Chemotherapy-Induced Upregulation of Small Extracellular Vesicle-Associated PTX3 Accelerates Breast Cancer Metastasis

Breast cancer is the most commonly diagnosed malignancy and the leading cause of cancer-associated death in women worldwide, and is responsible for nearly 15% of all new cancer cases and 40,000 deaths per year in the United States (1). Approximately 12% to 17% of newly diagnosed breast cancer cases are classified as triple negative (TNBC) due to the absence of cellular hormone and growth factor receptors (2). Despite the relatively low incidence, TNBC is associated with higher rates of recurrence and lower overall survival than receptor-positive tumors (3). Specifically, patients diagnosed with TNBC have an overall 77% five-year survival rate, compared with 93% for other breast cancer subtypes, and those with metastatic disease have a 22% five-year survival and 80% five-year recurrence rate (4).

Neoadjuvant chemotherapy treatment (NAC), which is typically composed of an anthracycline, alkylator, and taxane regimen, is the standard of care for patients with locally advanced TNBC (4). Despite the ubiquitous use of NAC, only ∼30% of TNBC patients achieve a pathologic complete response (pCR), defined as an absence of residual invasive disease, following treatment (4–6). Patients who do not achieve pCR have significantly decreased three-year overall and disease-free survival rates compared with patients who do achieve pCR (5), and evidence suggests that in a subset of patients, chemotherapy treatment itself may actually promote tumor metastasis (7–12). Therefore, an urgent need exists to understand the molecular mechanisms underlying TNBC metastasis, and how chemotherapy treatment accelerates metastasis, in order to facilitate the development of more effective therapeutic options.

Small extracellular vesicles (sEV) have recently emerged as important regulators of cancer progression and metastasis through the horizontal transfer of cellular material (13). These nanovesicles are secreted by almost all cell types and contain a variety of proteins, RNA, and DNA, which are packaged into intraluminal vesicles within the late endosome and released into the extracellular environment following fusion of the multivesicular body (MVB) with the plasma membrane (14). Uptake of tumor-derived sEVs in secondary tissues has been shown to promote tumor cell invasiveness, facilitate epithelial-to-mesenchymal transition, suppress the host immune system, and prime the premetastatic niche to enhance metastasis (15–20). Despite the well-defined and multifaceted role of sEVs in cancer progression, the effect of chemotherapy treatment on tumor-derived sEV secretion and function is poorly understood. In this study, we establish that the anthracycline doxorubicin (DXR) upregulates breast cancer cell sEV secretion and differentially regulates sEV protein content, and that sEVs derived from DXR-treated cells accelerate metastasis in models of both highly metastatic and poorly metastatic TNBC, demonstrating that sEV secretion is one mechanism by which chemotherapy treatment promotes metastasis. Moreover, we identify the glycoprotein pentraxin-related protein 3 (PTX3) as a novel regulator of sEV-induced breast cancer metastasis, which accelerates tumor progression through priming of the premetastatic niche. We further show that inhibition of both sEV uptake in secondary organs and secretion by primary tumor cells suppresses chemotherapy-induced metastasis, suggesting that targeting tumor-derived sEVs may be an effective therapeutic strategy to suppress metastasis in TNBC patients.

The human breast cancer cell lines MDA-MB-231 (ATCC HTB-26), MDA-MB-468 (ATCC HTB-132), and MCF7 (ATCC HTB-22) were purchased from ATCC. SV40 large T antigen-immortalized mouse embryonic fibroblasts were generated in our lab as previously described (21). Luciferase reporter cells were generated with a pCDH1-SV40-Luc2 construct and selected with hygromycin (200 μg/mL). CRISPR/Cas9-mediated PTX3 knockout (KO) was performed by subcloning three single-guide RNAs (gRNA) targeting PTX3 (5′-AGTGAACTTACGGTCCTCAG-3′, 5′-CCACCACAAACACTATGGAT-3′, 5′-GACGCGAGCCGACCTGCACG-3′) into pLenti-CRISPR-V2 (Addgene 52961). Lentiviral packaging in 293T cells and transduction of pLenti-CRISPR-PTX3 into MDA-MB-231 cells were performed as previously described (22). Puromycin selection (1 μg/mL) was used to isolate single clones, which were screened using immunoblotting for gene disruption, and four validated single clones were pooled. CRISPR/Cas9-mediated RAB27A KO was performed by transducing MDA-MB-231 cells with lentiCas9-Blast (Addgene 52962) and using Blasticidin selection (4 μg/mL) to create a Cas9-stable control cell line. Three gRNAs targeting RAB27A (5′-CCTCTGCCAGTGGCTCCATC-3′, 5′-CTCCAAATTTATCACAACAG-3′, 5′-TGCAGTTATGGGACACAGCA-3′) were subcloned into LRG (Addgene 65656). LRG-Rab27a was delivered by lentiviral transduction to MDA-MB-231 Cas9 cells followed by FACS sorting as previously described (22). Single clones were validated by immunoblotting for gene disruption and two clones pooled. All cell lines in the lab are passaged for less than 6 months before use and periodically authenticated by Mycoplasma testing, morphologic inspection, and STR analysis.

