Eribulin Activates the cGAS-STING Pathway via the Cytoplasmic Accumulation of Mitochondrial DNA

Microtubule-targeting agents (MTAs), including both microtubule stabilizers and destabilizers are highly effective chemotherapeutic drugs used in the treatment of solid tumors and hematologic malignancies. In addition to the shared ability of all MTAs to block cell cycle progression, growing evidence shows that different agents of this class can also have mechanistically distinct effects on nonmitotic microtubule-dependent cellular processes, including cellular signaling and transport. Herein, we test the biologic hypothesis that MTAs used in the treatment of triple-negative breast cancer (TNBC) can differentially affect innate immune signaling pathways independent of their antimitotic effects. Our data demonstrate that the microtubule destabilizer eribulin, but not the microtubule stabilizer paclitaxel, induces cGAS-STING–dependent expression of interferon-β in both myeloid and TNBC cells. Activation of the cGAS-STING pathway by eribulin was further found to be mediated by the accumulation of cytoplasmic mitochondrial DNA. Together, these findings provide mechanistic insight into how eribulin can induce innate immune signaling independent of its antimitotic or cytotoxic effects. SIGNIFICANCE STATEMENT Microtubule-targeting agents (MTAs) are often used in the treatment of breast cancer and have been used in combination with immune checkpoint inhibitors to improve efficacy. Although all clinically approved MTAs share an antimitotic mechanism of action, their distinct effects on interphase microtubules can promote differential downstream signaling consequences. This work shows 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
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). Although MTAs have long been thought to exert their anticancer effects primarily through inhibition of mitosis, growing evidence demonstrates that their nonmitotic effects contribute significantly to their anticancer efficacy (Komlodi-Pasztor et al., 2011, 2012Field et al., 2014;Kaul et al., 2019b). Mechanistically distinct MTAs can also differ from one another in their nonmitotic effects on oncogenic signaling pathways, and such differences could be important factors in their clinical efficacy (Karbowski et al., 2001;Dybdal-Hargreaves et al., 2017;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 classspecific immunomodulatory effects of MTAs, such as the ability of paclitaxel to directly activate toll-like receptor 4-dependent signaling (Rajput et al., 2013;Wanderley et al., 2018), have been extensively explored, many of the underlying mechanisms of how MTAs elicit distinct immunologic 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 biologic 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 (IFN) b 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.
siRNA Transfection, RNA Extraction, and qRT-PCR. Treatment with immune pathway inhibitors was initiated 4 hours before treatment with eribulin or HT-DNA followed by continuous treatment with the inhibitor for the indicated time. For IFNAR knockdown experiments, THP-1 cells were transfected for 48 hours with siRNAs (Sigma Aldrich) against IFNAR1 (SASI_Hs01_00121376, SASI_Hs02_ 00302973) and IFNAR2 (SASI_Hs01_00208506, SASI_Hs01_002 08505) and compared with scramble control (SIC002) using the RNAi-MAX 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 ÀDDCt method, in which GAPDH was used as the control. All treated and stimulated samples were compared with 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 biologic replicates. The range of the biologic replicate values is represented in every bar graph. Data were subject to ordinary one-way ANOVA or two-way ANOVA for statistical analysis as described in figure legends. All primers were ordered from Sigma Aldrich, and sequences can be found in Supplemental 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 500g for 5 minutes at 4 C to pellet both the adherent and nonadherent cells. Cells were mixed with Krishan's reagent [50 mg/ml propidium iodide, 1 mg/ml sodium citrate, 0.3% (v/v) IGEPAL CA-630, and 20 mg/ml ribonuclease A] and analyzed using a Guava Muse Cell Analyzer (Luminex, Austin, TX) and expressed as percentage of cells in G 1 , S, and G 2 /M based on their DNA content, where error bars represent the range among biologic duplicates. For caspase 3/7 cleavage, adherent treated cells were incubated at 37 C for 30 minutes in a 50-ml/ml Caspase 3/7 green reagent (Thermo-Fisher, Waltham, MA) solution. Cells were subsequently rinsed with PBS and analyzed using a Guava Muse Cell Analyzer (Luminex) and expressed as percentage of cells alive and apoptotic, where error bars represent the range in measured values in biologic duplicates.
