HIF-1–regulated TPM3 links hypoxia to motility and invasion beyond the hypoxic fraction in triple-negative breast cancer

Hypoxia, TNBC, TPM3, motility, Extracellular Vesicles .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Zhou et al., 2

Hypoxia is a defining feature of triple-negativ e breast cancer (TNBC), driving invasion, metastasis, and therapy resistance. Understanding the molecular effectors of hypoxia is essential to identify new therapeutic targets. Here, we investigated tropomyosin 3 (TPM3), an actin-binding protein that regulates filament stability. TPM3 is significantly upregulated in breast cancer, including in TNBC, where elevated levels correlate with poor overall survival. Using validated hypoxia signatures and TNBC cell models, we show that TPM3 is induced in physiologically relevant hypoxic conditions in a HIF-1–dependent manner. Both mRNA and protein levels of TPM3 increased in response to hypoxia, and TPM3 colocalised with F-actin, supporting cytoskeletal organisation. Functional assays demonstrated that depletion or inhibition of TPM3 impaired cell morphology, motility, and invasion in hypoxic TNBC cells, while not affecting viability. Notably, TPM3 inhibition synergised with Paclitaxel and Doxorubicin, enhancing therapeutic efficacy. In addition, TPM3 was incorporated into extracellular vesicles (EVs), with hypoxia increasing EV-mediated transfer of TPM3 to normoxic cells and promoting their motility. These findings establish TPM3 as a hypoxia- inducible, HIF-1–regulated effector of cytoskeletal dynamics and intercellular communication, underscoring its potential as a therapeutic target to limit TNBC aggressiveness and improve treatment outcomes. .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Zhou et al., 3

Hypoxia (conditions of insufficient oxygen) is a hallmark of the tumour microenvironment (TME) in solid cancers, arising as a cons equence of rapid tumour growth and inadequate or inefficient vasculature. Hypoxia is associated with therapy resistance and poor patient prognosis 1. The adverse prognostic impact of hypoxia has been particularly well documented in breast cancer and is even more pronounced in triple-negative breast cancer (TNBC), the most aggressive subtype which lacks targeted treatment options 2,3. Patients who succumb to TNBC do so because of metastatic disease and while a relationship between hypoxia and increased metastasis is well-known the underpinning mechanisms are less clear. Emerging evidence suggests t hat hypoxia-driven changes in gene expression, activation of hypoxia-inducible factors (HIFs) , epithelial-to-mesenchymal transition (EMT), metabolic reprogramming, and remodelling of the TME may all contribute to enhanced metastatic dissemination 4,5. The identification of hypoxic tu mour markers which predict poor patient outcome is therefore critical to identify novel therapeutic strategies. TPM3 (Tropomyosin 3) encodes a member of the tropomyosin family of actin-binding proteins which stabilise actin filaments and r egulate the cytoskeleton. TPM3 plays important roles in maintaining cell structure, intr acellular transport, and muscle contraction 6,7. While TPM3 function has been predominantly char acterised in muscle cells/myopathies, dysregulation or gene fusions involving TPM3 have been implicated in various cancers 7–13. Interestingly, the mRNA for TPM3 has been described in microvesicles generated by platelets and it has been suggested that TPM3 could be delivered to breast cancer cells and promote metastasis 14. However, a role for TPM3 has not been described in the hypoxic TME. Here, we found that TPM3 is a target of the HIF-1 transcription factor in a broad range of hypoxic conditions and controls the motility and invasion capacity of hypoxic TNBC cells. Most importantly, we identify TPM3 as an extracellular vesicle (EV) cargo protein which can increase the motility of normoxic cells. Together, these data suggest that the hypoxia- mediated induction of TPM3 contributes to the metastatic potential of both the hypoxic and oxygenated tumour fractions of TNBC. .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Zhou et al., 4

TPM3 expression was found to be significantly higher in breast cancer (BRCA) tissue, compared to normal breast tissue ( Figure 1A and Figure S1A ). Furthermore, TPM3 expression was found to be significantly higher in TNBC compared to normal tissue ( Figure 1B and Figure S1B ). Next, we asked how TPM3 expression correlated with TNBC patient survival and found that high TPM3 is associated with a poorer overall survival rate in TNBC (Figure 1C ). To further explore a relationship between TPM3 and hypoxia, we used two validated hypoxia signatures and found a significant correlation in patient samples, demonstrating that TPM3 expression is increased in more hypoxic TNBC ( Figure 1D and S1C) 15,16. Tumour hypoxia exists as a gradient of oxy gen tensions within a tumour and also includes transitions between levels, known as cyclic or intermittent hypoxia 17. It is important to consider a range of hypoxic conditions when in vestigating hypoxia-mediated biology as the biological response can differ 18. Here, we considered 2% O 2, <0.1% O 2 and cyclic conditions (transitions between 2 and <0.1% O 2). As expected, HIF-1 α was stabilised in all of the hypoxic conditions, however evidence of the DNA damage response was limited to <0.1% O2 and the unfolded protein response to <0.1% O 2 and cyclic conditions ( Figure 1E) 19. TNBC cell lines (MDA-MB-231 and MDA-MB-453) were exposed to the hypoxic conditions (2, <0.1% O 2 and cyclic) and changes in TPM3 mRNA determined. As expected, the well-validated HIF target, CAIX was induced in response to hypoxia ( Figure 1F ) 20. A significant increase in TPM3 mRNA was observed in all the hypoxic conditions tested in both cell lines demonstrating that TPM3 is hypoxia inducible in a broad range of hypoxic conditions relevant to TNBC (Figure 1G and S1D). This hypoxia-mediated induction in all three conditions and correlation with hypoxia signatures suggested the possibility that TPM3 was regulated by one of the hypoxia .