To make sEV-depleted FBS, heat-inactivated FBS was centrifuged twice at 120,000 RCF in an SW32Ti rotor (Beckman) for 12 hours at 4°C and filtered through a 0.2 μm filter. For sEV isolation, conditioned media were collected from cells cultured for 24 hours in sEV-depleted media with 400 nmol/L (MDA-MB-231 and MCF7) or 40 nmol/L (MDA-MB-468) DXR, or with a volume of DMSO equivalent to 10% the volume of DXR. Conditioned media were centrifuged at 500 RCF for 10 minutes at 4°C, filtered through a 0.2-μm syringe filter (VWR 28145-501), and concentrated using Pierce 100K MWCO protein concentrators (Thermo Fisher 88533) by centrifuging at 4,000 RCF and 4°C. The concentrated supernatant was centrifuged at 10,000 RCF for 20 minutes at 4°C and then centrifuged twice at 120,000 RCF for 2.5 hours at 4°C and once for 12 hours at 120,000 RCF and 4°C in an SW55Ti rotor (Beckman). The sEV pellet was resuspended in phosphate-buffered saline (PBS) and stored at −20°C. Density gradient ultracentrifugation was performed using a protocol adapted from Kowal and colleagues (23).

sEV quantification was performed using a NanoSight NS300 (Malvern Instruments Ltd.). EV samples were diluted to a volume of 1 mL (1:500 to 1:1,000) in particle-free water and five, 60-second videos were generated at 24.98 frames/sec in scatter mode, as previously described (24). Particle size and concentration were calculated using NanoSight software version 3.2 (Malvern Instruments Ltd.). To quantify vesicle secretion per cell, cells were treated with DMSO or DXR in sEV-depleted media for 24 hours and trypsinized using Trypsin-EDTA 0.05% without Ca or Mg (VWR Scientific 45000-660) at the time of conditioned media collection. Cells were counted using a Cellometer Auto 2000 Cell Viability Counter (Nexcelom Bioscience), and total vesicle concentration in each sample was divided by the final cell count.

sEV electron microscopy was performed using a protocol adapted from Rikkert and colleagues (25). Briefly, sEVs were diluted in PBS, and 10 μL sample was pipetted onto a 400 mesh copper grid with carbon-coated formvar film and imaged using a JEOL JEM1400 transmission electron microscope (JEOL USA Inc.).

For mass spectrometry (MS) analysis, 50 μg sEVs were resuspended in PBS and quantified using tandem MS at the University of Michigan with a TMT-10plex Isobaric Label Reagent Set (Thermo Fisher 90110) according to the manufacturer's instructions. Samples were analyzed using an Orbitrap Velos mass spectrometer (Thermo Fisher Scientific), and data processing was performed using Proteome Discover and SequestHT. Proteins were filtered against a human database using a false discovery rate of less than 5%. Pathway analysis was performed using Advaita iPathway Guide (Advaita Bioinformatics). The proteomic heat map was generated using the R complex-heat map package.

