Anti‐pyroptotic function of TGF‐β is suppressed by a synthetic dsRNA analogue in triple negative breast cancer cells

Development of innovative therapeutic modalities would address an unmet clinical need in the treatment of triple negative breast cancer (TNBC). Activation of retinoic acid‐inducible gene‐I (RIG‐I)‐like receptors (RLRs) such as melanoma differentiation‐associated gene 5 (MDA5) and RIG‐I in cancer cells is suggested to suppress tumor progression by inducing cell death. Transfection of polyI:C, a conventionally used synthetic double‐stranded RNA (dsRNA) analogue that activates RLRs, has been evaluated in clinical trials. However, detailed mechanisms of tumor suppression by RLRs, especially interactions with other signaling pathways, remain elusive. Here, we showed that transfection of polyI:C suppressed transforming growth factor‐β (TGF‐β) signaling in a MDA5‐ and RIG‐I‐dependent manner. We found that suppression of TGF‐β signaling by polyI:C promoted cancer cell death, which was attenuated by forced expression of constitutively active Smad3. More detailed analysis suggested that cell death by polyI:C transfection exhibited characteristics of pyroptosis, which is distinct from apoptosis. Therapeutic efficacy of polyI:C transfection was also demonstrated using a mouse model. These results indicated that intratumor administration of polyI:C and related dsRNA analogues may be promising treatments for TNBC through inhibition of the anti‐pyroptotic function of TGF‐β.


Introduction
Triple negative breast cancer (TNBC) is one of the most aggressive types of cancer and constitutes~16% of invasive breast cancer cases [1,2]. Although conventional cytotoxic chemotherapy was the only option for treating TNBC for many years, molecular target therapies such as poly (ADP-ribose) polymerase (PARP) inhibitors for patients with BRCA mutations and antibodies against programmed death-ligand 1 (PD-L1) or PD-1 for PD-L1-positive patients have been approved for use in recent years [3]. Because PD-L1 expression tends to be higher in patients with TNBC than those in other subtypes of breast cancer, immune checkpoint inhibitors are expected to have higher potential for treatment of TNBC patients [4,5].
Activating innate immune signaling in cancer cells, such as by intratumoral injection of ligands, is currently proposed to be an interesting strategy to combat cancer cells and is expected to be applied concomitantly with immune checkpoint inhibitors [6,7]. The group of receptors involved in activating innate immune signaling are called pattern recognition receptors (PRRs), which include Toll-like receptors (TLRs) and nucleotide-binding oligomerization domain (NOD)-like receptors (NLRs). PRRs recognize pathogenic components to protect hosts from pathogens or molecules released from damaged cells to notify surrounding cells of danger. Among PRRs, activation of retinoic acid-inducible gene-I (RIG-I)-like receptors (RLRs), which include melanoma differentiation-associated gene 5 (MDA5) and RIG-I in cancer cells, is suggested as beneficial strategy in the treatment of TNBC due to the fact that the RIG-I gene is rarely mutated or deleted [8]. RLRs can recognize cytosolic double-stranded RNA (dsRNA) to trigger the formation of a complex with mitochondrial antiviral signaling protein (MAVS). The complex then induces interferon (IFN) by phosphorylation of interferon regulatory factor 3 (IRF3) and informs the surrounding cells of a virus infection [9,10]. RLR agonists are also expected to be applicable to cancer treatment by inducing tumor cell death and increasing the immunogenicity of tumors by upregulating IFNs or proinflammatory cytokines [7,11]. This is supported by findings that some RIG-I and MDA5 agonists show promising effects in mouse tumor models when administered by intratumoral transfection [12]. Intratumoral transfection of these agonists is also being evaluated in clinical trials of aggressive solid tumors, including TNBC.
Transforming growth factor-b (TGF-b) signaling is one of the most important pathways in cancers. Upon TGF-b binding, a heterotetrameric complex of type I and type II TGF-b receptors (TbRI and TbRII) is formed to induce the phosphorylation of downstream components Smad2 and Smad3. Activated Smad2 and Smad3 then form a complex with Smad4, translocate to the nucleus, and trigger dramatic genome-wide changes in transcription [13]. It is important to note that TGF-b is believed to have both tumor-suppressive and tumor-promoting functions [14][15][16]. Antiproliferative effects and the promotion of cell death are tumorsuppressive facets of TGF-b, and cancers often evade TGF-b signaling by mutating TGF-b receptors or Smads [16]. On the other hand, with respect to the tumor-promoting aspect of TGF-b, cancers may acquire significant benefits from this signaling pathway. There are a number of reports that focus on the protumorigenic function of TGF-b using mammary carcinoma cells to investigate the cooperation of mutant RAS proto-oncogenes with TGF-b in the induction of epithelial-mesenchymal transition (EMT) or enhanced cellular invasion, the stabilization of mammalian target of rapamycin signaling to retain the stemness of cells by prolonged TGF-b exposure, and the collaboration with jun B proto-oncogene (JUNB) to enhance cell invasion [17][18][19][20]. Likewise, the promotion of migration, invasion, bone metastasis, and cell survival are characteristics of TGF-b in TNBC [21,22]. It is also reported that TGF-b reduces the expression of major histocompatibility class I and suppresses CD8+ T-cell-mediated killing of other types of cancers [23]. Taken together, there appears to be more potent characteristics of protumorigenic function of TGF-b signaling in TNBC, and suppression of TGF-b signaling is a well-founded strategy to combat TNBC.
Regarding the interaction between RLR signaling and TGF-b signaling, a previous report revealed that IRF3 activated by dsRNA ligand physically interacts with Smad3 to inhibit its binding to TbRI, which leads to the attenuation of Smad3 phosphorylation [24]. However, the effect of attenuating TGF-b signaling by RLR ligands on cancer cells is not well characterized, especially in context of the protumorigenic function of TGF-b signaling. Furthermore, the previous report did not evaluate dsRNA as a therapeutic ligand for cancer [24].
In the present study, we chose the synthetic dsRNA analogue polyI:C as an RLR agonist, which is commonly and classically used experimentally and has been evaluated in clinical trials of aggressive solid tumors [12]. We found that polyI:C suppressed TGF-b signaling in TNBC cells in an MDA5-and RIG-I-dependent manner, which led to attenuation of the prosurvival function of TGF-b signaling. We also explored the characteristics of cell death induced by polyI:C in TNBC cells and revealed a novel role for TGF-b as an anti-pyroptotic factor. These findings provide support for the efficacy of polyI:C as a promising drug for the treatment of TNBC partly mediated by the inhibition of TGF-b signaling.
2.4. RNA extraction and quantitative reverse transcription-polymerase chain reaction (qRT-PCR) Total RNA was extracted using an RNeasy Mini Kit (#74106; QIAGEN, Hilden, Germany) in accordance with the manufacturer's protocol. First-strand cDNAs were synthesized using a PrimeScript II 1st strand cDNA Synthesis Kit (#6210A; Takara Bio). The qRT-PCR was performed using FastStart Universal SYBR Green Master (Rox) (#04913914001; Roche Diagnostics, Basel, Switzerland) with a StepOnePlus Real-Time PCR System (#4376357; Thermo Fisher Scientific). PCR data were normalized to the expression of TBP. Primer sequences are provided in Table S1.

