SENTI-101, a Preparation of Mesenchymal Stromal Cells Engineered to Express IL12 and IL21, Induces Localized and Durable Antitumor Immunity in Preclinical Models of Peritoneal Solid Tumors

This article is featured in Highlights of This Issue, p. 1497

Multiple solid tumors, including gynecological and gastrointestinal tumors, have the capacity to metastasize and spread through the peritoneal cavity in a process known as peritoneal carcinomatosis (PC; ref. 1). One of the most frequent tumors that metastasize to the peritoneal cavity is advanced high-grade serous ovarian cancer (HGSOC; refs. 2, 3), which accounts for more than 20,000 new cases every year in the United States alone and is often associated with disease progression and low overall survival rates (<30% 5-year survival rate; ref. 4).

Current treatment approaches for HGSOC include cytoreductive debulking surgery combined with platinum-based chemotherapy, including intraperitoneal chemotherapy (5, 6). In the last year, PARP inhibitors have been approved as a second-line maintenance treatment for HGSOC (7, 8). Although current standard of care therapies prolong time to progression, they have limited impact on overall survival (9, 10). Thus, novel therapeutics are needed to address this unmet medical need.

The remarkable success of immunotherapy in the treatment of certain solid tumors, such as melanoma or lung cancer (11), has not effectively translated into the treatment of advanced HGSOC; recent clinical trials have shown a disappointing lower response rate (12–15) and, hence, an efficacious and durable response to treatment remains elusive for these patients. There have been clinical studies in HGSOC that attempted to alternatively engage the patient's immune system by using immune-stimulatory cytokines such as IL12 (16–20). Although cytokines are powerful modulators of the immune system, their therapeutic potential is limited by their short half-lives and systemic toxicities (21, 22).

A novel therapeutic modality that is currently being explored for the treatment of HGSOC consists of mesenchymal stromal cells [also referred to as mesenchymal stem cells (MSCs); refs. 23–26]. MSCs are a heterogeneous and undifferentiated cell population that can be sourced from bone marrow (BM-MSC), umbilical cord (UC-MSC), and adipose tissues (Ad-MSC), and have the ability to differentiate into osteoblast, adipocyte, and chondroblast lineages upon appropriate stimulation (25–26). MSCs are immune-privileged and lack the expression of HLA-II molecules, making them an attractive candidate for off-the-shelf allogeneic cell therapies (27–30). Furthermore, BM-MSCs have innate tumor-homing capacity (31, 32) and are known to preferentially interact with tumor extracellular matrix components (33), thus having the potential to selectively deliver potent drugs to the tumor microenvironment (34–36).

Here we present the preclinical development of SENTI-101, an allogeneic cell-based therapeutic consisting of human BM-MSCs engineered to overexpress two potent immune-stimulatory cytokines, IL12 and IL21, resulting in a multimodal and durable antitumor immune response in peritoneal solid tumors.

Murine MSCs were purchased from Cyagen (Balb/c MSCs #MUCMX-01001; C57Bl/6 MSCs #MUBMX-01001). Human bone marrow-derived MSCs were obtained from Rooster Bio (RUO and GMP-grade donors). MSCs were passaged at 70% to 90% confluency. Cells were trypsinized using TrypLE and subcultured at 3,000 to 5,000 cells/cm². Murine MSCs were cultured using minimum essential medium (MEM) alpha medium containing 10% MSC-FBS, 200 mmol/L L-glutamine, 1 μg/μL of murine FGF-basic, and penicillin/streptomycin. Cells were confirmed free of mycoplasma and tested routinely. Human MSCs were cultured with RoosterBasal-MSC and RoosterBooster-MSC-XF Supplement. Cells were passaged for a limited number of times (<6 passages) and were resuspended in 200 μL of PBS for in vivo administration at the doses listed in each figure legend.

Lenti-X 293T packaging cell line was cultured in DMEM, high-glucose, 1 mmol/L sodium pyruvate, 10% heat-inactivated fetal bovine serum (HI-FBS) and seeded at 72,700 cells/cm² prior to transfection. Lentiviral transfer vectors were cotransfected with psPAX2 packaging vector and pMD2.g envelope vector containing the VSV-G envelope protein using FuGene-HD. Viral supernatants were concentrated using Lenti-X concentrator. Concentrated vectors were added to MSCs seeded in 6-well plates 4 hours prior to viral transduction. Cells were spinoculated at 800 g for 1 hour at 32°C, followed by incubation at 37°C 5% CO₂. Four hours post-spinoculation, an additional 2 mL of complete media was added to each well. Control MSCs were generated in parallel to vector-engineered MSCs by transducing the cells with a sham lentiviral vector without the expression cassette.

