Eribulin activates the cGAS-STING pathway via the cytoplasmic accumulation of mtDNA

ABSTRACT

Microtubule targeted agents are often used in the treatment of breast cancer and, most recently, have been used in combination with immune checkpoint inhibitors to improve their efficacy. While all clinically approved MTAs share an antimitotic mechanism of action, their distinct effects on interphase microtubules can promote differential downstream signaling consequences. We show that the microtubule destabilizer eribulin, but not the microtubule stabilizer paclitaxel, activates the cGAS-STING innate immune signaling pathway through the accumulation of mitochondrial DNA in the cytoplasm.

INTRODUCTION 353/750
Microtubule targeting agents (MTAs) are a mainstay in the treatment of many malignancies, including breast cancer (Dumontet and Jordan, 2010). MTAs have been historically classified as microtubule stabilizers or destabilizers based on the biochemical mechanism by which they disrupt microtubule dynamics. Microtubule stabilizers, including the taxanes and epothilones, promote the net polymerization of microtubules whereas microtubule destabilizers, including the vinca alkaloids and eribulin, promote net microtubule depolymerization (Steinmetz and Prota, 2018). While MTAs have long been thought to exert their anticancer effects primarily through inhibition of mitosis, growing evidence demonstrates that their non-mitotic effects contribute significantly to their anticancer efficacy (Field et al., 2014;Kaul et al., 2019b;Komlodi-Pasztor et al., 2011;Komlodi-Pasztor et al., 2012).
Mechanistically distinct MTAs can also differ from one another in their non-mitotic effects on oncogenic signaling pathways and such differences could be important factors in their clinical efficacy (Dybdal-Hargreaves et al., 2018;Karbowski et al., 2001;Kaul et al., 2019b).
In addition to their impact directly on cancer cells, MTAs also have differential effects on both innate and adaptive immune cell populations implicated in antitumor immunity (Fong et al., 2019). Although some of the compound or class-specific immunomodulatory effects of MTAs, such as the ability of paclitaxel to directly activate TLR4-dependent signaling (Rajput et al., 2013;Wanderley et al., 2018), have been extensively explored, many of the underlying mechanisms of how MTAs elicit distinct immunological effects are unknown. A more complete interrogation of the effects of MTAs on immune signaling pathways is warranted given the recent clinical evaluations of MTAs, including taxanes and eribulin, in combination with immune checkpoint inhibitors to enhance the therapeutic response in metastatic triple-negative breast cancer (TNBC) (Mittendorf et al., 2020;Schmid et al., 2020;Tolaney et al., 2021).
Collectively these observations prompted us to test the overall biological hypothesis that MTAs have different effects on innate immune signaling pathways. Herein, we show that eribulin, but not paclitaxel, induces the cGAS-STING-dependent expression of interferon β (IFNβ) and downstream interferon stimulated genes (ISGs) in both immune and TNBC cells through the cytoplasmic accumulation of mitochondrial DNA (mtDNA) independent of mitotic arrest or the initiation of apoptosis.

Cells and reagents
BT-549 cells were obtained from the Georgetown University Lombardi Comprehensive Cancer Center, Washington, DC. CAL-51 cells were purchased from Creative Bioarray (Westbury, NY).