RNA sequence (RNA-seq) analysis
To prepare libraries from the cells for RNA-seq, an RNeasy Mini Kit with RNase-Free DNase Set (QIA-GEN), Dynabeads mRNA DIRECT Micro Purification Kit, and Ion Total RNA-Seq Kit v2 (Thermo Fisher Scientific) were used according to the manufacturers' protocols. cDNA libraries were quantified using an Ion Library TaqMan Quantitation Kit (Thermo Fisher Scientific). Sequencing was performed with an Ion Proton System (Thermo Fisher Scientific) using an Ion PI Hi-Q Chef Kit (Thermo Fisher Scientific) and Ion PI Chip Kit v3 (A26770, Thermo Fisher Scientific). Reads were aligned to the human genome (Genome Reference Consortium Human Build 37 (GRCh37)/hg19) as a reference genome using TopHat2 open-source bioinformatics tool. Fragments per kilobase of exon per million sequence reads (FPKM) values were calculated using the Cuffdiff function of Cufflinks. Raw sequence data have been deposited in the Gene Expression Omnibus (GEO) database (accession number GSE152414).

Gene set enrichment analysis (GSEA)
Mean FPKM values were calculated from duplicate biological samples for each condition. Genes used for GSEA were extracted using the threshold value of ≥ 10 FPKM for any of the conditions. Analysis was performed using default parameters. The chip platform 'Human_NCBI_Entrez_Gene_ID_MSigDB.v7.1.chip' was used as gene symbol data. Gene ontology gene sets 'M_UP' and 'IM_DOWN' were obtained from Lehmann et al. [25].