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Zhou et al., 5 inducible factors (HIFs). To investigate possible HIF dependence, we treated MDA-MB-231 cells with siRNA to HIF-1 β to prevent both HIF-1 and HIF-2 mediated signalling and determined the impact on TPM3 expression. Loss of HIF-1 β abrogated the induction of TPM3 and CAIX in hypoxia ( Figure 1H and S1E-G). To determine if TPM3 induction in hypoxia was mediated by HIF-1 or HIF-2, we used the matched colorectal cell lines RKO and RKOHIF-1α-/- and found that TPM3 induction in hypoxia was abrogated by loss of HIF-1 α (Figure 1I and S1H-J). Together, these data confirm that TPM3 is hypoxia-inducible in all the cell lines investigated and that this occurs in a HIF-1 dependent manner. MDA-MB-231 and MDA-MB-453 cells were then exposed to hypoxia (2% O 2) and western blotting carried out for TPM3. TPM3 protein was found to increase in response to hypoxia in both cell lines (MDA-MB-231 Figure 1J, K and MDA-MB-453 S2A, B ). Hypoxia-mediated induction of TPM3 was further confirmed in response to <0.1% O 2, cyclic conditions and a third TNBC cell line (BT-549) (Figure S2C-E). Next, we validated that TPM3 colocalised with F-actin (detected using phalloidin) 21. TPM3 appeared predominantly cytoplasmic with an apparent filamentous structure. Notably, a punctate signal near to the nucleus was also observed however this was unaffected by siRNA mediated knockdown of TPM3 and was therefore considered non-specific staining (Figure S2F-G ). A clear colocalisation of TPM3 and F-actin was observed which did not change in response to hypoxia ( Figure 2A, B ). Next, we analysed cell morphology with and without siTPM3 knockdown and observed an increase in elongated trailing edges following loss of TPM3, suggesting a loss of actin filam ent structure and a reduced ability to contract the trailing edge during migration. To confirm, we measured the circularity of cells and found a significant reduction in cell circularity when TPM3 was depleted, in both normoxic and hypoxic (2% O 2) conditions ( Figure 2C). As loss of TPM3 appeared to impair the ability of MDA-MB-231 cells to recoil their trailing edge, we hypothesised that TPM3 may also .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Zhou et al., 6 influence the leading edge, where F-actin stabilis ation in lamellipodia and focal adhesions is essential for generating the mechanical force required for trailing edge recoil 22. Loss of TPM3 significantly reduced the intensity of phalloidin staining at the leading edge under both hypoxic (2% O2) and normoxic conditions ( Figure 2D). Depletion of TPM3 had no effect on the width of the leading edge under normoxia but had a significant effect under hypoxia, further supporting a role for TPM3 in stabilising F-actin and cell motility ( Figure 2E, F ). Together, our data suggest the hypothesis that TPM3, through its role in stabilising actin filaments, could promote the motility of hy poxic cells and therefore drive metastatic progression in aggressive TNBC. To test the role of TPM3 in hypoxia-mediated motility, we carried out wound healing assays in MDA-MB-231 cells with siRNA-mediated depletion of TPM3. Before investigating motility, we verified that cell cycle distribution and viability were not altered in the hypoxic conditions used (Figure S3A-C). Furthermore, we demonstrated that loss of TPM3 had no significant effect on clonogenic survival of cells in any of the hypoxic conditions tested ( Figure S3D-E). Depletion of TPM3 was found to significantly slow down wound closure in hypoxic conditions but had no impact in normoxia (Figure 3A-C and S4A,B). A small molecule inhibitor of TPM3 has been described (ATM-3507) which we also used to test the role of TPM3 in hypoxia- mediated motility 23. Again, we first verified that ATM-3507 did not significantly impact viability in both normoxic and hypoxic conditions at the dose reported to inhibit TPM3 activity (Figure S4C ) 12. Cells exposed to ATM-3507 were significantly less motile in hypoxic conditions (Figure 3D, E and S4D, E). Metastatic potential is determined by both motility and the ability to invade, therefore we investigated the contribution of TPM3 to hypoxia-mediated invasion 24. As expected, hypoxic cells showed a greater invasion capacity compared to the normoxic however, this was significantly reduced when TPM3 was depleted ( Figure 3F-H and S4F, G). In conclusion, TPM3 has an important role in cell migration and invasion in hypoxic TNBC cells. .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Zhou et al., 7 The standard of care for TNBC patients includes Carboplatin, Doxorubicin, Paclitaxel and radiotherapy. Inhibition of TPM3 has been shown to combine effectively with Paclitaxel and Doxorubicin in normoxia, resulting in the reduction of cell viability in neuroblastoma and ovarian cancer cell lines 12,25. Here, we investigated the efficacy of combining TPM3 inhibition or siRNA-mediated loss with the standard of care for TNBC in physiologically relevant conditions. MDA-MB-231 cells were exposed to hypoxia (2% O 2) and a range of doses of ATM-3507 with Carboplatin, Doxorubici n or Paclitaxel followed by an assay for viability. The Highest Single Agent (HSA) model was used to investigate potential synergy and revealed a reduction of cell viability ac ross increasing concentrations of ATM-3507, Carboplatin, Doxorubicin and Paclitaxel ( Figure S5A-C). In addition, strong synergy (score >10) was observed with ATM-3507 combined with Doxorubicin or Paclitaxel, no antagonistic or additivity regions were indicated ( Figure 4A, B, C ). To determine any effect on radiosensitivity, MDA-MB-231 cells were treat ed with siTPM3 followed by irradiation in hypoxia (2% O 2). Loss of TPM3 had no impact on radiosensitivity in normoxia or hypoxia (Figure 4D and S5D). In addition, inhibition of TPM3 in normoxia or hypoxia (2% O 2) did not significantly impact radiosensitivity (Figure S5E). Together, these data suggest that inhibition of TPM3 could be combined with standard of care for TNBC to potentially reduce metastatic spread and improve patient prognosis. The hypoxic TME is known to impact neighbouring oxygenated cancer cells although a role for TPM3 in this has not been described 26. MDA-MB-231 cells were treated with siRNA TPM3, exposed to normoxia or hypoxia (2% O 2) and conditioned media collected and transferred to normoxic cells followed by a wound healing assay (Figure 5A, B). Conditioned media from hypoxic cells (donor cells) when added to normoxic cells (recipient cells) significantly increased the wound healing rate of the normoxic cells. However, when the conditioned media from TPM3 depleted donor cells was added to recipient cells no increase in the motility of the normoxic recipients was observed ( Figure 5C, D). This finding suggests that TPM3 contributes to factors present in the conditioned media or regulates the export of .