Interferon-b Flow Cytometry Assay. THP-1 or HCC1937 cells were treated with 1 ml/ml GolgiPlug (555029; BD Biosciences, Franklin Lakes, NJ) with 100 nM eribulin or the DMSO vehicle control for 2-6 hours. Cells were then collected, blocked with TruStain FcX block (BioLegend, San Diego, CA), and stained with Zombie NIR viability dye (BioLegend) for 15 minutes in the dark. After fixation and permeabilization, cells were incubated with anti-IFNb-fluorescein isothiocyanate(FITC) (PBL Assay Science, Piscataway, NJ) for 20 minutes in the dark. Intracellular IFNb staining in live cells was determined using the Cytek Aurora at the University of Texas Health Science Center San Antonio Flow Cytometry Facility and analyzed with FlowJo software (Becton, Dickinson and Company, Ashland, OR).
Microtubule Indirect Immunofluorescence and Mitochondrial Staining. BMDM cells grown on coverslips were fixed and permeabilized using cold 100% methanol and immunostained using anti-b-tubulin (1:400 dilution, T4026; Sigma Aldrich) as primary and anti-mouse FITC (1:200 dilution, F-3008; Sigma Aldrich) as a secondary antibody. HCC1937 cells were stained with 50 nM Mitotracker Red CMXRos (Thermo-Fisher) in serum-free RPMI 1640 for 30 minutes at 37 C, fixed with 4% paraformaldehyde, and permeabilized with 0.5% (v/v) Triton X-100 prior to immunostaining with anti-b-tubulin. Samples were imaged on a Nikon (Tokyo, Japan) Eclipse 80i fluorescence microscope with NIS elements using z-stacks at 1000Â total magnification. Subcellular Fractionation and Detection of Mitochondrial DNA. Detection of mtDNA from cytoplasmic and organelle-enriched fractions was performed as described previously (Wolf et al., 2016). Briefly, cells were scraped into 600 ml of cold disruption buffer (20 mM HEPES, 10 mM KCl, 1.5 mM MgCl 2 , 1 mM EDTA, and 250 mM sucrose) containing 1Â protease inhibitor cocktail (Sigma Aldrich) and placed on ice for 5 minutes. Samples were passed through a 22gauge syringe 30 times followed by a low spin to remove unbroken cells and subsequently subjected to centrifugation at 10,000g for 10 minutes at 4 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 Supplemental Table 1.
Data Analysis. Data are presented with individual data points representing biologic 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 biologic replicates, with technical replicates performed as an internal check for the rigor of each biologic replicate. The statistical analyses presented in this manuscript were conducted using GraphPad Prism and described in the respective figure legends. In summary, we used a one-way ANOVA with Dunnett's post hoc test compared with the control condition if only one experimental variable was being measured or two-way ANOVA with Tukey's post hoc test to compare among each condition if two experimental variables were being measured. If only two samples were being compared, we performed an unpaired t test analysis.

Results
Eribulin Stimulates the Expression of IFNb Independent of Mitotic Arrest or Cytotoxicity. We first determined the time-dependent effects of MTAs on nonadherent THP-1 human monocytic leukemia cells by quantifying the distribution of tubulin in the polymerized microtubule form and soluble tubulin heterodimers by nondenaturing 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 (Huizing et al., 1995;Goel et al., 2009;Dybdal-Hargreaves et al., 2017), microtubule depolymerization by eribulin and stabilization by paclitaxel were observed within 2 hours (Fig. 1A). Neither G 2 /M accumulation nor caspase cleavage were observed in MTA-treated cells at either 2 or 6 hours ( Fig. 1, B and C). However, 24 hours 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. 1, B and C).
At 24 hours, we also evaluated the effects of both drugs on cytokine expression and found that eribulin was distinct from paclitaxel, particularly in its induction of IFNb and ISGs in THP-1 cells, suggesting mechanistic differences between these MTAs that cannot be strictly attributed to their shared antimitotic and apoptotic effects at this time point (Supplemental Fig. 1A). A more detailed time course demonstrated that eribulin indeed promoted the induction of IFNb in THP-1 cells within 6 hours (Fig. 1D), a time when interphase microtubules were disrupted, but prior to mitotic arrest or caspase cleavage. Low expression of IFNb 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 IFNb protein 6 hours after drug treatment in live cells using a dual gating strategy (Fig. 1E). Together, these data demonstrate that eribulin can induce the expression of IFNb 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 eribulininduced IFNb expression, we further evaluated the concomitant expression of the ISG IFIT1 over time after the addition of eribulin to THP-1 cells and found that expression also increased significantly 24 hours after treatment (Supplemental Fig. 1B). Further analysis revealed a similar eribulin-mediated induction of another ISG, OAS1, with a more modest induction of the ISGs IFITm3 and ISG15, which were not observed in paclitaxel-treated cells (Supplemental Fig. 1C). 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 whether the eribulinmediated 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 (Supplemental Fig. 2A). Indeed, inhibition of JAK kinase activity with ruxolitinib or siRNA-mediated knockdown of IFNAR Eribulin Activates cGAS-STING via mtDNA Accumulation was sufficient to completely suppress eribulin-mediated expression of the ISG IFIT1 (Fig. 1, F and G) without inhibiting induction of IFNb expression (Supplemental Fig. 2, B-E). Together, these data demonstrate that either pharmacological or genetic inhibition of the canonical interferon receptor signaling pathway was sufficient to decouple the eribulin-mediated induction of IFNb from downstream ISG induction, further supporting that eribulin induces a functional interferon response.
