Stabilizing Tumor-Resident Mast Cells Restores T-Cell Infiltration and Sensitizes Sarcomas to PD-L1 Inhibition

Harnessing the immune system to fight cancer has led to significant clinical benefits in many tumor types. The most successful example of cancer immunotherapy is the use of immune-checkpoint inhibitors (ICI). However, ICI responses are limited to patients whose tumors are infiltrated by a sufficiently diverse repertoire of cytotoxic CD8⁺ T cells (1, 2). Apart from lymphocyte infiltration, other features of the tumor microenvironment (TME) are likely to obstruct antitumor immune responses and lead to treatment failure in sarcomas (3, 4). In fact, cancer-associated fibroblasts (CAF), which is the most common nonmalignant cell type in many aggressive solid tumors, can produce and maintain elevated levels of extracellular matrix (ECM) components, including collagen I and hyaluronan. Accumulation of these structural components causes the stiffening of the tumor and abnormal elevation of mechanical forces that compress tumor vessels, leading to their collapse (5–9). Notably, in tumors with an abundance of ECM, 95% of intratumoral blood vessels might be compressed and up to 80% are totally collapsed and lacking perfusion (10). Hypoperfusion, in turn, can induce hypoxia, resulting in increased immunosuppressive signaling, which hampers antitumor immunity (11–13). Many types of sarcomas are typical examples of these TME abnormalities, whereas restoring TME function to normal levels (i.e., TME normalization) has been proven to be an effective strategy to potentiate immunotherapy in stroma-rich tumors, such as breast and pancreatic cancers (10, 14–16). The effect of normalization strategies in sarcoma tumors has not been studied yet (3).

Indeed, current TME normalization strategies utilizing drugs that directly reprogram CAFs, such as TGFβ inhibitors, can overcome the ECM barrier to immune cell infiltration and enhance the delivery of ICIs (14, 15, 17–21). Nevertheless, resistance to immunotherapy is given not only by the inability of immunotherapy to reach and eliminate cancer cells but also by the immunosuppressive microenvironment which dampens the activation of cytotoxic T cells and prevents immunotherapy from establishing a lasting immune protection against tumor recurrence. This challenge calls for a sophisticated harnessing of immune cells, such as the tumor-resident mast cells (MC), whose biological function can be programmed. MCs infiltrate the TME with a not yet fully developed or matured phenotype that can be influenced by the surrounding milieu (22) to play a protumorigenic (23, 24) or an antitumorigenic role (25), which can involve interaction with CAFs (26). Upon activation, MCs release secretory granules containing mediators that influence adaptive immune responses, ECM remodeling, and angiogenesis (27–31). Increasing evidence suggests that human MCs release VEGF-A and IL6 in response to hypoxia (32, 33), which favors immunosuppression and poor response to immunotherapy (34–36). Thus, targeting MCs with clinically administered drugs such as antihistamines, MC stabilizers (e.g., ketotifen, cromolyn), or other MC-targeting drugs represents a promising approach to overcome resistance to current cancer treatment modalities (37–39). The functional plasticity of MCs makes them key orchestrators of the immune TME, and their inhibition may hold the key to breaking tumor resistance to immunotherapy.

In this study, we aimed to explore the role of tumor-resident MCs in modulating antitumor responses either as a core component of the host immune system or via their interaction with CAFs. MC–fibroblast interactions are well characterized in the context of wound healing and pulmonary fibrosis. In these settings, MCs are in close proximity to fibroblasts acting as potent regulators of fibroblast proliferation and collagen synthesis (40, 41). Nonetheless, their dynamic interaction with CAFs in tumors is not well studied. Understanding the dynamic cross-talk between MCs and CAFs could lead to new therapeutic strategies that reprogram MCs and CAFs to quiescent phenotypes, reducing levels of ECM, angiogenesis, and immunosuppression.

Here, we hypothesize that reprogramming of MCs can normalize tumor stiffness and perfusion in murine sarcoma models through their interaction with CAFs. Specifically, we investigate the effects of ketotifen, a second-generation histamine H1 blocker and mast cell stabilizer, on modulating the TME. We demonstrate that ketotifen can reprogram MCs toward effectively suppressing CAF-induced stiffness elevation, leading to increased perfusion in murine models of soft-tissue and bone sarcoma. Through this unexplored mechanism of MC and CAF reprogramming, we show that ketotifen's combination with chemotherapy and PD-L1 blockade overcomes the primary resistance of these tumor models to immunotherapy, resulting in extended survival and immunologic memory. Furthermore, as part of a phase II clinical trial (EU clinical trials register, EudraCT #: 2022-002311-39), we show that administration of ketotifen in patients with sarcoma recapitulates the reduction in tumor stiffness observed in the tumor mouse models for different sarcoma subtypes.

