LPAR4 mediates resistance to interferon-induced stress in soft tissue sarcoma Authors: Jin-Fen Xiao, Emily Y. Ko, Ashley Smith, Roberta Piras, Annaliese Fowler, Kristin Ishaya, Longqiang Wang, Bin Wang, Adriana Paulucci-Holthauzen, Sharon M. Landers, Angela D. Bhalla, Keila E. Torres, Diana Shamsutdinova, Khalida M. Wani, Alexander J. Lazar, Emily Z. Keung, Neeta Somaiah, Elise F. Nassif Haddad, Jlenia Guarnerio Published: 2026 ABSTRACT Soft tissue sarcomas are aggressive malignancies with limited responses to immunotherapy. Strategies that induce viral mimicry and type I interferon (IFN-I) signaling may enhance tumor immunogenicity but remain poorly explored in sarcoma. Here, we show that the CDK4/6 inhibitor abemaciclib triggers viral mimicry and robust IFN-I responses in sarcoma cells, leading to tumor growth suppression and remodeling of the tumor microenvironment toward an immune-permissive state. However, sarcoma cells activate resistance mechanisms that limit IFN-induced stress. We identify lysophosphatidic acid receptor 4 (LPAR4) as a key mediator of this adaptation. LPAR4 dampens IFN signaling, reduces oxidative stress, and promotes tumor cell survival under therapeutic pressure. Genetic silencing of LPAR4 enhances IFN responses, increases reactive oxygen species, and sensitizes sarcoma cells to apoptosis, resulting in delayed tumor growth. These findings identify LPAR4 as a central regulator of IFN-driven stress adaptation and a potential therapeutic target to enhance viral mimicry–based strategies in sarcoma. INTRODUCTION Soft tissue sarcomas (STS) are a heterogeneous group of mesenchymal malignancies classified by their cell of origin, degree of differentiation, and underlying genetic alterations (1, 2). Poorly differentiated subtypes, including dedifferentiated liposarcoma (DDLPS) and undifferentiated pleomorphic sarcoma (UPS), are among the most aggressive forms, with limited responsiveness to standard therapies and poor outcomes once the disease becomes recurrent or metastatic. Current treatment relies on surgical resection and radiotherapy, often combined with anthracycline-based chemotherapy, yet durable disease control remains uncommon (3). Immune checkpoint inhibitors (ICIs), such as antibodies targeting PD-1 or PD-L1, have revolutionized treatment for several solid tumors but have shown limited efficacy in STS (4–7). This lack of response is partly attributable to the low immune infiltration characteristic of most STS, which limits effective antitumor immune activation (5, 8). Consequently, there is a critical need for adjuvant strategies that enhance tumor immunogenicity and improve responses to existing therapies. One promising approach is the induction of viral mimicry responses within tumor cells (9). Agents that promote derepression of endogenous retroviruses (ERVs) leads to the accumulation of viral-like double-stranded RNAs (dsRNAs). These dsRNAs activate innate immune sensors, including the RIG-I/MDA5/MAVS pathway, triggering robust type I interferon (IFN-I) and inflammatory cytokine production within the tumor microenvironment (TME) (2). IFN-I signaling, mediated primarily by IFN-α and IFN-β, exerts potent antitumor effects through both tumor cell–intrinsic and tumor cell–extrinsic mechanisms (10). Within tumor cells, IFN-I signaling suppresses proliferation (11–13), perturbs metabolic homeostasis (14), alters differentiation (15) and senescence programs (16), and ultimately promotes apoptosis (17, 18). IFN-I also enhances tumor immunogenicity by regulating antigen processing and presentation through major histocompatibility complex (MHC) class I and II pathways in both tumor and antigen-presenting cells (19, 20). This includes formation of the immunoproteasome, composed of the PSMB8, PSMB9, and PSMB10 subunits, which generates peptides for presentation as tumor neoantigens (21). In parallel, IFN-I signaling in the TME drives antitumor immunity by activating dendritic cells (DCs) and promoting CD8+ T cell cytotoxicity and CD4+ TH1 (T helper 1) differentiation (22). Tumor-associated macrophages (TAMs), the most abundant immune population in sarcomas (23, 24), play a central role in shaping immune responses (25–27). Elevated IFN-I levels favor polarization of TAMs toward an inflammatory, antitumor phenotype while limiting immunosuppressive myeloid populations, thereby enhancing immune-mediated tumor control (28, 29). Several therapeutic agents can induce IFN-I responses through viral mimicry. In STS, CDK4/6 inhibitors—particularly palbociclib and the newer agent abemaciclib—are especially attractive because they potentially combine immune-modulatory action with direct targeting of cell cycle dysregulations common in sarcomas (30–35). CDK4 amplification in DDLPS and CCNE1 amplification or CDKN2A loss in UPS could render these tumors particularly susceptible to CDK4/6 inhibitors (2), which, accordingly, are currently being evaluated in clinical trials for both CDK4-dependent and CDK4-independent sarcomas (36, 37). Here, we show that abemaciclib induces robust IFN-I signaling in sarcoma cells, which is essential for remodeling the TME and restraining tumor growth. However, we also find that sarcoma cells engage mechanisms that mitigate IFN-induced stress and limit therapeutic efficacy of this drug. We identify lysophosphatidic acid receptor 4 (LPAR4) as a key mediator of this stress-resistant response. LPARs (LPAR1–6) are described as activating diverse signaling pathways, including PI3K (phosphatidylinositol 3-kinase)/AKT and MAPK (mitogen-activated protein kinase), and have been implicated in tumor progression, invasion, and immune suppression (38–41). Despite this, their roles in sarcoma remain largely unexplored, especially for LPAR4, which has been only marginally characterized in cancer. Our data demonstrate that LPAR4 limits IFN response and oxidative stress, thereby promoting tumor cell survival under therapeutic pressure. Genetic silencing of LPAR4 enhances IFN signaling, increases reactive oxygen species (ROS) accumulation, and sensitizes sarcoma cells to apoptosis. Together, these findings identify LPAR4 as a central regulator of stress adaptation in sarcoma and nominate its inhibition as a possible strategy to enhance therapies that rely on viral mimicry and IFN induction. RESULTS Activation of IFN signaling in sarcoma cells alters the TME and decreases tumor growth High expression of IFN-stimulated genes (ISGs) is associated with improved prognosis in patients with sarcoma (fig. S1A). To directly examine the role of IFN-I signaling in sarcoma growth and in shaping the tumor immune microenvironment, we used a genetically engineered mouse model of undifferentiated sarcoma driven by Ccne1 overexpression combined with Trp53 loss (model hereafter referred to as Ccne1). As previously reported, tumors arising in this model are largely immune excluded, with limited infiltration of T cells and myeloid cells into the tumor parenchyma (42). Notably, Ccne1 sarcoma cells grew more rapidly and formed larger tumors in IFN-insensitive hosts lacking the interferon-α/β receptor (IFNAR) 1 (Ifnar1-deficient mice) compared with wild-type controls (Fig. 1A). These findings suggest a critical role for IFN-I signaling in restraining sarcoma growth. Because this model exhibits low baseline immunogenicity yet remains highly sensitive to IFN-I, it provides an ideal system to dissect how IFN-I shapes antisarcoma immune responses and to evaluate strategies aimed at enhancing therapeutic efficacy. Abemaciclib induces IFN-1 repones and remodels the sarcoma microenvironment. (A) Ccne1 UPS tumors grew significantly larger in Ifnar1-deficient mice (n = 7) than in wild-type C57BL/6 mice (n = 6). Means ± SEM; Mann-Whitney U test. (B) Median fluorescence intensity (MFI) of J2 antibody–based dsRNA staining in Ccne1 cells following treatment with abemaciclib. NT, nontreated control. Box plot shows the median MFI and interquartile range (IQR); whiskers indicate 1.5 × IQR. Statistical significance was determined by the two-tailed Student’s t test; **P < 0.01. Each dot