Previous studies have paradoxically demonstrated a shared upregulation of IFNb 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;Mackenzie et al., 2017;Zierhut et al., 2019;Lohard et al., 2020). However, these findings were distinct from the rapid induction of IFNb 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 noncycling primary murine 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 hours ( Fig. 2A). However, the primary noncycling BMDMs were refractory to mitotic accumulation or caspase cleavage by eribulin even 24 hours after addition (Fig. 2, B and C). Similar to the THP-1 cells, eribulin-induced IFNb expression was observed within 2 hours in these noncycling cells with subsequent IFIT1 expression at 24 hours, further demonstrating that eribulin-induced IFNb production leads to a functional ISG response independent of mitotic arrest or the initiation of apoptosis (Fig. 2, D and E). Altogether, these data demonstrate that acute eribulin-dependent induction of IFNb 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 IFNb Is Dependent on the cGAS-STING-TBK1 Pathway. In mammals, the transcription of IFNb is primarily mediated by interferon regulatory or vehicle for 4 hours and then treated with DMSO or 100 nM eribulin for 24 hours with the inhibitor still present or (G) with siRNA to IFNAR1/2 (siIFNAR1/2) or a scrambled sequence (siCtrl) for 48 hours followed by treatment with DMSO or 100 nM eribulin for 24 hours. Significance was determined by two-way ANOVA (drug * inhibitor/siRNA) with Tukeys's post hoc test. qRT-PCR data are shown as individual points from two independent biologic replicates with error bars denoting range. *P < 0.05, **P < 0.01, ****P < 0.0001. Veh, vehicle.
factors 3 and 7, NF-jB, and the AP-1 complex (Honda et al., 2006). To preliminarily investigate the likelihood of involvement of each of these transcription factors in the eribulinmediated expression of IFNb, 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 (Supplemental Fig. 3, A-C). The TBK1 inhibitor BX795 was used to suppress the interferon regulatory factors 3 and 7 signaling pathway, the IjB kinase inhibitor TPCA-1 was used to inhibit NF-jB nuclear localization, and the c-Jun N-terminal kinase inhibitor SP600125 was used to inhibit transcription mediated by the AP-1 complex (Bennett et al., 2001;Podolin et al., 2005;Clark et al., 2009). We found that pretreatment of BMDM with the TBK1 inhibitor completely abrogated eribulin-mediated IFNb and IFIT1 induction and that this response was also partially inhibited by NF-jB, but not AP-1, inhibition (Fig. 3, A and B).
