Tumor-Infiltrating Nociceptor Neurons in Ovarian Cancer Treatment Resistance.

We thank the following funding sources for their support of this work. Institutional Development Award (IDeA) from the National Institute of General Medical Sciences of the National Institutes of Health grants P30GM145398 (PDV) & the Kansas Institute for Precision Medicine P20GM130423 (AKG), the National Cancer Institute grant R01CA260132 (AKG), University of Pennsylvania Basser Center for BRCA External Grant Program Innovation Award (PDV & AKG), National Institutes of Health/National Institute of Dental and Craniofacial Research grant R01DE032712 (PDV), National Institutes of Health P50 grant CA228991 (RD), Dr. Miriam and Sheldon G. Adelson Medical Research Foundation (RD), Penn ACC Translational Center of Excellence in Ovarian Cancer (RD), Honorable Tina Brozman Foundation for Ovarian Cancer Research (RD & AKG), a grant from the Ovarian Cancer Research Alliance (A.K.G.), and Predicine, Inc (AKG) as well as graduate student fellowship awards from the OVERRUN Ovarian Cancer Foundation (to CVT and ALT), Claneil Foundation (RD), Helene Ross Bogutz Fund for Ovarian Cancer Early Detection (RD), and the Marjorie S. Stanek and Lowell H. Dubrow Ovarian Cancer Research Center Endowed Fund (RD). AKG is the Chancellor's Distinguished Chair in Biomedical Sciences Endowed Professor. The Histology and Imaging Core (Sanford Research), supported by the National Institute of Health, National Institute of General Medical Sciences P30GM145398, provided their services and expertise towards this project.

Following Institutional Review Board (IRB) approval and patient consent, tumor samples were obtained from the University of Pennsylvania through the Ovarian Cancer Research Center Tumor BioTrust Collection ( https://www.med.upenn.edu/OCRCBioTrust/ ) (RRID: SCR_02287) or from the Sanford Health pathology department. Formalin‐fixed paraffin‐embedded (FFPE) ovarian cancer cases consisted of: High‐Grade Serous Ovarian Carcinoma (HGSOC, n = 19), Low‐Grade Serous Ovarian Carcinoma (LGSOC, n = 36), clear cell carcinoma ( n = 12), advanced clear cell carcinoma ( n = 4), mucinous ovarian carcinoma ( n = 7), and endometrioid ovarian carcinoma ( n = 7). Matched pre‐ and post‐treatment samples of HGSOC ( n = 18) were similarly obtained. Tissue sections were immunohistochemically (IHC) stained for βIII‐tubulin to identify tumor‐infiltrating nerves. Samples collected were from biopsy, initial debulking surgery, or post‐treatment resection. All patient samples were collected from females. Samples were collected 2–9 years prior to use in this study. Sections analyzed came from FFPE blocks containing the most amount of cancer tissue for each patient. Samples were shipped at room temperature. Immunohistochemical (IHC) staining and analysis was performed on entire tumor sections. All samples analyzed were primary tumors. Fresh tumor tissue (used for MEA analysis) was similarly collected following IRB approval and patient consent. Tissues were shipped overnight at 4°C in Tissue Storage Solution (Miltenyi Biotec, #130‐100‐008) from the University of Pennsylvania to Sanford Research and analyzed immediately upon arrival. Bright field microscopy was used to analyze βIII‐tubulin IHC‐stained tumor sections, which were scored as follows: score of 0, no positive staining; score of 1 indicates 10% positive staining; score of 2 indicates 30% positive staining; score of three indicates more than 30% positive staining. Five random fields/sample were analyzed; each field predominantly consisted of tumor. This scoring scheme follows what is typically used by pathologists [ 91 ]. Innervation in regions lacking tumor was not assessed. In addition, metastatic lesions were not analyzed for innervation. Immediately upon arrival, fresh tumors were sliced using a scalpel and the approximately 900‐µm‐thick tissue slices were kept in oxygenated artificial cerebrospinal fluid (ACSF; 119 m m NaCl, 2.5 m m KCl, 1 m m NaH 2 PO 4 , 26.2 m m NaHCO 3 , 11 m m glucose, 1.3 m m MgSO 4 , and 2.5 m m CaCl 2 ) at room temperature to preserve neural circuits within the slices. At least n = 3 slices/sample were analyzed; more slices were generated from larger samples. An MEA1060‐Inv‐BC microelectrode array system (Multichannel Systems) was used for MEA analysis. MEAs contain multiple microelectrodes that stimulate and record electrical activity from overlying cells or tissue slices [ 47 ]. We used a perforated MEA (pMEA100/30) to ensure ample contact between the tumor slice and electrodes. In addition, a low‐level vacuum was used to generate negative pressure between the perforations and the tissue slices to ensure that direct contact was made and maintained for the duration of recordings. The perforated MEAs used contain a 6 × 10 electrode grid in which 30 µm electrodes are spaced 100 µm apart. Electrical activity was recorded at room temperature; 25‐kHz sampling frequency was used along with the Butterworth 2nd‐order digital filter, which was set to high pass with a cutoff frequency of 10 Hz (to eliminate slow field potentials). A STG4000 stimulus generator (Multichannel Systems) was used for electrical stimulation which was biphasic voltage (−0.5 V and + 0.5 V each) for 100 µs, was repeated after a 23 ms interval and was applied to selected electrodes and evoked spike activities were recorded. Electrical activity was continuously recorded for 1–5 min/slice and analyzed using the MC_Rack 4.6.2 software from Multichannel Systems. Baseline recordings were followed by stimulation. Representative tracings are shown (Figure S4 ) and demonstrate: baseline activity, after which selected electrodes were stimulated resulting in an evoked response from the other electrodes. The stimulus was then turned off and electrical activity returned to baseline. Nearly all samples had absent or very low baseline activity; stimulation of one or more electrodes induced evoked responses in all slices. N = 5 HGSOC, n = 1 LGSOC, n = 2 endometrioid carcinomas and n = 2 normal ovaries were analyzed. The low incidence of LGSOC and endometrioid carcinoma resulted in the low number of samples obtained. Clear cell carcinoma is very rare in the United States, and we were unable to obtain any fresh clear cell carcinoma tissues for MEA analysis. All in vivo animal studies were performed within the Animal Resource Center (ARC) at Sanford Research, whose Animal Welfare Assurance is on file with the Office of Laboratory Animal Welfare. The Assurance number is