represents one technical replicate; data are representative of four independent experiments. (C) qPCR quantification of IFN-I and ISG expression in Ccne1 cells, expressed as log2 relative expression (compared to nontreated cells). Data are representative of three independent experiments. (D) Abemaciclib treatment induced up-regulation of the viral RNA sensors MDA5 and RIG-I proteins in Ccne1 cells. Data are representative of three independent experiments. (E) ELISA measuring secreted Ifnβ in the supernatant of Ccne1 sarcoma cells treated with abemaciclib. Means ± SD; Student’s t test; **P < 0.01. Results represent pooled data from three independent experiments. (F) Schematics of the experimental design. All tumors were harvested after 11 days of abemaciclib treatment. (G) Abemaciclib treatment significantly increased proportions of CD45+, CD11b+, macrophage, and monocyte populations within the Ccne1 TME. Each dot represents one mouse. Flow cytometry analysis shows means ± SEM; Mann-Whitney U test. (H) Representative crops from IF staining of CD3+ T cells (green) in Ccne1 tumors treated with abemaciclib or vehicle. Nuclei were stained with DAPI (blue). Scale bars, 600 and 200 μm. Quantification of CD3+ cell density (cells/mm2) in each tumor section is shown on the right. Means ± SEM; n = 6 tumors per group, one section quantified per tumor. For all panels, *P < 0.05; **P < 0.01. To trigger IFN responses in the tumor mass, we used the CDK4/6 inhibitor abemaciclib, which has been previously reported to activate viral mimicry–induced IFN responses in the context of breast cancer (36) but never fully assessed in this respect in STS. Abemaciclib treatment showed significant accumulation of dsRNAs in mouse (Fig. 1B) and human STS cell lines (fig. S1B), elevated expression of ISGs and IFN-I (Fig. 1C and fig. S1C), up-regulation of the viral RNA sensors RIG-I and MDA5, known mediators of viral dsRNA-induced IFN responses (30, 37, 43) (Fig. 1D), and secretion of IFN-β (Fig. 1E). Myeloid cells represent the major immune population within the sarcoma TME (24). IFNs profoundly affect the activation and polarization of these cells, typically increasing monocyte infiltration and promoting their differentiation into IFN-producing macrophages with antitumor properties (27). We next investigated whether IFN responses in the sarcoma cells affect the myeloid compartment of the TME. To trigger IFN responses in the tumor, abemaciclib treatment was initiated once the average size of all tumors reached or exceeded 100 mm3. Treatment was administered daily by oral gavage, and tumor growth was monitored by caliper measurements. The experimental endpoint was defined as the point at which the average tumor volume in the control group (vehicle-treated mice) reaching ∼1 cm3 (day 11 in the shown experiment) or the occurrence of tumor ulceration, in accordance with animal care guidelines. At this stage, all mice from both groups were euthanized simultaneously (Fig. 1F). Tumors were collected, weighed, and processed for downstream analyses, including flow cytometry and immunofluorescence (IF) staining. Flow cytometry revealed a marked increase in total immune (CD45+) infiltrates in abemaciclib-treated tumors. Within this population, myeloid cells (CD11b+), including monocytes (CD11b+Ly6C+) and macrophages (CD11b+F4/80+), were proportionally increased following treatment, whereas neutrophil (CD11b+Ly6C+Ly6G+) levels remained unchanged (Fig. 1G and fig. S1D). Changes in the myeloid compartment were accompanied by a significant increase in CD3+ T cell infiltration in abemaciclib-treated tumors (Fig. 1H). This increase was consistently observed both when comparing tumors at the same time point and in an independent experimental setting in which abemaciclib was administered for a longer duration (up to 31 days), with tumors from all groups collected at a matched size (∼1 cm3). CD4+ T cells, CD8+ T cells, and natural killer (NK) cells were all increased following abemaciclib treatment compared to vehicle-treated controls (fig. S1, E and F), suggesting a reshaping of the immune TME. These findings are consistent with prior preclinical studies in colorectal cancer, where prolonged abemaciclib monotherapy enhanced T cell inflammatory signatures within tumors and improved the efficacy of anti–PD-L1 therapy (44). Myeloid cells are highly plastic and can adopt a range of activation states that either support or inhibit tumor growth. TAMs with anti-inflammatory phenotypes are thought to promote tumor progression by suppressing T cell activity, whereas inflammatory, IFN-responsive macrophages enhance antitumor immunity. On the basis of previous findings that IFN-responsive TAMs restrain tumor growth (24, 28), we tested the hypothesis that tumor cell–derived IFNs reprogram myeloid cells toward an IFN-responsive/expressing, antitumor activation state. To test this, we profiled sarcoma-infiltrating myeloid cells using single-cell RNA sequencing (scRNA-seq) from vehicle-treated and abemaciclib-treated tumors (fig. S2A). Multiple myeloid subpopulations were identified (Fig. 2A and fig. S2B), including monocytes (high Plac8 and Ly6c2 expression), monocyte-derived DCs (moDCs; high expression of H2-Ab1, H2-Aa, H2-DMa/b1, and Cd74), and several TAM subtypes. Among these, MHC-II-high TAMs expressed antigen-presentation genes along with lower levels of complement-related genes (C1qc and C1qa), whereas distinct subsets of TAMs expressing either Arg1 or Mrc1 (CD206) were also detected (Fig. 2A). Abemaciclib treatment increased the proportions of monocytes and antigen-presenting macrophages while reducing Mrc1+ and Arg1+ macrophages (Fig. 2B). Differential expression and pathway enrichment analysis of myeloid cells (pseudobulked from scRNA-seq) revealed strong up-regulation of IFN-α and IFN-γ response pathways following treatment, whereas pathways associated with hypoxia and glycolysis were down-regulated (fig. S2C). To corroborate results from scRNA-seq, we used Mx1gfp IFN-stimulated response element (ISRE) reporter mice to trace IFN-responsive myeloid cells in vivo, where green fluorescent protein (GFP) expression, under the control of the Mx1 locus, reflects binding of ISREs downstream of IFN signaling (45) (Fig. 2C and fig. S2D). In this setting, we assessed that abemaciclib treatment increased the IFN-responsive monocyte fraction (Ly6C+ GFP+) of the myeloid compartment (CD11b+) ∼2.4-fold compared to vehicle-treated tumors, as shown in Fig. 2C and fig. S2E. IFN response is important for the tumor-suppressive effect of abemaciclib. (A) Dot plot showing top differentially expressed genes in each myeloid cluster identified by scRNA-seq. (B) Stacked bar plots showing the proportions of the five myeloid cell types in individual mice treated with abemaciclib or vehicle. (C) Left: Schematic of Mx1gfp reporter mice, illustrating the knock-in of an inducible GFP/OVA (ovalbumin) fusion gene at the translation start site (TSS) of the Mx1 (MX dynamin-like GTPase 1) locus. Right: Flow cytometry analysis showing significantly increased Mx1-GFP+ (IFN-responsive) monocytes in the CD11b+ compartment of abemaciclib-treated Ccne1 tumors (n = 8) compared with vehicle controls (n = 7). Mann-Whitney U test. Each dot represents one mouse. (D) Schematics of the experimental design is represented on the top panel. The bottom-left panel shows tumor volume measurements for Ccne1 tumors growing in C57BL/6 mice. Mice were randomized into treatment groups when the average tumor size is above 100 mm3. The bottom-right panel presents endpoint tumor weights in grams. Each dot represents one mouse. Statistical significance was assessed using the nonparametric Mann-Whitney U test. (E) Schematics of the experimental design is represented on the top panel. The bottom-left panel shows tumor volume measurements for Ccne1 tumors growing in Ifnar1−/− recipient mice. The bottom-right panel presents endpoint tumor weights in grams. Data represent means ± SD. Each dot represents one mouse; Student’s t test. (F) Model illustrating that abemaciclib treatment increases