Gene Ontology analysis
The reads per kilobase of exon per million sequence reads (RPKM) value of each gene in the cancer cell lines was obtained from Broad Institute Cancer Cell Line Encyclopedia (CCLE, 'CCLE_DepMap_18q3_R-NAseq_RPKM_20180718.gct'). Gene ontology of each gene list was analyzed using the analysis tool 'Enrichr' (https://amp.pharm.mssm.edu/Enrichr/) [26,27].

Annexin V and PI staining
After collecting all cells, including both the floating cells and attached cells, the dying or dead cells were detected using an eBioscience Annexin V Apoptosis Detection Kit APC (#88-8007-72; Thermo Fisher Scientific) according to the manufacturer's protocol. Flow cytometry was performed using a Gallios Flow Cytometer (Beckman Coulter, Brea, CA, USA), and the data were analyzed using FLOWJO software (BD, Franklin Lakes, NJ, USA).

Microscopy
To evaluate the cell state after polyI:C transfection and capture images, an IX70 microscope with DP72 microscope digital camera (Olympus, Tokyo, Japan) or a BZ-X710 microscope (Keyence, Osaka, Japan) was used.

Lactate dehydrogenase (LDH) release assay
To determine LDH values (%) after 60 and 72 h of transfection, Hs578T cells were seeded into 96-well plates at a density of 2 9 10 4 cells/well. To determine LDH values (%) after 36 and 48 h of transfection or of mock transfections, Hs578T cells were seeded into 96-well plates at a density of 1 9 10 4 cells/well to prevent the cells from being confluent. The % LDH values were calculated using a CytoTox 96 Non-Radioactive Cytotoxicity Assay (#G1780; Promega) according to the manufacturer's protocol. After 10 min of reaction time, stop solution was added and absorbance at 490 nm was measured using a Model 680 Microplate Reader (Bio-Rad, Hercules, CA, USA).

Measurement of caspase 3/7 activity
Hs578T cells expressing HA or constitutively active Smad3 (caSmad3) were seeded into 96-well plates at a density of 1 9 10 4 cells/well. A Caspase-Glo 3/7 Assay System (#G8090; Promega) was used to measure the activity of caspase 3/7 according to the manufacturer's protocol by using one plate. After 1 h of substrate incubation, luminescent signals were measured using a Mithras LB 940FP Microplate Reader (Berthold Technologies, Bad Wildbad, Germany). Another plate was used to count viable cells using trypan blue staining.

Chromatin immunoprecipitation (ChIP) sequence data of Smad2
Anti-Smad2 ChIP sequence data from Hs578T cells stimulated with activin A were obtained in our previous study (GEO, accession number GSM3301952) [28]. Smad2 binding at the target gene locus was shown using the UCSC Genome Browser on Human Dec. 2013 (GRCh38/hg38) Assembly.

In vivo polyI:C treatment
In vivo experiments were performed according to the policies of the Animal Ethics Committee of the University of Tokyo (approval number: Med.-P16-140). Mice were obtained from Japan SLC Inc. (Shizuoka, Japan). A total of 5 9 10 5 4T1 cells were inoculated into the mammary fat pad of 8-to 10-week-old BALB/c female mice that were anesthetized with isoflurane (#099-06571; FUJIFILM Wako Pure Chemical Corporation, Osaka, Japan). Tumor sizes were measured using a caliper, and tumor volumes (V) were calculated using the formula: V = (length) 9 (width) 2 /2. Intratumoral polyI:C delivery was performed using in vivo jetPEI transfection reagent with glucose solution (#201; Polyplus-transfection, Illkirch, France) in accordance with the manufacturer's protocol with an ionic balance of N/P = 6. A total volume of 100 lL reagent containing polyI:C (20 lg/mouse) or an equivalent amount of distilled water was intratumorally injected into the tumors of mice that were anesthetized with isoflurane. Tumor weight was measured at day 21 or at the day of endpoint [weight loss (> 20% or > 10%/week), mean diameter > 1.2 cm, or general appearance]. Tumor samples were snap-frozen and stored at À80°C until use.

