staining for NM-95, a human-specific nuclear marker, further demonstrated that stromal cell compartments observed in the tumors were nonhuman and therefore comprised of infiltrating murine stromal cells, not the injected preadipocytes (Supplementary Fig. S3F). These data are consistent with the lack of tumor development in the adipocyte-only tumor initiation studies shown in Supplementary Fig. S2F–S2G. The results of the limiting dilution led us to initially hypothesize that preadipocytes were increasing the population of TICs. TICs are classically characterized by quiescence, increased expression and activation of stem cell transcription factors SOX2 , OCT4 , and NANOG , increased aldehyde dehydrogenase (ALDH) activity, and expression of surface markers such as CD133 and CD117 (Supplementary Fig. S4A; refs. 18 , 23 , 31 , 32 ). Surprisingly, ACI-23 cells cocultured with preadipocytes had decreased expression of SOX2 and OCT4 (Supplementary Fig. S4B). Moreover, although CD133 surface expression was increased, there was no change in NANOG gene expression, ALDH activity, or spheroid formation (Supplementary Fig. S4B–S4D), indicating that classic TIC features are not induced by preadipocytes in vitro . We next hypothesized that preadipocytes could be activating pathways in cancer cells to enable survival under isolation stress ( 33 ). To study this possibility, we used several in vitro assays to model the phenotype we observed in our mouse model. High-grade serous ovarian cancer (HGSOC) is the most prevalent histological subtype ( 34 ), and to confirm the relevance of our findings across serous ovarian cancers, we expanded these assays using additional cell lines derived from HGSOC (CAOV4 and OVCAR8) as well as in SKOV3 cells, which are ascites-derived and therefore of interest in metastasis ( 35 ). Preadipocytes were cultured in an SFM for 48 hours to generate CM enriched with preadipocyte-secreted factors. At 3 days, CM enhanced cell viability in ACI-23, CAOV4, OVCAR8, SKOV3, and VCBF004, which is an ascites-derived primary cell line. This difference in viability was increased at 7 days and was significant in all cell lines tested ( Fig. 2A ). Coculture with preadipocytes additionally increased colony formation in all cell lines tested (ACI-23, CAOV4, and OVCAR8; Fig. 2B ). Secreted factors from omental preadipocytes enhance cell viability by regulating growth and survival under isolation and nutrient stress. A, Cell viability in ACI-23, CAOV4, OVCAR8, and SKOV3 cells as well as primary, ascites-derived VCBF004 cells in SFM and pAdipCM at 3- and 7-days posttreatment. Multiple unpaired t tests ( n = 3). B, Colony formation was measured in ACI-23, CAOV4, and OVCAR8 cells after 7-day coculture with omental preadipocytes, and values represent the number of colony-forming units (CFU). Unpaired t test ( n = 3). C, Apoptosis was measured via caspase-3/7 activity in cells cultured in SFM and pAdipCM in ACI-23, CAOV4, OVCAR8, and SKOV3. Multiple unpaired t tests ( n = 3). D, Proliferation of ACI-23, CAOV4, OVCAR8, and SKOV3 in SFM and pAdipCM was assessed by measuring incorporation of EdU in Hoechst-positive nuclei, and proliferation was calculated relative to a complete media control with 10% FBS. Scale bar, 10 μm. Two-way ANOVA with Sidak post hoc test ( n = 3). *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; ns, not significant. Given that increased viability could be due to either reduced apoptosis or increased proliferation, we measured apoptosis via caspase-3/7 activity and proliferation via EdU incorporation either in SFM or CM. In ACI-23 cells, apoptosis decreased in response to CM at 7 days; however, apoptosis was significantly increased in SKOV3 cells ( Fig. 2C ). A decrease in proliferation was consistent in all cell lines treated with SFM relative to growth in complete media, as expected, and treatment with CM partially rescued this phenotype in all cell lines at Day 3, and all HGSOC cell lines at Day 7 ( Fig. 2D ; Supplementary Fig. S5). To obtain a deeper understanding of transcriptional changes induced by preadipocytes that may lead to increased tumorigenic properties of cancer cells, we performed RNA sequencing of ACI-23 cells cocultured with either preadipocytes ( Fig. 3A ) or mature adipocytes ( Fig. 3B ). After coculture with preadipocytes there were 435 upregulated and 1,114 downregulated DEGs, with 347 and 980 hits specific to preadipocytes, respectively ( Fig. 3C ). Principle cluster analysis of gene expression from monoculture and coculture of preadipocytes (Supplementary Fig. S6A) and adipocytes (Supplementary Fig. S6B) showed that all biological replicates clustered similarly, suggesting differential gene expression is due to the signaling between the preadipocytes and cancer cells and not other culture conditions. Pathway enrichment via Metascape showed the most significant gene set hits for preadipocytes included ECM organization, regulation of IGF transport and uptake by IGF binding proteins (IGFBP), and cellular responses to stress ( Fig. 3D–G ). Analysis of gene ontology by Reactome and gene set enrichment analysis further confirmed that ECM genes were enriched in ACI-23 cells cocultured with preadipocytes (Supplementary Fig. S6C and S6D). Interestingly, preadipocyte cocultures were specifically enriched for the AVB3 integrin pathway (Supplementary Fig. S6E), which regulates cellular adhesion to ECM substrates ( 36 ). The top 25 upregulated genes in each of the top three Reactome gene sets were organized by their fold change (log 2 FC) and although there was some overlap between DEGs between preadipocytes and adipocytes, the profiles for each pathway were distinct between cocultures ( Fig. 3E–G ; Supplementary Fig. S6F–S6H). Additionally, several pathways were completely cell-type specific, with preadipocytes upregulating IGF regulatory genes, whereas adipocytes upregulated PPARA -associated genes ( Fig. 3F ; Supplementary Fig. S6G). Preadipocytes differentially regulate gene programs in ACI-23 that highlight ECM remodeling and IGF signaling. A and B, Volcano plot of ACI-23 cells after transwell coculture with human omental preadipocytes ( n = 4; A ) or differentiated adipocytes ( n = 4; B ) at a 1:40 ratio. Values are expressed as the log 2 value of the ratio of transcript reads from cocultured cells relative to monocultured controls. DEGs with a −log 10 P value below 0.1 and log 2 -fold change greater than 0.5 are significantly changed and are depicted in red. C, Venn diagram of DEGs with the number of upregulated (red) and downregulated (blue) genes. D, Gene ontology of ACI-23 after preadipocyte or differentiated adipocyte coculture. Lists of DEGs with an lFDR below 0.1 from ACI-23 cells cocultured for 6 days with human omental preadipocytes or differentiated adipocytes were queried in Metascape to search within the Reactome pathway database. Ontology terms are ranked alphabetically and graphed by the number of enriched genes within the gene set (gene ratio) and their log-transformed P value. E–G, Top 25 upregulated genes from Reactome in cancer cells after preadipocyte coculture were plotted, with significance shown by color scale and lFDR shown by dot size. E , Extracellular matrix organization gene list. F , Regulation of IGF transport and uptake by IGFBPs gene list. G , Cellular responses to stress gene list. H, gene expression of 1:40 cocultures in ACI-23, CAOV4, OVCAR8, and SKOV3 via qRT-PCR. Multiple unpaired t tests ( n = 3). *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; ns, not significant. Several of the genes identified in enriched pathways, including collagens 1A1, 1A2, and 6A2 ( COL1A1 , COL1A2 , and COL6A2 ), fibronectin ( FN1 ), IGF binding protein 5 ( IGFBP5 ), and matrix metallopeptidase 2 ( MMP2 ) were measured via qRT-PCR in ACI-23, CAOV4, OVCAR8, and SKOV3 ( Fig. 3H ). COL1A1 , COL1A2 , IGFBP5 , and MMP2 were elevated after 1:40 coculture in all cell lines tested, whereas FN1 and COL6A2 were cell line dependent ( Fig. 3H ). Interestingly, when the same genes are measured in cocultures with a 1:1 ratio the increased expression of collagens, MMP2 , and IGFBP5 is lost in most lines (Supplementary Fig. S6I). We seeded cancer cells into a direct coculture with preadipocytes at a 1:40 dilution to model the conditions of the limiting dilution. Preadipocytes were fluorescently labeled using CMFDA to enable specific analysis of the cancer cell population via immunofluorescence. Direct cocultures were then stained with antibodies against collagen 1, collagen 6, IGFBP5, fibronectin, and MMP2 to measure proteins identified in the RNA sequencing. Although it was expected that the genes upregulated in transwell coculture would translate to elevated protein expression in direct preadipocyte coculture, when ACI-23, CAOV4, OVCAR8, or SKOV3 were plated with CMFDA-labeled preadipocytes we observed either no change or a decrease in expression of all proteins except fibronectin ( Fig. 4A ; Supplementary Fig. S7A and S7B). Upregulation of ECM genes is an early event in ovarian cancer tumorigenesis and requires low cell density. A, Collagen 1 expression (red) in direct cocultures of ACI-23, CAOV4, OVCAR8, and SKOV3 cells with CMFDA-labeled omental preadipocytes (green) at a 1:40 ratio. Representative images show monoculture controls and direct cocultures, with colonies in coculture identified in white outline. Unpaired t test ( n = 4). Scale bar, 100 μm. B, Subcutaneous injection with ACI-23 and omental preadipocytes at a 1:40 ratio. “Early” tumors were resected at 24 days, remaining tumors were designated “Late” and were excised once the first tumor reached 1,000 mm 3 . C, Tumor growth rate (left) and size comparison at resection (right) of early and late tumors. Two-way ANOVA and unpaired t test, respectively ( n = 6). D, Gene expression in early and late tumors, measured via qRT-PCR relative to GAPDH. Unpaired t test ( n = 6, one outlier removed from each group). E, Correlations between COL1A1 and COL1A2 gene expression with tumor size at time of resection in early (red) and late (blue) tumors. *, P < 0.05; **, P < 0.01; ns is not significant. ( B, Created with BioRender.com .) An alternative hypothesis is that the ECM proteins are upregulated specifically under conditions in which cancer cells are in isolation stress, which is defined as the loss of appropriate ECM, cell–cell contact, and nutrient deprivation ( 37 ). Isolation stress would be expected in the indirect cocultures but not necessarily the direct cocultures where preadipocytes provide an ECM and cell-to-cell contact. We subcutaneously injected ACI-23 cells and omental preadipocytes at the 1:40 dilution and harvested tumors when they first seemed or once they were established ( Fig. 4B and C ) and measured changes in ECM gene expression via qPCR. We found that COL1A1 and MMP2 were significantly downregulated in established tumors relative to early tumors ( Fig. 4D ), which mirrors several of the decreases we observed in our direct cocultures. There was no change in the expression of COL1A2 , COL6A2 , or FN1 ( Fig. 4D ). Interestingly, when transcript levels of COL1A1 and COL1A2 were correlated with final tumor size at harvest in early and established tumors, positive trends were only observed in the early tumors, further supporting the notion that upregulation of collagen 1 is an important mediator of early preadipocyte-stimulated tumorigenesis ( Fig. 4E ). We next sought to determine which secreted factors were responsible for preadipocyte-induced ECM remodeling. A proteomic array of 62 adipokines was used to establish a secretory profile for ACI-23 cells alone, preadipocytes alone, and ACI-23 and preadipocyte cocultures. Preadipocytes representing two separate donors predominantly secreted IL6 and CCL2 whereas ACI-23 secreted IL8 and TGFβ, consistent with other reports on interleukins in the TME ( Fig. 5A ; Supplementary Fig. S8A and S8B; ref. 5 ) To test the effect of preadipocyte-specific secreted proteins, we stimulated ACI-23, CAOV4, and OVCAR8 cells with recombinant IL6 and CCL2 and measured changes in ECM gene expression; however, there were no consistent changes (Supplementary Fig. S8C and S8D). IGF1 is elevated in coculture and regulates ECM gene expression and alternative NF-κB. A, Representative adipokine secretory profiles from ACI-23: preadipocyte cocultures and monoculture controls in complete media. B, Quantification of IGFBPs from adipokine arrays. One-way ANOVA with Tukey post hoc test ( n = 4). C, IGF1 ELISA of supernatants from 6-day preadipocyte and cancer cell cocultures, with monoculture controls in complete media. Two-way ANOVA with Sidak post hoc test ( n = 3). D, IGF1 ELISA of supernatants from preadipocyte and cancer cell cocultures, with monoculture controls used for RNA sequencing, all in complete media. Two-way ANOVA with Sidak post hoc test ( n = 3). E, Gene expression of IGF1 (top) and IGF2 (bottom) in primary and omental tumors from three publicly available data sets relativized to the average of the primary tumors. Unpaired t test. F, ECM gene expression via qRT-PCR of ACI-23 cells