endogenous dsRNAs in tumor cells, which are sensed by MDA5 and RIG-I, leading to IFN-related genes and IFN-β up-regulation. IFN-β secretion from the sarcoma cells into the TME activates signaling pathways in the immune cells by binding the IFNAR [Created in BioRender. J. Xiao (2026) https://BioRender.com/rr0karh]. For all panels, *P < 0.05; **P < 0.01. Amplified IFN responses in tumor cells and in the TME significantly reduced tumor growth in mice (Fig. 2D). IFN signaling in the TME is key for the therapeutic efficacy; the antitumor effects of abemaciclib were largely abrogated in Ifnar1-deficient mice (Fig. 2E), suggesting that an IFN-sensitive TME is essential for restraining tumor growth. Collectively, these results show that increasing levels of IFN-I in the tumor mass can be therapeutically advantageous by promoting accumulation of IFN-responsive myeloid cells and T cells in the sarcoma microenvironment, even for tumors that are poorly infiltrated by the immune cells at baseline (Fig. 2F). Lpar4 expression is associated with sarcoma cell survival under abemaciclib treatment We next examined how viral mimicry responses—and the downstream IFN signaling they elicit in both tumor cells and the TME—shape the pathological features of sarcoma. Specifically, we investigated how tumor cell transcriptional programs are reprogrammed as cells adapt to elevated IFN levels and increased immune infiltration in the TME triggered by abemaciclib. To address this, we next analyzed ex vivo tumor cells from the scRNA-seq experiment reported in fig. S2A, and the growth curve is shown in Fig. 2D. This analysis revealed three transcriptionally distinct tumor cell clusters. One cluster was defined by markers of active proliferation (e.g., Ki67, Mcm3, and Mcm5), whereas the other two were characterized by either (i) high expression of extracellular matrix (ECM) components (Aspn, Ogn, Fbn1, and Cdh11) or (ii) a hypoxia-associated signature (Mt1/2, Hmox1, Bnip3, and Ero1l) (Fig. 3A). When comparing cluster proportions across treatment conditions, we found that both cycling and hypoxic-like tumor cells decreased after abemaciclib treatment, whereas ECM-high tumor cells were proportionally enriched (Fig. 3B and fig. S3, A and B). Given that ECM-rich sarcoma cells appeared relatively insensitive to abemaciclib, we focused on this population to identify distinguishing features beyond ECM gene expression. Differential gene expression analysis revealed that, as expected, these abemaciclib-insensitive cells (remaining after abemaciclib treatment) expressed not only significantly higher levels of ECM-related genes compared to treatment-sensitive cells but also higher levels of Lpar4, encoding LPAR4 (Fig. 3, C and D). Notably, higher LPAR4 expression is significantly correlated with poor prognosis in patients with sarcoma (Fig. 3E). In a clinical trial setting, LPAR4 protein expression was detected in human DDLPS samples collected after treatment with palbociclib, a CDK4/6 inhibitor precursor to abemaciclib (Fig. 3F). Sarcoma cells expressing high levels of Lpar4 resist abemaciclib treatment. (A) Dot plot showing top genes defining three tumor cell clusters: Proliferating tumor cells (tumor cycling), hypoxic-like tumor cells (tumor hypoxia), and ECM-enriched tumor cells (tumor ECM). (B) Stacked bar plot comparing the proportions of the three tumor cell clusters in vehicle-treated versus abemaciclib-treated tumors. (C) Volcano plot showing significantly altered genes in abemaciclib-treated tumors compared with vehicle controls. Lpar4 was among the top up-regulated genes following abemaciclib treatment. Cutoffs were defined as log2 (fold change) > 1 or log2 (fold change) < −1 and P < 0.01. (D) Violin plot depicting the expression of Lpar1 to Lpar6 in Ccne1 tumor cells treated with abemaciclib or vehicle. (E) High LPAR4 expression is associated with worse survival in patients with STS in The Cancer Genome Atlas (TCGA SARC dataset). Log-rank P value and hazard ratio (HR) are shown. (F) DAB immunohistochemical (IHC) analysis of LPAR4 expression in specimens of patients with DDLPS from a palbociclib clinical trial (33). Left: Representative case displaying high LPAR4 staining, with positive signals localized to the membrane and cytoplasm. Right: H-score quantification of LPAR4 IHC staining. Means ± SD. Each dot represents one patient sample. (G) Pathways significantly enriched in patients with LPAR4-high versus LPAR4-low STS in TCGA SARC or (H) in patients with LPAR4-low versus LPAR4-high. In humans and in mice, the LPAR gene family consists of six LPA receptors (LPAR1–6), whose expression is cell type specific (38). Examination of the six receptors across cell types in our scRNA-seq data revealed that tumor cells mainly express Lpar1 and Lpar4 (fig. S3C). However, whereas Lpar1 expression was unaffected by abemaciclib, Lpar4 was notably up-regulated (Fig. 3D), suggesting a potential association with tumor cell resistance. Our results indicate that abemaciclib treatment is associated with enrichment of tumor cells expressing high levels of ECM genes and LPAR4. Although the role of the ECM has been widely studied in cancer—particularly epithelial tumors—and sarcomas are known to be highly fibrotic (46), the role of LPAR4 in cancer remains poorly explored. A recent study in pancreatic cancer reported a link between LPAR4 expression and the ECM protein fibronectin-1, which contributes to tumor stemlike properties (47). Beyond this report, LPAR4 has received little attention in cancer biology, and its role in regulating IFN responses has not been previously examined. On the basis of these observations, we sought to determine the functional consequences of increased LPAR4 expression. We stratified patients with STS [from The Cancer Genome Atlas soft-tissue sarcoma (TCGA SARC)] according to LPAR4 expression and performed pathway analysis comparing LPAR4-high (expression z-score > 1) versus LPAR4-low (z < −1) samples, controlling for the STS subtype. LPAR4-high tumors exhibited up-regulation of pathways associated with cell proliferation, stem cell regulators (e.g., β-catenin and Hedgehog signaling), and ECM organization (Fig. 3G and fig. S3D). Conversely, pathways associated with oxidative phosphorylation, mitochondrial function, apoptosis, and IFN signaling were enriched in LPAR4-low samples (Fig. 3H and fig. S3E). To assess how LPAR4 expression changes in response to abemaciclib, we performed both short-term and long-term in vitro treatments. Short-term, low-dose exposure (125 mM for 7 days) led to a transient increase in Lpar4 mRNA between days 3 and 5, followed by a decline by days 6 and 7 (fig. S3F), consistent with an adaptive response to abemaciclib-induced stress. In contrast, long-term, high-dose treatment (1 μM) eliminated most cells, but remaining cells exhibited persistently elevated Lpar4 expression compared to the initial tumor cell population, with further increases upon prolonged exposure (3 and 8 weeks; fig. S3G). Together, these findings suggest that abemaciclib—or the IFN response it induces—promotes Lpar4 up-regulation. They also raise the possibility that treatment selects for preexisting Lpar4-high cells within the tumor population, contributing to resistance and tumor relapse. Consistent with prior studies of LPAR4 in pancreatic cancer, these data support a model in which LPAR4 functions as a stress-buffering factor. Given that elevated IFN signaling imposes cellular stress in tumors (10, 11, 22), increased LPAR4 expression may help protect tumor cells from IFN-induced stress. LPAR4 represses IFN responses and promotes tumor cell survival and growth pathways To define the functional role of Lpar4, we silenced its expression in sarcoma cells using short hairpin RNAs (shRNAs) (fig. S4A) and assessed the downstream effects by RNA sequencing and reversed-phase protein array (RPPA), both at baseline and following abemaciclib treatment. Consistent with our observations in patient samples, Lpar4 silencing enhanced IFN-related signaling, inflammatory pathways, and antiviral responses driven by ISGs (Fig. 