The MCA205 mouse fibrosarcoma cell line was purchased from Millipore (SCC173, Millipore) and cultured in expansion medium consisting of Roswell Park Memorial Institute medium (RPMI-1640, LM-R1637, Biosera) containing 2 mmol/L L-glutamine (TMS-002-C, Sigma), 1 mmol/L sodium pyruvate (TMS-005-C, Sigma), 10% fetal bovine serum (FBS, FB-1001H, Biosera), 1× nonessential amino acids (TMS-001-C, Sigma), 1% antibiotics (A5955, Sigma), and 1× β-mercaptoethanol (ES-007-E, Sigma). The K7M2 mouse osteosarcoma cell line (CRL283) and NIH3T3 fibroblast cell line (CRL-1658) were obtained from ATCC and cultured in Dulbecco's Modified Eagle's Medium (DMEM, LM-S2041, Biosera) supplemented with 10% FBS and 1% antibiotics. The MC/9 mast cell line was purchased from ATCC (CRL-830, ATCC) and cultured in RPMI medium supplemented with 10% FBS, 1% antibiotics, 1× β-mercaptoethanol, 100 ng/mL stem cell factor (SCF, 250–03, PeproTech), and 10 ng/mL IL3 (213–13, PeproTech). All cells were maintained in humidified incubators at 37°C with 5% CO₂ and routinely tested and confirmed to be free of mycoplasma contamination.

CAFs were isolated from MCA205 fibrosarcoma tumors with a size of 200 mm³. Briefly, resected tumors were washed in 1× PBS, finely minced, and incubated in Liberase solution for 30 minutes at 37°C. Enzymatic activity was ceased with a complete culture medium and cell suspension was passed through a 40-μm cell strainer to obtain single cells. Filtrate was centrifuged at 300 × g at 4°C for 5 minutes, the supernatant was removed completely, and the cell pellet was resuspended in a complete growth medium and transferred into a cell culture vessel at 37°C/5% CO₂. The cells attaching to the vessel bottom within 20 minutes were recognized as CAFs, and the floating cells were abandoned. This isolation process was repeated 3 times to get purified CAFs.

Ketotifen fumarate salt (K2628, Sigma) was dissolved in sterile normal saline (0.9% NaCl in ddH₂O, w/v). Mouse monoclonal antibody anti–PD-L1 (B7-H1, clone 10F.9G2, RRID: AB_10949073) and rat IgG2b isotype control, antikeyhole limpet hemocyanin (LTF-2, RRID: AB_1107780) were purchased from Bio X Cell and dissolved in the recommended InVivoPure pH 7.0 Dilution Buffer (IP0070). Doxorubicin hydrochloride was obtained from Nicosia General Hospital as a ready-made solution of 2 mg/mL. Anti–PD-L1 was administered via intraperitoneal injection at a dose of 10 mg/kg (42) and doxorubicin at a dose of 5 mg/kg (14).

Six-week-old C57BL/6OlaHsd (RRID:MGI:5656288) female mice were injected in the leg muscle with 2.5×10⁵ MCA205 fibrosarcoma cells. The osteosarcoma syngeneic tumor model was generated by orthotopic implantation of 1×10⁶ wild-type K7M2 cells in the proximal tibia of 6-week-old BALB/cOlaHsd (RRID:MGI:5653315) female mice as previously described (43). All mice were maintained in specific pathogen-free conditions in the animal facilities of the Cyprus Institute of Neurology and Genetics. They were housed in a controlled temperature/humidity (22°C/55%) environment on a 12-hour light–dark cycle and kept with free access to food and water throughout the whole experiment period. A sample size identical to that of previous studies, which measured the same output variables, was used. All in vivo experiments were conducted per the animal welfare regulations and guidelines of the European Union (European Directive 2010/63/EE and Cyprus Legislation for the protection and welfare of animals, Laws 1994–2013) under a license acquired and approved (CY/EXP/PR.L2/2018, CY/EXP/PR.L14/2019, CY/EXP/PR.L15/2019, and CY/EXP/PR.L03/2020) by the Cyprus Veterinary Services Committee, the Cyprus national authority for monitoring the welfare of animals in research.

Ketotifen at 1, 5, 10, and 25 mg/kg or an equal volume of diluent (control group) was administered daily by intraperitoneal injection (i.p.) for 8 days. Ketotifen treatment for MCA205 and K7M2 tumor models started when the average tumor volume reached 60 mm³ and 150 mm³, respectively, and was completed at 600 mm³. At the study endpoint, primary tumors were surgically excised and stored in 1× PBS at −80°C until further processing for immunofluorescence staining experiments.

Mice bearing MCA205 tumors were pretreated with daily ketotifen at the dose of 10 mg/kg (found to be optimal in this study) or an equal volume of diluent (control group) before the chemoimmunotherapy combination treatment. Specifically, ketotifen treatment commenced when the average volume of MCA205 tumors was 40 mm³. The first cycle of doxorubicin and anti–PD-L1 combination occurred when tumors reached an average size of 150 mm³ (day 7) and repeated for an additional two cycles with a three-day interval (i.e., on days 10 and 13 after cancer cell implantation). Ketotifen administration continued for an additional two days after the last cycle of combination treatment. Following completion of the study protocol, primary tumors having an average size of 700 mm³ (on day 16) were resected and processed accordingly for the flow cytometry experiment, and survival of mice was monitored daily until the rechallenge experiment.