treated with 100-ng/mL IGF1 for 72 hours in complete media; data are fold change of vehicle. Unpaired t test ( n = 3–5). G, NF-κB activity in OVCAR8-NF-κB-Luc cells treated with vehicle control or 100-ng/mL IGF1 for 4, 24, 48, and 72 hours in complete media. One-way ANOVA with Tukey post hoc ( n = 3). H, Gene expression of RELA and RELB in ACI-23 treated with vehicle control or 100-ng/mL IGF1 for 72 hours in complete media measured via qRT-PCR. Unpaired t test ( n = 3). I, RelB expression in cytosolic and nuclear fractions of ACI-23 cells treated with IGF1 for 72 hours in SFM. Lysate fractions were run on the same gel but displayed as separate panels for clarity. Western blots were also probed for nuclear and cytosolic housekeeping proteins lamin A/C, GAPDH, and α-tubulin ( n = 2). J, NF-κB activity in OVCAR8-NF-κB-Luc cells treated with SFM or CM for 4, 24, 48, and 72 hours. One-way ANOVA with Tukey post hoc ( n = 3). K, Gene expression of RELA and RELB in ACI-23 cocultured with preadipocytes for 72 hours in complete media measured via qRT-PCR. Unpaired t test ( n = 3). *, P < 0.05; **, P < 0.01; ****, P < 0.0001; ns, not significant. The IGFBPs were also differentially expressed by preadipocytes and cancer cells. ACI-23 secreted IGFBP2, IGFBP4, and IGFBP6, whereas preadipocytes secreted IGFBP3, IGFBP4, IGFBP6, and IGFBP7, and all were detected in coculture ( Fig. 5A and B ). Expression of these genes was enhanced with preadipocyte coculture ( Fig. 3F ). The IGFBPs belong to a highly conserved protein family that regulates IGF signaling to control cell proliferation, differentiation, viability, and migration ( 38 ). We measured IGF1 in supernatants collected from the cocultures subjected to RNA sequencing and found that cocultures had increased IGF1 ( Fig. 5C ). This finding was validated in additional cocultures using ACI-23, CAOV4, OVCAR8, and SKOV3 ( Fig. 5D ). Patient datasets retrieved from NCBI Gene Expression Omnibus (accession numbers GSE178913 , GSE98281 , and GSE137237 ) show IGF1 mRNA is elevated in omental relative to primary tumors, whereas IGF2 is not significantly changed ( Fig. 5E ). ACI-23 cells treated with exogenous IGF1 significantly increased expression of ECM genes COL1A2 and COL6A2 , elevated expression of FN1 ( P = 0.0595), and significantly decreased expression of IGFBP5 and MMP2 . There was no change in the expression of COL1A1 ( Fig. 5F ). IGF1 supports cell survival via IGF/AKT/NF-κB signaling in gastric cancer ( 39 ), and integrin-mediated adhesion to the ECM can enable cell survival via activation of NF-κB ( 40 ). Furthermore, our lab has demonstrated that NF-κB plays a role in ovarian cancer tumorigenesis ( 23 ). Therefore, we measured the activation of NF-κB using OVCAR8 cells stably transfected with an NF-κB luciferase reporter. Treatment with IGF1 activated NF-κB ( Fig. 5G ), and preadipocyte coculture increased gene expression of the alternative NF-κB transcription factor RELB but not the canonical factor RELA ( Fig. 5H ). To confirm nuclear localization of RelB, ACI-23 cells were treated with IGF1 and separated cytosolic and nuclear fractions were examined via immunoblot, which demonstrated that IGF1 stimulation increased nuclear RelB relative to vehicle treatment ( Fig. 5I ). This was confirmed using CM showing increased NF-κB activity was also observed after treatment with CM ( Fig. 5J ), and preadipocyte coculture increased RELB mRNA expression ( Fig. 5K ). To better understand the role of IGF1 in preadipocyte-induced tumor growth we repeated our subcutaneous tumor model in the presence of the IGF1 receptor (IGF1R) inhibitor NVP-AEW541 ( Fig. 6A ; ref. 41 ). Although there were no significant differences in tumor initiation between NVP-AEW541 and vehicle-treated mice, tumor growth was significantly impeded ( Fig. 6B and C ). Akt is a known downstream target of the IGF1R ( 38 ), and as expected, mice treated with NVP-AEW541 had decreased Akt phosphorylation relative to vehicle-treated mice ( Fig. 6D ). Alternative NF-κB signaling activation involves ubiquitin-mediated cleavage of p100 ( 42 ) and given that IGF1 activates alternative NF-κB, the abundance of p52 and p100 were measured in tumors, with a significant decrease in the ratio of p52: p100 in NVP-treated tumors ( Fig. 6D ). IGF1R inhibition decreases preadipocyte-induced tumor growth, AKT/NF-κB signaling, and intratumoral collagen 1. A, ACI-23 ovarian cancer cells and omental preadipocytes at a 1:40 ratio (cancer cells: preadipocytes) were subcutaneously injected into the left flank of Nu/J mice pretreated with NVP-AEW541 ( n = 6) at 50 mg/kg or vehicle control ( n = 5) and received daily treatment for the duration of the study. All mice were sacrificed once the first tumor reached a terminal endpoint. B, Tumor initiation (left) and growth (right) in subcutaneous injection model. Two-way ANOVA with Sidak post hoc test for growth ( n = 5–6). C, Tumor weight at resection (left) and tumor size in vivo (right) with NVP-AEW541 or vehicle control. Unpaired t test. D, Immunoblot of AKT and alternative NF-κB pathway activation. The ratios of phosphorylated to total AKT and cleaved p52 to uncleaved p100 were calculated using densitometry. Unpaired t test. E, IHC of collagen 1 expression within the vehicle and NVP-AEW541–treated subcutaneous tumors. Unpaired t test. Scale bar, 300 μm. F, Immunoblot of RelB knockdown in shNEG and shRELB OVCAR8 cells ( n = 3; left). Gene expression of COL1A1 and COL1A2 by OVCAR8-shNEG and OVCAR8-shRELB. Unpaired t test ( n = 3). G, Analysis of publicly available RNA-seq for COL1A1 and COL1A2 of matched primary and omental tumors. Paired t test. H, IHC of Ki67 expression in the vehicle and NVP-AEW541–treated subcutaneous tumors. Unpaired t test. Scale bar, 100 μm. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. ( A, Created wtih BioRender.com.) Exogenous IGF1 increased expression of COL1A2 in vitro ( Fig. 5F ) and staining via IHC revealed that inhibiting IGF signaling reduced intratumoral collagen 1 as quantified in tumor compartments of whole tissue scans ( Fig. 6E ). To confirm that RelB acts as a transcriptional regulator of collagen 1 in cell lines beyond ACI-23, gene expression of COL1A1 and COL1A2 was measured in OVCAR8 cells with stable knockdown of RelB. Knockdown was confirmed via immunoblot, and as anticipated, OVCAR8-shRELB cells expressed significantly less COL1A1 and COL1A2 , indicating that RelB positively regulates the expression of collagen 1 in HGSOC ( Fig. 6F ). In addition to being IGF1 inducible, COL1A2 was the most significantly upregulated ECM gene after preadipocyte coculture ( Fig. 3A ), and therefore we silenced COL1A2 via siRNA and repeated the cell viability assay with CM. In cells transfected with siCOL1A2, the increase in cell viability with CM treatment relative to SFM was reduced, whereas it remained significant in siNEG transfected cells (Supplementary Fig. S9A and S9B). Furthermore, analysis of publicly available RNA sequencing data of donor-matched primary tumors and omental metastases reveals an increase in COL1A1 and COL1A2 expression in 75% of omental tumors ( Fig. 6G ; n = 16). In addition to a reduction in collagen 1, tumors also expressed less Ki67 after treatment with NVP-AEW541, indicating that IGF1R inhibition decreases tumor cell proliferation ( Fig. 6H ). Given that collagen 1 facilitates ovarian cancer progression and survival by enabling cellular adhesion and proliferation ( 43 , 44 ), we employed siRNAs to COL1A2 and COL6A2 to test the ability of ovarian cancer cells to adhere to different substrates after knockdown and pretreatment with either SFM or CM. Adhesion was most efficient to fibronectin and Matrigel, with fewer cells adhered to gelatin-coated wells or uncoated TC plastic after 4 hours. Interestingly, in 3/4 cell lines, we observed a significant ( P = 0.15 for OVCAR8) increase in adhesion to fibronectin that was specific to CM stimulation combined with COL1A2 knockdown and not COL6A2 (Supplementary Fig. S9C and S9D). IGF1R inhibition via NVP-AEW541 decreased collagen 1 deposition in vivo ( Fig. 6E ) and silencing collagen 1 via siRNA enhanced adhesion in vitro (Supplementary Fig. S9A–S9C), and therefore, we tested NVP-AEW541 in an intraperitoneal model to determine if IGF1R inhibition affected the potential for cancer cells to adhere and form metastases. OVCAR8 cells were selected for this assay as they form more IP tumors relative to ACI-23 ( 45 ). OVCAR8 cells were engineered to express luciferase and RFP (OVCAR8-RFP-LUC) to enable quantitative single-organ resolution of metastatic burden and subsequent cell sorting ( Fig. 7A ). Cells were validated in vitro to confirm that differences in cell number could be detected via luciferase assay (Supplementary Fig. S10A). Omental adhesion is enhanced whereas long-term tumor growth is decreased with IGF1R inhibition. A, A total of 1 × 10 6 OVCAR8-RFP-LUC cells were injected intraperitoneally into Nu/J mice ( n = 4–6). Mice were pretreated with NVP-AEW541 at 50 mg/kg or vehicle control and received daily treatment for the duration of the study. Mice were sacrificed at 4 hours, 7 days, or 28 days postinjection for endpoints. B, Omental adhesion as measured by luciferase assay at 4 hours and 7 days postinjection. Multiple unpaired t tests ( n = 5–6, one outlier removed from NVP 7-day). C, Intraperitoneal adhesion to multiple organs measured at 7 days postinjection. Multiple unpaired t tests ( n = 5). D, Gene expression of cells isolated via FACS from the omentums of mice treated with vehicle or NVP-AEW541 ( n = 3) relative to mock ( n = 4). One-way ANOVA with Tukey post hoc test. E, Integrin expression within ACI-23 subcutaneous tumors treated with vehicle ( n = 5) or NVP-AEW541 ( n = 6) was assessed via immunoblot (left) and quantified via densitometry (right). Multiple unpaired t tests. F, Gene expression of integrins from ACI-23 early or late subcutaneous injection tumors via qRT-PCR relative to GAPDH. Multiple unpaired t tests, ( n = 5). G, ACI-23, CAOV4, OVCAR8, and SKOV3 cells were treated with NVP-AEW541 or vehicle control ( n = 3) and αVβ3 integrins were measured via flow cytometry. Two-way ANOVA with Tukey post hoc test. H, IVIS images (left) and quantifications (right) of individual mice with intraperitoneal tumor growth by OVCAR8-RFP-LUC cells treated daily with either NVP-AEW541 or vehicle control ( n = 6 mice in each group). Two-way ANOVA with Sidak post hoc test. I, Quantification of flow cytometry measuring OVCAR8-RFP-LUC cells digested from the omentum at day 28 ( n = 4). Unpaired t test. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; ns, not significant. ( A, Created with BioRender.com .) Interestingly, although NVP-AEW541 decreased subcutaneous tumor growth, it significantly increased the number of cancer cells adhered to the omentum 7 days post-injection ( Fig. 7B ). This phenotype was omentum-specific, with no differences in adhesion in any of the other intraperitoneal organs tested, including other visceral fat deposits ( Fig. 7C ). Cells that adhered to the omentum at the 7-day time point were removed and selected by sorting for RFP+ cells (Supplementary Fig. S10B) and analyzed for transcriptional changes of integrins known to bind fibronectin. Although ITGB3 expression was decreased in both vehicle and NVP-treated cells relative to mock, ITGAV expression was elevated in NVP-treated cells relative to mock ( Fig. 7D ). Taken with the increase in omental adhesion, these data suggest that IGF1R inhibition enhances omental adhesion by increasing adhesion to fibronectin. We next evaluated integrin expression in the subcutaneous tumors treated with NVP-AEW541 via immunoblot and found that NVP-AEW541 similarly increased integrin αV and decreased expression of integrin β3 relative to vehicle-treated tumors ( Fig. 7E ). There were no changes in expression of integrin α5 and integrin β1 ITGB1 , which are also putative fibronectin-binding integrins. We next asked whether the opposing changes in ITGAV and ITGB3 expression could be temporal, and thus probed the RNA from the early/late tumor model for gene expression of fibronectin-binding integrins. We found that ITGAV was significantly enriched in established (late) tumors, whereas ITGB3 was significantly enriched in early tumors ( Fig. 7F ). Importantly, when we measured the abundance of the αVβ3 heterodimer on the surface of NVP-AEW541-treated cells in vitro , we found that inhibiting IGF1R increased αVβ3 integrins in a dose-dependent manner, except for SKOV3 cells, which had the highest baseline expression ( Fig. 7G ; Supplementary Fig. S10C). NVP-AEW541 reduces cell viability at 10 µmol/L (Supplementary Fig. S10D), and as we observed, some increases in αVβ3 integrins at 2.5 µmol/L with better cell viability we additionally tested adhesion to fibronectin with 2.5-µmol/L NVP-AEW541. In all lines except for SKOV3, 2.5-µmol/L NVP-AEW541 significantly enhanced cell adhesion to fibronectin (Supplementary Fig. S10E). Lastly, when OVCAR8-RFP-LUC cells are injected intraperitoneally and growth is measured over an extended period, NVP-AEW541 treatment decreased intraperitoneal tumor growth and omental tumor burden after 28 days, as measured by IVIS imaging and RFP + cells ( Fig. 7H and I ), which caused a slight increase ( P = 0.12) in early tumor growth (Supplementary Fig. S10F and S10G), mirroring the earlier subcutaneous and intraperitoneal models. Taken together, our findings implicate preadipocyte-mediated tumorigenesis in ovarian cancer involving IGF1 activation of alternative NF-κB and subsequent ECM remodeling that inhibits adhesion during early stages of metastasis but promotes proliferation and tumor growth of established tumors.