All other cell lines were purchased from the American Type Culture Collection (ATCC, Manassas, VA). All TNBC lines were authenticated by STR-based profiling (Genetica DNA Laboratories, Cincinnati OH). Primary wild-type bone marrow-derived macrophages (BMDMs) were generated from BALB/c or C57BL/6 mice purchased from Envigo (Indianapolis, IN) as previously described (Hasan et al., 2012). C57BL/6 Sting gt/gt BMDMs were obtained from animals purchased from Jackson Labs (Bar Harbor, ME). BMDMs cells were maintained in DMEM media (Gibco, Grand Island, NY) supplemented with 10% FBS (Corning, Corning, NY) and 50 μg/mL gentamycin (LifeTechnologies, Carlsbad, CA). THP-1, HCC1937, HCC1806, BT-549, MDA-MB-436 and CAL-51 cells were maintained in RPMI 1640 media (Corning) supplemented with 10% FBS and 50 μg/mL gentamycin. HCC1937 Rho 0 cells were generated by culturing cells in 150 ng/mL ethidium bromide (EtBr) for 5 days as described (White et al., 7 2014 (SASI_Hs01_00208506, SASI_Hs01_00208505) and compared to scramble control (SIC002) using the RNAiMAX reagent (Invitrogen, Carlsbad, CA). Extraction of RNA from cells was done using TRIzol (Ambion, Austin, TX) as indicated by the manufacturer. cDNA was synthesized with iScript cDNA synthesis kit (Bio-Rad, Hercules, CA) and analyzed using a Bio-Rad CFX qRT-PCR using iTaq Universal SYBR Green Supermix (Bio-Rad). mRNA fold change was calculated using the 2 -ΔΔCt method where GAPDH was used as the control. All treated and stimulated samples were compared to their corresponding vehicle or mock control, which was set to a relative value of 1. All experiments were performed as two technical replicates with a variability cutoff of less than 1 Ct within each of two independent biological replicates. The range of the biological replicate values is represented in every bar graph. Data were subject to ordinary one-way ANOVA or two-way ANOVA for statistical analysis as 8 described in figure legends. All primers were ordered from Sigma Aldrich and sequences can be found in Supplementary Table 1.

Cell cycle and caspase 3/7 cleavage flow cytometry assays
For cell cycle analysis, treated cells were harvested on ice by collecting the media and scraping the adhered cells before centrifuging at 500 g for 5 min at 4 o C to pellet both the adherent and non-adherent cells. Cells were mixed with Krishan's reagent (50 µg/mL propidium iodide, 1 mg/mL sodium citrate, 0.3% (v/v) IGEPAL CA-630 and 20 µg/mL ribonuclease A) and analyzed (1:10000 dilution, T4026, Sigma Aldrich) as primary antibodies and species specific IRDye as secondary antibodies (1:5000 dilution, LI-COR biosciences Lincoln NE). Immunoblots were visualized using a LI-COR Biosciences Odessey® Fc Imager.

Intracellular tubulin polymerization assay
Assessment of intracellular tubulin polymerization in non-adherent THP-1 cells in response to MTAs was done as described previously (Giannakakou et al., 1997). In short, THP-1 cells were washed with 37 o C PBS and lysed at 37 o C for 5 min in a hypotonic lysis buffer (1 mM MgCl 2 , 2 mM EGTA, 0.5% (v/v) Nonident P-40 (Shell), 50 mM Tris-HCL pH 6.8 and 1x protease This article has not been copyedited and formatted. The final version may differ from this version. inhibitor cocktail). The lysate was then centrifuged at 14,400 g for 10 min at 37 o C to separate soluble tubulin heterodimers (supernatant) from polymerized microtubules (pellet). The pellet was resuspended in an equal volume to the supernatant and 10% of each fraction was subjected to SDS-PAGE and immunoblotting using β-tubulin (1:2000 dilution, T4026, Sigma Aldrich) as the primary antibody and anti-mouse IRDye as the secondary antibody (1:5000 dilution, LI-COR biosciences). Immunoblots were visualized using a LI-COR Biosciences Odessey® Fc Imager.

Sub-cellular fractionation and detection of mitochondrial DNA
Detection of mitochondrial DNA (mtDNA) from cytoplasmic and organelle enriched fractions was performed as described previously (Wolf et al., 2016). Briefly, cells were scraped into 600 µl of cold disruption buffer (20 mM HEPES, 10 mM KCl, 1.5 mM MgCl 2 , 1 mM EDTA, and 250 mM sucrose) containing 1x protease inhibitor cocktail (Sigma Aldrich) and placed on ice for 5 min. Samples were passed through a 22-gauge syringe 30 times followed by a low spin to remove unbroken cells and subsequently subjected to centrifugation at 10,000 g for 10 min at 4 o C. The pellet was designated the organelle enriched fraction (containing nucleus and mitochondria) and the supernatant the cleared cytoplasmic fraction as confirmed by the relative enrichment of VDAC (1:1000 dilution, 4866, Cell signaling Technologies), Lamin B1 (1:1000 dilution, ab16048, Abcam) and GAPDH (1:1000 dilution, D4C6R, Cell signaling Technologies) by immunoblotting. DNA was isolated from each fraction using a QiaAmp kit (Qiagen, Germantown, MD) and evaluated for genomic or mitochondrial DNA by qRT-PCR. Cell number was used for normalization between samples and equal percentages of cytoplasmic and organelle enriched fractions were compared for each sample to evaluate relative distribution. The primers used to assess mitochondrial genes were ordered from Sigma Aldrich and are in Supplementary   Table 1.