CT26.WT-Fluc-Neo were purchased from Imantis Life Sciences (Catalog no.: CL043; Lot no.: CL-IM147) and were cultured in DMEM media with 10% FBS and 0.4 mg/mL G418. B16F10-Fluc-Puro were purchased from Imantis Life Sciences (Catalog no.: CL052; Lot no.: CL-IM150) and cultured in RPMI media with 10% FBS and 1 μg/mL puromycin. Either 5e4 (CT26) or 1e5 (B16F10) cells were injected intraperitoneally into female Balb/c or C57Bl/6J mice (The Jackson Laboratory). Tumor burden was measured by bioluminescence imaging after injection of D-Luciferin substrate (15 mg/mL). For in vivo and ex vivo imaging of reporter cells (transduced with the nanoLuc reporter) 100 μL of nanoGlo substrate solution was injected into mice, and mice or tissues were immediately imaged. AMI and Lago imagers (Spectral Instruments) were used with prior calibration and appropriate negative controls. AMI software was used to quantify bioluminescence flux (photons/second) in a region of interest (ROI). Subcutaneous tumor volume was measured using a caliper.

Tumor burden, body weight, and mice overall health were assessed weekly, and mice were euthanized at scheduled time points or when reaching endpoint criteria for survival studies. On scheduled time points tissues were collected as follows: blood was collected in EDTA-coated tubes via submandibular bleeding or intracardiac puncture and processed for flow cytometry or collected in serum (clot activator) tubes and centrifuged for serum separation; peritoneal fluid was collected via peritoneal lavage; peritoneal tumors and other organs were collected, and flash frozen in liquid N₂ or fixed in formalin and embedded in paraffin for histologic evaluation.

All mouse studies were performed in compliance with Institutional Animal Care and Use Committee guidelines (IACUC protocol EB17-010-108).

HEK-Blue IL12 cells, an IL12 reporter cell line capable of sensing both mouse and human variants of IL12 (hkb-il12, from InvivoGen), were purchased and used used following manufacturer's instructions. Supernatants collected from MSC cultures were centrifuged at 400 × g for 5 minutes to remove cells prior to assay. HEK-Blue IL12 cells were seeded with sample supernatants in a 96-well plate, which were then incubated for 24 hours. A total of 180 μL of QUANTI-Blue were mixed with 20 mL of supernatants collected from the HEK-Blue IL12 cells. Absorbance was measured at 650 nm using BioTek SynergyH1 spectrophotometer. Recombinant mIL-12, mIL-21 and culture media were used as positive and negative controls for this assay.

Supernatants from MSC cultures or control samples were added in a 24-well plate with NK-92 cells. Cells were incubated for 30 minutes and fixed and permeabilized using Cytofix/Cytoperm Kit. Cells were stained with fluorescently labeled antibodies for pSTAT1 or pSTAT3. Recombinant IL12 (10 ng/mL) or IL21 (10 ng/mL) were used as negative or positive controls, respectively. Samples were analyzed using Cytoflex LX flow cytometer (Beckman Coulter) and Ultra-Bright counting beads for normalization.

Assay was performed by Eurofins DiscoverX using a reporter cell line (PathHunter U2OS IL-21R/IL-2RG Dimerization Cell Line) in which one cytokine receptor chain is tagged with a small peptide epitope ProLink (PK) and the other chain is tagged with enzyme acceptor (EA). Ligand (IL21) binding induces dimerization of the two receptors, facilitating complementation of PK and EA fragments. This interaction generates an active unit of β-galactosidase, which is detected using a chemiluminescent substrate.

MSCs were seeded in a 6-well plate with 2 mL of media and supernatants collected at 24 hours. Mouse and human ELISA kits for IL12 and IL21 were purchased from R&D Systems and were used following manufacturer's instructions. Plates were read at 540 nm (reference wavelength) and 450 nm using BioTek SynergyH1 microplate reader spectrophotometer and Gen5 software. 10-plex Murine ProcartaPlex Luminex kits were customized by Thermo Fisher Scientific(Assay ID MXXGRM9; Life Technologies) and used following manufacturer's instructions. Primary murine tissue samples (serum or intraperitoneal fluid) were diluted in sample assay buffer where multiple dilutions were used to ensure all analytes were within standard range. Samples that required less dilution (<10-fold) were diluted in ProcartaPlex Platinum Assay Buffer (Thermo Fisher Scientific, Catalog No. EPXP-1112-000) to mitigate bead aggregation, whereas more concentrated samples that required further dilution (>10-fold) were diluted with the universal assay buffer provided in the kit. Human IL12 and IL21 levels in xenograft models were quantified with MilliPlex (EMD Millipore) kits following manufacturer's instructions. Samples were analyzed in Prism 8 using a weighted-5PL curve fit of the standards to interpolate sample concentrations.