In this study, we sought to determine the role of preadipocytes in ovarian cancer tumorigenesis. Using primary omental preadipocytes and mature adipocytes, we assessed phenotypic, transcriptomic, and proteomic changes in cancer cells exposed to each adipocytic population to dissect novel mechanisms of omental metastasis. Our findings demonstrate that preadipocytes support ovarian cancer progression via upregulation of a prosurvival cell-autonomous matrisome regulated by an IGF signaling axis ( Fig. 8 ). Specifically, CM containing preadipocyte-secreted factors increases cell viability over time, enhances colony formation, and promotes adhesion to fibronectin through regulation of integrin expression. Moreover, our limiting dilution xenograft model demonstrates that preadipocytes support tumor initiation in vivo comparable to mature adipocytes, a novel finding that contributes to our understanding of the omental microenvironment. Omental preadipocytes mediate an IGF1-AKT-NF-KB signaling axis in ovarian cancer cells that inhibits integrin-mediated adhesion to fibronectin but promotes the proliferation and growth of established tumors. Inhibition of IGF1 signaling via NVP-AEW541 treatment leads to increased adhesion to omentum in the early stages of metastatic growth and decreased proliferation in the late stages of tumor growth. (Created with BioRender.com .) We initially performed the subcutaneous injections as a way to measure the tumor-initiating ability of cancer cells in the presence of preadipocytes as xenotransplantation limiting dilution assays have been the gold standard for assessing tumor-initiating cells ( 46 , 47 ). An important limitation of this model is the contribution of subcutaneous adipocytes, which have molecular profiles that are distinct from the visceral adipocytes associated with the ovarian TME ( 48 – 50 ) and as such could influence the phenotypes observed. Other studies found no difference in migration ability or chemotherapy resistance in ovarian cancer cells treated with subcutaneous derived factors relative to visceral derived ( 51 , 52 ). We did not observe any tumor growth with the low dilution of tumor cells injected alone in the subcutaneous model; however, we cannot rule out interference of subcutaneous adipocyte-derived factors in the tumors that developed when coinjected with preadipocytes. The discovery that preadipocytes enable tumor initiation in conditions that are otherwise nonpermissive builds on other studies demonstrating that stromal cells regulate tumorigenic potential. A study with cancer-associated mesenchymal stem cells demonstrated that cancer-associated mesenchymal stem cells increase cancer stem cell capacity, which we did not observe with preadipocytes, and suggests that these two populations of adipocyte precursor cells promote tumorigenesis via distinct mechanisms ( 53 ). It is also worth noting that MSCs are described as a tumor-infiltrating cell type, whereas preadipocytes are omentum-resident cells, and that MSCs acquire a protumorigenic phenotype after transformation into cancer-associated MSCs ( 53 ). As preadipocytes were required for tumor initiation at low dilutions but did not modulate other well-established stem cell phenotypes in vitro , preadipocytes may be influencing transient phenotypes required for tumor initiation, such as viability under stress. Cellular stress can be induced by hypoxic conditions, treatment with cytotoxic agents, nutrient deprivation, and isolation stress, where cells lack cell–cell contract or adhesion to a matrix. Colony formation is an in vitro model of isolation stress, and one of the first steps in this survival process is integrin-mediated cell adhesion ( 54 ). In pancreatic cancer, αVβ3 integrins confer stress tolerance by upregulating fibronectin to form a specialized niche, which is important in ovarian cancer due to the enrichment of fibronectin in metastasis, and was further supported by the increase in FN1 in coculture ( 33 , 55 ). Furthermore, NF-κB can mediate integrin expression and is a critical pathway in ovarian cancer for regulating survival and proliferation during tumorigenesis, both of which were affected by inhibition of the IGF1R ( 19 , 23 ). Previous studies have also shown that αVβ3 integrins can directly interact with the IGF1R ( 56 ). Increased collagen deposition is associated with increased metastatic burden in ovarian cancer and collagen remodeling signatures are associated with poor outcomes in patients with cancer ( 57 , 58 ). Collagens I and VI are elevated in the remodeled matrisome of advanced ovarian cancer, and collagen VI has additionally been implicated in cisplatin resistance ( 43 , 59 , 60 ). Our data demonstrating differential expression of collagen 1 can perhaps be explained by the role of fibronectin in the assembly of collagen 1 fibers. Treatment with IGF1 upregulates COL1A2 and COL6A2 in vitro , whereas inhibition of IGF1R in vivo decreases the deposition of collagen 1 and slows both subcutaneous and intraperitoneal tumor growth. These findings are corroborated by recent studies demonstrating that IGFR1 is a negative prognostic indicator in late-stage ovarian cancer treated with platinum-based chemotherapy ( 61 ). The finding that inhibition of IGF, although slowing the growth of tumors, promotes omental adhesion is paradoxical and perhaps reflects the dynamic and transient phenotypic states of cancer cells during the metastatic process. It is conceivable that cellular resources for proliferation are redirected toward the expression of integrins such as αVβ3 required for both survival and adhesion during metastasis to overcome isolation stress. Further studies investigating the role of preadipocytes in facilitating ECM production and efficient colonization of the omentum during metastatic seeding will provide new insights into ovarian cancer biology and potentially identify novel targets to prevent ovarian cancer progression.