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Zhou et al., 8 proteins and/or EVs from hypoxic cells, t hereby influencing the migratory behaviour of normoxic recipient cells. To investigate further, we again generated conditioned media from normoxic and hypoxic (2% O 2) cells and added it to wounded normoxic recipient cells ( Figure 5E). However, this time we also included the dynamin inhibitor, Dynasore, to prevent EV uptake in the recipient cells ( Figure 5F ). As previously, we saw a significant increase in wound healing with the addition of hypoxic conditioned media. However, when Dynasore was added a significant reduction in wound closure rate was observed ( Figure 5F, G and S6A ). This suggests EV uptake from the hypoxic conditioned media is critical to the impact on recipient cell motility in normoxia. To account for the known off-target effect of Dynasore in reducing basal motility, data were normalised to control cells treated with Dynasore alone 27. Importantly, we confirmed that Dynasore did not affect TPM3 expression validating that the reduction of wound healing in hypoxic conditioned media following Dynasore treatment was not due to TPM3 suppression in the migrating cells ( Figure S6B ). These data again confirm that hypoxia-induced TPM3 contributes to a condi tioned media which has the capacity to increase motility of normoxic cells and implicates EV transfer as the underlying mechanism. Therefore, we isolated and purified EVs to confirm if they were responsible for the impact on normoxic cell motility. First, we verified using transmission electron microscopy (TEM) that EV production could be seen in MDA-MB-231 cells in hypoxic conditions. Late endosomes and EVs were observed in normoxia and hypoxia with an apparent increase in vesicle release in hypoxia (2% O 2) (Figure 6A). A clear increase in the abundance of mitochondria following 2% O 2 exposure was also observed and has been reported previously 28. To investigate the apparent increase in EVs detected by TEM, we isolated EVs from the conditioned media of normoxic and hypoxic (2% O 2) MDA-MB-231 cells and quantified them using nanoparticle tracking analysis (NTA). The total concentration of EV s increased significantly in hypoxia (2% O 2), .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Zhou et al., 9 while the size distribution was similar in normoxic and hypoxic conditions ( Figure 6B-D ). Next to determine if EVs released by hypoxic cells impacted recipient cell migration, MDA- MB-231 cells were again exposed to hypoxia (2% O 2) and TPM3 induction confirmed (Figure 6E ). EVs were then isolated and added to wounded normoxic recipient cells to determine impact on would healing rate ( Figure 6F ). EVs isolated from hypoxic cells increased the motility of normoxic cells and this was abrogated in the presence of Dynasore (Figure 6G and S6C). These data led us to the non-mutually exclusive conclusions that TPM3 could play a role in EV production in hypoxic cells or that TPM3 could be transferred as cargo through EVs to recipient cells. MDA-MB-231 cells were treated with scramble siRNA or siTPM3 and exposed to hypoxia (2% O 2) followed by EV isolation and NTA. Again, EV production was increased in hypoxic conditions but loss of TPM3 did not significantly change EV production or size distribution in normoxia or hypoxia ( Figure 7A, B, C ). Next, we again isolated EVs from normoxic and hypoxic cells and used NTA to ensure equal loading for western blotting. ALIX is a well-characterised EV protein and was used as a control 29. Blotting for the mitochondrial protein PRDX3 and Golgi associated GM130 was also included to demonstrate the purity of the EVs 30. As expected, western blotting of whole cell lysates showed that ALIX, PRDX3 and GM130 were expressed equally in normoxic and hypoxic cells while TPM3 was induced in hypoxia ( Figure 7D ). TPM3 was found to be included in normoxic EVs and the levels increased in EVs generated from hypoxic cells therefore validating TPM3 as an EV cargo protein (Figure 7E, F).

In this study, we demonstrate that TPM3 is a hypoxia-inducible gene in TNBC, and that this occurs in a HIF-1-dependent manner. Under hypoxic conditions, TPM3 supports actin filament stability, maintaining cell shape and enabling efficient migration and invasion. Loss of TPM3 disrupted F-actin organisation at both the leading and trailing edges, reduced cell .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Zhou et al., 10 circularity, and impaired motility. Importantly, TPM3 is also incorporated into EVs released by hypoxic cells, which enhance motility in normoxic recipient cells. Together, these findings demonstrate that hypoxia-induced TPM3 enhances migration capacity across the tumour, extending its impact beyond the hypoxic fraction (Figure 7G). Our findings support that targeting TPM3 during treatment of TNBC could reduce metastatic burden, by reducing the migratory potentia l and invasiveness of residual hypoxic and normoxic cancer cells. Importantly, TP M3 is druggable and appeared well-tolerated in vivo as evidenced by preclinical studies with ATM-3507 and its precursor TR100 12,25,31. A previous report has shown that TPM3 mRNA is included in EVs generated by platelet cells therefore raising the question of whether this also occurs in TNBC cells and increases in hypoxia 14. Our study is the first to experimentally validate TPM3 as an EV cargo protein. Notably, this is supported by previous proteom ic datasets that also detected TPM3 in EV preparations, although these studies were not focused on TPM3 and did not pursue validation or functional investigation 32,33 . A previously reported proteomic analysis also determined an interaction between TPM3 and Right Open reading frame kinase 3 (RIOK3) in hypoxia (0.1% O 2) suggesting that in addition to the mechanism of hypoxia-induction described here, TPM3 could also be phosphorylated and potentially stabilised by RIOK3 in hypoxia 34. In conclusion, our study identifies TPM3 as a novel, HIF-1–regulated effector that links hypoxia to cytoskeletal remodelling, motility, and intercellular communication in TNBC. By demonstrating its incorporation into EVs and its functional contribution to both hypoxic and normoxic cell behaviour, we highlight TPM3 as a mediator that bridges the heterogeneity of the tumour microenvironment. Together, these findings position TPM3 as both a biomarker of hypoxic adaptation and a promising therapeutic target to constrain TNBC aggressiveness and improve patient outcomes. .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Zhou et al., 11