Data analysis
Data are presented with individual data points representing biological replicates and errors are shown as a range of these values. Sample sizes were determined in advance based on previous measurements of accuracy and precision for each of the methodologies used in the current study.
These parameters were defined by utilizing validated and literature grounded positive controls for each assay conducted. All statistical analysis was performed on biological replicates with technical replicates performed as an internal check for the rigor of each biological replicate. The statistical analyses presented in this manuscript were conducted using Graphpad Prism and described in the respective figure legends. In summary, we utilized a one-way ANOVA with Dunnett's posthoc test compared to the control condition if only one experimental variable was being measured or two-way ANOVA with Tukey's posthoc test to compare among each This article has not been copyedited and formatted. The final version may differ from this version. condition if two experimental variables were being measured. If only two samples were being compared, we performed an unpaired t-test analysis.

Eribulin stimulates the expression of IFNβ independent of mitotic arrest or cytotoxicity
We first determined the time-dependent effects of MTAs on non-adherent THP-1 human monocytic leukemia cells by quantifying the distribution of tubulin in the polymerized microtubule form (P) and soluble tubulin heterodimers (S) by non-denaturing lysis and centrifugation (Giannakakou et al., 1997). When THP-1 cells were treated with 100 nM eribulin or paclitaxel, clinically relevant concentrations of these drugs that promote a rapid microtubule disruption in breast cancer cells (Dybdal-Hargreaves et al., 2018;Goel et al., 2009;Huizing et al., 1995), microtubule depolymerization by eribulin and stabilization by paclitaxel was observed within 2 h (Fig. 1A). Neither G 2 /M accumulation nor caspase cleavage were observed in MTA treated cells at either 2 or 6 h ( Fig. 1B, C). However, 24 h after either 100 nM eribulin or paclitaxel treatment to THP-1 cells, there was a significant increase in G 2 /M accumulation and approximately 50% of cells had initiated the process of apoptosis as demonstrated by the activation of caspases 3/7 (Fig. 1B, C).
At 24 h, we also evaluated the effects of both drugs on cytokine expression and found that eribulin was distinct from paclitaxel, particularly in its induction of IFNβ and ISGs in THP-1 cells, suggesting mechanistic differences between these MTAs that cannot be strictly attributed to their shared anti-mitotic and apoptotic effects at this time point (Fig. S1A). A more detailed time course demonstrated that eribulin indeed promoted the induction of IFNβ in THP-1 cells within 6 h (Fig. 1D), a time when interphase microtubules were disrupted, but prior to mitotic arrest or caspase cleavage. Low expression of IFNβ was observed with paclitaxel treatment at each of the time points evaluated with no significant difference from the vehicle control (Fig.   1D). Intracellular flow cytometry further demonstrated an eribulin-mediated increase in IFNβ protein 6 h after drug treatment in live cells using a dual gating strategy (Fig. 1E). Together, these data demonstrate that eribulin can induce the expression of IFNβ at both the mRNA and protein levels in THP-1 cells prior to mitotic accumulation and apoptosis, through a mechanism distinct from paclitaxel.