All animal studies were performed according to guidelines established and approved by the Institutional Animal Care and Use Committee at the Penn State College of Medicine. An orthotopic xenograft breast cancer model was generated by injecting 1.0 × 10⁶ cells into the left fifth inguinal mammary fat pad of 6- to 8-week-old female NOD SCID gamma (NSG, Jackson 005557) or female NOD Rag gamma (NRG, Jackson 007799) mice. Cells were injected in a 50:50 mixture of PBS and Matrigel basement membrane matrix (Fisher CB-40234). For sEV treatment experiments, mice received tail-vein injections of 10 μg purified sEVs in 100 μL PBS, or an equivalent volume of PBS, three times per week, with treatments beginning 1 week after xenograft. For chemotherapy treatment experiments, mice received tail-vein injections of 5 mg/kg DXR (NRG and NSG mice) or 2.5 mg/kg DXR (NSG mice) as indicated, or an equivalent volume of PBS on days 7 and 12 after tumor xenograft. As indicated, chemotherapy-treated mice received daily subcutaneous injections of 0.2 mg/kg heparin in 100 μL PBS, or an equivalent volume of PBS, beginning 1 week after tumor xenograft and continuing for the duration of the experiment. Metastatic tumor growth was tracked by imaging mice on a weekly basis using an IVIS Lumina III in vivo bioluminescent imaging system. Mice were anesthetized using 2.5% isoflurane (IsoSol Isoflurane, USP, VEDCO) and injected with 5 μL/g body weight of 30 mg/mL D-luciferin dissolved in PBS 5 minutes prior to imaging. Photon flux was calculated using region of interest (ROI) measurements around the primary tumor site or the anterior thoracic area for lung and liver metastasis. Primary tumor volume was measured using calipers and calculated as volume = π(length*width²)/6. At the experiment endpoint, mice were humanely euthanized and tumor, lung, and liver tissue were harvested for ex vivo analysis and histology.

A metastatic model was established by injecting female 6- to 8-week-old NSG mice via the tail vein with 10 μg purified sEVs in 100 μL PBS, or an equivalent volume of PBS, on days 0, 2, and 4, followed by tail-vein injection of 2.5 × 10⁵ cells on day 5. Metastatic tumor growth was quantified by weekly IVIS imaging. Whole-body ROI measurements were used to quantify metastatic tumor growth. At the experimental endpoint, mice were euthanized, and lung and liver tissue were harvested for ex vivo analysis.

GraphPad Prism 7.0 was used for statistical analysis of data. Two-tailed Student t tests were used for single comparisons, and group differences were evaluated using one-way or two-way ANOVA. Statistical significance was set to P < 0.05.

To confirm that chemotherapy treatment accelerates breast cancer metastasis, we generated orthotopic xenografts of MDA-MB-231 cells expressing a luciferase reporter in the mammary fat pad of female NSG mice, and then treated mice twice over the course of 5 days with a low dose of the topoisomerase II inhibitor DXR or an equivalent volume of PBS, in line with established treatment models (Supplementary Fig. S1A; ref. 26). Bioluminescent imaging of the anterior thoracic region, which masks the luminescent signal produced by the primary tumor in order to more clearly visualize metastatic radiance in the pulmonary and hepatic regions, confirmed that mice treated with DXR demonstrated a significant acceleration in tumor cell metastasis compared with mice treated with PBS (Supplementary Fig. S1B and S1C). Primary tumor volume was significantly decreased in mice treated with DXR compared with PBS (Supplementary Fig. S1D and S1E), confirming that DXR decreases primary tumor growth while promoting metastasis.

We hypothesized that chemotherapy treatment altered sEV secretion from primary tumors, and that sEVs derived from chemotherapy-treated tumor cells promote metastasis. In order to determine the effect of chemotherapy treatment on sEV secretion, we selected concentrations of DXR that inhibit less than 10% of cell viability in vitro in the highly metastatic MDA-MB-231 and the poorly metastatic MDA-MB-468 human TNBC cells lines (Supplementary Fig. S1F). Differential ultracentrifugation was used to isolate sEVs from the conditioned media of cells treated for 24 hours with DXR or DMSO, the vehicle control. Using electron microscopy, immunoblotting, and nanoparticle tracking analysis (NTA), we confirmed that isolated particles displayed the characteristic size, morphology, and surface markers of sEVs (Fig. 1A–C). We found that low-dose DXR treatment significantly upregulated both the number of sEVs and the amount of sEV protein secreted per cell from both breast cancer cell lines in vitro (Fig. 1D and E). However, the amount of protein secreted per sEV was not significantly affected by DXR treatment (Fig. 1F). Importantly, we found that the number of circulating sEVs was significantly upregulated in the plasma of tumor-bearing mice treated with DXR compared with mice treated with PBS (Fig. 1G), demonstrating that DXR upregulates sEV secretion both in vitro and in vivo.