Statistical analysis
Student's t-test was used for comparison between two samples, and the Tukey-Kramer test was used for comparison of multiple samples. b signaling, we first confirmed the activation of RLR signaling in TNBC cells. We chose the widely used TNBC cell line Hs578T in which the TGF-b signaling pathway is retained and the phosphorylation of Smad3 by TGF-b is reported to occur [31]. Strong C-terminal phosphorylation of IRF3 at Ser386, which is critical for its activation [32], was observed following the transfection of 1 lgÁmL À1 of polyI:C in Hs578T cells (Fig. 1A). As previously reported, the amount of total IRF3 was downregulated by activation of PRR signaling [33,34]. We then evaluated the effect of polyI:C on TGF-b signaling and found that polyI:C suppressed the phosphorylation of Smad3 with strong induction of MDA5 and RIG-I, which is a known positive feedback mechanism (Fig. 1B). We also found that increased expression of the known target genes of TGF-b, ZEB1, LRRC15 [35], and PMEPA1 was suppressed by polyI:C (Fig. 1C). Of note, expression of SERPINE1, which is also a well-known target of both TGF-b and inflammatory response, was markedly increased by polyI:C even in the absence of TGF-b stimulation, and SMAD7 is not significantly induced by TGF-b in this experimental setting (Fig. S1A). The above data suggested that polyI:C attenuated TGF-b signaling and the expression of its target genes by suppressing the phosphorylation of Smad3.

Results
We next examined whether the attenuation of TGFb signaling by polyI:C was dependent on RLRs. Gene inactivation of both MDA5 and RIG-I using lentiviral expression of Cas9 and gRNAs targeting MDA5 and RIG-I suppressed the polyI:C-induced attenuation of Smad3 phosphorylation, while polyI:C-induced phosphorylation of IRF3 was strongly reduced by both MDA5 and RIG-I gRNAs (Fig. 1D). In addition, suppression of TGF-b-induced target gene expression by polyI:C was partly attenuated by MDA5 and RIG-I gRNAs (Fig. 1E). The gene inactivation efficiency of MDA5 and RIG-I was also confirmed by the downregulation of IFNB1 expression by MDA5 and RIG-I gRNAs, which is a common target of activated IRF3. We observed better inactivation efficiency using pair #1 gRNAs compared with that using pair #2 gRNAs (Figs 1D and S1B). These results suggested that polyI: C suppressed TGF-b signaling in an MDA5-and RIG-I-dependent manner.
We also evaluated the potential for inhibition of TGF-b signaling by cell-intrinsic activation of RLR signaling. Gene ontology analysis of genes extracted from the CCLE database that negatively correlated with MDA5 expression in TNBC cell lines showed that the only enriched ontology was TGF-b signaling from the dataset 'Panther_2016', while the most positively correlated gene set was 'Toll receptor signaling pathway' as expected (Fig. S2A). Using another ontology dataset, 'WikiPathways_2019_Human', we confirmed that the ontology of TGF-b signaling was significantly enriched in genes inversely correlated with MDA5 expression in TNBC, while ontologies of positively correlated gene sets were associated with interferon signaling (Fig. S2B). We also performed RNAseq analysis of Hs578T cells with or without TGF-b stimulation to determine the global target of TGF-b signaling in these TNBC cells (Table S3). Using the RNA-seq data, GSEA was performed for the genes that inversely correlated with MDA5. We found that many of the genes inversely correlated with MDA5 expression were targets of TGF-b signaling (Fig. S2C). We also studied the molecular subtype of TNBC [25]. Many upregulated genes in the mesenchymal subtype of TNBC, in which activation of TGF-b signaling has been reported, were also upregulated by TGF-b stimulation (Fig. S2D). On the other hand, many downregulated genes in the immunomodulatory subtype that are characterized by the activation of immune cell signaling were also increased by TGF-b stimulation of Hs578T cells (Fig. S2E). These results suggested the possibility that endogenous activation of RLR signaling can also suppress TGF-b signaling.

Attenuation of TGF-b signaling by polyI:C promoted cell death
Because the induction of cell death is one of the marked phenotypes of polyI:C transfection [7,11], we examined whether the attenuation of TGF-b signaling by polyI:C would accelerate cell death in Hs578T cells.
To investigate the biological effect of polyI:C-induced suppression of TGF-b signaling, we established Hs578T cells stably expressing caSmad3 in which the C-terminal phosphorylation site of Smad3 was changed from SSVS to DDVD [30] (Fig. S3A). Exogenous expression of caSmad3 was expected to activate the signaling pathway at a point downstream of signal inhibition by polyI:C. PolyI:C transfection induced the damaged floating cells in the culture, while the stable expression of caSmad3 markedly counteracted the effect of polyI:C and increased the intact cells attached to the plate ( Fig. 2A). We then evaluated polyI:C-induced cell death using flow cytometry and staining with annexin V and propidium iodide (PI). We found that the expression of caSmad3 decreased the proportions of annexin V + /PI + cells, which were remarkably increased by polyI:C transfection (Fig. 2B,C). Of note, stimulation with TGF-b did not halt cell death possibly because the amount of phosphorylated Smad3 was suppressed by polyI:C and not enough for protection from cell death. We additionally stimulated Hs578T cells with TGF-b before polyI:C transfection, and found that pretreatment with TGF-b very weakly suppressed polyI:C-induced cell death, although the effect of TGF-b was not significant (Fig. S3B). These results suggested that the activation of the downstream signaling pathway by expression of caSmad3 in Hs578T cells partly rescued the cells from polyI:C-induced cell death.