In the K7M2 tumor model, administration of daily ketotifen started when the average tumor size reached 70 mm³ (day 18) and continued until completion of the study protocol. The first cycle of doxorubicin and anti–PD-L1 combination occurred when tumors reached an average size of 150 mm³ (day 22) and repeated on days 25, 28, and 31. The study ended when tumors reached a mean volume of 550 mm³ (day 33). Mice were sacrificed and tumors were collected for immunofluorescence staining experiments. Planar dimensions (x, y) of the tumor were monitored every 2 to 3 days using a digital caliper, and tumor volume was estimated from the volume of a sphere with a diameter equal to the average of planar dimensions.

Mice were considered long-term survivors if no tumor was detected 80 days after the surgical resection of MCA205 primary tumors. Five long-term survivors from each treatment group were rechallenged with a subcutaneous injection of 2.5 × 10⁵ MCA205 cells at the flank region. Naïve mice of the same age were injected in parallel to serve as controls.

The clinical study of ketotifen (EudraCT #: 2022-002311-39) was approved in the European Union/European Economic Area under the Clinical Trials Directive 2001/20/EC and conducted in accordance with the Declaration of Helsinki regarding ethical principles for medical research involving human subjects. A written informed consent form was obtained from each participant included in the study and the study was approved by the Cyprus National Bioethics Committee (ΕΕΒΚ/ΕΠ /2022/36) and the Cyprus Pharmaceutical Services (ΚΔ 3/22).

Ultrasound measurements of elastography from patients were collected as a component of an ongoing proof-of-concept phase II clinical trial (EudraCT #: 2022-002311-39), which aims to explore the impact of ketotifen on reducing tumor stiffness and improving the efficacy of chemotherapy in patients with sarcoma. Written informed consent was obtained from all participants.

Patients with different sarcoma subtypes were subjected to shear wave elastography (SWE) imaging before initiating ketotifen treatment to establish the baseline stiffness of the tumor. Each imaging session involved capturing five SWE images at different planes, and the average value of these images was considered. Subsequently, ultrasound imaging was conducted at regular intervals to monitor potential changes in stiffness induced by ketotifen administration.

Statistical analysis was performed using GraphPad Prism 9.0 software (RRID:SCR_002798). For tumor growth curves and ultrasound measurements involving multiple time points, two-way analyses of variants (ANOVA) followed by Dunnett multiple comparison were used. For other experiments including toxicity assessment, histology, and flow cytometry, one-way ANOVA or unpaired parametric Welch t test was used depending on the number of groups being compared. Only statistically significant differences along with the exact P values are displayed in the figures. A P value of ≤0.05 was considered statistically significant. Sample size and repetition of each experiment are indicated in corresponding figure captions. Sample size was based on previously published studies measuring the same output variables. Data are presented as means with standard errors.

All other data supporting the findings of this study are available from the corresponding author and the clinicians responsible for running the clinical study upon request due to patient privacy. Additional methods used in the study can be found in Supplementary Data.

The presence of MCs in sarcoma tumors was confirmed by CD117 (c-Kit) and tryptase immunofluorescence staining of human tissue microarrays. Tryptase is a serine protease unique to MCs with a reported role in myofibroblast proliferation and collagen synthesis (41, 44, 45). MCs were found in all sarcoma subtypes tested (Supplementary Fig. S1A) and they were at a closer distance to myofibroblasts in tumor tissue as compared with healthy tissue (Fig. 1A). To this end, we sought to investigate the consequences of MC activation on myofibroblasts by using a coculture assay whereby murine MC/9 MCs and NIH3T3 fibroblasts were separated by a transwell chamber with micropores that only allowed chemical communication (Fig. 1B). In this system, we determined MC degranulation efficiency by measuring the concentration of β-hexosaminidase (β-hex) released in the supernatant (46). MC degranulation was induced through exposure to compound 48/80 (C48/80), which binds to the MRGPRB2 receptor (the orthologue of the human G-protein–coupled receptor MRGPRX2; Supplementary Fig. S2A; ref. 47). Transforming growth factor β (TGFβ) was used as a fibroblast activator due to its capacity to induce the myofibroblast phenotype of NIH3T3 cells (48). MC/9 cells were cultivated with NIH3T3 cells for 24 hours and then treated with either ketotifen or TGFβ for an additional 24 hours in a low serum medium, followed by a 30-minute sensitization with C48/80. We found that ketotifen inhibits degranulation in a dose-dependent manner, with the concentration of 100 μg/mL causing more than 50% reduction in β-hex release compared with C48/80 treatment (Fig. 1C; Supplementary Fig. S2B and S2C). The presence of TGFβ in the coculture did not affect MC degranulation (Fig. 1C). To further explore the MC–fibroblast interaction, we monitored the differentiation process of NIH3T3 cells in the coculture system by assessing the levels of αSMA and Collagen I, which are two markers of the myofibroblast phenotype (Fig. 1D–F). A significant increase in both αSMA and Collagen I protein expression occurred only in the presence of both TGFβ and activated MCs, whereas no change was observed in the presence of TGFβ alone. On the contrary, ketotifen impaired this myofibroblast differentiation (Fig. 1D–F). The effect of MCs on myofibroblast activity was further confirmed in CAFs isolated from murine fibrosarcoma tumors (MCA205) and cocultured with MC/9 cells (Supplementary Fig. S1B). RNA isolation from CAFs and assessment of fibrosis-related genes, such as Col3A1 and CTGF, indicated a significant upregulation in transcript levels upon exposure to MCs, whereas the addition of ketotifen inverted this effect. MCs, however, did not have any impact on regulating the expression of αSMA (ACTA2) and collagen I (Col1A1; Supplementary Fig. S1C–S1F). Therefore, our findings demonstrate that ketotifen can directly affect CAF function by antagonizing H1 receptor binding and indirectly by stabilizing MCs and impeding their degranulation.