Ovarian cancer is the most lethal gynecological cancer mainly due to its late diagnosis and limited therapeutic options ( 1 ). Ovarian cancer is exacerbated by its complex tumor microenvironment (TME) via exposure to a variety of signals from within the peritoneal space that support disease progression. Many of these signaling factors are secreted by cells of the omentum, a dynamic endocrine organ comprised of lipid-storing mature adipocytes, fibroblastic preadipocytes, immune cells, and other stromal cells ( 2 , 3 ). Importantly, in the clinic, the omentum is the primary site of metastasis for ovarian cancer cells and the organ with the largest tumor burden in metastatic disease ( 4 , 5 ). During peritoneal metastasis adipocytes secrete adipokines that encourage cancer cell homing to the omentum, which is lined by mesothelial cells that facilitate cancer cell adhesion via mechanisms that rely on MMP2 and fibronectin, as well as CD36 ( 4 , 6 – 9 ). Following invasion through the mesothelium, mature adipocytes of the omentum provide fuel to ovarian cancer cells in the form of secreted fatty acids. Delipidation of mature adipocytes that occurs during initial ovarian cancer colonization causes dedifferentiation and a concomitant increase in preadipocyte like cells in the omentum ( 10 ), which may play a critical role in perpetuating metastasis. A similar phenomenon has been described as transdifferentiation into a fibroblast like population that is protumorigenic ( 11 ). Although mature omental adipocytes provide metabolic support, the role of the fat-precursor preadipocytes in metastatic tumorigenesis is understudied and warrants further investigation. It has been shown in other models that preadipocytes secrete a variety of cytokines, chemokines, and growth factors that are distinguishable from mature adipocytes ( 12 , 13 ). Preadipocytes make up 15% to 50% of the stromal cell compartment of visceral fat ( 14 ), and preadipocytes derived from omental adipose tissue produce more collagens, matrix metallopeptidases and other proteins associated with extracellular matrix (ECM) remodeling, both relative to preadipocytes derived from subcutaneous fat and relative to differentiated adipocytes from the same fat compartment ( 15 , 16 ). Moreover, the matrisome of ovarian cancer is responsive not only to signals from the TME but also in response to therapeutic pressure ( 17 ), which suggests that an improved understanding of the ovarian cancer matrisome response to external signals could reveal new avenues to combat therapy resistance. Here, we investigated the role of preadipocytes in supporting tumorigenic properties of ovarian cancer cells using in vitro coculture experiments, RNA sequencing analyses, and mouse models including a limiting dilution subcutaneous model and intraperitoneal models, which taken together indicate that preadipocyte-induced signaling programs support tumorigenesis and progression.

CAOV4 (HTB-76), OVCAR8 (CVCL_1629), and SKOV3 (HTB-77) were authenticated by and obtained directly from the ATCC and maintained in RPMI medium supplemented with 10% FBS and 1% penicillin–streptomycin (Gibco) at 37°C and 5% CO 2 . ACI-23 cell line was provided by the National Cancer Institute via Material Transfer Agreement and maintained in Dulbecco Modified Eagle Nutrient Mixture F-12 (DMEM/F12) medium (Gibco) supplemented with 10% FBS and 1% penicillin–streptomycin (Gibco) at 37°C/5%CO 2 . Human omental preadipocytes were purchased from Zen-Bio Inc. and maintained in omental preadipocyte media (OM-PM). Omental preadipocytes were differentiated into mature adipocytes according to manufacturer’s instructions and maintained in omental adipocyte media (OM-AM). Briefly, differentiation was achieved using 1 week in differentiation media followed by at least 1 week in OM-AM. Cells were not lifted for the duration of the differentiation and subsequent experiments. When applicable, cells were starved in a serum-free media (SFM) consisting of DMEM/F12, 0.1% BSA, and 1% penicillin–streptomycin. Preadipocyte-CM was made by washing preadipocytes and replenishing them with low-volume SFM for 48 hours. CM was spun at 3,000 rpm for 10 minutes and aliquots were stored at −20°C. For indirect cocultures, preadipocytes were seeded and differentiated in the bottom well of transwell cell culture plates for 2 weeks prior to coculture. Preadipocytes (bottom well) and cancer cells (top well) were seeded the day coculture began. Indirect cocultures with preadipocytes were seeded in a 12-well transwell plate both at 1:40 and 1:1 cancer cell: preadipocyte ratios and maintained for 6 days in a combination of DMEM/F12:OM-PM or RPMI:OM-PM (ACI-23: DMEM; CAOV4, OVCAR8, SKOV3: RPMI). Mature adipocytes were generated in the bottom well of a transwell plate and cocultured with ovarian cancer cells seeded in the upper well at a 1:40 cancer cell: adipocyte ratio. Cultures were maintained for 6 days in a combination of DMEM/F12:OM-AM. Direct cocultures were plated onto 0.2% gelatin-coated coverslips in a 24-well plate with a low ratio of cancer cells to 5-chloromethylfluorescein diacetate (CMFDA)–labeled preadipocytes (1:40) and grown for 4 days. Cells were cultured to a confluency of 60% to 70%; 30-nmol/L pools of four siRNAs (Horizon) targeting COL1A2 , COL6A2 , or a nontargeting control (siNEG) were transfected using Lipofectamine RNAiMAX (Thermo Fisher Scientific) per the manufacturer’s instructions. Gene silencing was confirmed via qRT-PCR. Cells were harvested using Cellstripper (Corning), counted, and 5 × 10 5 to 1 × 10 6 cells were stained with secondary-conjugated primary antibodies (Supplementary Table S1). ALDH1 activity was measured using the AldeFluor assay kit (Stemcell Technologies) per the manufacturer’s instructions, as previously described ( 18 ). Flow cytometry was performed either on a BDFACS Melody or a BDFACS Aria. Colony formation was measured after coculture in a six-well transwell plate for 7 days. Five hundred cancer cells/well were seeded in complete