4A). Notably, these pathways were activated both under baseline conditions (Fig. 4A) and in the setting of abemaciclib treatment (fig. S4B), suggesting a synergistic interaction between Lpar4 silencing and the drug exposure. To further evaluate this, we curated IFN-related genes from the pathways identified in Fig. 4A and fig. S4B and visualized their expression as a heatmap. This analysis showed that, although either Lpar4 silencing or abemaciclib alone induced modest IFN responses, their combination resulted in a markedly enhanced transcriptional response (Fig. 4B). These findings were validated in independent experiments, where reverse transcription quantitative polymerase chain reaction (RT-qPCR) confirmed a synergistic increase in IFN-responsive genes upon Lpar4 silencing and abemaciclib treatment (Fig. 4C), accompanied by elevated protein levels of RIG-I, MDA5, and phosphorylated STAT1 (Fig. 4, D and E). In addition, activation of the immunoproteasome (PSMB8 and PSMB9) was observed, consistent with heightened IFN signaling and cellular stress (Fig. 4E). LPAR4 regulates IFN responses in sarcoma cells. (A) Pathway analysis of differentially expressed genes in Lpar4-silenced Ccne1 cells (shLpar4) compared to control cells (shSCR). (B) Heatmap showing that Lpar4 silencing increased ISG expression, both alone and even further in combination with abemaciclib treatment. (C) RT-qPCR analysis confirmed that silencing Lpar4 enhanced the abemaciclib-induced up-regulation of IFN-I and ISGs in Ccne1 cells. Means ± SD. Data are representative of three independent experiments. (D and E) Lpar4 silencing (shLpar4) increased the expression of RIG-I, MDA5, PSMB8, PSMB9, STAT1, and phosphorylated STAT1 (p-STAT1). The increased expression is further pronounced when Lpar4 silencing is combined with abemaciclib treatment. (F) MFI of J2 antibody–based dsRNA staining in Lpar4-silenced Ccne1 cells following treatment with abemaciclib (250 nM). Data are shown as means ± SD. Each dot represents one technical replicate. Student t test; *P < 0.05; ***P < 0.001. Results are representative of three independent experiments. (G) MFI of J2 antibody–based dsRNA staining in Lpar4-overexpressing Ccne1 cells following treatment with abemaciclib (500 nM). Data are shown as means ± SD. Each dot represents one technical replicate. Student t test; *P < 0.05; **P < 0.01. Results are representative of two independent experiments. (H) Schematic representation of the model by which LPAR4 represses dsRNAs and IFN responses in the tumor cells [Created in BioRender. J. Xiao (2026) https://BioRender.com/rr0karh]. Together, these results align with our patient data (Fig. 3, G and H) and support a model in which low Lpar4 expression permits excessive IFN signaling in the context of IFN-inducing therapies such as abemaciclib, potentially increasing tumor cell stress. To determine how Lpar4 silencing enhances IFN responses in tumor cells, we tested the hypothesis that Lpar4 regulates viral mimicry responses by modulating intracellular dsRNA levels. To this end, we performed J2 dsRNA staining in Lpar4-silenced cells cultured with or without abemaciclib. Consistent with the observed enhancement in ISG expression (Fig. 4, B to E), Lpar4 silencing significantly increased dsRNA accumulation, which was further augmented by abemaciclib treatment (Fig. 4F). Conversely, Lpar4 overexpression (fig. S4C) reduced the level of dsRNA induced by abemaciclib (Fig. 4G) and attenuated the expression of IFN-I and ISGs triggered by treatment (fig. S4D), supporting a protective role for LPAR4 (Fig. 4H). Besides IFN responses, proteomic analysis revealed metabolic and mitochondrial dysfunction in the sarcoma cells upon Lpar4-silencing, both in the untreated and abemaciclib-treated contexts. RPPA showed that the metabolic enzyme ACSS2, which supports acetate utilization and metabolic adaptation under stress (48, 49), was markedly reduced in Lpar4-deficient compared to Lpar4-proficient cells (Fig. 5A). Similarly, ATP5 (adenosine triphosphate synthase, also known as mitochondrial complex V), which is crucial for producing ATP and is directly affected by metabolic stress (50) decreased upon Lpar4 silencing (Fig. 5B). Phosphorylation of phospholipase C–γ (PLC-γ) is a key signaling event that occurs in response to metabolic stress, especially oxidative stress. Thus, the activation of PLC-γ is a protective mechanism that helps cells survive stress-induced damage (51). PLC-γ expression was also reduced upon Lpar4-silencing (Fig. 5C), suggesting that, in the absence of Lpar4, sarcoma cells may be less able to buffer and recover from metabolic stress. Moreover, insulin receptor substrate 2 (IRS2) was reduced upon Lpar4 silencing (Fig. 5D), which can affect oxidative phosphorylation, ATP production, and apoptosis in a context-dependent manner (52, 53). Last, there was reduced phosphorylation of BAD protein (Bcl-2–associated agonist of cell death), suggesting its promotion of apoptosis (via reduced ability of the phosphorylated form to bind to and inactivate survival proteins like Bcl-2 and Bcl-xL) (54, 55) (Fig. 5E). In the same line, Lpar4 silencing reduced the expression levels of fatty acid synthase (FAS), where FAS inhibition is associated with mitochondrially primed apoptosis (56) (Fig. 5F). LPAR4 regulates oxidative phosphorylation in sarcoma cells. (A to F) RPPA quantification of proteins involved in metabolic and mitochondrial function, including ACSS2 (A), ATP5A (B), phosphorylated PLC-γ (pPLC-γ; C), IRS2 (D), phosphorylated BAD (pBAD; E), and FAS (F). Means ± SD; each dot represents one biological replicate. (G) Cellular ROS levels were significantly increased when Lpar4 silencing was combined with abemaciclib treatment (shLpar4 Abema) compared with abemaciclib treatment alone (shSCR Abema). Quantification is shown on the left, whereas representative flow histograms are shown on the right. FITC, fluorescein isothiocyanate. Means ± SD. Student’s t test. Each dot represents one biological replicate. (H) mtSOX levels measured were significantly increased when Lpar4 silencing was combined with abemaciclib treatment (shLpar4 Abema) compared with abemaciclib treatment alone (shSCR Abema). Quantification is shown on the left, whereas representative flow histograms are shown on the right. Means ± SD. Each dot represents one biological replicate. Student’s t test. (I) Cellular ROS levels following abemaciclib treatment were significantly lower in Lpar4-overexpressing sarcoma cells (Lpar4-OE Abema) compared to Lpar4 wild-type cells (CTR Abema). Quantification is shown on the left, whereas representative flow histograms are shown on the right. Means ± SD. Student’s t test. Each dot represents one biological replicate. (J) mtSOX levels were significantly lower in Lpar4-overexpressing sarcoma cells (Lpar4-OE Abema) compared to Lpar4 wild-type cells (CTR Abema). Quantification is shown on the left, whereas representative flow histograms are shown on the right. Means ± SD. Each dot represents one biological replicate. Student’s t test. (K) Schematic highlighting the protective function of LPAR4 in restraining abemaciclib-induced IFN-I responses, cellular ROS accumulation, and mitochondrial stress [Created in BioRender. J. Xiao (2026) https://BioRender.com/rr0karh]. For all the panels in (G) to (J), results are representative of two or more independent repetitions; *P < 0.05; **P < 0.01; ****P < 0.0001. Together, these data demonstrate that loss of Lpar4 disrupts mitochondrial integrity and stress adaptation, resulting in reduced metabolic flexibility and enhanced oxidative stress. To further investigate this, we measured total cellular ROS and mitochondrial superoxide (mtSOX) in Lpar4-silenced sarcoma cells following abemaciclib treatment, using fluorogenic indicators CellROX and MitoSOX, respectively. Both ROS and mtSOX levels were markedly elevated upon combined Lpar4 silencing and abemaciclib exposure (Fig. 5, G and H); this was verified with a second shRNA against Lpar4 (shLpar4-2; fig. S5, A and B). Correspondingly, mitochondria displayed enlarged