A‐4568‐01. Sanford Health is also a licensed research facility under the authority of the United States Department of Agriculture (USDA) with USDA certificate number 46‐R‐009. AAALAC, Intl has also accredited the Sanford Health Animal Research Program. The ARC is a specific pathogen‐free facility where mice are maintained in IVC Tecniplast Green line Seal Safe Plus cages which are opened only under aseptic conditions in an animal transfer station. Aseptic technique is used to change animal cages every other week. All cages have individual HEPA filtered air and animal rooms are maintained at 23.9°C, 30%–70% humidity, have a minimum of 15 air changes per hour, and have a 14:10 light/dark cycle. Corncob bedding and nesting materials, both autoclaved prior to use, are maintained in all cages. Animals are fed irradiated, sterile food (Envigo) and provided acidified water (pH 2.8–3.0) ad libitum. There is a maximum of 5 mice/cage and they are observed daily (technicians and research staff looking for abnormal behavior, signs of illness or distress, the availability of food and water and proper husbandry). All animal experiments were performed under an approved Sanford Research IACUC protocol (2023‐0105), within institutional guidelines and comply with all relevant ethical regulations. Control (C57BL/6J wildtype) animals were at least 6 weeks old. C57BL/6J mice were each approximately 20–24 gm in weight and purchased from The Jackson Laboratory. Double transgenic (TRPV1 cre ::DTA fl/wt ) animals were generated by crossing TRPV1‐Cre (The Jackson Laboratories, #017769; RRID:IMSR_JAX017769) with Rosa26‐DTA mice (The Jackson Laboratory, #009669; RRID: IMSR_JAX:009669) [ 92 ]. The resulting animals (TRPV1 cre ::DTA fl/wt ) express DTA (diphtheria toxin fragment A) under control of the TRPV1 promoter resulting in genetic ablation of all TRPV1‐expressing cells (including neurons) throughout development. These transgenic mice have been previously characterized and validated [ 50 , 67 ]. C57BL/6J or nociceptor neuron ablated TRPV1 cre ::DTA fl/wt female mice, were subcutaneously implanted in the hindlimb with 1 × 10 5 cells. This non‐orthotopic site was selected to enable accurate and repeated tumor growth measurements. Tumor growth was monitored weekly by caliper measurements and tumor volume was calculated using the formula: ½ (length x width) [ 2 ]. Once a palpable tumor was present (approximately day 10 post tumor implantation), animals received weekly intraperitoneal injection with 50 mg/kg carboplatin in 100 µl. The carboplatin dose was based on a previous publication and is clinically relevant [ 93 ]. Anti‐β‐III Tubulin (Abcam, Cat# ab78078, 1:250, RRID:AB_2256751). Antibody utilized for Immunofluorescent staining of Dorsal Root Ganglia (DRG) Anti‐β‐III tubulin (Abcam, Cat# ab18207, 1:500, RRID:AB_444319). Anti‐CD9 (Abcam, Cat# ab92726, 1:500, RRID:AB_10561589) Anti‐CD81 (Santa cruz, Cat# sc‐166029, 1:500, RRID:AB_2275892) Anti‐CD63 (BD Biosciences, Cat#556019, clone H5C6, 1:1000, RRID:AB_396297) Anti‐TRPV1 (Alomone Labs, #ACC‐030, 1:500, RRID:AB_2313819). FFPE tissues were cut at 5 µm thickness. We utilized the Discovery Ultra automated slide staining system (Ventana Medical Systems, Inc.) for deparaffinization and antigen retrieval. The antigen retrieval step was performed using the Ventana CC1 solution, which is a basic pH tris‐based buffer. Tissue was incubated with the antibody cocktail for 1 h at 37°C. Tissues were rinsed with TBS and incubated with chromogen Betazoid DAB (Biocare) for 5 min. Slides were counterstained with hematoxylin, dehydrated, cleared, and coverslipped. Samples were analyzed using an Olympus BX‐51 bright‐field inverted microscope. Immortalized human (female) ovarian carcinoma cell lines representing the following subtypes: HGSOC, Clear Cell, Mucinous, and Endometrioid were grown under normal conditions of 5% CO 2 at 37°C. All cell lines were a kind gift from Dr. Ronny Drapkin (University of Pennsylvania) and Dr. Andrew K. Godwin (University of Kansas Medical Center) and were obtained between 2015 and 2017. All cell lines tested are free of mycoplasma and have been authenticated. None of the cell lines appear on the database of misidentified cells. Media requirements for each cell line are as follows: OVCAR‐3 (RRID# CVCL_0465), Fu‐OV‐1 (CVCL_2047), OVSAHO (RRID# CVCL_3114), Kuramochi (RRID# CVCL_1345): DMEM:HAMS F12 at 1:1 with 10% fetal calf serum (FCS). OVCAR‐4 (RRID# CVCL_1627), TOV‐112D (RRID# CVCL_3612), MCAS (RRID# CVCL_3020), OVTOKO (RRID# CVCL_3117), OVMANA (RRID# CVCL_3111), OVISE (RRID# CVCL_3116): RPMI 1640 with 10% FCS. JHOS‐4 (RRID# CVCL_4649), JHOC‐5 (RRID# CVCL4640), JHOC‐7 (RRID# CVCL_4641): DMEM:HAMS F12 at 1:1 with 10% FCS and 1% L‐Glutamine+ MEM NEAA. TOV‐21G (RRID# CVCL_3613): MCDB105 and Medium 199 at 1:1 with 15% FCS. CaOV3 (RRID# CVCL_0201): DMEM with 10% FCS. Mouse HGSOC cells (MOSEC Trp53 −/− ;Pten −/− DKO clone 4) (RRID# CVCL_ B6JA) were maintained in Alpha‐MEM with ribonucleosides, deoxynucleosides and L‐glutamine (Gibco #12571‐048), 10 µg/mL insulin‐transferrin‐sodium selenite (Roche #11074547001), 20 pg/mL β‐estradiol (Sigma #E8875), 10% FCS [ 23 ]. FT190 (RRID# CVCL_UH57), FT194 (RRID# CVCL_UH58), FT246 (RRID# CVCL_UH61). FT cell lines were cultured with DMEM:HAMS F12 at 1:1 with 2% Ultroser G (Crescent Chemical Company, #67042). PEO1 (RRID# CVCL_2686) and PEO4 (RRID# CVCL_2690) cell lines were cultured with RPMI‐1640 with 10% FCS. A2780 (RRID# CVCL_0134), A2780/CP70 (RRID# CVCL_0135), and A2780/C30 (RRID# CVCL_F639) cell lines were cultured with RPMI‐1640 with 10% FCS, 2 m m L‐glutamine, 0.2 units/mL human insulin. To purify sEVs, cells were seeded onto two 150 mm dishes (Fisher, FB012925 ) at 40% confluency and allowed to attach overnight. The following day, the seeding media was removed, cells were washed with sterile phosphate‐buffered saline (PBS, Fisher, MT21040CV) and fed 8 mL of media containing sEV‐depleted fetal calf serum (FCS). sEVs were depleted from FCS as follows. The FCS was spun in an ultra‐centrifuge at 110 000 × g overnight. The following day, the supernatant was collected (leaving behind the sEVs in the pellet), filter sterilized with 0.22 µm syringe filter (EMD/Milli, SLGL0250S), and heat‐inactivated at 56°C for 30 min. sEV‐depleted FCS was added to the growth media at 10% and used to feed cells utilized for sEV purification. The cells were returned to the incubator until 90% confluence was reached at which time the media (containing all cell‐released sEVs) was