Following the confirmation of the presence of MCs in human sarcoma samples and the in vitro characterization of MC–fibroblast interaction, we hypothesized that inhibition of MC degranulation would normalize ECM in murine sarcoma models. We assessed the impact of MC degranulation by treating MCA205 fibrosarcoma and K7M2 osteosarcoma tumors with ketotifen. Ketotifen was administered in four different doses (i.e., 1, 5, 10, and 25 mg/kg), once a day by intraperitoneal injection for a period of 8 days (refs. 49–51; Fig. 2A). Although no dose exhibited an antitumor effect (Supplementary Fig. S3A–S3D; ref. 52), the highest dose caused splenomegaly (Supplementary Fig. S4). To confirm the effect of ketotifen on MC degranulation, we stained ketotifen-treated tumors for tryptase (41, 44, 45) and found reduced levels after treatment with 10 mg/kg ketotifen in both sarcoma models (Supplementary Fig. S3E–S3G).

We also monitored the effect of ketotifen on tumor stiffness in vivo using ultrasound SWE (Fig. 2B). We performed SWE measurements before, after 3 days (for MCA205) or 4 days (for K7M2) of ketotifen treatment, and upon study completion. Our results demonstrate that the 10 mg/kg dose reduced tissue stiffness most effectively in MCA205 fibrosarcoma tumors, with elastic modulus values reaching 20 kPa (Fig. 2C). K7M2 osteosarcoma tumors similarly benefited from 10 mg/kg of daily ketotifen (Fig. 2D). The pertinent reduction in tumor stiffness was further validated ex vivo at the microscopic tissue scale using atomic force microscopy (AFM) of tumor samples. AFM analysis indicated a reduction in the average value of elastic modulus of K7M2 tumors (Supplementary Fig. S3H), with a shifted distribution toward lower values, which suggests a decrease in collagen density (Supplementary Fig. S3I; ref. 53). Notably, the effect of ketotifen is transient as 14 days following the last day of ketotifen administration, the tumor acquires the stiffness level of the control group (Supplementary Fig. S3J).

Next, we sought to assess the clinical value of ketotifen in reprogramming the TME and reducing stiffness. To achieve this, we analyzed preliminary data from the ongoing phase II clinical trial investigating the use of ketotifen to augment therapy in sarcoma patients. A total of 8 patients with various sarcoma subtypes, including epithelioid leiomyosarcoma, myxofibrosarcoma, atypical lipomatous tumor, leiomyosarcoma, pleomorphic leiomyosarcoma, synovial sarcoma, and desmoid tumor, underwent initial ultrasound imaging before ketotifen treatment to establish the baseline tissue stiffness. Subsequently, elastography measurements were taken throughout the course of ketotifen administration for patients who either received systemic therapy or patients who underwent surgery alone (Fig. 2E).

Ketotifen was administered at the recommended dose of 1 mg twice daily to all patients. Notably, in all experimental cases, tumor stiffness significantly decreased upon ketotifen treatment, even though the baseline tissue stiffness was notably lower compared with the mouse studies (Fig. 2F–G).

As tissue stiffening is a consequence of collagen abundance in the intratumoral space (54), and based on our in vitro study on MC–fibroblast cross-talk, we hypothesized that the activity of CAFs will also be affected by ketotifen intervention, either directly or indirectly. We found that only the concentration of 10 mg/kg ketotifen reduced both CAF density and proliferation, as measured by the expression of αSMA and Ki67, respectively (Fig. 3A and B; Supplementary Fig. S5A and S5B). We observed that the administration of the highest dose of ketotifen did not affect the proliferation of CAFs. This suggests that the tumor may trigger secondary responses in an effort to compensate for and maintain the fibrotic stroma. The reduction of αSMA was also confirmed on transcript level along with a decrease in collagen mRNA and protein levels (Fig. 3C–F). Besides collagen, CAFs synthesize hyaluronic acid, suggesting that changes in CAF activation and proliferation could result in inhibition of hyaluronan production. To address this, we performed immunofluorescence staining for hyaluronan binding protein 1 (HABP1), which demonstrated a suppression in hyaluronan levels at the optimum dose of ketotifen (Fig. 3G and H; Supplementary Fig. S5C and S5D).

Increased tumor stiffness is indicative of the accumulation of intratumoral mechanical forces that compress blood vessels and, thus, induce hypoperfusion (55–57). There is also evidence that MCs are angiogenic mediators, as they increase vascular permeability (58, 59) and neovascularization (60, 61), which could also reduce blood flow. Accordingly, we hypothesized that MC inhibition could restore the function of tumor blood vessels by both reducing vessel permeability and decompressing the vessels. Indeed, administration of ketotifen at 5, 10, and 25 mg/kg increased the pericyte coverage of the tumor vasculature that fortifies the vessel walls, as indicated by the colocalization of the NG2 pericyte marker and CD31 endothelial protein, compared with the lowest dose and untreated control. Notably, the vascular density was not affected (Fig. 4A–C). Surprisingly, we did not observe any reduction in VEGF transcripts. Instead, we observed an upregulation in IFNγ transcription levels in MCA205 tumors following 10 mg/kg ketotifen treatment (Fig. 4D). IFNγ is an angiostatic protein whose concentration in the TME has been directly related to vessel normalization and reduced vessel leakiness (62–64). Furthermore, the reduction in IFP levels with ketotifen treatment provides another evidence of vascular normalization (Fig. 4E). In line with this, ketotifen increased the fraction of vessels with open lumen, a proof of vessel decompression (Fig. 4F).