Cell lines and reagents MDA-MB-231 (TNBC, provided by Dr Amanda Coutts, University of Oxford), MDA-MB-453 (TNBC, provided by Dr Isabel Pires, University of Manchester) and BT-549 (TNBC, provided by Prof. Katherine Vallis, University of Oxford). Colorectal RKO and RKO HIF-1α-/- cancer cells were grown in DMEM 35. Cell culture media was supplemented with 10% FBS and cells were maintained in an incubator set at 37 /i1 C and 5% CO₂ . All cell lines were verified mycoplasma free using a MycoAlertTM mycoplasma detection kit (Lonza). Inhibitors/drugs used were ATM- 3507 (Sigma-Aldrich, 1861449-70-8), Dynasore (Med Chem Express, 304448-55-3). For siRNA-mediated knockdowns, MDA-MB-231 cells were transfected with siRNA to a final concentration of 50 nM using Lipofectamine RNAiMAX (Invitrogen) following the manufacturer’s protocols. siRNA sequences are provided in Table SI. Hypoxia A Bactron II anaerobic chamber (Shel Labs) was used for hypoxic treatment at <0.1% O 2. A Whitley M35 Workstation (Don Whitley Scientific) was used for 2% O 2 or cyclic hypoxia. Cycling conditions were <0.1% O 2 for 2 h followed by 2% O 2 for 2 h as described previously 18. For all experiments except MTT assays, cells were seeded on glass dishes and harvested inside the chambers with equilibrated reagents. Western blotting Samples were lysed in UTB (9 M Urea and 75 mM Tris-HCl pH 7.5 supplemented with 0.15 M β -mercaptoethanol prior to use). EV samples were lysed in 1× RIPA lysis buffer (Millipore, 20-188) supplemented with protease inhibitor cocktail (Roche, 11873580001). After a brief sonication, proteins were separated on a 4-20% polyacrylamide gel (Bio-Rad) and transferred onto a nitrocellulose membrane (B io-Rad). Primary antibodies used were: TPM3 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Zhou et al., 12 (Abcam, ab113692), β-actin (Santa Cruz, sc-69879), HIF-1 α (BD Biosciences, 610958), GRP78 (BD Biosciences, 610979), KAP1 (Bethyl, A300-274A), KAP1-S824 (Bethyl, A300- 767A), ALIX (Abcam, ab275377), PRDX3 (Abcam, ab73349), GM130 (Abcam, ab52649). Secondary antibodies: IRDye 680RD goat anti-mouse IgG (LI-COR, 926-68070), IRDye 800CW Goat anti-rabbit IgG (LI-COR, 926-32211), Goat anti-mouse IgG HRP (Invitrogen, 31430), Goat anti-rabbit IgG HRP (Invitr ogen, 31460). Images were acquired by chemiluminescence using Odyssey Infrared Imaging (LI-COR Biosciences) or ChemiDoc XRS+ Gel Imaging System (Bio-Rad). RT-qPCR Trizol (Invitrogen) was used to isolate RN A, and the Verso enzyme kit (Thermo Fisher Scientific) to reverse transcribe RNA. SYBR Green PCR Master Mix kit (Applied Biosystems) was used, and the reaction was carried out on a StepOne Real-Time PCR System (Thermo Fisher Scientific) with v2.0.5 software (Applied Biosystems). RNA fold change was measured using a 2 -ΔΔ Ct method relative to the 18S endogenous control gene. Data shown are the mean of three biological replicates ± SEM. All primers sequences are available in supplementary Table S2. Immunofluorescence Cells were seeded onto autoclaved cover slips (Menzel-Glaser). Cells were fixed in 4% (w/v) paraformaldehyde in PBS. Samples were permeabilised in 0.1% PBS-Triton X-100 and blocked with 5% (w/v) BSA (Thermo Fisher) in PBS. A LSM710 confocal microscope (Carl Zeiss Microscopy Ltd) was used for imagi ng. Antibodies/reagents used: TPM3 (Abcam, ab113692), β-actin (Santa Cruz, sc-69879), Alexa Fluor 488-conjugated goat anti-mouse (Invitrogen, A32723), Alexa Fluor Plus 647 Phalloidin (Invitrogen, A30107). 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-2H-tetrazolium bromide (MTT) assay .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Zhou et al., 13 Cells were seeded in flat bottomed 96 well plates. Each experiment was carried out in triplicate. MTT reagent (5 mg/mL, Invitrogen) was added to cells in the dark for 3 hours (37°C, 5% CO 2). After MTT containing culture media was removed, DMSO was added to cells and incubated in the dark for 15 minutes (37 °C). The plate was then read using a POLARstar Omega plate reader (BMG Labtech) (absorbance 570 nm). Cell viability was measured relative to the untreated control. Wound healing assay Cells were seeded and grown to 95-100% confluency before treatment. At least five parallel wounds were made by scratching the cell monolayer with a 20 μL pipette tip. After rinsing in PBS, the cells were incubated in 0.5% FBS containing culture media. EVOS M5000 (Thermo Fisher) was used to image the wounds immediately after scratching and over time. The area of each wound was measured by ImageJ (National Institutes of Health). The wound closure % was calculated using the following formula where Α is the area of the wound: Wound Closure % /g3404 /g3436 /g1827 t= 0 /g3398/g1827 t= D t /g1827 t= 0 /g3440 /g3400 100% Invasion assay Cells were seeded in an 8 μ m pore size BioCoat TM Matrigel Invasion Chamber (Corning). Trypsinised cells (5×10 5) were added to the upper chambers of a 24-well plate in DMEM containing 0.1% FBS. After treatment, the motile cells at the bottom of the filter were stained with crystal violet (0.5% w/v in 50% MeOH and 20% EtOH). The number of cells which had invaded was measured by counting the stained cells using EVOS M5000 (Thermo Fisher). Colony survival assay Cells were seeded at the appropriate density for siRNA transfection/drug treatment under normoxia or hypoxia for indicated periods, each experiment was carried out in triplicate. Colonies were allowed to grow for 8-10 days in a standard humidified incubator at 37°C and .