collected. The conditioned media was centrifuged at 2000 × g for 20 min, the supernatant collected and spun again at 10 000 × g for 30 min to remove any cell debris. The supernatant was removed and put into pre‐cleaned ultracentrifuge tubes (Fisher, 03‐126). Supernatants were ultracentrifuged at 110 000 × g in a Sorval WRX Ultracentrifuge with a Surespin630 rotor for 2 h. The supernatant was discarded, the pellet washed with 10mL of sterile PBS, and spun for another 2 h at 110 000 × g. After this final spin, the supernatant was discarded, and the pellet (containing purified sEVs) was resuspended in 400 uL of sterile PBS. Samples were aliquoted and stored at 4°C short term or −80°C long term. This purification methodology has been previously validated by our group [ 20 ]. Validation of purified sEVs was completed using the Nanosight NS300 Nanoparticle Tracking Analysis (Malvern Panalytical). This instrument utilizes Brownian motion to capture video of particle movement which generates high‐resolution particle size and concentration measurements. Prior to analysis, samples were diluted by adding 10 uL of the purified sEV sample into 750 uL of sterile PBS. Particle size was verified to be within the range for sEVs (50–150 nm). sEVs were also validated by western blotting for CD81, CD9, and/or CD63, accepted sEV markers [ 53 ]. Purified sEVs were lysed with 1% Triton‐X‐100 for 5 min at room temperature. Using a modified Pierce BCA protein assay, the protein concentrations were determined. 30 ug of total protein were separated by SDS‐PAGE and transferred to an Immobilon‐FL transfer membrane (Millipore). The membrane was blocked with 5% bovine serum albumin for 1 h and then incubated in primary antibody overnight at 4°C. Following washes, membranes were incubated in secondary Donkey anti‐Rabbit 800CW and Donkey anti‐mouse 680RD IR Dyes (LI‐COR) or rabbit‐anti‐mouse‐HRP and visualized with a LI‐COR Odyssey FC viewer. DRG isolation was performed on 6–12‐week‐old C57BL/6J mice. Mice were euthanized via 3 L/min CO 2 for 5 min. Animals were then perfused transcardially with 12 mL HBSS (Fisher MT21022CV) using a 26‐gauge needle (Fisher, 14‐826‐15) and a 12 mL syringe (Fisher, 22‐008035). Following perfusion, a segmented dorsal laminectomy, as described by Malin et al. [ 94 ], was completed to expose the spinal cord. Using a dissection microscope and micro‐scissors, the DRG were isolated starting at the lumbar region moving up to the cervical region. DRG were placed in a 35 mm dish (Fisher, FB012920 ) containing 2 mL of HBSS to prevent them from drying out following excision. Dissection was limited to 1 h to improve DRG viability. DRG were plated on an eight‐well chamber slide (Fisher, PEZGS0816) coated with 20 uL of low growth factor/phenol red‐free Culturex (R&D Systems, 3433‐010‐01) and allowed to solidify for 3 min. Next, using micro‐scissors, four DRG were plated per well into the Culturex and incubated at 37°C for 35 min. 200 uL of pre‐warmed media consisting of Hams F12 (Fisher, MT10080CV) containing 10% sEV‐depleted FCS and 1% pen/strep (Fisher, MT300002Cl) was added to each well containing DRG. The DRG were then incubated overnight, and neurite outgrowth assessed the following day as a measure of DRG viability. 3 µg of purified sEV (based on BCA protein assay) were added to viable DRG. DRG treated with 2 ng of recombinant Nerve Growth Factor (NGF) served as a positive control, while treatment with sterile PBS served as a negative control. DRG were returned to the incubator and neurite outgrowth analyzed 48 h later. Forty‐eight hours after sEV treatment, the media was removed from each well, the DRG were washed with Hank's Balanced Salt Solution (HBSS) then fixed with 4% paraformaldehyde (PFA) in HBSS overnight. The following morning, the 4% PFA was removed and DRG were rinsed with HBSS 3 times for 5 min. DRG were permeabilized with 1% Triton‐X‐100 in HBSS (Fisher PI28314) for 8 h and blocked overnight with 10% Normal Goat Serum (Fisher 50‐588‐35) with 1% Triton‐X‐100 in HBSS. The blocking solution was then removed and the DRG were rinsed with HBSS 3 times for 5 min. DRG were then incubated with primary antibody in HBSS overnight. The following day, the DRG were rinsed with HBSS 3 times for 5 min and incubated with secondary antibody Goat anti‐Rabbit AlexaFluor 647 (Fisher A21245) diluted 1:250 in HBSS overnight. DRG were next rinsed with HBSS 3 times for 5 min and coverslips mounted with Vectashield and Dapi (Fisher NC9524612). All solutions used were pre‐warmed to 37°C to prevent DRG dislodgment from the Culturex. Washes and staining were completed in an incubator at 37°C with 5% CO 2 . DRG staining was visualized and imaged on a Nikon A1R TIRF Confocal microscope at a magnification of 4x. Images of vehicle‐ or sEV‐treated DRG were analyzed with Image J using the plugin Neurite J to complete Sholl analysis. Concentric circles were set to 25 µm apart to determine number of neurites and distance of neurite outgrowth. GraphPad was used to analyze ImageJ data using a repeated‐measure ANOVA comparing sEV treatment versus vehicle (sterile PBS treatment). The maximum number of intersections (nMax) or the maximum neurite outgrowth distance (dMax) are reported. To determine if carboplatin treatment of cancer cells impacts sEV‐induced neurite outgrowth, a dose‐response curve of carboplatin was completed to determine the appropriate in vitro treatment dose. Since only HGSOCs are densely innervated, we assessed only HGSOC cell lines including: OVCAR‐4, OVSAHO, and CaOV3. Five thousand cells/well were plated in a 96‐well plate and allowed to attach overnight. The following day, the cells were treated with fresh media containing 1, 5, 10, or 25 ug/mL of carboplatin (Teva 00703‐4244‐01) or vehicle for 24, 48, and 72 h. CellTiter 96 AQueous One Solution Cell Proliferation Assay (Promega, G3580) was used to determine the number of viable cells. Having determined that 10 ug/mL is the optimal dose, 5 × 10 6 cells of each line were plated on 150 mm dishes. The following day, the cells were treated with either 10 ug/mL of carboplatin or vehicle for 48 h. Following treatment, media was changed, and cells remained undisturbed for an additional 48 h after which time the conditioned media was collected and sEVs purified and validated as described. GraphPad Prism 10.4.2 was utilized for all statistical analyses. Details of the analysis (including outliers, sample size, statistical test, post‐hoc adjustments, p values) are included in each figure legend and below.