Next, we sought to investigate whether restoration of these tumor blood vessel abnormalities was sufficient to reestablish tumor blood perfusion and improve oxygenation. Contrast-enhanced ultrasound imaging shows an increase in the area of the tumor covered by the contrast agent (Fig. 4G). Our findings indicate that 10 mg/kg ketotifen was the most effective dose at increasing vessel perfusion in both fibrosarcoma and osteosarcoma tumors (Fig. 4H and I). Similarly, by measuring the intratumoral levels of hypoxia using pimonidazole, we observed that 10 mg/kg ketotifen alleviated hypoxia in both tumor models (Fig. 4J–L). We conclude that ketotifen at a daily dose of 10 mg/kg can effectively reprogram MCs and CAFs, normalizing both the tumor ECM and vasculature and thus improving perfusion and oxygenation.

Subsequently, we set out to investigate the ability of ketotifen to improve chemoimmunotherapy in the same sarcoma models. To mimic the clinical therapeutic approach for patients with sarcoma, we used doxorubicin chemotherapy and the immune-checkpoint inhibitor, anti–PD-L1 antibody, after confirming its expression in tumor tissue (Supplementary Fig. S6A and S6B). First, we pretreated mice bearing MCA205 tumors with daily 10 mg/kg ketotifen for 3 days to restore vessel functionality and thus assist the delivery of anticancer treatment. The anticancer treatment administered was three cycles of 5 mg/kg doxorubicin and/or 10 mg/kg anti–PD-L1, every three days. In the K7M2 study, mice were administered ketotifen for four days, followed by four cycles of chemoimmunotherapy (Fig. 5A). Neither anti–PD-L1 nor doxorubicin monotherapy exhibited an antitumor activity as indicated by the growth curves of MCA205 and K7M2 tumors, whereas combination with ketotifen induced a significant antitumor response (Fig. 5B and C). The greatest decrease in tumor volume was observed when ketotifen was coadministered with the combination of doxorubicin and anti–PD-L1 antibody (Fig. 5B and C).

To support our hypothesis that the increased efficacy of cytotoxic treatment was attributed to ketotifen effects, we compared the average tumor elastic modulus measured with SWE with the average tumor growth rate from the first cycle of chemoimmunotherapy until the end of the experiment in both MCA205 and K7M2 tumor models, and we examined their clustering pattern (65). We used an unsupervised learning methodology to process the data, which separated them into four clusters: a cluster with low elastic modulus and low tumor growth rate (cluster 1), a cluster with low elastic modulus and high growth rate (cluster 2), a cluster with high elastic modulus and low tumor growth rate (cluster 3), and a cluster with high elastic modulus and high tumor growth rate (cluster 4; Fig. 5D; Supplementary Fig. S6C). Clusters 1 and 2 consist solely of mice treated with ketotifen either with or without anticancer therapy. Specifically, ketotifen monotherapy group corresponds to 4% of total mice in cluster 1, whereas in cluster 2, ketotifen monotherapy accounts for 69%, suggesting that ketotifen alone is less likely to cause an antitumor effect, but the combination of ketotifen, doxorubicin, and anti–PD-L1 belongs solely to cluster 1 with a low tumor growth rate. Interestingly, there are cases in cluster 1 that have a negative tumor growth rate. Half of mice comprising cluster 3 received anti–PD-L1 plus doxorubicin combination, whereas most mice in cluster 4 correspond to the control group followed by anti–PD-L1, doxorubicin monotherapies or combination.

Furthermore, a linear regression analysis was performed to generate a model capable of predicting the average tumor growth after a specific treatment or combination of treatments (Fig. 5E; Supplementary Tables S1 and S2). The model is a function of the elastic modulus (E), initial volume at the beginning of treatment ( |${{V}_0}$|⁠), tumor type and treatment ( |${{V}_r}\ = \ a\ E + b\ {{V}_0} + c\ E\ {{V}_0} + d$|⁠), where the parameters a, b|$,$| and c depend on tumor type and d on treatment and tumor type. Parameter d (mm³/day) directly affects tumor growth rate and can be used as a means of comparison for every group that has similar E and |${{V}_0}$|⁠. For instance, parameter d of doxorubicin monotherapy in MCA205 and K7M2 tumor models is reduced by 10% and 21.1%, respectively, compared with the control groups ( |$d\ = $| −162.6 for MCA205, |$d\ = \ 57.7$| for K7M2). Upon comparison of control with anti–PD-L1 monotherapy with regard to parameter d, the changes are minor, with a 2% increase and 7.5% decrease for the MCA205 and K7M2 tumor models, respectively. On the contrary, by comparing parameter d of the ketotifen group ( |$d\ = $| −35.1 for MCA205, |$d\ = $| 62.7 for K7M2) with that of the ketotifen + doxorubicin + anti–PD-L1 group, we observed a 138.4% and 74.36% reduction for MCA205 and K7M2 tumor models, respectively. According to the ANOVA test, the coefficient dd of anti–PD-L1 and doxorubicin are significantly different than zero, whereas the coefficient d of ketotifen alone does not differ from zero. However, the coefficient d of ketotifen + doxorubicin is significantly different than zero, which indicates additive effects. The capability of ketotifen to remodel the tumor stroma and subsequently to promote the efficacy of drugs, such as ICIs, is also evident when measuring the intratumoral delivery of anti–PD-L1 antibodies conjugated with the fluorescence probe Atto 680 (Fig. 5F). K7M2 tumor-bearing mice pretreated with ketotifen exhibit a higher accumulation of anti–PD-L1 in the tumor rather than the control mice.