Characterization of polyI:C-induced cell death
To understand the therapeutic efficacy of polyI:C transfection on TNBC, we further characterized the polyI:C-induced cell death. As caspase dependency is an important factor for characterizing cell death, we treated polyI:C-transfected Hs578T cells with the pan-caspase inhibitor Z-VAD-FMK. We found that Z-VAD-FMK treatment dramatically decreased the proportion of annexin V + /PI + cells that were markedly increased by polyI:C, suggesting that polyI:C-induced cell death was largely dependent on caspase function (Fig. 3A,B). A previous report suggested that RLR-induced caspase-dependent cell death is mediated by intrinsic apoptosis, extrinsic apoptosis, and pyroptosis [11]. Although the definition of 'pyroptosis' has changed with time, it is currently believed that pyroptosis is a kind of programmed necrotic cell death induced by gasdermins [36,37]. Therefore, we examined whether the polyI:C-induced cell death was pyroptosis. After transfection of Hs578T cells with polyI:C, we observed cell swelling with large bubbles (Fig. 3C), the release of cellular contents quantified by LDH (Fig. 3D), and the absence of annexin V + and PIearly apoptotic population (Figs. 2B and 3A), which are important phenotypes of pyroptosis. In addition, we evaluated gasdermin cleavage by blotting the N terminus of GSDMD and GSDME. GSDMD and GSDME of the gasdermin family are known to be endogenously cleaved and contribute to pyroptosis. While we did not detect GSDMD cleavage after polyI:C transfection of Hs578T cells, cleavage of GSDMD was detected in THP-1 cells following the combination of LPS and nigericin, which was used as a positive control (Fig. 3E). In contrast, N-terminal-cleaved GSDME was detected in polyI:C-transfected Hs578T cells (Fig. 3E). These results suggested that polyI:C transfection induced pyroptosis in Hs578T cells.

Attenuation of TGF-b signaling by polyI:C transfection accelerated caspase 3-dependent pyroptosis
Inhibition of cell death by Z-VAD-FMK and detection of GSDME cleavage suggested that activation of the cell death pathway by polyI:C in Hs578T cells was predominantly executed by caspase 3, which is the only known caspase to process and activate GSDME. Therefore, we first used the caspase 3-specific inhibitor Z-DEVD-FMK and found that polyI:C-induced cell death was almost entirely prevented by Z-DEVD-FMK treatment (Fig. 4A). Z-DEVD-FMK treatment also inhibited GSDME cleavage induced by polyI:C in Hs578T cells (Fig. 4B). These results suggested that polyI:C-induced pyroptosis in Hs578T cells was largely dependent on caspase 3 activation. As described above, attenuation of TGF-b signaling by polyI:C accelerated Hs578T cell death. Furthermore, the forced expression of caSmad3 in Hs578T cells decreased the proportion of annexin V + /PI + cells and directly increased the number of annexin V -/PIcells without the appearance of early apoptotic annexin V + /PIcells. Therefore, the next question we asked was whether caSmad3 expression would attenuate polyI:C-induced pyroptosis in Hs578T cells. To address this, we measured caspase 3/7 activity and found that caSmad3 expression suppressed caspase 3/7 activation, which was elevated by polyI:C transfection (Fig. 4C). In addition, caSmad3 expression also abrogated polyI:C-induced GSDME cleavage and the release of LDH (Fig. 4D,E). Expression of GSDME is important for switching from apoptosis to pyroptosis [37]. Although expression of GSDME is lower in all types of breast cancer than that in other types of cancer cells (Fig. S4), we found that GSDME expression in TNBC was significantly higher than that in non-TNBC [25,38] (Fig. 4F), supporting a previous report [39]. The above results suggested that TNBC-specific GSDME expression appeared to contribute to polyI: C-induced caspase 3-mediated pyroptosis, which was inhibited by caSmad3 expression.
To examine whether suppression of TGF-b signaling by polyI:C and its pyroptosis-promoting effect are universally applicable in TNBC, we additionally used another TNBC cell line, BT-549, in which TGF-b also shows tumor-promoting effects [22]. Transfection of polyI:C suppressed phosphorylation of Smad3 (Fig. S5A) and the induction of TGF-b target genes PMEPA1, SERPINE1, and SMAD7. We have not observed significant induction of ZEB1 and LRRC15 (Fig. S5B) that we have seen in Hs578T cells (Fig. 1C), possibly reflecting the context dependency of the TGF-b targets [40]. We also checked caspase-dependent cell death using Z-VAD-FMK. PolyI:C transfection notably induced annexin V + /PI + populations, and Z-VAD-FMK treatment dramatically suppressed these populations, suggesting caspase-dependent cell death by polyI:C was also observed in BT-549 cells (Fig. S5C). Significant production of N-terminal cleaved GSDME, not GSDMD, was also observed in BT-549 cells by polyI:C transfection (Fig. S5D). Thus, we analogously produced BT-549 cells stably expressing caSmad3, and evaluated polyI:C-induced cell death. Expression of caSmad3 prominently suppressed cell death (Fig. S5E-G), and TGF-b pretreatment showed a trend to partially abrogate polyI:C-induced cell death (Fig. S5H). Cleavage of GSDME was also attenuated by caSmad3 (Fig. S5I).