media (RPMI-1640 or DMEM-F12), and preadipocytes (Zen-Bio) were seeded with 2 × 10 4 cells in OM-PM in the transwell. On Day 7, the transwells were removed and cells were washed with dPBS. Cells were stained with 0.5% crystal violet for 10 minutes at room temperature and washed until all residual crystal violet was removed. Plates were dried overnight, imaged with a high-content scanner, and quantified manually. Cells were seeded into 96-well, white-wall clear-bottom plates and grown under specified conditions. Viability was assessed with CellTiter-Glo 2.0 (Promega) per the manufacturer’s instructions on a SpectraMax iD3 plate reader (Molecular Devices). Cells were seeded into 96-well, white-wall clear-bottom plates and allowed to adhere overnight in complete media. Cells were then starved in SFM for 4 to 6 hours. SFM was aspirated and replaced with either fresh SFM or CM. Cells were cultured for 3/7 days, and apoptosis was measured using the Caspase-Glo 3/7 Assay (Promega) per the manufacturer’s instructions. Plates designated for a 7-day experiment had their media replenished on Day 3. Luminescence was measured on a SpectraMax iD3 plate reader. Cells were washed in cold PBS, fixed in 4% paraformaldehyde for 10 minutes and blocked in 10% goat serum. Primary antibodies were incubated at 4°C overnight in a blocking solution using antibodies to collagen 1, collagen 6, fibronectin, IGFBP5, and MMP2 (Supplementary Table S1). Slides were incubated with a secondary antibody conjugated to AlexaFluor 555 (Thermo Fisher Scientific) at room temperature for 1 hour. Nuclei were stained with 1-µg/mL Hoechst. Slides were imaged on a Primo Star HAL/LED microscope (Zeiss) using CellSens software. Tumors were resected, fixed in 10% neutral buffered formalin, and stored in 70% ethanol. Tumors were embedded in paraffin and sectioned at 5 μm. Antigen retrieval was performed in citrate buffer and quenched with hydrogen peroxide. Slides were incubated with primary antibodies (Supplementary Table S1) overnight at 4°C followed by an HRP-linked secondary for 1 hour at room temperture and processed using DAB (Vector, SK-4100). Slides were imaged on a Primo Star HAL/LED Microscope (Zeiss) using CellSens software. Digital quantification of DAB staining was performed on four randomly chosen fields per slide using ImageJ with a FIJI deconvolution package as described previously ( 19 ). RNA was harvested from ACI-23 cells, omental preadipocytes, and differentiated adipocytes using the Direct-zol RNA Miniprep Plus Kit (Zymo). Final RNA concentration was determined with a SpectraMax QuickDrop (Molecular Devices) and purified using RNeasy MinElute Cleanup Kit (Qiagen). Sequencing was performed by HiSeq 4000 at the IGM Genomics Center at the University of California, San Diego UCSD. Quality control, alignment, normalization, differential gene expression by DEseq2, and gene set enrichment analyses were completed by the Center for Computational Biology and Bioinformatics at University of California, San Diego. Metascape was used for gene annotation and ontology. Differential genes (adjusted P 0.5) were ranked for pathway enrichment in the Reactome Database in Metascape. Reactome gene ontology terms or genes from selected Reactome pathways (ranked by log 2 FC > 1) were visualized as dot plots using ggplot2. Secreted factors from cocultures and monoculture controls were measured using the Proteome Profiler Adipokine Array (ARY024) from R&D Systems according to manufacturer’s instructions. The NF-κB Luciferase Cignal Lenti Reporter Assay (Qiagen) was transduced per the manufacturer’s protocol at MOI of 50, selected with 2-μg/mL puromycin, and expanded to generate stably expressing reporter cell lines. Cells were plated in 96-well, white-wall clear-bottom plates at 1,000 cells/well with treatments added the following day. Cells were treated for 15 minutes, 1 hour, 4 hours, and 24 hours before analysis using the ONE-Glo EX Luciferase Assay System (Promega) per the manufacturer’s protocol. Luminescence was read on a SpectraMax iD3 plate reader. Ninety-six-well black-wall plates were coated with 250-µg/mL fibronectin (Sigma-Aldrich), Matrigel (Corning), and 0.2% gelatin (Sigma-Aldrich) overnight at 4°C and 60 minutes at 37°C, followed by aspiration of substrates before plating. Cells were pretreated with SFM or CM overnight, and single-cell suspensions were generated using Cellstripper (Corning) and plated at a density of 1,000 cells/well in SFM for 4 hours at 37°C. Plates were washed twice with dPBS and adherent cells were fixed with 4% paraformaldehyde (methanol free) for 10 minutes at room temperature and then washed again. For quantification of cell attachment, fixed cells were permeabilized and blocked with 0.2% Triton X-100% and 3% BSA in PBS for 30 minutes at room temperature and then washed. Permeabilized cells were incubated with Hoechst (5 μg/mL) and Phalloidin-AF647 (10 μg/mL) in 0.02% Triton X-100% and 0.3% BSA in PBS at 4°C overnight. The following day, plates were washed and maintained with 100 μL/well of PBS before imaging in the Pico ImageXpress. Cell number was quantified as the number of f-actin/Hoechst dual positive cells. Oil Red O was purchased (Sigma-Aldrich) and used per the manufacturer’s instructions. Spheroid formation was measured in cells treated with SFM/CM as previously described ( 19 ). Briefly, cells were plated with 300 cells per well in an untreated, black-walled plate in 50:50 stem cell media: SFM/CM. Spheroid formation was measured after 7 days. IGF1 was measured using human IGF1 DuoSet sandwich enzyme-linked immunosorbent assay (ELISA; R&D Systems) per the manufacturer’s instruction using media harvested from cocultures. Cells were lysed in NP-40 with a 1X Halt protease/phosphatase inhibitor cocktail (Thermo Fisher Scientific). Nuclear and cytosolic fractions were collected using Active Motif Nuclear Extract Kit per the manufacturer’s instructions. Protein concentration was determined using Rapid Gold BCA Protein Assay (Thermo Fisher Scientific). Samples