To determine the effect of sEVs derived from DXR-treated breast cancer cells on tumor growth and progression, we generated orthotopic xenografts of MDA-MB-231 and MDA-MB-468 cells in NSG mice. One-week after xenograft, we began treating mice three times per week with tail-vein injections of 10 μg purified sEVs derived from DMSO- or DXR-treated breast cancer cells, or with an equivalent volume of PBS (Fig. 2A). Bioluminescent imaging of the anterior thoracic region demonstrated that treatment with DXR-EVs significantly accelerated metastatic tumor growth compared with treatment with DMSO-EVs or PBS in both the highly metastatic and poorly metastatic xenograft models (Fig. 2B and C; Supplementary Fig. S2A and S2B). Histologic scoring of lung metastatic tumor burden using hematoxylin and eosin–stained lung sections demonstrated that DXR-EV treatment significantly enhanced pulmonary metastasis, while quantification of total liver tumor area demonstrated a slight, but not statistically significant, increase in liver metastasis in mice treated with DXR-EVs (Fig. 2D–F; Supplementary Fig. S2C–S2E). Notably, sEV treatment did not significantly affect primary tumor growth in either cell line, as demonstrated by bioluminescent imaging (Fig. 2G; Supplementary Fig. S2F and S2G), weekly caliper measurements (Fig. 2H), and quantification of tumor volume ex vivo (Supplementary Fig. S2H–S2K). These results suggest that DXR-EVs accelerate pulmonary metastasis without affecting primary breast tumor growth.

We hypothesized that DXR-EVs accelerate breast tumor metastasis either by promoting primary tumor cell migration or invasion or by priming the premetastatic niche in secondary organs. To test the hypothesis that DXR-EVs enhance primary tumor cell movement, we performed migration and invasion assays in vitro using an IncuCyte Live-Cell Analysis System. MDA-MB-231 and MDA-MB-468 cells treated with 10 μg/mL DXR-EVs demonstrated no significant increase in migration or invasion in vitro compared with cells treated with DMSO-EVs or PBS (Supplementary Fig. S3A–S3D), suggesting that uptake of DXR-EVs by primary tumor cells may not significantly enhance metastasis.

To test the hypothesis that DXR-EVs prime the premetastatic niche, we first used whole-organ fluorescent imaging to demonstrate that both MDA-MB-231– and MDA-MB-468–derived Vybrant DiD-labeled DMSO- and DXR-EVs can localize to the lungs and liver in vivo within the first 12 to 24 hours after injection (Supplementary Fig. S4A–S4J). To determine how uptake of sEVs in these organs affects tumor progression, we experimentally modeled metastatic dissemination by injecting female NSG mice three times over the course of 5 days with 10 μg DXR-EV or DMSO-EVs, or an equivalent volume of PBS, followed by tail-vein injection of MDA-MB-231 or MDA-MB-468 cells (Fig. 3A). We found that whole-body metastatic tumor growth was significantly enhanced in mice pretreated with DXR-EVs compared with mice pretreated with DMSO-EVs or PBS in both tumor models (Fig. 3B and C; Supplementary Fig. S5A and S5B). Of note, bone marrow metastatic bioluminescent signaling was increased in mice treated with MDA-MB-468 DXR-EVs, highlighting differences in EV organotropism between the two cell lines. Lung section scoring confirmed that tumor burden was significantly enhanced in mice pretreated with DXR-EVs compared with DMSO-EVs or PBS, while the total liver tumor area was not statistically different between treatment groups, although an increase trend in liver metastasis was observed in mice pretreated with MDA-MB-231 DXR-EVs (Fig. 3D–F; Supplementary Fig. S5C and S5D). These results suggest that sEVs derived from DXR-treated breast cancer cells promote pulmonary metastasis by priming the premetastatic niche.