Subsequently, we investigated if this immune response is preserved against rechallenge of long-term survivors with the same cell line. Mice were considered long-term survivors if no tumor was detected 47 days after the surgical resection of MCA205 primary tumors. Five long-term survivors from each treatment group were rechallenged with a subcutaneous injection of 2.5 × 10⁵ MCA205 cells at the flank region, whereas naïve mice of the same age were injected in parallel to serve as controls. Interestingly, no tumor growth occurred in the ketotifen + doxorubicin + anti–PD-L1 group, suggesting the development of an adaptive memory response, which is triggered upon encountering tumor antigens. On the contrary, all naïve and ketotifen-treated mice developed progressively growing tumors. Likewise, large tumors were formed in three mice of the doxorubicin + anti–PD-L1 group and in four mice of the anti–PD-L1 group. In the ketotifen + doxorubicin and ketotifen + anti–PD-L1 group, only one mouse out of five developed fibrosarcoma (Fig. 5G). In agreement with the hierarchy of efficacy observed in tumor growth measurements, the combination of ketotifen + doxorubicin + anti–PD-L1 demonstrated increased immunologic memory compared with the doxorubicin + anti–PD-L1 group.

Taken together, these findings demonstrate that pretreatment with ketotifen can effectively establish favorable conditions within the TME toward enabling doxorubicin and anti–PD-L1 antibody treatment to confer immunologic memory.

To further evaluate the contribution of ketotifen in stimulating immune responses, we assessed the presence of T-cell populations in MCA205 tumors using flow cytometry. Total T-cell population was primarily defined by the expression of CD3⁺ protein, a pan T-cell marker, and then distinguished into the cytotoxic CD8⁺ or the regulatory T-cell subtype (CD3⁺CD4⁺CD25^hiCD127^loFoxp3⁺; Fig. 6A; Supplementary Fig. S8A). We observed a statistically significant increase in total T-cell recruitment of the doxorubicin + anti–PD-L1 and ketotifen + doxorubicin + anti–PD-L1 treatment groups. In K7M2 tumors, however, only the groups of mice receiving ketotifen prior to cytotoxic treatment exhibited an increase in T-cell infiltration as indicated by immunofluorescence staining for CD3 protein (Supplementary Fig. S7D). Moreover, we found that the ratio of CD8⁺ T cells to immunosuppressive Tregs was significantly increased after combining ketotifen with doxorubicin and immunotherapy (Fig. 6B). Immunofluorescence staining for Ki67 and CD8⁺ confirmed the activation status of recruited CD8⁺ T cells in osteosarcoma tumors (Fig. 6C and D). Considering that ketotifen upregulates IFNγ, which is a key mediator of antitumor immunity with vascular normalization activity (66–69), we hypothesized that accumulation of CD8⁺ T cells achieved by ketotifen combination with either doxorubicin or PD-L1 blockade is the result of tissue reoxygenation. To address this, we stained K7M2 tissue sections for pimonidazole adducts as a measure of hypoxia formation and found that only the groups receiving ketotifen exhibited a significant reduction of hypoxic areas allowing for adequate tumor oxygenation (Supplementary Fig. S7A and S7B). We further explored if restoration of intratumoral T-cell infiltration occurs via an upregulation of immune cell adhesion to the wall of perfused vessels. Ketotifen treatment increased mRNA expression of the immune cell adhesion molecules ICAM and P-selectin (P-sel) (Supplementary Fig. S7C). We confirmed these findings by immunofluorescence staining of CD3 and CD31 endothelial markers demonstrating that all combination treatments of ketotifen exhibited more association between T lymphocytes and endothelial cells, without affecting the endothelial cell area fraction (Supplementary Fig. S7E). Consistent with previous studies highlighting the effect of ICI on improving vascular perfusion (70), we found that anti–PD-L1 monotherapy and doxorubicin combination increase the colocalization of CD3⁺ T cells and CD31 (Fig. 6E and F). Furthermore, as hypoxia fosters the accumulation of myeloid stroma and polarization into suppressive cell populations, such as myeloid-derived suppressor cells (MDSC) and tumor-associated macrophages, we confirmed that disruption of hypoxic zones by ketotifen has an immediate effect on potentiating the antitumor immune responses of doxorubicin and immunotherapy by reducing MDSC density in MCA205 tumors (Supplementary Fig. S8B and S8C). Unlike other studies supporting that doxorubicin monotherapy decreases the number of MDSCs, leading to enhanced numbers of CD4⁺ and CD8⁺ cells (71, 72), here, we did not observe such a reduction and the immunologic benefit of the doxorubicin + ICI combination was relatively modest and did not counterweight that induced upon ketotifen combination.