To determine the downstream functionality of eribulin-induced IFNβ expression we further evaluated the concomitant expression of the interferon stimulated gene (ISG) IFIT1, over time after the addition of eribulin to THP-1 cells and found that expression also increased significantly 24 h after treatment (Fig. S1B). Further analysis revealed a similar eribulinmediated induction of another ISG, OAS1, with a more modest induction of the ISGs IFITm3, and ISG15, which were not observed in paclitaxel treated cells (Fig. S1C). Although ISG induction coincided with the mitotic arrest and apoptotic effects of eribulin, it is important to note that these shared phenotypes were not sufficient to promote ISG induction in paclitaxel treated cells. To further test if the eribulin-mediated induction of ISGs was driven by canonical interferon signaling, which is mediated by JAK-STAT signaling downstream of the binding of interferon to the interferon receptor (IFNAR), we took both genetic and pharmacological approaches (Fig. S2A). Indeed, inhibition of JAK kinase activity with ruxolitinib or siRNA mediated knockdown of IFNAR was sufficient to completely suppress eribulin-mediated expression of the ISG IFIT1 (Fig. 1F, G), without inhibiting induction of IFNβ expression ( Fig.   S2B-E). Together, these data demonstrate that either pharmacological or genetic inhibition of the canonical interferon receptor signaling pathway was sufficient to decouple the eribulin-This article has not been copyedited and formatted. The final version may differ from this version. Previous studies have paradoxically demonstrated a shared upregulation of IFNβ in response to either microtubule stabilization or destabilization driven by cytoplasmic nuclear genomic DNA (gDNA) originating from micronuclei formed after prolonged mitotic arrest and slippage (Harding et al., 2017;Lohard et al., 2020;Mackenzie et al., 2017;Zierhut et al., 2019).
However, these findings were distinct from the rapid induction of IFNβ by eribulin, but not paclitaxel, seen in our experiments that occurs prior to mitotic arrest or caspase cleavage. This prompted us to further evaluate the eribulin-mediated IFN response in non-cycling primary murine bone marrow-derived macrophages (BMDMs). Similar to the results in the THP-1 human cell line, treating murine BMDM with 100 nM eribulin was sufficient to promote depolymerization of the microtubule network within 2 h ( Fig. 2A). However, the primary noncycling BMDMs were refractory to mitotic accumulation or caspase cleavage by eribulin even 24 h after addition (Fig. 2B, C). Similar to the THP-1 cells, eribulin induced IFNβ expression was observed within 2 h in these non-cycling cells with subsequent IFIT1 expression at 24 h, further demonstrating that eribulin-induced IFNβ production leads to a functional ISG response independent of mitotic arrest or the initiation of apoptosis (Fig. 2D, E). Altogether, these data demonstrate that acute eribulin-dependent induction of IFNβ and downstream ISG signaling in both primary murine and immortalized human immune cells is independent of antimitotic or apoptotic effects, providing a rationale to further explore the mechanistic underpinnings of this induction.

Eribulin stimulation of IFNβ is dependent on the cGAS-STING-TBK1 pathway
This article has not been copyedited and formatted. The final version may differ from this version. In mammals, the transcription of IFNβ is primarily mediated by interferon regulatory factors 3 and 7 (IRF3/7), nuclear factor-κB (NF-κB), and the AP-1 complex (Honda et al., 2006).
To preliminarily investigate the likelihood of involvement of each of these transcription factors in the eribulin-mediated expression of IFNβ, BMDMs were pretreated with pharmacological inhibitors of each of these pathways prior to eribulin treatment at concentrations that did not promote apoptosis within the timeframe evaluated ( Fig. S3A-C). The tank binding kinase 1 (TBK1) inhibitor BX795 was utilized to suppress the IRF3/7 signaling pathway, the IκB kinase inhibitor TPCA-1 was used to inhibit NF-κB nuclear localization, and the c-jun n-terminal kinase (JNK) inhibitor SP600125 was used to inhibit transcription mediated by the AP-1 complex (Bennett et al., 2001;Clark et al., 2009;Podolin et al., 2005). We found that pretreatment of BMDM with the TBK1 inhibitor completely abrogated eribulin-mediated IFNβ and IFIT1 induction and that this response was also partially inhibited by NF-κB, but not AP-1, inhibition ( Fig. 3A, B).