To establish how uptake of DXR-EVs primes the pulmonary premetastatic niche, we performed unbiased proteomic analysis of MDA-MB-231 DMSO- and DXR-EVs. Of the 1,758 EV proteins found in our screen, we identified 40 significantly upregulated and 167 significantly downregulated proteins in DXR-EVs compared with DMSO-EVs (Fig. 4A–C). We were particularly interested in the inflammatory glycoprotein PTX3, which was one of the most highly upregulated proteins in our screen. High PTX3 levels are associated with a significantly decreased probability of relapse-free survival in breast cancer patients who have received NAC (Fig. 4D), making it a promising target. To further confirm that PTX3 is a sEV-associated protein, we performed density-gradient ultracentrifugation and found strong colocalization between PTX3 and the sEV marker CD63 (Supplementary Fig. S6A). We used CRISPR/Cas9 to KO PTX3 (crPTX3) in MDA-MB-231 cells (Fig. 4E) and confirmed that PTX3 KO did not affect tumor cell migration in vitro compared with CRISPR nontargeting (crNT) MDA-MB-231 cells (Supplementary Fig. S6B). We isolated DMSO- and DXR-EVs from crNT and crPTX3 cells and used immunoblotting to confirm that PTX3 is absent in crPTX3 sEVs, and that DXR treatment upregulates PTX3 expression approximately 2-fold in crNT sEVs (Fig. 4F and G). This phenotype is unlikely unique to the MDA-MB-231 cell line, as PTX3 was also upregulated approximately 1.5-fold in DXR-EVs derived from MDA-MB-468 and MCF7 cells (Supplementary Fig. S6C and S6D). We further characterized crNT and crPTX3 sEVs using electron microscopy and NTA to confirm that PTX3 KO did not affect sEV size or morphology (Fig. 4H and I). NTA demonstrated that DXR significantly increased sEV secretion per cell in both crNT and crPTX3 cells compared with DMSO treatment, and no significant differences in sEV secretion were found between crNT and crPTX3 cells (Fig. 4I and J). We used dot plots with or without detergent treatment, which disrupts the vesicle membrane to expose internal proteins, to confirm that PTX3 is upregulated in DXR-EVs and to demonstrate that PTX3 localizes to the sEV surface (Fig. 4K).

To determine how sEV-associated PTX3 affects breast tumor progression, we first confirmed that MDA-MB-231 cell migration and invasion were not enhanced by treatment with crPTX3 sEVs compared with crNT EVs or PBS (Supplementary Fig. S6E and S6F). To establish whether sEV-associated PTX3 accelerates breast cancer metastasis, we repeated our premetastatic niche model by treating female NSG mice three times with 10 μg crNT DXR-EVs or crPTX3 DXR-EVs, or an equivalent volume of PBS, followed by tail-vein injection of MDA-MB-231 cells (Fig. 5A). As expected, bioluminescent imaging confirmed that mice pretreated with crNT DXR-EVs demonstrated a significant acceleration in whole-body metastasis compared with PBS (Fig. 5B and C). However, mice pretreated with crPTX3 DXR-EVs showed a drastic reduction in tumor burden compared with crNT DXR-EV pretreatment, with levels of metastatic progression equivalent to mice pretreated with PBS (Fig. 5B and C). Histologic scoring of lung tumor burden corroborated this finding, despite no significant differences in liver tumor burden between treatment groups (Fig. 5D–F). We found no significant differences between crNT and crPTX3 EV uptake in cells in vitro (Supplementary Fig. S6G), suggesting that differences in EV uptake are unlikely responsible for the observed decrease in metastasis.

To confirm that PTX3 is involved in chemotherapy-induced metastasis, we performed orthotopic xenografts of crNT and crPTX3 MDA-MB-231 cells in female NRG mice. One week after xenograft, on day 0, we treated mice with DXR or PBS followed by a second treatment on day 5 (Fig. 5G). We found that anterior thoracic metastatic tumor growth was significantly increased in mice with crNT primary tumors treated with DXR compared with PBS, while mice with crPTX3 primary tumors treated with DXR did not show a comparable increase in metastasis (Fig. 5H and I; Supplementary Fig. S6H). Importantly, we found that sEVs isolated from the plasma of crNT tumor-bearing mice treated with DXR demonstrated a significant increase in PTX3 compared with sEVs isolated from mice treated with PBS, confirming that DXR treatment upregulates circulating sEV-associated PTX3 in vivo (Fig. 5J; Supplementary Fig. S6I). Consistently, crNT and crPTX3 tumors treated with DXR were significantly smaller than crNT tumors treated with PBS at the experiment endpoint (Supplementary Fig. S6J–S6M). Taken together, these results demonstrate that DXR treatment significantly upregulates sEV-associated PTX3, which is critically involved in chemotherapy-induced breast cancer metastasis.