Detailed mechanism of protection from polyI:C-induced cell death by caSmad3
We then investigated the mechanisms of caSmad3-triggered cell survival of Hs578T cells following polyI:C transfection. RNA-seq analysis revealed that TGF-b stimulation did not dramatically alter antiapoptotic Bcl-2 factor expression in Hs578T cells (Fig. S6A). Furthermore, treatment with the TbRI inhibitor LY364947 did not increase cell death (Figs. 5A and S6B), suggesting that not the TGF-b signaling alone but its interference with cell death pathway triggered by polyI:C exhibited cell survival in Hs578T cells. One of the known mechanisms of RLR-induced cell death is explained by Noxa, a BH3-only protein. We found that polyI:C dramatically induced Noxa expression, which was suppressed by the gene inactivation of MDA5 and RIG-I (Fig. 5B). Stimulation of the cells with TGF-b did not affect the expression of Noxa, which was in accordance with the result observed in the flow cytometry in Fig. 2C. Next, we measured Noxa expression after polyI:C transfection with or without caSmad3 expression. We found that caSmad3 partly suppressed the induction of Noxa mRNA and protein expression (Figs 4D and 5C). Decreased expression of Noxa with caSmad3 expression was also observed in BT-549 cells (Fig. S5I). To investigate the possibility of transcriptional repression by the Smad pathway downstream of TGF-b, we reanalyzed the Smad2 ChIP sequence data in Hs578T cells stimulated with the TGF-b family ligand activin A [28]. We identified a Smad2 binding site in the promoter region of the PMAIP1 gene encoding Noxa in Hs578T cells (Fig. 5D). A peak in the vicinity of this site was also confirmed in an anti-Smad2/3 ChIP-seq data obtained from TGF-b-stimulated Hs578T cells, though the peak   was weak (data not shown) [19]. These results suggest that Smad-mediated transcriptional repression may be associated with Noxa downregulation, which then led to the attenuation of polyI:C-induced cell death by caSmad3 expression. However, the contribution from inhibition of the Smad pathway to the strong induction of Noxa by polyI:C was limited. Our results therefore suggested that additional mechanisms were involved in the induction of pyroptosis through inhibition of the TGF-b signaling. Therefore, we examined the association between p38 and the caSmad3-mediated reduction of cell death. We found that caSmad3 expression attenuated polyI:C-induced phosphorylation of p38 in Hs578T cells, which could be caused by induction of DUSP1 by TGF-b, a known p38 phosphatase (Fig. S7A,B). We used two p38 inhibitors and found that SB203580 partly decreased polyI:C-induced cell death, but the effect of SB202190 was less (Fig. S7C,  D). Together, these results suggested that caSmad3 possibly attenuated polyI:C-induced pyroptosis through its multiple downstream factors.