morphology, consistent with dysfunctional, ROS-releasing mitochondria (fig. S5C). Conversely, Lpar4 overexpression reduced cellular ROS and mtSOX accumulation, confirming its protective role (Fig. 5, I and J). Together, these findings suggest that, in the absence of Lpar4, abemaciclib-induced IFN signaling is further amplified, leading to mitochondrial dysfunction and ROS accumulation (Fig. 5K). Having established that Lpar4 mitigates stress accumulation in tumor cells, we next investigated its functional contribution to tumor cell survival and growth. Given the elevated levels of ROS and mtSOX observed upon combined Lpar4 silencing and abemaciclib treatment, we hypothesized that, in the abemaciclib treatment setting, Lpar4 silencing would further increase treatment-induced apoptosis. Consistent with this prediction, annexin V staining revealed that Lpar4-silenced sarcoma cells, compared to Lpar4-proficient cells, exhibited significantly higher levels of apoptosis when treated with abemaciclib in vitro (Fig. 6A and fig. S6A), whereas Lpar4 overexpression reduced apoptotic cell death (Fig. 6C and fig. S6B). These changes in apoptotic susceptibility translated into corresponding differences in tumor cell survival as measured by crystal violet absorbance—Lpar4 silencing sensitized tumor cells to abemaciclib treatment in a dose-dependent manner (Fig. 6B), whereas Lpar4 overexpression promoted their survival (Fig. 6D). Lpar4 knockdown combined with abemaciclib synergistically suppresses tumor and prolongs survival. (A) Lpar4 knockdown combined with abemaciclib treatment (shLpar4 Abema) significantly increased the proportion of annexin V+ apoptotic cells compared with abemaciclib-treated scramble control (shSCR Abema). Means ± SD; n = 5. Student’s t test. Data are representative of three independent experiments. (B) Lpar4 silencing sensitized Ccne1 cells to abemaciclib treatment compared with scramble control cells. Means ± SD; n = 3; Student’s t test. Data are representative of four independent experiments. (C) Forced expression of Lpar4 in Ccne1 cells (Lpar4-OE) significantly reduced the abemaciclib-induced proportion of annexin V+ apoptotic cells. CTR, empty vector control. Means ± SD. Each dot represents a biological replicate. Student’s t test. Data represent pooled results from four independent repeats. (D) Ccne1 cells overexpressing Lpar4 (Lpar4-OE) exhibited reduced sensitivity to abemaciclib relative to empty vector controls (CTR). Means ± SD; n = 3; Student’s t test. Data are representative of three independent experiments. (E) Lpar4 knockdown combined with abemaciclib treatment (shLpar4 Abema) led to a significant reduction in Ccne1 tumor volumes compared with abemaciclib-treated scramble controls (shSCR Abema). Each dot represents one mouse. The top panel shows a scheme of the experimental design. (F) Combined Lpar4 inhibition and abemaciclib treatment synergistically prolonged survival in mice. Survival significance was determined using the log-rank test. (G and H) Lpar4 overexpression (Lpar4-OE) diminished the tumor-suppressive efficacy of abemaciclib. The top panel in (G) shows a scheme of the experimental design. The bottom panel in (G) shows tumor volume progression. (H) Weights (in grams) at endpoint. Mann-Whitney U test. Each dot presents one mouse. (I) Schematic of LPAR4-mediated tumor cell survival in the context of abemaciclib treatment and attendant cellular stresses [Created in BioRender. J. Xiao (2026) https://BioRender.com/rr0karh]. For all the panels, *P < 0.05; **P < 0.01; ****P < 0.000. These findings were also validated in vivo; Lpar4 silencing combined with abemaciclib treatment led to a marked reduction in tumor growth and a significant improvement in survival (Fig. 6, E and F, and fig. S6C). Conversely, Lpar4 overexpression abrogated the growth reduction that was originally conferred by abemaciclib in the Lpar4 wild-type (CTR) setting (Fig. 6, G and H). Together, these results demonstrate that Lpar4 promotes tumor cell survival by limiting oxidative and IFN-induced stress, especially in the IFN-rich setting of abemaciclib therapy. Therapeutic targeting of LPAR4, through inhibitors or blocking antibodies, could potentiate the efficacy of abemaciclib and other agents that induce IFN, immunogenic, or metabolic stress in sarcomas (Fig. 6I). DISCUSSION Acute induction of IFN-I by therapies that trigger viral mimicry responses exerts potent antitumor effects through both tumor cell–intrinsic and tumor cell–extrinsic mechanisms (57, 58). At the tumor cell–intrinsic level, sustained IFN-I signaling disrupts mitochondrial function, increases ROS, and perturbs metabolic homeostasis, ultimately overwhelming survival pathways and promoting apoptosis. Extrinsically, IFN-I is a central regulator of tumor immunogenicity, enhancing antigen processing and presentation and promoting cytotoxic CD8+ T cell activation (59–61). These properties have motivated the clinical and preclinical evaluation of IFN-I–inducing strategies, alone or in combination with immunotherapy, across multiple cancer types (36, 62). However, despite the poor responsiveness of STS to current immunotherapies, the therapeutic potential of IFN-I–based approaches in this disease remains largely unexplored. CDK4/6 inhibitors, such as abemaciclib, are promising agents for STS as they target cell cycle deregulation while also inducing viral mimicry through DNMT1 deregulation and DNA hypomethylation, which ultimately increase dsRNAs and IFN-I responses in the TME (36). Consistent with findings in breast cancer (63, 64), abemaciclib increases T cell infiltration in sarcoma and alters the activation state of myeloid cells, the most abundant immune population in the UPS TME (24). Although monocyte recruitment and differentiation into TAMs are often linked to tumor progression, our data support a more nuanced view in which IFN signaling can reprogram these cells toward antigen-presenting or inflammatory phenotypes. A central finding of this study is that, although viral mimicry and IFN-I signaling suppress sarcoma growth, they also impose cellular stress that drives the activation of tumor-intrinsic survival mechanisms. We identify LPAR4 as a key mediator of this adaptive response. Lpar4 is up-regulated following abemaciclib treatment, and elevated LPAR4 levels reduce dsRNA accumulation induced by the therapy, thereby attenuating IFN-driven stress responses. Consistent with this model, loss of LPAR4 amplifies viral mimicry and IFN signaling, exacerbates mitochondrial dysfunction and oxidative stress, and increases apoptosis in vitro, ultimately resulting in delayed tumor growth in vivo. LPAR4 remains poorly characterized in cancer biology, with only one prior study in pancreatic cancer implicating LPAR4 in chemotherapy-induced stress adaptation, where it promoted fibronectin deposition and stemlike features that supported tumor growth (47). Consistent with this report, we find that LPAR4-high sarcoma cells—across both mouse and human models—exhibit elevated expression of ECM-related genes. Our data extend these observations by linking LPAR4 to IFN responses and positioning LPAR4 as a key modulator of tumor cell fate in the context of viral mimicry–inducing therapies. Whether IFN activation and mitochondrial dysfunction are mechanistically coupled downstream of LPAR4, or whether LPAR4 regulates mitochondrial integrity and IFN-I signaling independently, remains an open question. Supporting the former possibility, IFN-I activation has been widely linked to mitochondrial dysfunction and ROS accumulation across cancer types. Persistent activation of the cGAS–STING–IFN-I axis suppresses NRF2 activity, increases oxidative stress, and promotes aerobic glycolysis, contributing to chronic inflammation and tissue pathology (18). Conversely, mitochondrial damage itself can amplify innate immune signaling; stress-induced permeabilization of the mitochondrial membrane can release mitochondrial DNA and dsRNA into the cytosol via BAX/BAK pores, activating RIG-I/MDA5/MAVS, PKR, and cGAS-STING pathways and further