After confirming that sEVs derived from DXR-treated tumor cells accelerate metastasis, we sought to determine whether inhibiting sEV uptake and secretion would suppress chemotherapy-induced metastasis. Heparan sulfate proteoglycans on the surface of secondary cells serve as receptors for sEV binding and internalization, and heparin treatment has been shown to block sEV transfer between cells in vitro (27, 28). We first confirmed that heparin treatment suppresses EV uptake in vivo by treating mice with Vybrant DiD-labeled EVs preincubated with or without 0.2 mg/kg heparin and performing whole-organ fluorescent imaging 24 hours after injection (Fig. 6A–C). After imaging we digested whole lung tissue and used flow cytometry to confirm that heparin pretreatment suppressed EV uptake in vivo (Supplementary Fig. S7A and S7B), and used the IncuCyte system to confirm that heparin treatment does not suppress tumor cell migration in vitro (Supplementary Fig. S7C).

To determine whether heparin treatment suppresses chemotherapy-induced metastasis in vivo, we performed orthotopic xenografts of MDA-MB-231 cells in female NSG mice and treated mice twice over the course of 5 days with tail-vein injections of DXR or PBS. Mice treated with DXR received daily subcutaneous injections of 0.2 mg/kg heparin or an equivalent volume of PBS (Fig. 6G). Quantification of anterior thoracic bioluminescent imaging confirmed that DXR treatment accelerated metastasis compared with treatment with PBS (Fig. 6H and I). Significantly, mice treated with daily subcutaneous doses of heparin demonstrated a complete suppression of DXR-induced metastasis (Fig. 6H and I). Of note, both groups of mice treated with DXR showed a decrease in primary tumor growth compared with mice treated with PBS, with no difference in tumor growth between the two groups, indicating that DXR significantly decreased primary tumor growth and that heparin treatment suppressed chemotherapy-induced metastasis without affecting primary tumor growth (Supplementary Fig. S7D and S7E).

To validate that primary tumor-derived sEVs are involved in chemotherapy-induced metastasis, we used CRISPR/Cas9 to KO RAB27A, a Rab-GTPase involved in shuttling the MVB to the plasma membrane for sEV secretion, in MDA-MB-231 cells (crRab27a; Fig. 6D). We confirmed that RAB27A KO does not affect MDA-MB-231 proliferation in vitro (Supplementary Fig. S7F), and used NTA and electron microscopy to demonstrate that RAB27A KO decreases sEV secretion (Fig. 6E and F). We performed xenografts of crRab27a MDA-MB-231 cells in female NSG mice in parallel with control MDA-MB-231 xenografts, and treated mice twice with DXR (Fig. 6G). Using bioluminescent imaging, we confirmed that genetic inhibition of sEV secretion significantly decreased metastatic tumor growth compared with control tumors treated with DXR (Fig. 6H and I; Supplementary Fig. S7G) and found that DXR decreased both control and crRab27a primary tumor size compared with PBS treatment (Supplementary Fig. S7D and S7E). These results demonstrate that sEVs derived from chemotherapy-treated primary tumor cells are critically involved in promoting breast cancer metastasis.