To define the innervation status of the five histological subtypes of ovarian cancer, FFPE cases of HGSOC ( n = 19), LGSOC ( n = 37), clear cell carcinoma ( n = 9), mucinous carcinoma ( n = 7), and endometrioid carcinoma ( n = 7) were collected. Sections were IHC‐stained for β‐III tubulin, a pan‐neuronal marker. Given that clear cell carcinoma is generally diagnosed at early stage, we also analyzed n = 4 cases of advanced clear cell carcinoma to determine whether innervation increases with advanced disease (Figure  1 ; treatment and histopathological patient information in Figures S1 – S3 ). Staining was quantified by a scorer who was blinded to the tumor type. Interestingly, only HGSOC (Figure  1B,G ) emerged as densely innervated. LGSOC was sparsely innervated (Figure  1A,G ) and the remaining histological subtypes were predominantly lacking tumor‐infiltrating nerves (Figure  1C–G ). Cases of advanced clear cell carcinoma were not more innervated than their early‐stage counterparts suggesting that, at least for this subtype of ovarian cancer, tumor innervation density does not increase with disease stage (Figure  1G ). Immunohistological staining of human ovarian cancers for β‐III‐tubulin. Representative photomicrographs of A) Low‐Grade Serous Ovarian Carcinoma (LGSOC), n = 37 cases; B) High‐Grade Serous Ovarian Carcinoma (HGSOC), n = 19 cases; C) Clear Cell Ovarian Carcinoma, n = 9 cases; D) Advanced Clear Cell Ovarian Carcinoma, n = 4 cases; E) Mucinous Ovarian Carcinoma, n = 7 cases; and F) Endometrioid Ovarian Carcinoma, n = 7 cases. Scale bar, 200 µm. (G) Quantification of β‐III tubulin scoring. Statistical analysis by one‐way ANOVA with multiple comparisons. **, p < 0.01; ****, p < 0.0001. To validate these IHC findings, we turned to micro‐electrode arrays (MEAs). MEAs have been extensively used in the field of neuroscience to study neural circuits in slices of mouse brain [ 47 , 48 , 49 ]. We have previously published that MEA analysis accurately measures electrical activity of fresh tumor slices, that this activity correlates with the density of tumor innervation, and further, that MEA activity is significantly attenuated when tumors are grown in mice genetically ablated of tumor‐infiltrating neurons [ 50 ]. Here, MEA analysis showed that HGSOC slices harbor increased electrical activity as compared to LGSOC, endometrioid carcinoma, and normal ovary (Figure S4 ). Given the rarity of LGSOC and endometrioid carcinoma, very few cases of fresh tumor were available for analysis. Moreover, since the incidence of clear cell carcinoma in the United States is very low, we were unable to obtain fresh clear cell tumor tissue for MEA analysis. While the limited sample size precluded statistical analysis between the groups, the MEA data are consistent with the IHC findings suggesting that HGSOCs are more densely innervated than the other histological subtypes of ovarian cancer. Together, these findings are consistent with our previously published study in which publicly available human datasets were analyzed (OncoLnc, Gepia2, Oncomine). We showed that ovarian tumors with highly expressed neuronal genes suffered a worse survival as compared to those whose tumors harbored a low expression of neuronal genes [ 50 ]. These findings are consistent with other published studies showing that patients with densely innervated tumors have poor survival [ 17 , 51 , 52 ]. Given the stark differences in tumor innervation evident between the histological subtypes of ovarian cancer, we wondered whether this reflected differences in the neurite outgrowth capabilities of their sEVs [ 20 , 21 ]. Therefore, we purified sEVs from n = 7 HGSOC, n = 6 clear cell carcinoma, n = 1 mucinous and n = 1 endometrioid cell lines. These sEVs were validated by western blot for CD9 and CD81, accepted sEV markers (representative full western blot in Figure S5A ) [ 53 ]. In addition, all purified sEVs were assessed by Nanosight particle tracking analysis and found to be of the size (50–150 nm) consistent with sEVs (Figure S5B ). Following validation, sEVs were tested for neurite outgrowth activity on cultured dorsal root ganglia (DRG). Sensory DRG were chosen as we have previously defined that sensory nerves (specifically those expressing TRPV1) infiltrate ovarian cancers [ 23 ]. In this in vitro assay, 3 ug of purified sEVs are added to cultured DRG and neurite outgrowth is quantified 48 h later. sEVs from five HGSOC cell lines (OVCAR‐3, OVCAR‐4, CaOV3, OVSAHO, and Kuramochi) demonstrated robust and significant neurite outgrowth activity similar to that of the positive control (NGF). sEVs from JHOS‐4 and Fu‐OV‐1 did not harbor neurite outgrowth activity (Figure  2A and Figure S6A–G ). These findings are consistent with our IHC analysis of patient HGSOC samples (Figure  1B,G ) demonstrating that the majority of HGSOC cases are highly innervated. Given that the cell of origin of HGSOC is the fallopian tube secretory cell [ 30 , 34 , 54 , 55 ], we used sEVs purified from immortalized human fallopian tube (FT) secretory cells as controls [ 56 ]. Here, the sEVs from three different FT cell lines harbored minimal neurite outgrowth activity (Figure  2B and Figure S7A–C ). The finding that FT cells still promoted low‐level neurite outgrowth suggests that sEVs from non‐malignant cells may have neurite outgrowth activity. In fact, Schwann cells release sEVs that promote regeneration of axons following injury in the peripheral nervous system [ 57 ]. In the tumor microenvironment, recruitment of nerves by non‐malignant cells like cancer‐associated fibroblasts can occur via secretion of soluble neurotrophic factors (e.g., nerve growth factor, NGF) [ 58 ], however, a contribution of non‐malignant sEVs to nerve recruitment remains unknown. sEV‐mediated neurite outgrowth in ovarian cancer. Pie charts indicating the number of cell lines whose sEVs did (blue) or did not (gray) harbor significant neurite outgrowth activity from DRG relative to the negative control. Neurite outgrowth activity was analyzed by Sholl analysis. Positive control was DRG treated with nerve growth factor (NGF, 2 ng); negative control was DRG treated with phosphate‐buffered saline (PBS). N≥3 DRG analyzed per condition. The following groups are shown: A) High‐grade serous ovarian carcinoma (HGSOC) cell lines (OVCAR‐3, OVCAR‐4, CaOV3, OVSAHO, Kuramochi, JOHS‐4, Fu‐OV‐1), B) Fallopian tube (FT) cell lines (FT190, FT194, FT246), and C) Clear cell carcinoma cell lines (JHOC‐7, OVISE, TOV‐21G, OVMANA, OVTOKO, JHOC‐5). Statistical analysis by two‐way ANOVA in which interactions between neurite bifurcations and neurite length were compared to the PBS negative control. Most clear cell carcinomas cases (early stage and advanced) demonstrated little to no tumor‐infiltrating nerves (Figure  1C,G ). This was reflected in the minimal neurite outgrowth activity