reinforcing IFN-I signaling. Together, these studies suggest a feedforward loop in which oxidative stress and IFN signaling mutually reinforce one another, driving sustained metabolic stress (65). Alternatively, LPAR4 may exert a more direct role in regulating mitochondrial homeostasis. In this respect, other LPARs, including LPAR2 and LPAR3, have been implicated in mitochondrial regulation in multiple myeloma (66) and in noncancer contexts such as Hutchinson-Gilford progeria syndrome (67), respectively. Notably, LPAR3 has been shown to regulate expression of ANT2, a mitochondrial protein critical for maintaining membrane potential and bioenergetic function (67). Although the precise molecular pathways remain to be defined, it is plausible that altered metabolism and heightened IFN signaling are potential, nonmutually exclusive mechanisms underlying the elevated ROS and enhanced RIG-I/MDA5–IFN responses observed upon Lpar4 targeting. Future studies will be required to dissect these mechanisms and determine how LPAR4 integrates metabolic and inflammatory signaling under stress. Beyond its mechanistic role, our findings highlight the functional importance of LPAR4 in buffering IFN-induced stress and enabling sarcoma cell survival during abemaciclib treatment. These results nominate LPAR4 as a potential therapeutic vulnerability in sarcoma. Although selective LPAR4 inhibitors are not currently available, pharmacologic strategies targeting its ligand, LPA, have been explored. However, because LPA signals through multiple receptor subtypes with distinct and sometimes opposing functions, broad inhibition may be suboptimal or associated with toxicity. Our data show that abemaciclib selectively increases LPAR4, but not other LPARs, suggesting that receptor-specific targeting may offer improved therapeutic precision. It will also be important to determine whether LPAR4 functions in a ligand-dependent manner or can signal independently of LPA, as has been suggested in pancreatic cancer (47), and to define whether LPA is primarily derived from tumor cells or the surrounding stroma. Last, defining the contexts in which LPAR4 becomes essential for tumor survival will be critical for its therapeutic translation. Many anticancer therapies induce IFN signaling through viral mimicry (e.g., dsRNA accumulation), and our data indicate that LPAR4 restrains IFN responses by limiting dsRNA levels. This raises the possibility that LPAR4 represents a common mechanism of resistance across diverse treatment modalities. Comparative analyses of patient tumors before and after therapy will be important to determine whether LPAR4 up-regulation constitutes a general adaptive response that promotes tumor survival and relapse. Moreover, although our data show that abemaciclib induces transcriptional up-regulation of Lpar4 in vitro, additional mechanisms may also contribute. For example, Lpar4-high tumor cells may preexist within the tumor population and become selectively enriched following treatment due to increased resistance, thereby driving tumor regrowth. Although we did not directly test this possibility, it will be important to address in future studies. In summary, our study identifies LPAR4 as a central stress-adaptation node that integrates IFN and mitochondrial responses to promote sarcoma cell survival. These findings position LPAR4 as a promising therapeutic target to enhance the efficacy of treatments that induce IFN-I, viral mimicry, and metabolic stress and support the development of rational combination strategies in sarcoma and potentially other malignancies. MATERIALS AND METHODS Cell lines Human sarcoma cell lines, including fibrosarcoma HT1080 (CCL-12), liposarcoma 93T449 (CRL-3043), uterine sarcoma MESSA (CRL-1976), and human embryonic kidney (HEK) 293T (CRL-3216), were obtained from the American Type Culture Collection (ATCC). Human liposarcoma cell lines LPS863, LPS224, Lipo815, and LPS246 were obtained from the MDACC Cytogenetics and Cell Authentication Core. The human UPS cell line UPS271.1 was previously established in the lab of K.E.T. (68). Ccne1 cells were generated in our laboratory as described previously (42, 69). All cell lines were maintained in high-glucose Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% heat-inactivated fetal bovine serum (FBS) (A5256701, Gibco) and 1% penicillin-streptomycin (10378016, Gibco). Mycoplasma negativity was routinely verified (LT07-118, Lonza). Cells were cultured in a humidified incubator at 37°C with 5% CO2. IFN-β enzyme-linked immunosorbent assay For enzyme-linked immunosorbent assay (ELISA), 40,000 Ccne1 cells in 1 ml of culture media were seeded to a 6-well plate. Cells were treated either with vehicle or 500 nM abemaciclib for 96 hours before conditioned media were collected. Conditioned media were spun at 12,000g for 3 min to remove any debris. ELISA for quantifying IFN-β levels in conditioned media was measured and analyzed with the Mouse IFN-beta ELISA Kit (MIFNB0, R&D Systems) following the manufacturer’s protocol. Plasmids and virus production Lpar4 overexpression plasmid pLV-Puro-CMV-mLpar4 was purchased from VectorBuilder. shRNA plasmids for Lpar4 were generated using pLKO.1-hygro (#24150, Addgene) and pLKO.1 puro (#8453, Addgene). Primers for shRNA sequences are listed in Table 1. Lentivirus were packaged by cotransfecting psPAX2 (#12260, Addgene) and pMD2.G (#12259, Addgene) into HEK293T cells with the jetPRIME transfection reagent (101000027, Polyplus). Table 1.Primers for shRNA sequences.Fw, forward; Rv, reverse.shLpar4_Fw1CCGGATGGGTGACAGAAGATTTATTCTCGAGAATAAATCTTCTGTCACCCATTTTTTGshLpar4_Rv1AATTCAAAAAATGGGTGACAGAAGATTTATTCTCGAGAATAAATCTTCTGTCACCCATshLpar4_Fw2CCGGTATTCATCCTGGGTCTAATAACTCGAGTTATTAGACCCAGGATGAATATTTTTGshLpar4_Rv2AATTCAAAAATATTCATCCTGGGTCTAATAACTCGAGTTATTAGACCCAGGATGAATAshSCR FwCCGGAACGTACGCGGAATACTTCGACTCGAGTCGAAGTATTCCGCGTACGTTTTTTTGshSCR RvAATTCAAAAAAACGTACGCGGAATACTTCGACTCGAGTCGAAGTATTCCGCGTACGTT Primers for shRNA sequences. Fw, forward; Rv, reverse. RNA extraction and RT-qPCR Cells were lysed with TRIzol reagent (15596026, Thermo Fisher Scientific), followed by phase separation with chloroform (J67241.AP, Thermo Fisher Scientific). RNA was precipitated with isopropanol (I9516, Sigma-Aldrich) and washed with 70% ethanol (BP2818100, Fisher Scientific). Two micrograms of RNA was used for first-stand cDNA synthesis (4374967, Applied Biosystems). Five to 10 ng of RNA was used for each PCR using Tbp/TBP (TATA box–binding protein) or Actb/ACTB (β-actin) as the internal control. Primers for semiquantitative PCR are listed in Table 2. Gene expression relative to baseline was calculated using the comparative Ct method (ΔΔCt). Table 2.Primers for semiquantitative PCR.Mouse Lpar4 FwAACCTGGCCCTCTCTGATTT Lpar4 RvCGATCGGAAGGGATAGACAA Ifit3 FwGCTCAGGCTTACGTTGACAAGG Ifit3 RvCTTTAGGCGTGTCCATCCTTCC Mx1 FwTGGACATTGCTACCACAGAGGC Mx1 RvTTGCCTTCAGCACCTCTGTCCA Isg15 FwAAGAAGCAGATTGCCCAGAA Isg15 RvTCTGCGTCAGAAAGACCTCA Rig-I(Ddx58) FwAAGGAAAACTGGCCAAAGGT Rig-I(Ddx58) RvTGGTTTCAATGGGCTGTGTA Cxcl10 FwGTGAGAATGAGGGCCATAGG Cxcl10 RvTTTTTGGCTAAACGCTTTCAT Ifi35 FwGATAACCACAGGGTGCTCGT Ifi35 RvCTGGCCAATCTGGCATAAGT Ifitm2 FwCACTCTTCTTCAACGCCTGCTG Ifitm2 RvGGAGCTGATATTCAGGCACTTGG Ifna1 FwGGATGTGACCTTCCTCAGACTC Ifna1 RvACCTTCTCCTGCGGGAATCCAA Ifnb1 FwGCCTTTGCCATCCAAGAGATGC Ifnb1 RvACACTGTCTGCTGGTGGAGTTC Tbp FwCTACCGTGAATCTTGGCTGTAAAC Tbp RvAATCAACGCAGTTGTCCGTGGC Actb FwCATTGCTGACAGGATGCAGAAGG Actb RvTGCTGGAAGGTGGACAGTGAGGHuman MX1 FwGGCTGTTTACCAGACTCCGACA MX1 RvCACAAAGCCTGGCAGCTCTCTA IFIH1 FwGCTGAAGTAGGAGTCAAAGCCC IFIH1 RvCCACTGTGGTAGCGATAAGCAG CXCL10 FwGGTGAGAAGAGATGTCTGAATCC CXCL10 RvGTCCATCCTTGGAAGCACTGCA RIG-I(DDX58) FwCACCTCAGTTGCTGATGAAGGC RIG-I(DDX58) RvGTCAGAAGGAAGCACTTGCTACC IFITM2 FwGGCTTCATAGCATTCGCGTACTC IFITM2 RvAGATGTTCAGGCACTTGGCGGT IFNA1 FwAGAAGGCTCCAGCCATCTCTGT IFNA1 RvTGCTGGTAGAGTTCGGTGCAGA IFNB1 FwCTTGGATTCCTACAAAGAAGCAGC IFNB1 RvTCCTCCTTCTGGAACTGCTGCA ACTB FwCACCATTGGCAATGAGCGGTTC ACT B RvAGGTCTTTGCGGATGTCCACGT TBP FwTGTATCCACAGTGAATCTTGGTTG TBP RvGGTTCGTGGCTCTCTTATCCTC Primers for semiquantitative PCR. Western blot