TBK1 plays a pivotal role in activating the transcriptional expression of IFNβ during the process of intracellular DNA sensing by the cGAS-STING pathway (Fig. S4A) (Shang et al., 2019). Therefore, we sought to determine if the cGAS-STING pathway was involved in the eribulin-mediated stimulation of IFNβ. To test this, we pretreated BMDM and THP-1 cells with a covalent STING inhibitor, H-151 (Haag et al., 2018), prior to eribulin treatment using a concentration that inhibited TBK1 phosphorylation (Fig. S4B) but was not toxic to BMDM or THP-1 cells (Fig. S4C, D). Inhibition of STING signaling by H-151 was sufficient to attenuate the induction of IFNβ and IFIT1 by either eribulin (Fig. 3C-F (Sauer et al., 2011). Consistent with the results observed with the STING inhibitor, Sting gt/gt derived BMDMs failed to upregulate the expression of IFNβ and concomitant IFIT1 when treated with eribulin (Fig. 3G, H) or exogenous HT-DNA (Fig. S4I, J). Taken together, these studies demonstrate that the eribulin-mediated induction of IFNβ and concomitant IFIT1 expression occurs through a mechanism dependent on the cGAS-STING-TBK1 pathway.
To further interrogate the importance of an intact DNA sensing pathway for eribulinmediated expression of IFNβ, we capitalized on the previous finding that CAL-51 TNBC cells are deficient in cytoplasmic DNA sensing due to a lack of cGAS, which is required for the synthesis of the second messenger cGAMP that functions as a STING agonist (Fig. S4A) (Zierhut et al 2019). Immunoblot analysis of a molecularly diverse panel of TNBC cell lines confirmed that CAL-51 cells express STING, but do not express cGAS (Fig. 4A). As expected, each of the cGAS expressing TNBC cell lines evaluated were capable of inducing IFNβ expression when challenged with exogenous HT-DNA, but cGAS deficient CAL-51 cells were deficient in this response (Fig. 4B). Importantly, ectopic expression of RFP-cGAS (CAL-51-RFP-cGAS) (Fig. 4A) restored the ability of CAL-51 cells to respond to HT-DNA stimulation without introducing gross cytotoxicity ( Fig. 4B and Fig. S5A). Consistent with the role of the cGAS-STING pathway in the eribulin-mediated induction of IFNβ in immune cells, we found that eribulin also promoted the expression of IFNβ in cGAS and STING expressing TNBC cell lines as well as the CAL-51 line reconstituted with RFP-cGAS, but was unable to promote IFNβ expression in cGAS deficient CAL-51 cells or low cGAS expressing BT-549 cells (Fig. 4C, D).
Consistent with the lack of eribulin-mediated IFNβ induction, cGAS deficient CAL-51 cells also failed to upregulate IFIT1 when treated with eribulin, although the expression of this ISG was rescued by expression of RFP-cGAS (Fig. 4E). Using flow cytometry, we further demonstrated that live HCC1937 TNBC cells produced IFNβ protein, which was able to promote downstream IFIT1 expression within 6 h of treatment (Fig. 4F, G). The finding that IFNβ expression occurs in these TNBC cells within 6 h, before mitotic accumulation or cytotoxicity are observed ( Fig.   S5B-D), further demonstrates the independence of eribulin-mediated interferon expression and downstream ISG expression from its antimitotic mechanism of action. These results demonstrate that eribulin-mediated induction of IFNβ and downstream ISGs in immune cells as well as in a molecularly diverse panel of TNBC cells occurs in a cGAS-STING dependent manner.