of clear cell carcinoma sEVs. Only two clear cell‐derived sEVs harbored neurite outgrowth activity (Figure  2C ; Figure S8A,B ) while the majority did not (Figure S8C–F ). Due to the scarcity of cell lines representing mucinous and endometrioid ovarian carcinoma [ 59 ], we were only able to assess the neurite outgrowth activity of sEVs from one of each of these subtypes. sEVs purified from the conditioned media of MCAS cells, a mucinous cell line, harbor modest neurite outgrowth activity (Figure S8G ) while those from TOV‐112D, an endometrioid cell line, harbor minimal neurite outgrowth activity (Figure S8H ). While conclusions cannot be drawn with such limited data, these findings are similar to those of IHC‐stained patient samples demonstrating a scarcity of tumor innervation in these subtypes (Figure  1E–G ). While advancements in cancer therapy continue to emerge, carboplatin remains a powerful and widely utilized chemotherapeutic drug for the treatment of ovarian cancer [ 60 ]. Standard‐of‐care for HGSOC includes combination therapy of carboplatin with a taxane (usually paclitaxel). While the superiority of combination therapy over carboplatin monotherapy is dependent on the patient populations, dual therapy is generally preferred [ 61 , 62 , 63 ]. Paclitaxel stabilizes microtubules, which are required for the release of sEVs [ 64 , 65 ]. Thus, given our interest in the contribution of sEVs to neurite outgrowth and tumor innervation, our in vitro and in vivo studies utilized carboplatin alone. We tested whether carboplatin treatment of cells influences their sEV‐mediated neurite outgrowth activity. Based on their robust neurite outgrowth activity (Figure  2A and Figure S6A–E ), only HGSOC sEVs were tested. Here, cells were exposed to carboplatin at a dose (10 µg/mL) and time (48 h) where cellular proliferation is blocked but prior to the presence of extensive cell death (Figure S9A–C ). In the HGSOC sEVs tested (OVCAR‐4, OVSAHO, CaOV3, and Kuramochi), the significance of carboplatin treatment was found to independently impact the number of neurite intersections (bifurcations) and the neurite length. When comparing treated vs non‐treated groups, significant differences in the neurite length for all carboplatin‐treated cells lines were observed (OVCAR‐4, p < 0.0001; OVSAHO and CaOV3, p < 0.01; Kuramochi, p < 0.05) (Figures S10A,B and S11A,B ). The interaction between two factors (bifurcations and length) was not significant (Figure S12 ). We can be certain that the change evident in the neurite length is due to the treatment of the cells and not residual carboplatin, because when DRG are directly treated with carboplatin only, there is no significant difference in neurite length when comparing treated DRG vs untreated DRG (Figures S13 and S14 ). These data suggest that carboplatin treatment may modulate the cargo of sEVs such that they impact neurite outgrowth. Chemotherapy‐mediated alterations of sEV cargo have been previously identified [ 66 ]; our findings suggest these changes may affect tumor innervation. We have previously characterized that HGSOCs are innervated by sensory nerves [ 23 ]. Briefly, tumor‐infiltrating nerves in HGSOC patient samples were immunohistochemically stained for β‐III tubulin, TRPV1 (sensory marker), tyrosine hydroxylase (sympathetic marker), and VIP (parasympathetic marker). HGSOCs stained positively predominantly for TRPV1. Thus, to determine the impact of tumor‐infiltrating TRPV1‐expressing nociceptor neurons on disease progression and survival, we utilized a syngeneic model of HGSOC, MOSEC Trp53 −/− ;Pten −/− DKO (referred to as p53 −/− Pten −/− ), cells that has previously been validated [ 23 ]. These cells were implanted into either C57BL/6J or nociceptor neuron‐ablated (TRPV1 cre ::DTA fl/wt ) female mice and two experiments were performed simultaneously. TRPV1 cre ::DTA fl/wt transgenic animals lack TRPV1‐expressing nociceptive neurons and have been previously validated [ 50 , 67 ]. We confirmed ablation of TRPV1‐expressing neurons in the TRPV1 cre ::DTA fl/wt transgenic animals by IHC‐staining dorsal root ganglia for TRPV1 (Figure S15 ). In the first experiment, tumor growth in C57BL/6J ( n = 10 mice) or nociceptor‐ablated ( n = 15 mice) animals was monitored weekly. Given that chemotherapy remains standard‐of‐care for ovarian cancer patients, we also tested whether nociceptors influence treatment response. Therefore, in the second experiment, tumor‐bearing C57BL/6J and nociceptor‐ablated animals ( n = 20 mice/group) were treated weekly with carboplatin. The absence of nociceptors resulted in a modest, yet significant, decrease in tumor growth (Figure  3A ) but no significant improvement in survival (Figure  3B ). When tumor‐bearing animals were treated with carboplatin (50 mg/kg weekly, intraperitoneal, beginning on day 10 post‐tumor implantation), there was a measurable and significant reduction in tumor growth (Figure  3C ) and now a significant improvement in survival emerged with the nociceptor‐ablated animals (Figure  3D ). Importantly, the tumor growth curves for each group were nearly identical to the untreated counterparts suggesting that this model of HGSOC is treatment resistant. This is more easily observed when each group is graphed with its carboplatin‐treated equivalent (Figure  4A,B ). In C57BL/6J animals, inclusion of carboplatin treatment has no effect on survival (Figure  4C ). However, ablation of nociceptors when combined with carboplatin treatment results in a significant improvement in survival (Figure  4D ) suggesting that tumor‐infiltrating nociceptor neurons contribute to treatment resistance. Tumor‐infiltrating nerves promote disease progression. C57BL6/J or TRPV1 cre ::DTA fl/wt female mice were implanted with syngeneic Trp53 −/− Pten −/− cells (1 × 10 5 cells/mouse). A second group of mice were simultaneously implanted with tumor and treated with carboplatin (50 mg/kg); treatments were given every 5 days beginning at day 10 post‐tumor implantation (treatments indicated with arrows). Tumor growth (A, C) and survival (B, D) were analyzed. Statistical analysis for tumor growth was by two‐tailed Student's t ‐test (**, p < 0.01); statistical analysis for survival by log‐rank test. Ns, not significant. Indicates a censored mouse that died due to conditions unrelated to the study. Error bars, standard deviation. Tumor‐infiltrating nerves promote resistance to carboplatin. Tumor growth curves and survival graphs of carboplatin and no treatment groups were compared. Carboplatin treatment (50 mg/kg) did not impact tumor growth in either C57BL/6J or TRPV1 cre ::DTA fl/wt animals (A, B). Carboplatin treatment improved survival only in TRPV1 cre ::DTA fl/wt mice (C, D). Statistical analysis for tumor growth by two‐tailed Student's t‐test and for survival by log‐rank test. Ns, not significant. Error bars, standard