Literature reports that chemotherapeutic drugs can decrease primary tumor growth while promoting metastasis (8, 26, 29, 30). In this study, we investigated the role of chemotherapy-induced, primary tumor-derived sEVs in cancer progression. We first confirmed that DXR treatment accelerated metastasis in a mouse xenograft model of human TNBC and found that low doses of DXR significantly enhanced sEV secretion in TNBC cell lines in vitro as well as sEV circulation in tumor-bearing mice in vivo, corroborating literature accounts that sEV secretion is upregulated in response to cellular stress (31, 32). We found that DXR-EV treatment accelerated pulmonary metastasis in mouse breast cancer xenograft models compared with control sEV treatment, without affecting primary tumor growth. In agreement with previous reports (33), we found that both DMSO- and DXR-EVs are taken up by lung and liver tissue in vivo and saw no significant differences in the level of sEV uptake between these two groups, suggesting that the sEV cargo, rather than differences in uptake, was primarily responsible for the increase in pulmonary metastasis. We observed a significant acceleration in whole-body and pulmonary metastatic tumor growth in mice that had been pretreated with DXR-EVs compared with DMSO-EVs prior to tumor cell tail-vein injection, demonstrating that DXR-EVs prime the premetastatic niche to accelerate metastasis. While liver metastasis was not significantly affected by sEV treatment, we did observe a slight increase in liver metastasis following MDA-MB-231 DXR-EV treatment in both our primary tumor and premetastatic niche models, suggesting that DXR-EVs may prime the hepatic premetastatic niche to a lesser extent than the pulmonary niche. Similarly, mice pretreated with MDA-MB-468 sEVs demonstrated an increase in bone marrow uptake, highlighting differences in EV organotropism between the two breast cancer cell lines. Proteomic analysis of MDA-MB-231 sEVs revealed over 200 differentially expressed proteins in DXR-EVs. We focused on the upregulated TNF-inducible pattern recognition molecule PTX3, which we found predicts decreased event-free survival when elevated in breast cancer patients following NAC. Mice pretreated with crPTX3 DXR-EVs showed reduced pulmonary metastasis compared with mice pretreated with crNT DXR-EVs prior to tail-vein injection of MDA-MB-231 cells, demonstrating a critical role for PTX3 in sEV-mediated metastasis. Strikingly, while DXR treatment enhanced metastasis in crNT xenografts, metastasis was attenuated in crPTX3 xenografts treated with DXR. Taken together, these data suggest a model whereby DXR treatment upregulates both the secretion of primary breast tumor sEVs and the amount of PTX3 sEV cargo. Uptake of these tumor-derived DXR-EVs in secondary tissues, primarily the lung, primes the premetastatic niche, creating a favorable environment for metastatic breast tumor cell colonization.

One limitation of our study is that we utilized immunodeficient NSG and NRG mice to study the effect of sEVs on human TNBC metastasis. Although chemotherapy treatment weakens the host immune system, immunodeficient mice are not a truly representative model of breast cancer patients, and future studies are necessary to determine how sEV-associated PTX3 regulates metastasis in an immunocompetent model. However, in their 2019 paper, Keklikoglou and colleagues reported that chemotherapy treatment is able to enhance metastasis in both immunodeficient and immunocompetent transgenic mouse models of breast cancer (26). They found that paclitaxel and DXR-treated breast cancer cells release sEVs enriched in annexin A6 (ANXA6), which they found promotes lung metastasis through the activation of endothelial cells, induction of Ccl2, and expansion of monocytes within the premetastatic niche (26). Although we identified ANXA6 in our proteomic screen, it was not significantly enriched in our DXR-EVs, suggesting that different chemotherapeutic drugs likely regulate sEV protein cargo in a dose- and context-dependent manner. Because sEVs contain thousands of proteins and RNAs, this also suggests that ANXA6 is not the only sEV protein responsible for accelerating pulmonary metastasis, and that sEVs derived from chemotherapy-treated cells are likely capable of promoting tumor progression through multiple pathways, either in parallel or in concert. In our model, we found that PTX3 expression is upregulated approximately 2-fold in DXR-EVs derived from the highly metastatic MDA-MB-231 cell line, which expresses a high level of PTX3 mRNA, and upregulated approximately 1.5-fold in DXR-EVs derived from the poorly metastatic MDA-MB-468 and MCF7 cell lines, which express low levels of PTX3 mRNA. While this demonstrates that upregulation of sEV-associated PTX3 is a common mechanism of DXR treatment, it remains an interesting question whether highly metastatic cell lines are more reliant on PTX3-mediated priming of the premetastatic niche than poorly metastatic lines. Indeed, literature demonstrates that different chemotherapeutic agents upregulate various prometastatic sEV cargo, including heparanase (34) and miR-126a (30), highlighting the need to fully investigate the various mechanisms by which different drug-induced cargo accelerate metastasis.