Eribulin activates the cGAS-STING pathway by inducing the accumulation of cytoplasmic mitochondrial DNA
The dependence of cells on the cytoplasmic cGAS-STING pathway for the induction of IFNβ in response to eribulin suggests that eribulin could be promoting the accumulation of DNA in the cytoplasm. Since mammalian genomic DNA (gDNA) and mitochondrial DNA (mtDNA) serve as the major endogenous cGAS-STING DNA ligands (Härtlova et al., 2015;West et al., 2015) we hypothesized that gDNA and/or mtDNA sequences were enriched in the cytoplasm of cells after treatment with eribulin. To test this hypothesis, we performed differential centrifugation in HCC1937 cells to generate an organelle-enriched (OE) fraction that contained mitochondria and nuclei as evidenced by the enrichment in VDAC and lamin B1, respectively, as well as a cytoplasmic fraction that was enriched for GAPDH (Fig. S6A). We then quantified distinct mtDNA and gDNA sequences (6 each) that were enriched in the cytoplasmic fraction of HCC1937 cells after a 6 h treatment with eribulin by qRT-PCR and found a significant increase in the enrichment of mtDNA sequences in the cytoplasm (Fig. 5A). Furthermore, we found that the cytoplasmic enrichment of the COX-1 mtDNA sequence specifically occurred in eribulin treated HCC1937 and BMDM cells as compared to vehicle or paclitaxel treatment (Fig. 5B, C).
These results suggest that the release of mtDNA into the cytoplasmic space could initiate eribulin-mediated induction of the cGAS-STING DNA sensing pathway.
We hypothesized that microtubule destabilization could lead to the release of mitochondrial DNA into the cytoplasm because of the critical role microtubules play in maintaining mitochondrial homeostasis. Intriguingly, microtubule stabilizing and destabilizing agents have previously been shown to elicit differential effects on mitochondrial biogenesis that increase or reduce mitochondrial mass, respectively (Karbowski et al., 2001) and eribulin can effectively disrupt mitochondrial membrane potential within 30 min of treatment (Sampson et al., 2016). Therefore, we speculated that eribulin could disrupt mitochondrial homeostasis in a manner that could lead to the cytosolic accumulation of mtDNA. We thus examined the morphology of both microtubules and mitochondria in HCC1937 cells treated for 4 h with eribulin or paclitaxel. These cells were chosen due to their large size and fine resolution of the microtubule and mitochondrial network. As previously shown for other microtubule destabilizers (Okatsu et al., 2010), we found that eribulin-treated cells acquired a somewhat fragmented mitochondrial phenotype, a feature not observed in paclitaxel treated cells (Fig. S6B). Based on these findings, we hypothesized that other drugs that destabilized microtubules could also promote IFNβ and ISG expression. Indeed, when we treated THP-1 cells with increasing concentrations (10 -1000 nM) of the five different MTAs used in the treatment of breast cancer, we found that only the microtubule destabilizers (eribulin and vinorelbine) but not stabilizers (ixabepilone, paclitaxel, and docetaxel) upregulated IFNβ expression (Fig. 5D). Additionally, this IFNβ induction by vinorelbine was associated with downstream IFIT1 expression (Fig. 5E). This article has not been copyedited and formatted. The final version may differ from this version. These data support our hypothesis that destabilization of the interphase microtubule network leads to disruption of mitochondrial homeostasis and subsequent release of mtDNA to drive innate immune activation by these agents.
To directly test the hypothesis of whether the eribulin-mediated release of mtDNA was indeed responsible for the increase in IFNβ expression, we took advantage of the fact that ethidium bromide (EtBr) depletes mtDNA while preserving mitochondria themselves (Armand et al., 2004). This method has been used by others to evaluate mtDNA-mediated activation of the cGAS-STING pathway (Hu et al., 2020;White et al., 2014;Yamazaki et al., 2020). We therefore cultured HCC1937 cells in EtBr for 5 days to generate mtDNA depleted (rho 0 ) HCC1937 cells as noted by the depletion of mitochondrial genes in these cells as compared to the untreated controls (Fig. 5F). Strikingly, when mitochondrial deficient HCC1937 rho 0 cells were treated with eribulin, there was a complete loss of eribulin-mediated IFNβ induction (Fig. 5G) that coincided with the absence of eribulin-induced accumulation of mtDNA in the cytoplasm (Fig.   5H). Importantly, HCC1937 rho 0 cells retained the ability to induce IFNβ expression in response to exogenously added HT-DNA (Fig. S6C), indicating that these cells retain a functional cGAS-STING pathway. Collectively, these studies demonstrate that cellular mitochondria are critical for mediating the eribulin-dependent accumulation of cytoplasmic mtDNA, which promotes innate immune activation via the cGAS-STING pathway to drive expression of IFNβ and ISGs.