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Zhou et al., 14 5% CO2. Once the colonies formed ( ≥50 cells), crystal violet (0.5% w/v in 50% MeOH and 20% EtOH) was used for staining. Colonies were counted using an automated colony counter GelCountTM (Oxford Optronix) with GelCount (version 1.2) software, or a manual cell counter (Stuart Scientific). The survival fr action was calculated by number of colonies counted/number of cells seeded × PE, where PE is the plating efficiency of the untreated control (number of colonies counted/number of cells seeded). Colonies with radiation treatment in hypoxic conditions were carried out as previously described 36. Transmission electron microscopy (TEM) Cells were seeded onto glass coverslips and cultured until approximately 70% confluency was reached. Following treatment, cells were fixed in 2.5% glutaraldehyde and 4% PFA in 0.1 M PIPES buffer (pH 7.2) followed by washes in 0.1 M PIPES buffer, incubated in 50 mM glycine/PIPES for 15 minutes, washed once in 0.1 M PIPES, embedded in low-melting-point agarose, chilled, trimmed into 1–2 mm blocks, and returned to buffer. Samples were then treated with 1% osmium tetroxide and 1.5% potassium ferrocyanide in 0.1 M PIPES buffer for 1 hour at 4°C. Samples were washed in Milli-Q water and incubated overnight in 0.5% uranyl acetate at 4°C in the dark, followed by washes in Milli-Q water. Dehydration was performed on ice using a graded ethanol series (30%, 50%, 70%, 80%, 90%, 95%, and 3 × 100%) at room temperature. Samples were inf iltrated with Taab low-viscosity epoxy resin via ethanol:resin series (3:1 for 1 hour, 1:1 for 1.5 hours, and 1:3 for 1 hour), then 100% resin at room temperature overnight. Resin was refreshed twice the next day, with centrifugation (12,000 rpm, 3 minutes) between changes. Agarose-embedded blocks were transferred to Beem capsules containing fresh resin and pol ymerised at 60°C for a minimum of 24 hours. Ultrathin sections (~90 nm) were prepared using a Diatome diamond knife on a Leica UC7 ultramicrotome, mounted onto 200 mesh copper grids, and post-stained using Reynolds’ lead citrate for 5 minutes at room temperature, followed by washes in Milli-Q water. Sections .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Zhou et al., 15 were imaged using a JEOL 1400 transmission electron microscope equipped with a Gatan Rio CMOS detector. EV isolation and purification After treatment, culture media (FBS-free DMEM) was collected and transferred to a 50 mL Falcon tube for EV isolation. EV-containing media was centrifuged at 111.8 × g for 5 minutes at 4°C (Jouan CR4i Centrifuge, Thermo Electr on Corporation). Supernatant was collected, transferred to a new Falcon tube, and centrif uged at 1,006.2 × g for 10 minutes at 4 °C. Supernatant was collected, transferred to a new Falcon tube and centrifuged at 1,788.8 × g for 30 minutes at 4°C. The supernatant was transferred to an ultracentrifuge tube (Ultra- Clear centrifuge tubes, Beckman Coulter). Ultracentrifugation was conducted in a pre-cooled SW 32.1 Ti Swinging-Bucket rotor in an Opti ma XPN-80 Ultracentrifuge (Beckman Coulter) at 87,945 × g for 140 minutes at 4°C. Following ultracentrifugation, the media was removed, and the pellet resuspended in co ld, sterile PBS. Samples we re ultracentrifuged again at 87,945 × g for 140 min at 4°C. Supernatant was carefully aspirated leaving the purified EVs, which were resuspended in sterile PBS or fresh 10% FBS supplemented DMEM for subsequent experiments. Nanoparticle Tracking Analysis (NTA) Isolated EVs were resuspended in equal volumes of 1 × PBS across all conditions. EV number and size distribution were assessed usi ng ZetaView® (Particlemetrix) according to the manufacturer’s instructions. Stock EV samples were diluted in 1 × PBS to working concentrations ranging from 1:10,000 to 1:100,000 and loaded into the flow cell. Measurements were recorded as videos from 8-11 separate positions across the flow cell. EV number and size were estimated from this using the ZetaView analysis software Statistical analysis .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Zhou et al., 16 A two-tailed, paired Student’s t-test was used for the comparison of two means and a two- way analysis of variance (ANOVA) with Tukey’s multiple comparisons or Šídák’s multiple comparison test were used for the comparison of more than two means.

We are grateful to Raman Dhaliwal and Dr Charlotte Melia at the Sir William Dunn School of Pathology Electron Microscopy Facility. Funding CZ was supported by Breast Cancer Now (2022FebPR1492 awarded to EMH and EEP). SAT and EMH thank the EPSRC for the support of programme grant EP/S019901/1. GB was supported by a UNIQ+ Research Internship (University of Oxford). Contributions CZ, JTC, KF, PS, GB, EEP and EMH were responsible for collecting, analysing, and interpreting data. EMH and CZ drafted the manuscript. EMH was responsible for study conceptualisation and study oversight. JTC, CZ and EMH were responsible for figure creation. All authors critically reviewed the manuscript and approved the final version. Competing interests All authors declare no financial or non-financial competing interests. Figure legends Figure 1. TPM3 is induced in hypoxia in a HIF-1-dependent manner A. TPM3 mRNA levels in BRCA and normal breast tissue generated using TPM3 mRNA expression from TCGA-BRCA. num(BRCA)=1082; num(Normal)=514 B. TPM3 mRNA levels in TNBC and normal breast tissue generated using TPM3 mRNA from Genotype-Tissue Expression (GTEx). num(TNBC)=171; num(Normal)=514 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Zhou et al., 17 C. Kaplan-Meier curve of overall survival in TNBC patients with high or low TPM3 expression, generated using TCGA-BRCA (dichotomised at the median). Statistical significance was assessed by log-rank test. D. Correlation between the Buffa hypoxia signature and TPM3 mRNA expression from Metabric dataset. Statistical analysis was deter mined using simple linear regression test and Pearson correlation. E. MDA-MB-231 cells were exposed to a range of hypoxic conditions for 16 h followed by western blotting for the proteins indicated, β-actin was used as a loading control. F. MDA-MB-231 cells were exposed to the range of hypoxic conditions for 16 h followed by RT-qPCR to determine CAIX mRNA level. Statistical testing was done using a paired