PTX3 has a complex role in tumor progression (35) and has been shown to promote tumor metastasis and chemoresistance following treatment with the chemotherapeutic drugs cisplatin and 5-fluorouracil (36). As a pattern-recognition molecule induced by inflammation, PTX3 plays a critical role in host defense against infectious pathogens, tissue repair, and tumor-related inflammation (37). PTX3 has been shown to interact with Factor H and C1q to regulate complement activation, leading to leukocyte recruitment, proinflammatory cytokine production, and angiogenesis within the tumor microenvironment (38). Although the role of PTX3 in tumor inflammation is well characterized, our model of human TNBC utilizes immunodeficient NSG and NRG mice, which lack T, B, and NK cells, and have deficient cytokine signaling pathways, suggesting that cytokine-mediated inflammation may not be the primary mechanism by which DXR-EV-associated PTX3 accelerates metastasis. In addition to its role in complement activation, PTX3 is also associated with vascular endothelial dysfunction in a variety of pathologies (39, 40). PTX3 has been shown to induce morphologic changes that lead to dysfunction in the endothelial layer through the P-selectin and matrix metalloproteinase-1 (MMP-1) pathway, another protein that was found to be highly upregulated in DXR-EVs in our proteomic screen (41, 42). PTX3 is able to bind to P-selectin on the surface of vascular endothelial cells, leading to reduced nitric oxide production, inhibition of cell proliferation, increased expression of MMP-1, and endothelial cell dysfunction (41). Vessels exposed to PTX3 exhibit broken plasma membranes, diluted cytoplasm, and increased vascular leakage (41). We found that sEV-associated PTX3 localizes to the surface of the vesicle, suggesting that DXR EV-associated PTX3 may bind to P-selectin on the surface of vascular endothelial cells, promoting endothelial dysfunction and vascular leakiness that would enhance the ability of circulating tumor cells (CTC) and prometastatic immune cells to extravasate into the lungs. Chemotherapy treatment has previously been shown to promote metastasis by enhancing the ability of CTCs to intravasate into the vasculature (8), but the role of PTX3 in creating microenvironments of metastasis is unknown. Future research is necessary to determine the role of DXR-EVs on vascular dysfunction in both immunodeficient and immunocompetent mouse models, and to determine whether PTX3 may be a suitable therapeutic target to suppress chemotherapy-induced metastasis and enhance the efficacy of NAC.

Finally, we hypothesized that if DXR-EVs accelerate metastasis, inhibition of sEV secretion from the primary tumor or uptake in secondary tissues would suppress metastasis. The anticoagulant heparin has been shown to inhibit the binding and uptake of sEVs in secondary cells, so we hypothesized that heparin treatment could suppress sEV-mediated metastasis (28). Importantly, we found that mice cotreated with DXR and daily subcutaneous injections of heparin demonstrated an almost complete suppression in metastasis compared with mice treated with DXR and subcutaneous injections of PBS, suggesting that inhibition of primary tumor-derived DXR-EV uptake in secondary tissues does decrease metastasis. To confirm that inhibition of sEV uptake, rather than an off-target effect of heparin, was responsible for the observed suppression in metastasis, we knocked out RAB27A, a Rab-GTPase necessary for sEV secretion, in MDA-MB-231 cells. We found that mice xenografted with crRab27a primary tumors did not demonstrate an increase in metastasis following DXR treatment compared with mice bearing control tumors, reinforcing the finding that primary tumor-derived DXR-EVs play a critical role in chemotherapy-mediated metastasis. Heparin has long been thought to suppress metastasis in various cancer models, but its clinical use has been limited due to the increased risk of patient bleeding (43–45). Our research demonstrates a novel mechanism by which heparin is able to suppress chemotherapy-induced metastasis through the inhibition of DXR-EV uptake in secondary tissues, and suggests that further studies are warranted to determine whether heparin and its derivatives could be utilized as a therapeutic tool for high-risk TNBC patients.

H.G. Wang reports grants from NIH during the conduct of the study. No disclosures were reported by the other authors.

C.A. Wills: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. X. Liu: Data curation and investigation. L. Chen: Resources, data curation, investigation, and methodology. Y. Zhao: Investigation and methodology. C.M. Dower: Conceptualization, resources, and methodology. J. Sundstrom: Resources, software, supervision, investigation, and methodology. H.-G. Wang: Conceptualization, resources, supervision, funding acquisition, project administration, writing–review and editing.

We acknowledge all members of the Wang and Sundstrom labs, the Penn State College of Medicine Department of Pediatrics, Department of Ophthalmology, Department of Comparative Medicine, Bioluminescent Imaging Core, Transmission Electron Microscopy Core, and Flow Cytometry Core, as well as the University of Michigan Proteomics Resource Facility. This work was supported in part by the Lois High Berstler Research Endowment Fund, the Four Diamonds Fund, and the NIH Grant 5T32CA060395.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.