deviation. Together, the in vitro and in vivo studies suggest that carboplatin treatment potentiates the neurite outgrowth activity of tumor‐released sEVs. When TRPV1‐expressing neurons are genetically ablated (i.e., TRPV1 cre ::DTA fl/wt transgenic animals), an increase in tumor innervation fails to occur despite carboplatin treatment and thus, survival is improved. Defining the sEV cargo component(s) altered by carboplatin could identify key driver(s) of treatment resistance. While beyond the scope of this study, our in vitro data suggest that carboplatin impacts the cargo of sEVs released by tumor cells. However, altered sEV cargo from other cells within the tumor microenvironment may also contribute to treatment resistance. In fact, published studies show that sEVs from cancer‐associated fibroblasts [ 68 , 69 ], as well as tumor‐associated macrophages contribute to treatment resistance [ 70 , 71 ]. The in vivo studies suggesting a contribution of tumor‐infiltrating nociceptors to treatment resistance prompted us to examine tumor innervation in cases of HGSOC and pay specific attention to the treatment status of each patient. While standard‐of‐care therapy for ovarian cancer patients consists of chemotherapy, the main differences in treatment regimens involve the timing of this therapy: chemotherapy is either initiated before (neo‐adjuvant) or after surgical de‐bulking. We wondered whether these differences influenced tumor innervation. To assess this possibility, a previously blindly scored cohort of patient samples were now separated based on naïve (no chemotherapy prior to surgery) and neo‐adjuvant status and analyzed again [ 23 ]. Strikingly, naïve samples ( n = 12) were overwhelmingly low scoring for nerve twigs while neo‐adjuvant treated samples ( n = 18), that is residual disease, were high scoring (Figure  5A ). While compelling, these data were generated from unmatched samples (i.e., from different patients). Thus, to more stringently assess the potential contribution of chemotherapy to tumor innervation, analysis of matched samples was completed. Four matched cases (from the same patient) of pre‐ and post‐treatment samples were IHC‐stained for β‐III tubulin and scored by four independent scorers who were blinded to the conditions. Consistent with the above finding, pre‐treatment samples were low scoring for tumor‐infiltrating nerves while matched post‐treatment samples (residual disease) were high scoring (Figure  5B ). Representative photomicrographs demonstrate the differences in tumor‐infiltrating nerves in matched samples (Figure  5C and Figure S16A,B ). A second cohort of matched pre‐ and post‐treatment samples ( n = 10 cases) was also assessed with similar results. These data indicate that residual disease is densely innervated and suggest that chemotherapy contributes to this phenotype. Innervation differences pre‐ and post‐treatment. Unmatched cases of HGSOC ( n = 30) were immunohistochemically stained with β‐III tubulin and innervation quantified by a scorer who was blinded to the conditions. Blue, innervation score of 0‐1; orange, innervation score of 1‐2; gray, innervation score of 2‐3. The higher the score, the greater the extent of innervation. (A) Pie chart showing the prevalence of innervation and the sample status (pre‐ or post‐treatment). (B) Matched cases of pre‐and post‐treatment HGSOC cases were similarly stained and scored for innervation by four independent scorers. Statistical analysis by Student's t ‐test. N= 4 matched cases. The analysis was repeated with an additional n = 10 matched HGSOC cases with similar results. Statistical analysis by unpaired, two‐tailed t ‐test with Welch's correction. Error bars, standard deviation. (C) Representative photomicrographs of matched pre‐ and post‐treatment sample immunohistochemically stained for β‐III tubulin. Scale bars, 200 µm. While recurrent ovarian disease is common with HGSOC, in many instances, it remains initially sensitive to chemotherapy. Ultimately, however, patients experience treatment resistance. The patient tumor data demonstrating the presence of nerves in residual disease, together with our murine in vivo data showing a contribution of tumor‐infiltrating nerves to treatment resistance, suggest that a minimum density of tumor‐infiltrating nerves is necessary to convert treatment‐sensitive, innervated residual disease to treatment‐resistant disease. Tumor‐released sEVs lure nerves to the tumor bed [ 20 , 21 ]; the present study suggests that chemotherapy may potentiate tumor innervation. Mechanistically, we hypothesized that chemotherapy alters sEV cargo, endowing it with robust neurite outgrowth activity. To test this, we turned to a previously isolated set of isogenic high‐grade serous cell lines in which the parental cell line (PEO1) is platinum‐sensitive, while the PEO4 daughter cell line is platinum‐resistant. These cells were isolated from the same patient; the PEO1 cells were harvested earlier in the disease when the patient was still sensitive to platinum while the PEO4 cells were collected after platinum resistance was established [ 72 , 73 , 74 ]. Equal amounts of purified sEVs were tested on DRG as previously described (western blot validation of sEVs in Figure S17A ) [ 20 , 22 ]. While no significant difference in neurite length was evident between the groups (Figure  6A ), sEVs from platinum‐resistant PEO4 cells resulted in neurite complexity, measured as nMax, that was similar to that of the NGF control while the nMax resulting from the platinum‐sensitive PEO1 sEVs was instead similar to the PBS negative control (Figure  6B ). This intriguing difference suggested something had changed in the sEVs when treatment resistance was attained. To further assess this, we purified, validated, and tested sEVs from another set of isogenic treatment‐sensitive (A2780) and treatment‐resistant (A2780/CP70 and A2780/C30) cell lines [ 75 , 76 ]. Neurite outgrowth of DRG was similarly analyzed. While sEVs from the A2780/CP70 treatment‐resistant line induced significantly more neurite outgrowth from DRG than sEVs from the treatment‐sensitive parental A2780 line, sEVs from the second treatment‐resistant line, A2780/C30, did not (Figure S17B–D ). Importantly, the A2780 cell line and its derivatives represent endometrioid adenocarcinoma, while the PEO1 and PEO4 cells are high‐grade serous [ 59 , 77 ]. The similar findings of sEV‐mediated alterations in neurite outgrowth in the treatment‐resistant cells suggest that this phenomenon may not be limited to HGSOC. Chemotherapy and sEV‐mediated neurite outgrowth. (A) sEVs were purified from PEO1 (treatment sensitive) and PEO4 (treatment resistant) isogenic cell lines. DRG were treated with 3 ug of sEV from each of the indicated cell lines or with NGF (2 ng, positive control) or PBS (negative control) and neurite outgrowth assessed by Sholl analysis. Outliers were removed with the ROUT test with Q set at 1%. The number of intersections and neurite length for each condition is graphed. (B) The nMAX, maximum of intersections, a measure of neurite complexity, is shown in a bar graph for clarity. Statistical analysis by two‐way ANOVA (row factor x column factor). The column factor had a significant main effect (F(3, 15) = 3.407, p = 0.0453), whereas the row factor did not 9F(7,15) = 1.180, p = 0.3704. N = 6–8 DRG analyzed/group. Error bars, standard error. The experiment was repeated n = 2 times with similar results. Statistical details in Figure S18 .