t-test. G. MDA-MB-231 cells were exposed to the range of hypoxic conditions for 16 h followed by RT-qPCR for TPM3. Statistical testing was done using a paired t-test. H. MDA-MB-231 cells were treated with siRNA to HIF-1β followed by exposure to 21% or 2% O2 for 16 h. TPM3 mRNA level was determined and shown relative to the normoxic control. Statistical testing was done by two-way ANOVA with Šídák's multiple comparisons test. I. RKO and RKOHIF-1/i1-/- cells were exposed to hypoxia (2% O2) for 16 h followed by RT-qPCR for TPM3. Statistical testing was done by two-way ANOVA with Tukey’s multiple comparisons test. J. MDA-MB-231 cells were exposed to 2% O 2 for the times shown followed by western blotting for the proteins indicated. K. Quantification of TPM3 levels in cells treat ed as in part J. Statistical testing was done by paired t-test. Data shown from three separate experiments ( n = 3) are displayed with mean ± standard error of the mean (SEM) unless specified otherwise. Statistical testing was carried out as indicated. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p 0.05. Figure 2. TPM3 affects F-actin organisation at the leading edge under hypoxia. .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Zhou et al., 18 A. Representative images of immunofluorescence of MDA-MB-231 cells . Cells were stained for F-actin (phalloidin; red), TPM3 (green) and DAPI (blue). Scale bar is 50 μ m. B. MDA-MB-231 cells were exposed to hypoxia (2% O 2) for the times indicated followed by staining for TPM3 and F-actin. Quantificati on of colocalisation of TPM3 and F-actin was carried out using Costes’ thresholded Pearson’s coefficient. C. MDA-MB-231 cells were treated with a scramble siRNA or siTPM3 and then exposed to 21% or 2% O2 for 16 h. cells were then stained for TPM3, F-actin and DAPI. Quantification of cell circularity (a.u.) was determi ned and shown. Data collected from ≥ 100 cells per condition. Statistical testing was done by one-way ANOVA with Tukey’s multiple comparisons test. D. Cells were treated as in part A. Quantificat ion of maximum phalloidin intensity at the leading edge was determined. Data collected from ≥ 100 cells per condition. Statistical testing was done by one-way ANOVA with Tukey’s multiple comparisons test. E. Cells treated as in part A. Quantification of leading-edge width is shown. Data collected from ≥ 100 cells per condition. Statistical testing was done by one-way ANOVA with Tukey’s multiple comparisons test. F. Representative immunofluorescence microscopy images showing TPM3 (green), F-actin (phalloidin; red), and DAPI (blue) at the leading edge of MDA-MB-231 cells, transfected with scramble control (siSCR) or TPM3 siRNA in 21% or 2% O2. Scale bar is 50 μ m. Data shown from three separate experiments ( n = 3) are displayed with mean ± standard error of the mean (SEM) unless specified otherwise. * p < 0.05, **** p 0.05. Figure 3. Hypoxia-mediated migration and invasion is TPM3 dependent. A. MDA-MB-231 cells were treated with siTPM3 or a scramble siRNA followed by exposure to 21% or 2% O 2 for 8 h. The % of wound closure is shown in each condition. Statistical testing was done by two-way ANOVA with Šídák’s multiple comparison test. .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Zhou et al., 19 B. Knockdown of TPM3 in cells part A was validated by western blotting, β-actin was used as a loading control. C. Representative images of wound healing assay in part A. Scale bars are 200 μ m. D. MDA-MB-231 cells were treated with ATM-3507 (6 µM) for 1 h prior to and during exposure to 21% or 2% O 2 for 8 h. The % of wound closure is shown in each condition. Statistical testing was done using a two-way ANOVA with Šídák’s multiple comparison test. E. Representative images of wound healing assay in part D. Scale bars are 200 μ m. F. MDA-MB-231 cells were treated with siTPM3 or a scramble siRNA (siSCR) followed by exposure to 21% or 2% O 2 for 16 h. The invasion fold change relative to normoxic control is shown in each condition. Statistical testing was done by two-way ANOVA with Šídák’s multiple comparison test. G. Knockdown of TPM3 in cells used in part F was validated by western blotting, β-actin was used as a loading control. H. Representative images of invasion assay in part F. Data from three separate experiments ( n = 3) are displayed with mean ± standard error of the mean (SEM) unless specified otherwise. * p < 0.05, ** p < 0.01, **** p 0.05. Figure 4. Inhibition of TPM3 combines effectively with standard of care for TNBC A. MDA-MB-231 cells were treated with Carboplatin (1 - 64 μM) for 48 h and ATM-3507 (1.5- 12 μM) for 17 h. For the 16 h before an MTT assay was carried out the cells were in hypoxia (2% O 2). B. MDA-MB-231 cells were treated with Doxorubicin (1 - 64 μM) for 24 h and ATM-3507 (1.5- 12 μM) for 17 h. For the 16 h before an MTT assay was carried out the cells were in hypoxia (2% O2). .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Zhou et al., 20 C. MDA-MB-231 cells were treated with Paclitaxel (1 - 64 nM) for 72 h and ATM-3507 (1.5- 12 μM) for 17 h. For the 16 h before an MTT assay was carried out the cells were in hypoxia (2% O2). In A, B and C Highest Single Agent (HSA) synergy score was assessed by an interactive platform Combenefit, scores > 0 represent sy nergism (blue) and scores < 0 represent antagonism (red). Dose-response matrixes wi th drug concentrations on axes and combination effects as heatmap overlay are shown. Statistical testing was carried out using the built-in analysis algorithm 37. D. MDA-MB-231 cells were treated with siTPM3 or a scramble siRNA (siSCR) followed by exposure to 21% or 2% O 2 for 16 h prior to irradiation using the doses indicated. Cells in hypoxia were irradiated in hypoxic conditions (shown schematically). After irradiation, all cells were returned to 21% O2 and a colony survival assay carried out. Data from three separate experiments ( n = 3) are displayed with mean ± standard error of the mean (SEM). Figure 5. Conditioned media from hypoxic cells increases migration in normoxic cells in a TPM3-dependent manner A. Schematic representation of the wound assay used to determine the impact of TPM3 loss in donor cells on recipient cell migration. B. MDA-MB-231 (donor) cells were treated with siTPM3 or a scramble siRNA and exposed to 21% or 2% O 2 for 16 h, followed by western blotting to validate TPM3 knockdown and HIF-1α stabilisation. C. Representative images of wound healing assay from cells treated as in part D. Scale bars, 500 μ m. D. MDA-MB-231 donor cells were treated with siTPM3 or a scramble siRNA (siSCR) followed by exposure to 21% or 2% O 2 for 16 h. Conditioned media was then collected and applied to wounded MDA-MB-231 recipient cells in 21% O 2 for 16 h. Quantification of .