Suppression of TGF-b signaling by polyI:C contributed to tumor suppression
Finally, we addressed the efficacy of polyI:C suppression of TGF-b signaling in vivo. We used murine breast cancer 4T1 cells, which have been previously used as a model of TNBC [41]. PolyI:C transfection in 4T1 cells attenuated the phosphorylation of Smad3 following TGF-b stimulation (Fig. 6A). We then evaluated polyI:C-induced cell death in 4T1 cells. Cell swelling, direct transition of cellular annexin V -/PIpopulations to annexin V + /PI + populations, and cleavage of GSDME protein were all observed in 4T1 cells after polyI:C transfection, suggesting that polyI:C also promoted pyroptosis in 4T1 cells (Fig. 6B-E). In addition, we also examined the effect of caSmad3 overexpression on polyI:C-induced cell death. Forced expression of caSmad3 partially inhibited polyI:C-induced cell death (Fig. 6C,D) and GSDME cleavage (Fig. 6E). Treatment with Z-VAD-FMK also partially suppressed polyI:C-induced cell death (Fig. 6C,D). These results suggested that both caspase-dependent and caspase-independent mechanisms of cell death induced by polyI:C were involved in the 4T1 cells and that this was a critical reason for the partial suppression of polyI:C-induced cell death by caSmad3. We then evaluated the effect of intratumoral polyI:C transfection on tumor suppression using 4T1 cells orthotopically transplanted into BALB/c mice. Although polyI:C transfection did not significantly affect tumor volume, tumor weight was significantly reduced by polyI:C transfection (Fig. 6F,G) and phosphorylation of Smad3 in tumor samples was significantly reduced after polyI:C transfection (Fig. 6H,I). These results suggested that transfection of polyI:C induced pyroptosis in 4T1 cells, leading to the suppression of tumor growth, which was partially mediated by attenuation of TGF-b signaling.

Discussion
The activation of cytosolic dsRNA signaling in cancer cells has been a widely recognized and may be a promising strategy for treating cancers. In the present study, we demonstrated that activation of RLR signaling by transfection of TNBC cells with a synthetic dsRNA analogue suppressed TGF-b signaling. Furthermore, this process provided additional therapeutic benefit in that attenuation of TGF-b signaling led to enhancement of cell pyroptosis.

Activation of RLR singling or other PRR signaling in cancer cells
To prevent virus replication, hosts have a system that results in the death of infected cells [42]. It is important to note that this cell death pathway also exists in many cancer cells because many reports have suggested the importance of RLR signaling in tumor suppression. Overexpression of MDA5 [43], endogenous retroviruses [44,45], cytosolic induction of polyI:C [12,[46][47][48], and other ligands recognized by RLR [8] have been reported to increase cell death and tumor repression. Therefore, it is not surprising that polyI:C is being evaluated in a clinical trial of cancer treatment, including TNBC. In the present study, we showed that intracellularly administered polyI:C was recognized by MDA5 and RIG-I, which resulted in the attenuation of TGF-b signaling and promotion of cell death. Based on our present findings, other RLR ligands for cancer treatment may show similar inhibitory effects on TGF-b signaling through the same RLR signaling pathway.
In the context of TNBC treatment, recently approved PARP inhibitors for patients carrying BRCA mutations strikingly inhibit DNA-repairing system, resulting in breakage of DNA, release of small DNA called 'micronuclei' in the cytoplasm, and triggering cyclic GMP-AMP synthase (cGAS) stimulator of interferon genes (STING) DNA-sensing pathway that causes activation of TANK-binding kinase 1 (TBK1) and IRF3 [49,50]. Therefore, there is a possibility that PARP inhibitor simultaneously suppresses TGF-b signaling through production of phosphorylated IRF3 by TBK1, and this can be one of the mechanisms of cancer cell death by PARP inhibitor. In addition, activation of nucleic acid sensing pathway such as by cytosolic dsRNA administration or PARP inhibition (with activation of cGAS-STING signaling) upregulates immune checkpoint proteins [51][52][53]. Therefore, the combination therapy of anti-PD-1/PD-L1 antibody with PARP inhibitor or polyI:C is also suggested to suppress TNBC progression partly through the inhibition of TGF-b pathway. Of note, PARP also negatively regulates TGF-b-Smad signaling through ADPribosylation of Smads [54]. Further investigations will be needed regarding the antitumor effect and mode of action of each PARP inhibitor, in relation to its effect on the cGAS-STING signaling and TGF-b signaling.