Sarcomas comprise a heterogenous group of malignancies, of more than 100 different subtypes, arising from mesenchymal tissue. Despite the recent advances in cancer research, the application of immunotherapy in daily clinical practice for the treatment of sarcoma remains limited. Clinical studies have yielded conflicting results demonstrating that blockade of the PD-1/PD-L1 axis can benefit only specific histologic subtypes (4), thus highlighting the need for alternative treatment strategies. Given the role of the TME in determining the behavior and response to treatment, the current study exploited the potential of stabilizing the intratumoral MCs using the common antihistamine drug ketotifen. We primarily demonstrated that ketotifen is capable of inhibiting MC degranulation by stabilizing their membranes and, thus, making them less likely to rupture and release their content. We showed that blocking of granular content can subsequently impact fibroblast phenotype by reducing the expression of αSMA and collagen synthesis as indicated in the coculture experiments. Although the exact mechanisms are not fully elucidated, MCs can influence fibroblast differentiation through the release of various mediators and direct cell-to-cell interactions within the TME. Furthermore, we showed that ketotifen can exert a direct effect on CAFs by antagonizing H1 receptor binding on the cell surface and causing the subsequent inhibition of intracellular signaling events leading to the downregulation of ECM-related genes. In addition to histamine, MCs might release additional mediators, including TGFβ, IL13, and tryptase, which contribute to tumor fibrosis (73). TGFβ can promote fibroblast migration, proliferation, collagen formation, and fibroblast differentiation into CAF. IL13 is also a profibrotic mediator, which can promote fibrosis via TGFβ-dependent and independent mechanisms. Tryptase stimulates fibroblast proliferation and synthesis of collagen and fibronectin via the activation of the protease-activated receptor PAR-2 (74). Although not yet confirmed in the context of tumors, CCL2 also produced by MCs has been reported to attract fibrocytes to regions of injury (75).

Furthermore, we note that in this study, we focused on myofibroblastic CAFs, which are the major contributors to fibrosis, whereas a more thorough analysis comparing the single-cell transcriptome of the untreated and the ketotifen-treated groups is required to appreciate the full spectrum of fibroblast heterogeneity in solid tumors and to develop therapies that specifically target tumor-promoting CAFs. In line with this, a recent analysis of the TME of pancreatic tumors has demonstrated a distinct CAF subpopulation (CD141⁺MHCII⁻) as the responsible cell population of desmoplasia and a potential therapeutic target. Indeed, the CD141⁺MHCII⁻ CAF population resembled the myofibroblast subtype with regard to gene expression and active collagen production but also shared markers with other subtypes including the inflammatory CAFs (76). Another study suggests that for invasive squamous tumors of the pancreas exhibiting a higher proportion of activated stromal cells, blockade of the activated stroma using FAP inhibitors should be considered in the treatment regimen for augmented therapeutic outcome (77). Thus, a therapeutic strategy can be designed based on the characteristics of CAF subtypes.

The MC–CAF cellular cross-talk was further validated in murine sarcoma models. By titrating the ketotifen dose, we found that the dose of 10 mg/kg had the most prominent effect on tissue stiffness via reducing CAF proliferation and density. Consistent with the in vitro results, collagen and hyaluronan synthesis was also suppressed, suggesting that ketotifen exerts its TME reprogramming properties by affecting CAFs. Importantly, the resulting alleviation of stiffness enhanced vessel decompression, allowing for improved perfusion and oxygen delivery to the tumor site. Interestingly, ketotifen was also able to fortify the tumor vessel walls by increasing pericyte coverage.

Approved drugs that can reprogram/normalize the TME have already been repurposed and entered clinical trials, holding great promise in broadening the accessibility and increasing the efficacy of immunotherapies (78–81). Recent findings indicated that patients of multiple tumor types, who took H1-antihistamines during immune-checkpoint blockade treatment had a highly significantly reduced death rate compared with those who did not, whereas antihistamines had a minimal effect on the survival of chemotherapy-treated patients, which further supports the immunologic effects of ketotifen rather than any direct antitumor effects (82, 83). Accordingly, in our attempt to improve the responsiveness of fibrosarcoma and osteosarcoma tumors to conventional therapies, we combined the reprogramming properties of ketotifen with the cytotoxic activities of doxorubicin and anti–PD-L1. We found that pretreatment with ketotifen increases the sensitivity of MCA205 and K7M2 tumors to chemotherapy and checkpoint inhibition, whereas the combination of the three augments antitumor response as indicated by the regression of tumor volume. Doxorubicin acts by intercalating within the DNA and directly killing cancer cells, whereas antibodies that block the signal from T-cell coinhibitory receptors exert their antitumor responses by resuscitating the host immune system and, thus, leading to a more durable response (84–86). Another possible mechanism by which ketotifen may sensitize cancer cells to doxorubicin is via inhibition of exosome release. Although not explored in this study, ketotifen administration can inhibit calcium entry into the cell and in turn reduce exosome release. Consequently, increased cellular accumulation of doxorubicin contributes to cytotoxicity and enhanced antitumor effects (87).