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Zhou et al., 21 normalised wound healing % is shown as fold change relative to control media. Statistical testing was done by paired t-test. E. MDA-MB-231 donor cells were exposed to 21% and 2% O 2 for 16 h, followed by western blotting to indicated proteins. F. Schematic representation of the wound healing assay used to determine the impact of Dynasore. G. MDA-MB-231 donor cells were exposed to 21% or 2% O 2 for 16 h. Conditioned media was then collected and added to wounded MDA-MB-231 normoxic recipient cells +/- Dynasore (50 μ M) for 16 h. Quantification of normalised wound healing % is shown as fold change relative to control media. Statistical testing was done by paired t-test. Data from three separate experiments ( n = 3) are displayed with mean ± standard error of the mean (SEM). * p < 0.05, ** p < 0.01. Figure 6. EVs generated by hypoxic cells increase migration of normoxic cells A. MDA-MB-231 cells in 21% O 2 and 2% O 2 (12 h) were processed and imaged by TEM. Images of late endosomes (enlarged in i, ii), EVs (enlarged in iii, iv) and mitochondria (enlarged in v, vi) are shown. Scale bars are indicated on each image. B. MDA-MB-231 cells were exposed to 21% or 2% O 2 for 24 h followed by EV isolation. Nanoparticle tracking analysis (NTA) was then carried out to determine the total concentration of EVs. Statistical testing was done by paired t-test. C. Representative images of NTA from part B. D. Particle diameter was determined for the EVs isolated in part B using NTA. E. MDA-MB-231 cells were exposed to 21% or 2% O 2 for 24 h, followed by western blotting of whole cell lysates for the indicated proteins. F. Schematic representation of the wound healing assay to determine the impact of addition of EVs from normoxic or hypoxic cells on normoxic recipient cells. G. MDA-MB-231 donor cells were exposed to 21% or 2% O 2 for 24 h. Conditioned media was then collected and EVs isolated. EVs were then resuspended in FBS free normoxic .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Zhou et al., 22 culture media and added to MDA-MB-231 normoxic recipient cells +/- Dynasore (50 μ M) for 16 h. Quantification of normalised wound healing % is shown as fold change relative to control media. Statistical testing was done by paired t-test. Data from three separate experiments ( n = 3) are displayed with mean ± standard error of the mean (SEM). * p < 0.05, ** p < 0.01, and **** p < 0.0001. Figure 7. TPM3 is an EV cargo protein A. MDA-MB-231 cells were treated with siTPM3 or a scramble siRNA and exposed to 21% or 2% O2 for 24 h. EVs were isolated, followed by nanoparticle tracking analysis (NTA). The total concentration of EVs from each condition is shown. Statistical testing was done by paired t-test. B. Particle diameter distribution of the EVs isolated in part A. Statistical testing was done using a paired t-test. C. Western blotting of lysates from cells in part A to verify TPM3 knockdown and HIF-1 α stabilisation. D. MDA-MB-231 cells cultured under normoxic (21% O 2) or hypoxic (2% O 2) conditions for 24 h followed by western blotting of whole cell lysates (WCL) as indicated. E. Western blotting of purified EVs isolated from cells in part D followed by NTA quantification to ensure equal loading. ALIX was used as an EV marker and PRDX3/GM130 used to validate EV purity. F. Quantification of TPM3 protein expression in EVs from part E. TPM3 band intensity was normalised to ALIX (loading control) and expressed relative to 21% O2. G. A summary schematic. TPM3 is induced in a broad range of physiologically relevant hypoxic conditions in a HIF-1 dependent manner. The accumulation of TPM3 (shown in green) contributes to F-actin (shown in red) stabilisation at the leading edge and increased cell motility/invasion. EVs produced by the hypoxic cells contain TPM3, in addition to many other proteins, and when transferred to cells in normoxia (oxic) conditions also increases their motility. .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Zhou et al., 23 * p < 0.05, ** p 0.05.

1. Liao, C., Liu, X., Zhang, C. & Zhang, Q. Tumor hypoxia: From basic knowledge to therapeutic implications. Semin Cancer Biol 88, 172–186 (2023). 2. Tutzauer, J. et al. Breast cancer hypoxia in relation to prognosis and benefit from radiotherapy after breast-conserving surgery in a large, randomised trial with long-term follow-up. Br J Cancer 126, 1145–1156 (2022). 3. Comprehensive molecular portraits of human breast tumours. Nature 490, 61– 70 (2012). 4. Chen, Z., Han, F., Du, Y ., Shi, H. & Zhou, W. Hypoxic microenvironment in cancer: molecular mechanisms and therapeutic interventions. Signal Transduct Target Ther 8, 70 (2023). 5. Schito, L. & Rey-Keim, S. Hypoxia signaling and metastatic progression. Semin Cancer Biol 97, 42–49 (2023). 6. Gunning, P . W., Hardeman, E. C., Lappalainen, P. & Mulvihill, D. P. Tropomyosin – master regulator of actin filament function in the cytoskeleton. J Cell Sci (2015) doi:10.1242/jcs.172502. 7. Zhao, Y.-C. et al. TPM3: a novel prognostic biomarker of cervical cancer that correlates with immune infiltration and promotes malignant behavior in vivo and in vitro. Am J Cancer Res 13, 3123–3139 (2023). 8. Yang, J. et al. ALK-TPM3 rearrangement in adult renal cell carcinoma: a case report and literature review. Diagn Pathol 14, 112 (2019). 9. Tian, Z., Zhao, J. & Wang, Y. The prognostic value of TPM1–4 in hepatocellular carcinoma. Cancer Med 11, 433–446 (2022). 10. Zhu, Y. et al. Dual drive coexistence of EML4‐ ALK and TPM3‐ ROS1 fusion in advanced lung adenocarcinoma. Thorac Cancer 9, 324–327 (2018). 11. Lam, C. Y . et al. Identification and Characterization of Tropomyosin 3 Associated with Granulin-Epithelin Precursor in Human Hepatocellular Carcinoma. PLoS One 7, e40324 (2012). 