DISCUSSION 873/1500
MTAs, including the microtubule stabilizer paclitaxel and the microtubule destabilizer eribulin, are some of the most effective agents used in the treatment of metastatic TNBC (Dumontet and Jordan, 2010). Historically, their clinical success has been attributed to their This article has not been copyedited and formatted. The final version may differ from this version. shared ability to suppress mitosis leading to the apoptosis of rapidly dividing cancer cells.
However, accumulating evidence in both patients and preclinical models demonstrates that the anticancer effects of MTAs cannot solely be explained by their shared antimitotic effects, and that inhibition of interphase microtubule dynamics significantly contributes to their antitumor properties (Bates and Eastman, 2017;Field et al., 2014;Komlodi-Pasztor et al., 2012). Moreover, different MTAs can elicit distinct effects on cellular oncogenic signaling pathways as well as on mitochondrial homeostasis that may underlie unappreciated clinical effects between these drugs (Dybdal-Hargreaves et al., 2018;Karbowski et al., 2001;Kaul et al., 2019a). Herein, we show that the microtubule destabilizer eribulin induces the cGAS-STING pathway-mediated expression of IFNβ in innate immune cells as well as in TNBC cells within 2 -6 h by promoting the accumulation of cytoplasmic mtDNA.
One of our most surprising and important findings is that eribulin is distinct from paclitaxel in its ability to promote cytoplasmic accumulation of mtDNA leading to cGAS-STING-dependent interferon and downstream ISG induction. Our results demonstrate that the activation of this immune pathway occurs within 2 -6 h of drug addition and is not dependent on mitotic arrest or the initiation of apoptosis. This is different from studies demonstrating that both microtubule stabilizers and destabilizers can promote activation of the cGAS-STING pathway specifically during extended periods of mitotic arrest when genomic DNA is released into the cytoplasm (Mackenzie et al., 2017;Zierhut et al., 2019). Importantly, this previously reported genomic DNA release was dependent on mitosis and not observed in non-cycling cells. Our current finding that eribulin rapidly and specifically activates cGAS-STING through mtDNA release in both TNBC and immune cells including those that are terminally differentiated is significant particularly in solid tumors that have a much lower mitotic index than cancer cells in These findings are timely as they may contribute to a mechanistic understanding behind the efficacy of combinations of molecularly distinct classes of MTAs, including eribulin, with immune checkpoint inhibition in TNBC patients (Schmid et al., 2020;Tolaney et al., 2021). The cGAS-STING pathway has been shown to be indispensable for immune checkpoint inhibitors to exert their antitumor effects (Wang et al., 2017) and agents that activate the cGAS-STING pathway have been shown to enhance the efficacy of immune checkpoint therapy in metastatic breast cancer models primarily by priming the immune system to acquire an antitumor phenotype (Chandra et al., 2014;Cheng et al., 2018). These findings have prompted the discovery and development of pharmacological STING agonists; however, our data demonstrate that eribulin and likely a subset of currently approved microtubule destabilizers that innately possess this activity are already in clinical use. In agreement with our findings that another microtubule stabilizer, vinorelbine, promotes innate immune activation, preclinical studies have also demonstrated an increased response to immune checkpoint inhibitors when combined with vinorelbine (Orecchioni et al., 2018). We speculate that the mitotic-independent activation of the cGAS-STING pathway by eribulin and other microtubule destabilizers could provide an advantage over stabilizers in activating this antitumor associated immune signaling pathway.