Cells were treated with abemaciclib methanesulfonate (CAS 1231930-82-7, Cayman Chemical) for 72 to 96 hours before collection for Western blot analysis. Cell pellets were lysed with radioimmunoprecipitation assay (RIPA) lysis buffer (89900, Thermo Fisher Scientific), supplemented with protease inhibitor cocktail (11836170001, Roche) and phosphatase inhibitor (78420, Thermo Fisher Scientific). Protein concentrations were determined using the Bradford protein assay kit (5000202, Bio-Rad). Equal amounts of protein were mixed with 1× SDS sample buffer (Bioland Scientific), boiled at 95°C for 15 min, and resolved on gradient NuPAGE Bis-Tris gel (NP0329BOX, Thermo Fisher Scientific) for protein separation. Antibodies used in the study were obtained from Cell Signaling Technology: RIG-I (3743), MDA-5 (5321), β-Actin (4970), PSMB8 (13635), PSMB9 (87667), STAT1(14994), and p-STAT1(9167). dsRNA staining To assess dsRNA levels following abemaciclib treatment, cells were treated with abemaciclib for 96 hours before collection, at doses of 125, 250, or 500 nM. dsRNA detection was performed as previously described (70). Briefly, cells were harvested and then fixed and permeabilized using the Cyto-Fast Fix/Perm Buffer Set (426803, BioLegend) according to the manufacturer’s instructions. dsRNA was probed with 2 μg of SCICONS J2 anti-dsRNA antibody (10010500, Exalpha) diluted in Dulbecco’s phosphate-buffered saline (DPBS) containing 2% FBS for 2 hours at room temperature, followed by staining with Alexa Fluor 488–conjugated goat anti-mouse IgG (H + L) secondary antibody (A32723, Invitrogen/Thermo Fisher Scientific). Crystal violet staining of living cells Cells were seeded into 12-well plates and treated with 125, 250, or 500 nM abemaciclib for a total of 7 days. Fresh abemaciclib was replenished every 2 days. On day 7, cells were fixed with 10% neutral buffered formalin (28600-1, Statlab) for 10 min and subsequently stained with crystal violet (AC405830250, Acros Organics). After drying, the bound crystal violet dye was solubilized using 10% acetic acid (A38-500, Fisher Chemical) and absorbance was measured at 595 nm using the Multiskan SkyHigh Microplate Spectrophotometer. Mitochondria staining Ccne1 cells were seeded 50,000 cell per well into a 6-well plate and treated with or without 125 nM abemaciclib for 72 hours. Cells were detached, seeded to a glass cover slide pretreated with poly-d-lysine (0.1 mg/ml; A003E, Millipore) for 1 hour, and continued to culture for another 24 hours. For mitochondria staining, cells were incubated with 500 nM MitoTracker Deep Red FM (8778, Cell Signaling Technology) for 30 min at 37°C. After incubation, cells were fixed with 100% methanol (MX04751, MilliporeSigma) for 15 min at −20°C, followed by 4′,6-diamidino-2-phenylindole (DAPI) (1 μg/ml; 62248, Invitrogen) in DPBS for 5 min. Slides were then mounted using the ProLong Gold Antifade Mountant (P36930, Invitrogen). Images were acquired using a Nikon Eclipse Ti2 SoRa spinning disk confocal microscope, processed, and analyzed with the NIS-Elements software and QuPath. Image acquisition and analysis Images of mitochondria were acquired using a Nikon SoRa Super-Resolution Spinning Disk equipped with an ORCA-FUSION BT SCMOS camera. A 60× objective was used in combination with the SoRa disk and with additional magnifying lenses of 2.8×. The DAPI signal was collected by exciting the probe with a 405-nm laser, and emission was collected with an ET455/50M filter. The MitoTracker signal was excited with a 640-nm laser, and emission was collected with an ET705/72m emission filter. The quad dichroic mirror used on the excitation/emission pathway was a Nikon part 89101M EM FLTR DAPI/FITC/TRITC/CY5, 25 mm. The exposure time for both channels is 100 ms. Laser powers were 10% for 640 nm and 15% for 405 nm. Flow cytometry detecting MitoSOX and CellROX MitoSOX Green (M36006, Thermo Fisher Scientific) staining was carried out according to the manufacturer’s instructions. Briefly, 50,000 cells were seeded on a 6-well tissue culture plate and treated with abemaciclib (250 nM) for 3 days. Cells were washed with PBS and were disassociated with 0.5% trypsin buffer, then washed with Hanks’ balanced salt solution (HBSS), and centrifuged twice at 1250 rpm for 5 min. Subsequently, cells were stained with 0.5 μM MitoSOX reagent for 30 min at 37°C, protected from light, before flow cytometry analysis. Data were analyzed with FlowJo.v10. CellROX Green (C10492, Thermo Fisher Scientific) staining was carried out according to the manufacturer’s instructions. Briefly, 50,000 cells were seeded on a 6-well tissue culture plate and treated with abemaciclib (250 nM) for 3 days. Cells were washed with PBS and were disassociated with 0.5% trypsin buffer, then washed with PBS, and centrifuged twice at 1250 rpm for 5 min. Subsequently, cells were stained with 5 μM CellROX reagent for 45 min at 37°C, protected from light, before flow cytometry analysis. Data were analyzed with FlowJo v10. RPPA analysis For RPPA analysis, Ccne1 shSCR and shLpar4 cells were treated with or without 125 nM abemaciclib for 72 hours. Protein lysates were prepared and diluted in a twofold, five-point serial dilution series (from undiluted to 1:16) and then spotted onto nitrocellulose-coated slides. Samples were printed in a 4 × 12–pin–11 × 11 subarray format, accommodating 1056 serially diluted samples along with replicate vertical controls on each slide. Target proteins were detected by probing with specific primary antibodies, followed by signal amplification using a tyramide amplification system, and visualization by diaminobenzidine (DAB) colorimetric reaction. Each slide was stained with a single antibody using a GenPoint-based staining kit (Agilent). Digital slide images were acquired with a Huron TissueScope scanner, generating 16-bit TIFF files. Spot intensities from the TIFF files were quantified using the Array-Pro Analyzer software. Mean net intensities from all 1056 arrayed samples were used to perform curve fitting for each slide using the RPPA Super Position and Concentration Evaluation (SPACE). Relative protein abundance for each sample was determined by converging its dilution series to a single EC50 (median effective concentration) and interpolating the corresponding value in RPPA SPACE. Annexin V staining Cells were treated with abemaciclib for 96 hours before analysis. Both adherent cells and cells in the culture medium were collected for annexin V staining. Staining was performed according to the manufacturer’s instructions (640926 and 640930, BioLegend), followed by flow cytometry analysis. Animal studies All animal studies were conducted in accordance with protocols approved by the Institutional Animal Care and Use Committee (IACUC) at Cedars-Sinai Medical Center (IACUC protocol 008118) and MD Anderson Cancer Center (IACUC protocol 00002526). Six-week-old C57BL/6 (000664), Ifnar1−/− [B6(Cg)-Ifnar1tm1.2Ees/J, 028288], and Mx1gfp (B6.Cg-Mx1tm1.1Agsa/J, 033219) mice were obtained from the Jackson Laboratory. For Ccne1 tumors, female mice were inoculated either with 150,000 Ccne1 cells using three-dimensional (3D) scaffolds (CS1-0502-25, DSM Biomedical) as previously described (42) or subcutaneous injection with 750,000 cells combined with Matrigel. Tumor volumes (V) were measured using calipers to determine the tumor length (L) and width (W) and calculated using the formula V = 0.5 × W2 × L. When tumors reached an average volume of ∼100 mm3, mice were randomized into treatment groups and administered abemaciclib methanesulfonate (HY-16297, MedChemExpress) at 75 mg/kg daily by oral gavage. The study endpoint was defined as either tumor volume exceeding 1000 mm3 or ulceration, whichever occurred first. Except for survival study (Fig. 5F) and lymphocyte flow cytometry (fig. S1E), all tumors were harvested when the control group reached endpoint volume and subjected to flow cytometry analysis or fixed for immunohistochemical staining. For survival study and lymphocyte study, tumors were harvested upon individually exceeding 1000 mm3 (or ulceration). Tissue