TBK1 plays a pivotal role in activating the transcriptional expression of IFNb during the process of intracellular DNA sensing by the cGAS-STING pathway (Supplemental Fig.  4A) (Shang et al., 2019). Therefore, we sought to determine whether the cGAS-STING pathway was involved in the eribulin-mediated stimulation of IFNb. 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 (Supplemental Fig. 4B) but was not toxic to BMDM or THP-1 cells (Supplemental Fig. 4, C and D). Inhibition of STING signaling by H-151 was sufficient to attenuate the induction of IFNb and IFIT1 by either eribulin (Fig. 3, C-F) or exogenous HT-DNA (Supplemental Fig. 4, E-H) in both primary BMDMs and in THP-1 cells. To further examine the involvement of STING in mediating eribulin-induced expression of IFNb, we generated BMDMs from Sting gt/gt mice that contain a missense mutant allele that renders them defective for STING-induced INFb expression (Sauer et al., 2011). Consistent with the results observed with the STING inhibitor, Sting gt/gt derived BMDMs failed to upregulate the expression of IFNb and concomitant IFIT1 when treated with eribulin (Fig. 3, G and H) or exogenous HT-DNA (Supplemental Fig. 4, I and J). Taken together, these studies demonstrate that the eribulin-mediated induction of IFNb 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 eribulin-mediated expression of IFNb, we capitalized on the previous finding that CAL-51 TNBC cells are deficient in cytoplasmic DNA sensing because of a lack of cGAS, which is required for the synthesis of the second messenger cyclic GMP-AMP that functions as a STING agonist (Supplemental Fig. 4A) (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 IFNb 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; Supplemental Fig.  5A). Consistent with the role of the cGAS-STING pathway in the eribulin-mediated induction of IFNb in immune cells, we found that eribulin also promoted the expression of IFNb in cGAS-and STING-expressing TNBC cell lines, as well as the CAL-51 line reconstituted with RFP-cGAS, but was unable to promote IFNb expression in cGAS-deficient CAL-51 cells or low-cGAS-expressing BT-549 cells (Fig. 4, C and D). Consistent with the lack of eribulin-mediated IFNb induction, cGAS-deficient CAL-51 cells also failed to upregulate IFIT1 when treated with eribulin, although the expression of this Fig. 2. Eribulin-mediated interferon-b expression occurs independently of mitotic arrest and caspase cleavage. (A) Immunofluorescence images of microtubules (green) and DNA (blue) in BMDM cells treated with 100 nM eribulin (ERB) for 2, 6, or 24 hours as compared with DMSO (Veh). Scale bar, 10 mm. (B) Cell cycle analysis of BMDM cells treated with 100 nM eribulin for 2, 6, or 24 hours as compared with DMSO as percentage of cells in G 1 (black), S (gray), or G 2 /M (white) from two independent experiments, with error bars denoting range. (C) Caspase 3/7 cleavage in wild-type BMDM cells treated with 100 nM eribulin for 2, 6, or 24 hours compared with DMSO as percentage of live (white) or apoptotic (gray) cells from two independent experiments with errors denoting range. IFNb (D) and IFIT1 (E) mRNA in BMDM cells treated with 100 nM ERB for 2, 6, or 24 hours compared with DMSO. qRT-PCR data are shown as individual points from two independent biologic replicates with error bars denoting range. Significance was determined by vehicle-compared oneway ANOVA with Dunnett's post hoc test. **P < 0.01, ****P < 0.0001. Veh, vehicle.
Eribulin Activates cGAS-STING via mtDNA Accumulation ISG was rescued by expression of RFP-cGAS (Fig. 4E). Using flow cytometry, we further demonstrated that live HCC1937 TNBC cells produced IFNb protein, which was able to promote downstream IFIT1 expression within 6 hours of treatment (Fig. 4, F and G). The finding that IFNb expression occurs in these TNBC cells within 6 hours, before mitotic accumulation or cytotoxicity are observed (Supplemental Fig. 5, B-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 IFNb 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 IFNb in response to eribulin suggests that eribulin could be promoting the accumulation of DNA in the cytoplasm. Since mammalian gDNA and mtDNA serve as the major endogenous cGAS-STING DNA ligands (H€ artlova 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 organelleenriched (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 (Supplemental Fig. 6A). We then quantified distinct mtDNA and gDNA sequences (six each) that were enriched in the cytoplasmic fraction of HCC1937 cells after a 6-hour 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 with vehicle or paclitaxel treatment (Fig. 5, B and 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 minutes 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 hours with eribulin or paclitaxel. These cells were chosen because of 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 (Supplemental Fig. 6B). Based on these findings, we hypothesized that other drugs that destabilized microtubules could also promote IFNb and ISG expression. Indeed, when we treated THP-1 cells with increasing concentrations (10-1000 nM) of 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 IFNb expression (Fig. 5D). Additionally, this IFNb induction by vinorelbine was associated with downstream IFIT1 expression (Fig. 5E). These data support our hypothesis that destabilization of the interphase microtubule network leads Fig. 3. Eribulin-dependent upregulation of interferon-b and interferon-stimulated genes is dependent on the STING pathway. IFNb (A) and IFIT1 (B) mRNA in BMDM cells pretreated with 1 mM of the indicated inhibitor for 4 hours and then treated with 100 nM eribulin (ERB) for 2, 6, or 24 hours as compared with DMSO with the inhibitor still present from two independent experiments with error bars denoting range. IFNb (C and E) and IFIT1 (D and F) mRNA from BMDM and THP-1 cells pretreated with 1 mM H-151 (STINGi) for 4 hours and then treated with 100 nM eribulin for 2, 6, or 24 hours as compared with DMSO with the inhibitor still present. IFNb (G) and IFIT1 (H) mRNA in wild-type and Sting gt/gt BMDM cells treated with 100 nM eribulin for 2, 6, or 24 hours. Data are shown as individual points from two independent biologic replicates with error bars denoting range. Significance was determined by two-way ANOVA (time * inhibitor) using a Tukey's post hoc test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Veh, vehicle.