While our studies demonstrate that the interphase effects of eribulin disrupt mitochondrial localization and promote the release of mtDNA into the cytoplasm and that mtDNA is required for eribulin-mediated activation of the cGAS-STING pathway, the mechanism by which eribulin promotes these mitochondrial effects is implicated from previous studies. The cytoplasmic release of mtDNA under cellular stress can be mediated by the formation of macropores in the mitochondrial outer membrane either through oligomerization of This article has not been copyedited and formatted. The final version may differ from this version. the Bcl-2 associated proteins BAX and BAK (Galluzzi and Vanpouille-Box, 2018) or the formation of voltage-dependent anion channels (VDACs) oligomers (Kim et al., 2019). Previous studies have demonstrated that increases in cytoplasmic free tubulin heterodimers, similar to those we observe within 2 h of eribulin treatment (Figs. 1A, 2A), can directly interact with VDAC channels and disrupt mitochondrial membrane potential (Carre et al., 2002;Maldonado et al., 2010;Rovini, 2019). Additionally, other microtubule destabilizers, including the vinca alkaloids and combretastatin A-4, induce the expression of NOXA, which promotes the release and activation of BAX and BAK to form mitochondrial pores in non-cycling cells within the same timeframe where we observe an increase of cytoplasmic mtDNA with eribulin (Bates et al., 2013). Therefore, there is a strong precedence and rationale for how microtubule destabilization can alter mitochondrial permeability through multiple pathways to lead to the innate immune signaling observed in the current study. A rigorous determination of the relative impact of these mechanisms on the release of mtDNA by eribulin will require further investigation.
Together, this work sets the stage for investigation of other innate immune pathways that are specifically modulated by MTAs downstream of their effects on microtubule dynamics and structure. It also specifically prompts interrogation of the adaptive immunological events downstream of the activation of these innate immune sensing pathways, including whether activation of the cGAS-STING pathway within the tumor microenvironment can serve as a predictive biomarker of response to eribulin, particularly in combination with immunotherapy in TNBC.

ACKNOWLEDGMENTS
We thank Dr. Bruce A. Littlefield, Dr. Taro Semba, and members of the Risinger and Mooberry laboratories for their valuable comments and suggestions. We are grateful to Dr. Juli Bai and Dr.
Feng Liu for providing us with Sting gt/gt mice. We also thank Dr. Nan Yan for kindly sharing the THP-1 cells. We are very appreciative of Dr. Wouter Koek for his assistance in the statistical analysis of our data. Flow Cytometry data was generated in the University of Texas Health Significance was determined by 2-way ANOVA (drug * inhibitor/siRNA) with Tukeys's posthoc test. Quantitative RT-PCR data are shown as individual points from two independent biological replicates with error bars denoting range. *p < 0.05, **p < 0.01, ****p < 0.0001. wild type BMDM cells treated with 100 nM eribulin for 2, 6, or 24 h compared to DMSO as percentage of live (white) or apoptotic (grey) cells from two independent experiments with errors denoting range. IFNβ (D) and IFIT1 (E) mRNA in BMDM cells treated with 100 nM eribulin (ERB) for 2, 6, or 24 h compared to DMSO. Quantitative RT-PCR data are shown as individual points from two independent biological replicates with error bars denoting range. Significance was determined by vehicle compared 1-way ANOVA with Dunnett's posthoc test. **p < 0.01, ****p < 0.0001. Data are shown as individual points from two independent biological replicates with error bars denoting range. Significance was determined by 2-way ANOVA (time * inhibitor) using a Tukey's posthoc test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. test. Data are shown as individual points from two independent biological replicates with error bars denoting range. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Significance was determined by 2-way ANOVA (Rho status * drug) with Tukey's posthoc test.
(G) IFNβ mRNA in control and Rho 0 HCC1937 cells treated with 100 nM eribulin for 2 or 6 h as compared to DMSO. Significance was determined by 2-way ANOVA (Rho status * drug) with Tukey's posthoc test. (H) Cytoplasmic mtCOX-1 DNA present in control and Rho 0 HCC1937 cells treated with 100 nM eribulin or paclitaxel for 6 h as compared to DMSO. Significance was determined by 2-way ANOVA (Rho status * drug) with Tukey's posthoc test. Data are shown as individual points from two independent biological replicates with error bars denoting range. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.