12. Currier, M. A. et al. Identification of Cancer-Targeted Tropomyosin Inhibitors and Their Synergy with Microtubule Drugs. Mol Cancer Ther 16, 1555–1565 (2017). 13. Chen, S. et al. TPM3 mediates epithelial-mesenchymal transition in esophageal cancer via MMP2/MMP9. Ann Transl Med 9, 1338–1338 (2021). 14. Yao, B. et al. Delivery of platelet TPM3 mRNA into breast cancer cells via microvesicles enhances metastasis. FEBS Open Bio 9, 2159–2169 (2019). 15. Ye, I. C. et al. Molecular Portrait of Hypoxia in Breast Cancer: A Prognostic Signature and Novel HIF-Regulated Genes. Molecular Cancer Research 16, 1889–1901 (2018). 16. Buffa, F. M., Harris, A. L., West, C. M. & Miller, C. J. Large meta-analysis of multiple cancers reveals a common, compact and highly prognostic hypoxia metagene. Br J Cancer 102, (2010). 17. Bader, S. B., Dewhirst, M. W. & Hammond, E. M. Review cyclic hypoxia: An update on its characteristics, methods to measure it and biological implications in cancer. Cancers (Basel) 13, (2021). .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Zhou et al., 24 18. Bader, S. B. et al. Replication catastrophe induced by cyclic hypoxia leads to increased APOBEC3B activity. Nucleic Acids Res (2021) doi:10.1093/nar/gkab551. 19. Ma, T. S. et al. Hypoxia-induced transcriptional stress is mediated by ROS- induced R-loops. Nucleic Acids Res 51, 11584–11599 (2023). 20. Ledaki, I. et al. Carbonic anhydrase IX induction defines a heterogeneous cancer cell response to hypoxia and mediates stem cell-like properties and sensitivity to HDAC inhibition. Oncotarget 6, 19413–19427 (2015). 21. Kumari, R. et al. Focal adhesions contain three specialized actin nanoscale layers. Nat Commun 15, 2547 (2024). 22. Nagano, M., Hoshino, D., Koshikawa, N., Akizawa, T. & Seiki, M. Turnover of Focal Adhesions and Cancer Cell Migration. Int J Cell Biol 2012, 310616 (2012). 23. Janco, M. et al. Molecular integration of the anti-tropomyosin compound ATM- 3507 into the coiled coil overlap region of the cancer-associated Tpm3.1. Sci Rep 9, 11262 (2019). 24. Li, Y . et al. Invasion and metastasis in cancer: molecular insights and therapeutic targets. Signal Transduct Target Ther 10, 57 (2025). 25. Xu, X. et al. Targeting the actin/tropomyosin cytoskeleton in epithelial ovarian cancer reveals multiple mechanisms of synergy with anti-microtubule agents. Br J Cancer 125, 265–276 (2021). 26. Meng, S.-S. et al. LincRNA-p21 promotes mesenchymal stem cell migration capacity and survival through hypoxic preconditioning. Stem Cell Res Ther 9, 280 (2018). 27. Zang, J. L. et al. CCSer2 gates dynein activity at the cell periphery. J Cell Biol 224, (2025). 28. Roy, S. et al. Hypoxic tumor microenvironment: Implications for cancer therapy. Exp Biol Med 245, 1073 (2020). 29. Willms, E. et al. Cells release subpopulations of exosomes with distinct molecular and biological properties. Sci Rep 6, 22519 (2016). 30. Gomes, F., Turano, H., Haddad, L. A. & Netto, Luis. E. S. Human mitochondrial peroxiredoxin Prdx3 is dually localized in the intermembrane space and matrix subcompartments. Redox Biol 78, 103436 (2024). 31. Kee, A. J. et al. On-target action of anti-tropomyosin drugs regulates glucose metabolism. Sci Rep 8, 4604 (2018). 32. Zhang, Q. et al. Supermeres are functional extracellular nanoparticles replete with disease biomarkers and therapeutic targets. Nat Cell Biol 23, 1240–1254 (2021). 33. Hurwitz, S. N. et al. Proteomic profiling of NCI-60 extracellular vesicles uncovers common protein cargo and cancer type-specific biomarkers. Oncotarget 7, 86999–87015 (2016). 34. Singleton, D. C. et al. Hypoxic regulation of RIOK3 is a major mechanism for cancer cell invasion and metastasis. Oncogene 34, 4713–4722 (2015). 35. Dang, D. T. et al. Hypoxia-Inducible Factor-1α Promotes Nonhypoxia-Mediated Proliferation in Colon Cancer Cells and Xenografts. Cancer Res 66, 1684– 1693 (2006). 36. Pires, I. M. et al. Targeting radiation-resistant hypoxic tumour cells through ATR inhibition. Br J Cancer 107, (2012). 37. Di Veroli, G. Y . et al. Combenefit: an interactive platform for the analysis and visualization of drug combinations. Bioinformatics 32, 2866–2868 (2016). .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Zhou et al., 25 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Figure 1 HIF-1α GRP78 KAP1-S824 21 2 <0.1 Cyclic β-actin KAP1 % O2 4 8 16 24 TPM3 β-actin HIF-1α Time (h) 0 A B C D E F G H I J K Survival Probability Overall Survival (Months) Low TPM3 expression High TPM3 expression Log-rank p = 0.039 HR = 2.6 (95% CI: 1.02 -6.68) p (HR) = 0.0463 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Figure 2 F-actin TPM3 DAPI Merge Merge siSCR siTPM3siTPM3 siSCR Normoxia Hypoxia A B C D E F Normoxia TPM3 F-actin DAPI .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Figure 3 siTPM3 siSCR 21% O2 2% O2 TPM3 β-actin A B C D E F G H TPM3 β-actin siTPM3 21% O2 2% O2 siSCR 21% O2 2% O2 ControlATM-3507 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Figure 4 A B C D Carboplatin ATM-3507 ATM-3507 (µM) 1 2 4 8 16 32 64 1 3 6 12 Paclitaxel (nM) Carboplatin (µM) ATM-3507 (µM) 1 3 6 12 1 2 4 8 16 32 64 Doxorubicin (µM) ATM-3507 (µM) 1 2 4 8 16 32 64 1 3 6 12 2% O221% O2 Doxorubicin ATM-3507 2% O221% O2 Paclitaxel ATM-3507 2% O221% O2 2% O221% O2 21% O2 IR .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Figure 5 221 TPM3 β-actin HIF-1α - + - + % O2 siTPM3 TPM3 β-actin HIF-1α 221 % O2 Conditioned media Wound assay DONOR cells RECIPIENT cells 21% or 2% O2 +/- TPM3 21% O2 Conditioned media Wound assay DONOR cells RECIPIENT cells 21% or 2% O2 21% O2 +/- Dynasore A B C D F G E Norm cond- media Hyp cond- media siScrsiTPM3 Control .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Figure 6 TPM3 β-actin HIF-1α 221 % O2 EV collection Wound assay DONOR cells RECIPIENT cells 21% or 2% O2 21% O2 +/- Dynasore A B C D F E G 21% O2 2% O2 21% O22% O2 vi viiii ivii .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint Figure 7 A B C D E F TPM3 ALIX PRDX3 21 2 % O2 β-actin WCL TPM3 ALIX PRDX3 21 2 % O2 EVs HIF-1 TPM3 Increased motility Cancer cell in hypoxic fraction of tumour Cancer cell in oxic fraction of tumour G 221 TPM3 β-actin HIF-1α - + - + % O2 siTPM3 GM130 GM130 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 9, 2026. ; https://doi.org/10.64898/2026.01.08.698356doi: bioRxiv preprint