generally comprising 10 to 30 µg of protein were mixed with 1X bond-breaker TCEP solution (Thermo Fisher Scientific) and 1X NuPAGE LDS sample buffer (Thermo Fisher Scientific), boiled for 5 minutes at 95°C, and then subjected to SDS-PAGE using 4% to 12% Bis-Tris gels with MOPs running buffer followed by transfer to a polyvinylidene difluoride membrane for immunoblotting analysis. Primary antibodies were incubated overnight at 4°C using dilutions described in Supplementary Table S1. Secondary antibodies were incubated for 1 hour at room temperature. Blots were imaged using an iBright CL1000 Imaging System. Densitometry analysis was performed using iBright Analysis Software. Total RNA was isolated using the NucleoSpin RNA Plus Purification Kit or, in the case of cocultures, Direct-zol RNA Miniprep Plus Kit (for RNA sequencing) or E.Z.N.A. MicroElute Total RNA Kit (for qPCR) per the manufacturer’s instructions. RNA concentration was determined with a SpectraMax QuickDrop and converted into cDNA using the High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific). Quantitation of gene expression was performed with Taqman Fast Advanced Master Mix and Taqman probe assays (Thermo Fisher Scientific; Supplementary Table S2). Unless otherwise noted, genes are reported relative to GAPDH as a control using the comparative threshold cycle method. qPCR reactions were run and analyzed with the QuantStudio 3 machine and design software. For the limiting dilution subcutaneous xenografts, 8 × 10 5 , 2 × 10 4 , or 500 ACI-23 CD19-tagged cells either with or without 2 × 10 4 omental preadipocytes in 200 μL of 1:1 Matrigel and PBS were subcutaneously injected into the left flank of 8-week-old female athymic Nu/J mice. ACI-23-CD19 cells were generated as previously described ( 20 ). Mice were weighed, and tumors were measured twice weekly by researchers blinded to treatment. Mice were sacrificed once tumors reached 20 mm in length or after 90 days of growth. To investigate tumor initiation, the number of days for palpable tumors to seem (time to tumor initiation) was averaged. Cells for the transplantation assay were isolated via FACS by sorting for CD19 + cells and cells from five mice were used to inject two with preadipocytes and two without. To investigate early/late transcriptional events in preadipocyte-induced tumorigenesis, “early” tumors were randomized and resected once all mice had formed tumors, and the “late” time point was selected at the measurement at which the first tumor grew to 20-mm length. Tumors for RNA were snap-frozen in LN 2 and homogenized in RNA lysis buffer using a bead homogenizer. For the NVP-AEW541 subcutaneous model, cells were prepared in the same manner as the 1:40 arm of the limiting dilution. To ensure suppression of host IGF signaling at early time points, mice were pretreated with 50-mg/kg NVP-AEW541 in 4% DMSO:30% polyethylene glycol 300:5% Tween 80 or with the vehicle control via oral gavage 24 hours prior to cell injection. Thereafter, mice were treated once daily for the duration of the study. Tumors were measured in a blinded fashion as described above, and all tumors were resected once the first tumor reached 20-mm length. Mice that received intraperitoneal injection were treated with NVP-AEW541 or vehicle as described above, with an injection of 1 × 10 6 OVCAR8-RFP-LUC in dPBS 24 hours after pretreatment. After 4 hours and 7 days, organs were resected and assayed using 150-µg/mL IVISbrite D-Luciferin Potassium Salt Bioluminescent Substrate (Caliper). Cells were removed from the 7- and 28-day omentum and splenic fat and isolated via FACS for RFP. For 28-day experiments, live images were acquired twice a week under sedation with isoflurane from XGI-8 Gas Anesthesia Unit with an in vivo imaging system IVIS spectrum imaging system (Caliper Life Sciences, Perkin-Elmer) after 150-mg/kg IVISbrite D-Luciferin Potassium Salt Bioluminescent Substrate (Caliper). Living image software (Perkin-Elmer) was used for region of interest analysis of spectral images, with the same region of interest area and shape used for all animals. Statistics were generated using Prism 8.4.3 with data acquired from at least three independent biological replicates. Results are presented as mean ± SD. Significance was calculated using either a Student t test for comparisons of two means or ANOVA for comparisons of three or more means with post hoc test to identify differences between groups as described in figure legends. Differences between means are considered statistically significant at the 95% level ( P < 0.05). No stars or a label of ns is not significant. All animal studies were approved by the SDSU Animal Care and Use Committee (protocol approval number 21-05-003H). RNA sequencing data has been deposited to the NCBI Gene Expression Omnibus under accession number GSE262869 . The expression of MMP2 , COL6A1 , COL6A2 , FN1 , COL1A1 , COL1A2 , and IGFBP5 was downloaded from the Ovarian Serous Cystadenocarcinoma in The Cancer Genome Atlas on the Cancer Genomics Portal website ( 21 ). Of the 585 total cases, 300 had mRNA normalized expression. The Spearman correlation between genes and correlation matrix visualization was completed using the R package “PerformanceAnalytics” ( https://cran.r-project.org/web/packages/PerformanceAnalytics/index.html ). COL1A1 and COL1A2 expression for donor-matched primary and omental metastatic patient samples was obtained via Geo2R from the following accession numbers: GSE178913 , GSE98281 , and GSE137237 . All other raw data are available upon request from the corresponding author.

Characterization of preadipocytes gene expression and tumor promoting properties of preadipocytes versus mature adipocytes Analysis of resected tumors Analysis of classic TIC markers Proliferation assay PCA and GSEA Direct coculture images Antibody information, Taqman probe information, and Gene expression data IL-6 and CCL-2 stimulation knockdown of COL1A2 and COL6A2 Validation of luciferase labeled cells used in ip model