314
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 eribulinmediated release of mtDNA was indeed responsible for the increase in IFNb expression, we took advantage of the fact that 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 (White et al., 2014;Yamazaki et al., 2020;Hu et al., 2021). 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 with the untreated controls (Fig. 5F). Strikingly, when mitochondria-deficient HCC1937 rho 0 cells were treated with eribulin, there was a complete loss of eribulin-mediated IFNb 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 IFNb expression in response to exogenously added HT-DNA (Supplemental Fig.  6C), indicating that these cells retain a functional cGAS-ST-ING pathway. Collectively, these studies demonstrate that cellular mitochondria are critical for mediating the eribulindependent accumulation of cytoplasmic mtDNA, which promotes innate immune activation via the cGAS-STING pathway to drive expression of IFNb and ISGs.

Discussion
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 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 (Komlodi-Pasztor et al., 2012;Field et al., 2014;Bates and Eastman, 2017). 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 (Karbowski et al., 2001;  (Veh) or 100 nM ERB for 24 hours. Significance was determined by two-way ANOVA (cGAS * drug) with Tukey's post hoc test. (F) HCC1937 IFNb intracellular protein in live cells treated with DMSO or 100 nM eribulin for 6 hours. Significance determined by an unpaired twotailed t test. (G) Human IFIT1 mRNA in HCC1937 cells treated with 100 nM eribulin for 2, 6, or 24 hours compared with DMSO. Significance determined by vehicle-compared one-way ANOVA with Dunnett's post hoc test. Data are shown as individual points from two independent biologic replicates with error bars denoting range. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Veh, vehicle.
Eribulin Activates cGAS-STING via mtDNA Accumulation Dybdal-Hargreaves et al., 2017;Kaul et al., 2019a). Herein, we show that the microtubule destabilizer eribulin induces the cGAS-STING pathway-mediated expression of IFNb in innate immune cells as well as in TNBC cells within 2-6 hours 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 hours 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 noncycling 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 culture (Komlodi-Pasztor et al., 2011).
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 patients with TNBC (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 Significance determined by vehicle-compared two-way ANOVA (drug * concentration) with Tukey's post hoc test compared with vehicle. (E) IFIT1 mRNA in THP-1 cells treated with 10, 100, or 1000 nM ERB, VNR, or PTX for 24 hours. Significance determined by vehicle-compared two-way ANOVA (drug * concentration) with Tukey's post hoc test compared with vehicle. (F) COX-1 mRNA in control and ethidium bromide cultured (Rho 0 ) HCC1937 cells treated with 100 nM eribulin for 2 or 6 hours as compared with DMSO. Significance was determined by two-way ANOVA (Rho status * drug) with Tukey's post hoc test. (G) IFNb mRNA in control and Rho 0 HCC1937 cells treated with 100 nM eribulin for 2 or 6 hours as compared with DMSO. Significance was determined by two-way ANOVA (Rho status * drug) with Tukey's post hoc test. (H) Cytoplasmic COX-1 DNA present in control and Rho 0 HCC1937 cells treated with 100 nM eribulin or paclitaxel for 6 hours as compared with DMSO. Significance was determined by two-way ANOVA (Rho status * drug) with Tukey's post hoc test. Data are shown as individual points from two independent biologic replicates with error bars denoting range. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Veh, vehicle; ctrl, control.

316
currently approved microtubule destabilizers that innately possess this activity are already in clinical use. Consistent with our findings that the microtubule destabilizer vinorelbine promotes STING activation, preclinical studies have shown 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.
Although 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 the Bcl-2-associated proteins Bax and Bak (Galluzzi and Vanpouille-Box, 2018) or the formation of VDAC oligomers (Kim et al., 2019). Previous studies have demonstrated that increases in cytoplasmic free tubulin heterodimers, similar to those we observe within 2 hours of eribulin treatment (Figs. 1A and 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 noncycling cells within the same timeframe in which we observe an increase of cytoplasmic mtDNA with eribulin (Bates et al., 2013). Therefore, there is a strong precedent 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 immunologic 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.