We show that HGSOCs are densely innervated, while the other histological subtypes of ovarian cancer (low‐grade serous, clear cell, endometrioid, and mucinous ovarian carcinomas) are not. Consistent with this, we show that sEVs released from human HGSOC cell lines harbor robust neurite outgrowth activity. Since HGSOCs account for the majority of cases and deaths from ovarian cancer, these data suggest that dense tumor innervation is a poor prognostic factor. We previously demonstrated that tumor‐released sEVs recruit nerves to head and neck [ 20 ] and cervical tumors [ 22 ]. Here, we demonstrate a similar function for sEVs in ovarian cancer. Importantly, only HGSOC sEVs harbor this activity suggesting that sEV cargo and its neurite‐outgrowth activity is disease subtype specific. Ovarian cancer patients nearly always receive chemotherapy treatment. We found that in vitro carboplatin treatment of HGSOC cell lines significantly increases the neurite outgrowth activity of their sEVs as compared to untreated cells. Moreover, while carboplatin treatment of p53 −/− Pten −/− tumor bearing C57BL/6J animals had no effect on tumor growth or survival, when implanted into nociceptor‐ablated mice, carboplatin treatment significantly improved survival suggesting that tumor‐infiltrating nerves contribute to treatment resistance. Perhaps the most intriguing discovery from this study is the contribution of chemotherapy to tumor innervation and residual disease. Using matched and unmatched patient samples, we show that neo‐adjuvant chemotherapy correlates with densely innervated, residual disease. When sEVs from two sets of independently generated platinum‐sensitive and resistant ovarian cancer cell lines were tested on DRG, the platinum‐resistant variants (PEO4 and A2780/CP70) demonstrated increased neurite complexity (PEO4) or neurite outgrowth (A2780/CP70) activity; DRG treated with sEVs from the treatment‐sensitive cells (PEO1 and A2780) remained similar to negative (PBS) control. Interestingly, sEVs from the A2780/C30 treatment‐resistant variant of A2780 did not induce robust neurite outgrowth from DRG. Understanding how these two treatment‐resistant cell lines were derived sheds light on these data. The A2780/CP70 cell line was generated by intermittent exposure to increasing concentrations of cisplatin; the A2780/C30 cell line was instead produced by continuous drug exposure [ 76 , 78 ]. Given the side effects of chemotherapies, cancer patients do not receive continuous chemotherapy; instead, treatment regimens typically consist of a period of drug infusions followed by a defined “rest” (off drug) period and this pattern is repeated for a number of cycles. Thus, generation of the A2780/CP70 drug‐resistant cell line closely mimics clinical patient treatment protocols. While this approach is necessary, clinical trials have repeatedly demonstrated that shortening the duration between chemotherapy infusions provides a survival advantage [ 79 , 80 ]. Moreover, recurrent disease is more prevalent in patients that are neo‐adjuvant treated as opposed to those that receive up‐front surgical de‐bulking [ 81 ]. Importantly, one study shows that neo‐adjuvant therapy increases the risk of platinum‐resistant recurrent disease at late stage [ 82 ]. These findings together with our data suggest that chemotherapy modulates sEV cargo such that robust innervation of residual disease ensues. Enhanced innervation may ultimately contribute to platinum‐resistant, recurrent disease. Changes in sEV cargo induced by chemotherapeutic agents have been previously documented and support a bystander effect of chemotherapy on sEVs that ultimately contributes to disease progression [ 75 , 83 , 84 ]. However, the specific chemotherapy‐induced changes in sEV cargo and the mechanistic details linking them to increased tumor innervation remain unknown. In addition to chemotherapy‐induced changes in sEV‐mediated neurite outgrowth activity, published studies indicate that tumor‐infiltrating neurons harbor a nerve injury transcriptional profile. Specifically, these nerves demonstrate increased expression of Atf3 , a master transcription factor induced upon nerve injury. Following activation, Atf3 initiates a program of axonal repair which includes stimulation of neurite outgrowth. Elevated Atf3 expression in tumor‐infiltrating nerves has been documented in cancer models even in the absence of chemotherapy [ 85 , 86 , 87 , 88 ]. We speculate that, in addition to impacting sEV cargo, chemotherapy directly or indirectly inflicts additional nerve injury. This injury robustly initiates the Atf3 regenerative repair program thereby potentiating neurite outgrowth. This may be an additional mechanism explaining how post‐chemotherapy tumors are more densely innervated as compared to pre‐treatment. Taken together, our data suggest that chemotherapy modulates sEV cargo potentiating its tumor innervation capabilities, driving treatment resistance and disease progression. If correct, this hypothesis predicts that the time to treatment resistance and disease progression will be shorter in patients receiving neo‐adjuvant therapy as compared to those that have primary de‐bulking surgery. Published clinical trials support this prediction [ 82 , 89 ]. We further validate this hypothesis with our syngeneic carboplatin‐resistant ovarian cancer model; we show that simply removing TRPV1‐expressing cells (TRPV1 cre ::DTA fl/wt mouse) is sufficient to sensitize tumors to carboplatin therapy and improve survival. While our findings require additional validation, they suggest that patients receiving neo‐adjuvant chemotherapy may benefit from the addition of pharmacological agents that block sEV release and/or local nerve sprouting. While not currently clinically available, high‐throughput screening of FDA‐approved drugs has already identified agents with inhibitory sEV release activity [ 90 ]. These drugs hold great promise for combination therapeutic approaches in oncology. Similarly, as our understanding of the neural composition of cancer expands and key neurotransmitters, channels, and neurotrophic factors are identified, it is likely that FDA‐approved neurological drugs can be successfully repurposed for use in oncology.