Furthermore, in agreement with previous studies, we confirmed that a combination of antihistamines with either doxorubicin and/or immune-checkpoint inhibition equips the host with strong immunologic memory (15, 18). The prominent immune response in animals receiving the combination treatment is attributed to the improved T-cell trafficking and particularly the increased ratio of effector CD8⁺ T-cell population to suppressive Tregs compared with the groups with no or limited memory response. Although the cytotoxic nature of doxorubicin has been reported to boost T-cell immunity, here we did not observe any alteration in T-cell infiltration in either tumor model. In K7M2, however, we found that doxorubicin synergizes with immunotherapy to promote CD8⁺ T-cell proliferation and T-cell attachment to vessels, independently of hypoxia alleviation, suggesting that other mechanisms may underline such antitumor responses. Consistent with previous studies (62, 88), our findings support an IFNγ-dependent mechanism by which ketotifen facilitates vascular normalization and, subsequently, increases perfusion, which in turn upregulates adhesion molecules necessary to support T-cell extravasation, increases recruitment, and doubles the fraction of CD8⁺ T cells proliferating in these tumors, while decreasing the polarization of the immature myeloid cell population into MDSCs (89–92). MDSCs concentrate in hypoxic cores and form a potent barrier to tumor immunity (71, 93–95). Besides MDSCs, M2-like macrophages have been associated with immunosuppression. It has been previously described that the inhibition of the histamine receptor H1 on M2-like macrophages blocks the membrane translocation of immune-checkpoint VISTA, thereby revitalizing T-cell cytotoxic function and restoring response to anti–PD-1 treatment (82). Nevertheless, further research including studies in MC-deficient mice is required to conclude whether these immunologic responses are MC- or ketotifen-driven.

Finally, we explored the translational possibility of monitoring tissue elastic properties using ultrasound SWE as a noninvasive and reliable biomarker of responsiveness to immune-checkpoint inhibition. Like our studies in murine breast tumor models (12, 18, 96), we show here that the reduction of tumor elastic modulus increases the efficacy of immunotherapy and chemotherapy in sarcoma tumors, highlighting further the potential of tumor stiffness to serve as a predictive marker of response. Future studies assessing perfusion status during the course of treatment using contrast-enhanced ultrasound could be used to test the hypothesis that mechanical biomarkers can be developed to predict the response of tumors to immunotherapies and chemotherapies. Ultrasound SWE is already used in the clinic for the evaluation of liver fibrosis and has been investigated in patients and animal models with various cancer types as a marker of response to chemotherapy (97–99).

Our findings from the clinical evaluation of ketotifen in sarcoma patients further demonstrate that measuring and monitoring tumor stiffness using noninvasive, clinically applied ultrasound SWE can be used as a predictive biomarker, particularly for tumors with a low incidence rate and of different histologic subtypes, such as sarcomas. Stabilizing MCs holds great promise for overcoming resistance to immunotherapy in solid tumors, which are associated with high levels of MCs or their products and have a poor prognosis, and warrants further investigation.

In conclusion, in this study, we primarily describe a previously unexplored interaction between tumor-resident MCs and CAFs and demonstrate the ability of the approved antihistamine and mast cell stabilizer, ketotifen, to restore patho-physiologic abnormalities of the TME. As cancer might be considered a TME disease, we show that pretreatment with ketotifen establishes favorable conditions within the TME toward potentiating chemoimmunotherapy and establishing durable immune responses in murine models of sarcoma.

J.D. Martin reports personal fees from NanoCarrier Co., Ltd outside the submitted work. In addition, J.D. Martin has a patent for cancer treatment using ketotifen in combination with a checkpoint inhibitor pending and licensed to Materia Therapeutics Inc and a patent for cancer treatment using bosentan in combination with a checkpoint inhibitor pending and licensed to Materia Therapeutics Inc; Dr. Martin owns equity in Materia Therapeutics Inc. T. Stylianopoulos reports a patent for cancer treatment using ketotifen in combination with a checkpoint inhibitor pending and licensed to Materia Therapeutics Inc. No disclosures were reported by the other authors.

M. Panagi: Conceptualization, resources, investigation, methodology, writing–original draft. F. Mpekris: Investigation, methodology. C. Voutouri: Data curation, software, formal analysis. A.G. Hadjigeorgiou: Data curation, software, formal analysis. C. Symeonidou: Validation, visualization, methodology. E. Porfyriou: Project administration. C. Michael: Resources, investigation, methodology. A. Stylianou: Formal analysis, investigation. J.D. Martin: Conceptualization, writing–review and editing. H. Cabral: Conceptualization, supervision. A. Constantinidou: Conceptualization, supervision. T. Stylianopoulos: Conceptualization, supervision, funding acquisition, writing–review and editing.

We thank the patients and all the investigators who participated in these studies. The authors gratefully recognize the contributions of staff from the Bank of Cyprus Oncology Center in supporting the centralized screening of patients. The authors were fully responsible for all content and editorial decisions, were involved at all stages of manuscript development, and approved the final version.