processing and staining Harvested tumors were fixed in 10% formalin for 48 hours and then transferred into 70% ethanol before paraffin embedding. Tumor tissues were sectioned at 5 μM. Slides were deparaffinized at 65°C for 20 min, followed by rehydration and antigen retrieval with tris-EDTA or citrate buffer. For CD3 IF staining, tyramide signal amplification was performed using OPAL reagents (Akoya Biosciences). Primary anti-CD3 antibody (ab16669, Abcam) was diluted to 1:200 and incubated for 1.5 hours, followed by incubation with horseradish peroxidase (HRP)–conjugated polymer secondary antibody (Abcam) 15 min. After washing, slides were incubated with OPAL fluorophores (OPAL 520) at 1:200 dilution for 10 min at room temperature. Tissue sections were dehydrated and mounted with the ProLong Gold Antifade Mountant with DAPI (P36931, Invitrogen). Images were acquired by an ECHO Revolution upright microscope equipped with an sCMOS Mono camera and analyzed with QuPath. For DAB immunohistochemical staining of LPAR4 in patients with DDLPS, slides were treated with 3.7% hydrogen peroxide (L14000AP, Thermo Fisher Scientific) for 15 min before antigen retrieval with citrate buffer. Primary anti-LPAR4 antibody (HPA054951, Sigma-Aldrich) was diluted 1:45 and incubated overnight at 4°C in a humidified chamber. HRP-conjugated secondary antibody (Abcam) was applied for 30 min at 4°C, followed by DAB substrate development (ab64238, Abcam) for 5 min. Slides were counterstained with Gill2 hematoxylin diluted 1:10 (6765007, Epredia) for 7.5 min, cleared in 1% acetic acid, and blued with a bluing reagent (6769001, Epredia). Last, slides were dehydrated through graded ethanol and cleared in xylene and mounted (SP15-100, Fisher Chemical). Flow cytometry Flow analysis was performed with Symphony A5 (BD, Pharmingen) and CytoFLEX LX (Beckman Coulter) instruments. Cell sorting was carried out on FACSAria III (BD, Pharmingen) and CytoFLEX SRT (Beckman Coulter) systems. FlowJo (V10) was used for data analysis. Cells were stained with 1 μl of an antibody per 100 μl of staining buffer (DPBS with 2% FBS) for 15 min at room temperature. Antibodies used in this study were purchased from BioLegend: CD45 (30-F11), CD11b (M1/70), CD4 (GK1.5), CD8a (53-6.7), NK1.1 (S17016D), Ly6C (HK1.4), Ly6G(1A8), and F4/80 (BM8). Dead cells in the samples were excluded using Ghost Dye Red 710 (SKU 13-0871-T100, Cytek). For tumor dissociation, harvested tumors were minced into small pieces and digested in collagenase II (1 mg/ml; 17100015, Gibco) in staining buffer for 1 hour at 37°C with shaking. Single-cell suspensions were obtained by passed digested tissue through 70-μm strainers (Bioland Scientific LLC, Paramount, CA). Red blood cells were removed with ACK lysis buffer (A1049201, Gibco). scRNA-seq sequencing For scRNA-seq experiments, tumors were dissociated using the Mouse Tumor Dissociation Kit (130-096-730, Miltenyi Biotec) in conjunction with the gentleMAC Octo Dissociator with Heaters (Miltenyi Biotec). The resulting cell suspensions were filtered through 70-μm strainers, and red blood cells were removed using ACK lysis buffer. Dead cells were labeled with Ghost Dye Red 710, after which samples were hashed using antibody-conjugated oligo barcodes (TotalSeq-B hashtag reagent, BioLegend). Live cells were then isolated by flow sorting for downstream library preparation. scRNA-seq library preparation and sequencing Sorted live cells were processed using the Chromium Next GEM Single Cell 3′ v3.1 kit (10x Genomics) according to the manufacturer’s protocol. Gel bead-in-emulsion (GEM) were generated on a Chromium Controller, followed by reverse transcription, cDNA amplification, and library construction. Hashtag oligo (HTO) libraries were prepared in parallel. Final libraries were pooled and sequenced at the Applied Genomics, Computation & Translational Core at Cedars-Sinai Medical Center. scRNA-seq data analysis scRNA-seq data were analyzed as previously described (42). Briefly, gene expression counts matrices were obtained by aligning and counting FASTQ reads with the Cell Ranger Count pipeline (10x Genomics). For mouse tumor identification, a modified reference genome that includes the sequence for the red fluorescent protein (RFP) transgene was used for read alignment. Resultant gene counts were preprocessed in Seurat v4 (71) along with R packages for ambient RNA filtering [SoupX (72)] and for doublet detection [scDblFinder (73) and Seurat::HTODemux]. Cells with low gene complexity, high mitochondrial transcript percentage, or extreme unique molecular identifier (UMI) counts (indicative of dead or poor-quality cells or potential doublets) were removed, yielding a total of 16,677 cells passing quality control. Data were scaled with SCTransform, and variable features were identified using the FindVariableFeatures function. The dataset was then scaled and subjected to principal components analysis (PCA) for dimensionality reduction, followed by clustering with a shared nearest-neighbor (SNN) graph. Uniform manifold approximation and projection (UMAP) embeddings were generated for visualization, and cluster markers were identified using the FindMarkers function. Cell clusters were initially annotated on the basis of marker genes as follows: sarcoma cells (RFP), monocytes (Ly6c2) and macrophages (Fcgr2b and Csf1r), granulocyte/MDSC (myeloid-derived suppressor)–like cells (Hdc and S100a8), mast cells (Hdc and Gata2), DC and B cells [Cd74 with Ccr7, Siglech, or Ms4a1 (encoding CD20)], NK/T cells (Cd3e or Nkg7), endothelial cells (Pecam1 and Cdh5), and fibroblasts (Dpt, Lum, and Thy1). Last, tumor cell clusters and monocytes/macrophages were selected for further subclustering; subclusters were annotated as described in the text. Bulk RNA-seq and analysis For bulk RNA-seq analysis of mouse UPS, Ccne1 shSCR and shLpar4 cells were cultured with or without 125 nM abemaciclib for 72 hours before harvest. Total RNA was extracted using the TRIzol method. RNA samples were submitted to Novogene for mRNA library preparation using poly-A enrichment, followed by sequencing on the NovaSeq X Plus platform (PE150). Bulk RNA-seq FASTQ files were first subjected to quality control using FastQC to assess read quality, adapter contamination, and overrepresented sequences. Adapter trimming and removal of low-quality bases/reads were then performed (e.g., with Trim Galore or an equivalent tool), and only high-quality reads were retained for downstream analysis. Cleaned reads were pseudoaligned and quantified at the transcript level using Salmon (74) in quasimapping mode with default settings. Transcript-level abundances were imported into R and summarized to the gene level using the tximport package (75). Differential expression analysis was carried out with DESeq2 (76). Lowly expressed genes were filtered before modeling, and differentially expressed genes were identified with a negative binomial generalized linear model. P values were adjusted for multiple testing using the Benjamini-Hochberg procedure to control the false discovery rate. Differentially expressed genes (adjusted P value < 0.1) were used as input for overrepresentation analysis (ORA) using the R package clusterProfiler (77). For bulk RNA-seq analysis of human UPS, gene expression and sample metadata from TCGA SARC (2) was processed as above. We identified genes whose expression differs between patients with LPAR4-high (z-scored expression > 1) and LPAR4-low (z-score < −1), controlling for tumor subtype (ONCOTREE_CODE). Differentially expressed genes were used for ORA. Statistics Statistical analyses were performed using GraphPad Prism. Data are presented as means ± SD or means ± SEM, as specified in the figure legends. Two-tailed Student’s t tests were used when the compared groups contained equal sample sizes, whereas Mann-Whitney U tests were applied for unequal sample sizes or nonparametric comparisons. Survival analyses were conducted using the log-rank (Mantel-Cox) test. Significance was defined as follows: *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.