Tumor innervation refers to the sprouting of neurites from the central or peripheral nervous systems into tumor tissue; a growing body of evidence supports its importance in cancer progression [ 1 , 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 , 10 , 11 , 12 , 13 ]. As additional molecular details of this process emerge, the functional contributions of tumor‐infiltrating nerves will be defined and the ability to therapeutically target the neural components of malignancies will strengthen the arsenal of tools available to patients. To date, several cancers have been documented to harbor tumor‐infiltrating nerves including prostate [ 7 , 10 , 14 , 15 ], pancreatic [ 16 , 17 ], breast [ 18 , 19 ], head and neck [ 20 , 21 ], melanoma [ 13 ], cervical [ 22 ], ovarian [ 23 ], lung [ 24 ], and colorectal [ 25 , 26 , 27 ]. In most cases, dense tumor innervation correlates with poor outcomes suggesting an active neural contribution to disease progression. With the ever‐growing molecular and histological characterization of cancers, it becomes critical to define whether tumor innervation differences exist amongst cancer subtypes. This may be most critical for malignancies that are predominantly diagnosed at late stage and currently lack targeted therapies as defining their neural composition may provide therapeutic options that are currently quite limited. Epithelial ovarian cancer, the fifth most common cancer in women, is one such cancer. It remains the most lethal gynecologic malignancy [ 28 ]. Worldwide, nearly 250 000 women are diagnosed with this disease each year; 150 000 of which will succumb to it within the first year [ 29 ]. Ovarian cancer is a heterogeneous disease with multiple histologic subtypes that display unique biologic and clinical behaviors. High‐grade serous ovarian carcinoma (HGSOC), the most common form, accounts for over 70% of cases and the majority of deaths. The primary cell of origin for HGSOC is the fallopian tube secretory cell [ 30 , 31 , 32 , 33 , 34 ]. These malignancies nearly universally harbor mutations in TP53 as well as copy number alterations (gains and/or losses) with a majority of tumors being defective in homologous DNA recombination [ 35 , 36 , 37 ]. Low‐grade serous ovarian carcinoma (LGSOC), a rare histological sub‐type, accounts for approximately 2%–10% of epithelial ovarian cancer cases. Unlike HGSOC, women diagnosed with LGSOC are younger and their malignancy remains TP53 wildtype [ 38 ]. Like LGSOC, patients with mucinous ovarian cancer are also young. Most of these patients (65%–80%) are diagnosed with early‐stage disease which significantly improves their 5‐year overall survival which exceeds 90% [ 39 ]. However, when diagnosed at late‐stage, patients with mucinous ovarian cancer suffer a poor prognosis with an overall median survival of only 12–33 months [ 39 , 40 , 41 ]. Clear cell ovarian carcinomas represent less than 5% of all ovarian cancers; this subtype is particularly resistant to conventional platinum‐based chemotherapeutics. Given that this remains standard of care, patients with clear cell ovarian carcinoma are left with very few treatment options. Since these patients typically present at early stage, novel therapies (e.g., nerve targeting) would offer hope as interventions could commence early in the disease. Endometriosis is associated with endometrioid carcinoma of the ovary with an incidence of 21% [ 42 ]. Endometrioid carcinomas account for 13%–15% of all ovarian cancers [ 43 ]. Like clear cell carcinoma patients, those with endometrioid carcinoma are also diagnosed at a young age and have a better prognosis [ 44 ]. Standard treatment for ovarian cancer is surgical de‐bulking followed by platinum‐based chemotherapy; however, neoadjuvant chemotherapy is increasingly being utilized [ 45 , 46 ]. Though these therapies are initially effective, the vast majority of patients progress and succumb to recurrent, chemotherapy‐resistant disease [ 28 ]. Whether platinum chemotherapeutics impact tumor‐infiltrating nerves is currently undefined. We systematically assessed tumor innervation in the five major histological subtypes of ovarian cancer and found that only HGSOC is densely innervated. We and others have demonstrated that tumor‐released small extracellular vesicles (sEVs, lipid bilayer particles ranging in size from 50 to 150 nm) promote tumor innervation, that tumors with dense nerve infiltration grow faster and are more aggressive than sparsely innervated cancers and that sEVs directly contribute to this phenotype [ 20 , 21 ]. We found that sEVs from HGSOC cell lines harbored the greatest ability to induce neurite outgrowth in vitro. Carboplatin treatment of ovarian cancer cell lines impacted the neurite outgrowth activity of their sEVs in vitro, and significantly prolonged survival of treatment‐resistant syngeneic HGSOC tumors in vivo when implanted in transgenic animals lacking tumor‐infiltrating nerves. These data suggest that tumor‐infiltrating nerves contribute to treatment resistance. Consistent with this, IHC staining of matched pre‐ and post‐treatment HGSOC patient samples shows that pre‐treatment tumors are sparsely innervated and post‐treatment tumors are densely infiltrated with nerves. Taken together, these studies emphasize the importance and influence of tumor‐infiltrating nerves on HGSOC and define them as potential therapeutic targets.

R.D. serves on the advisory board of Repare Therapeutics and VOC Health. Daniel Vermeer has a licensing agreement with Nant for an HPV vaccine. A.K.G is a co‐founder of Sinochips Diagnostics, serves as a scientific advisory board member to Biovica, Clara Biotech, EXOKĒRYX, VITRAC Therapeutics, and Sinochips Diagnostics, and receives research funding from Predicine and VITRAC Therapeutics.

Supporting File : adbi70108‐sup‐0001‐SuppMat.docx.