Suppression of TGF-b signaling by activation of RLR signaling
Signaling interactions sometimes reveal important biological phenotypes. One study showed that knockdown of IRF3 accelerates TGF-b-induced EMT in HaCaT normal epidermal keratinocytes, and suggested that activated IRF3 suppresses TGF-b signaling by binding to Smad3 and blocking its activation by the receptors [24]. This is supported by the findings that the structure of C-terminal region of IRF3 is similar to that of the MH2 domain of Smad3, which is important for protein-protein interactions [55][56][57][58]. Considering our findings and the fact that TGF-b is also reported to be important for maintaining cancer stem cells that show resistance to chemotherapeutic drugs in breast cancer, polyI:C transfection may also target caner stem cells.

Prosurvival function of TGF-b signaling
It is noteworthy that TGF-b contributes to two contrasting aspects of cell fate. It is reported to exhibit both antisurvival and prosurvival functions, which are dependent on the context and cell types. For example, TGF-b suppresses antiapoptotic Bcl-2 and Bcl-X L expression to induce cell death [59][60][61], while in other situations, it induces their expressions to promote cell survival [62][63][64]. We focused on the prosurvival function of TGF-b in this study, which is also observed in microglia, pancreatic b cells, and regulatory T cells by induction of FLICE-inhibitory protein (FLIP or CFLAR), which suppresses the function of caspase 8 [65][66][67]. In the context of cancer, TGF-b induces BHLHE40, which encodes deleted in esophageal cancer 1 (DEC1) to promote survival of mammary carcinoma cells [21]. However, we were unable to detect upregulation of Bcl-2, Bcl-X L , or FLIP by TGF-b stimulation ( Fig. S6A and data not shown) or the strong induction of DEC1 by caSmad3 expression in Hs578T cells (data not shown). These results indicated that other prosurvival mechanisms of TGF-b signaling were present in TNBC cells. We suggested that expression of caSmad3 suppressed polyI:C-induced Noxa expression, which is reported to be important for polyI:C-induced cell death in cancer cells [46,48]. This is believed to be mediated through transcriptional suppression by Smads [68]. However, caSmad3-mediated reduction of Noxa expression is not drastic, leading to the possibility of other systems also being involved in the prosurvival function of TGF-b in polyI:C-triggered cell death, such as attenuation of p38 phosphorylation (Fig. S7).

Pyroptosis
The term 'pyroptosis' was originally defined as caspase 1-dependent cell death accompanied by inflammation [69]. However, the definition was changed after the identification of gasdermins as executioners in the permeabilization of plasma membranes [36,[70][71][72]. During the late stage of cell death, gasdermins are cleaved by activated caspases or other cytotoxic proteases, releasing the N-terminal domain, which can trigger pyroptosis by forming oligomers and pores in the cell membrane. This mechanism explains the characteristics of pyroptotic cells we observed. Recent findings suggest that pyroptosis is closely related to cancer progression. Pyroptosis occurs in cancer cells, changing the tumor immune microenvironment from 'cold' to 'hot' due to the accumulation and activation of antitumor immune cells, such as cytotoxic T cells or natural killer cells. This is believed to be caused by the release of damage-associated molecular patterns, such as HMGB1, from dead cells [73,74]. Given that activation of the tumor immune microenvironment is associated with immunotherapy efficacy [75] and GSDME is highly expressed in TNBC [39] (our data), polyI:C-induced pyroptosis may provide more potent therapeutic efficacy than we expected for the treatment of TNBC. In addition, we showed that caSmad3 expression attenuated polyI:C-induced pyroptosis through diminishing caspase 3 activity. This may be one reason TGF-b exhibits immune suppressive activity in the tumor microenvironment.

Conclusion
Transfection of polyI:C promoted pyroptosis, which was enhanced by the suppression of TGF-b signaling and its anti-pyroptotic function. The findings in the present study thus help elucidate a potential molecular targeting strategy associated with the use of synthetic dsRNA analogues for patients with TNBC.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. MDA5-and RIG-I-mediated suppression of TGF-b signaling. Fig. S2. Potential of cell-intrinsic activation of RLR signaling and attenuation of TGF-b signaling. Fig. S3. caSmad3 expression in Hs578T cells and weak inhibition of polyI:C-induced cell death by the pretreatment of TGF-b. Fig. S4. GSDME expression in various types of cancer cells. Fig. S5. Anti-pyroptotic effect of TGF-b is suppressed by polyI:C in BT-549 cells. Fig. S6. Mechanisms of caSmad3-mediated cell survival. Fig. S7. Attenuation of p38 phosphorylation by caS-mad3 partially inhibits polyI:C-induced cell death. Table S1. Sequences of the primers. Table S2. Sequences of the gRNAs. Table S3. Top 100 differentially expressed genes by TGF-b stimulation in Hs578T cells.