Mice were depleted of CD8, CD4, and NK1.1⁺ cells using 200 μg in 200 μL of depleting antibodies administered intraperitoneally every 3 days for a total of four doses. One week after start of the treatment, immune cell depletion was confirmed by collecting blood via submandibular bleeding and performing flow cytometry to quantify CD4, CD8, and NK cells. Checkpoint inhibitor antibody anti-PD1, clone RMP1–14 was administered intraperitoneally in three doses of 200 mg/kg every 3 days, starting 1 day prior to treatment with MSCs.

Mice were euthanized and tissues (tumor, spleen, peritoneal sacral and lumbar lymph nodes, peritoneal fluid, and blood) were collected. Tumors were digested using MACS Mouse Tumor Dissociation Kit and the gentle MACS Octo Dissociator (Miltenyi Biotec). Single cell suspensions from all organs were obtained using a 70 μm strainer. Red blood cells were lysed using RBC lysis buffer. Cell suspensions were stained with fluorescently labeled antibodies (BioLegend and BD Biosciences, see supplementary table of reagents), adding Brilliant Stain Buffer. For intracellular staining, cells were permeabilized and fixed using the Foxp3 fixation and permeabilization kit following manufacturer's instructions. For tetramer staining, samples were minimally processed prior to incubation with H-2Ld MuLV gp70 Tetramer-SPSYVYHQF antibody (MBL) and membrane antibodies. Dead cells were excluded using FVS780-fixable viability dye (BD Biosciences). Samples were analyzed using the Cytoflex LX flow cytometer (Beckman Coulter) and FlowJo was used to analyze the data.

Tumors from control and treated mice were fixed in formalin and embedded in paraffin for histology. Slides were deparaffinized in an oven chamber for 30 minutes at 37°C and then for 30 minutes at 60°C. Following deparaffinization, slides were rehydrated in Slide Brite, then in decreasing percentages of ethanol. After rehydration, antigen retrieval was performed using Rodent (pH 6) or Nuclear (pH 9.5) decloacker solution in NxGen Decloaking chamber at 95°C for 15 minutes. Slides were blocked using Rodent Block M and incubated for 30 minutes with primary antibodies, covered from light at RT. After primary antibody staining, slides were incubated with a secondary HRP-Polymer using the Opal 4-Color Anti-Rabbit Manual IHC Kit or Rat-on-Mouse HRP-Polymer for Rat antibodies (depending on the primary antibody), followed by opal fluorophore staining (Opal520, Opal570, or Opal690; Akoya Biosciences). DAPI staining was used to counterstain nuclei.

Universal Negative Control reagent was used instead of the primary antibody as a negative control. Slides were mounted using ProLong Diamond Antifade Mountant and scanned using Vectra Polaris microscope (Akoya Biosciences) and Phenochart software was used for visualization.

Detailed descriptions of reagents including catalog numbers are shown in Supplementary Materials Table.

MSCs have been previously shown to home into sites of inflammation and tumors, including intraperitoneal tumors (31, 36). We confirmed that BM-MSCs engineered to express a bioluminescent reporter (firefly luciferase) preferentially localized into tumor tissues compared with other peritoneal healthy organs when delivered intraperitoneally (Fig. 1A and B; Supplementary Figs. S1A and S1B). In terms of kinetics, IP-delivered MSCs were detected up to 3 days after administration, whereas the signal decreased by day 10 in syngeneic models of peritoneal carcinomatosis (B16-F10 model; Supplementary Fig. S1C). The innate tumor-tropism and kinetics of MSCs makes them an ideal candidate to deliver potent immunomodulatory agents to the tumor microenvironment, aimed to locally polarize and activate the immune system and thus avoid potential systemic immunotoxicities. To further evaluate this novel MSC-based immunomodulatory therapy concept, we engineered murine (Balb/c and C57Bl/6) BM-MSCs to express different payloads that have known immune stimulatory function (including cytokines, chemokines, and tumor microenvironment modifiers) and empirically determined their abilities to induce antitumor responses as single agents or in combination using two different syngeneic mouse models of intraperitoneal carcinomatosis with distinct immune landscapes: CT26-ip and B16F10-ip (37, 38). As part of the characterization of these syngeneic IP tumor models, we analyzed their immune infiltrates by flow cytometry and IHC (Fig. 1C; Supplementary Figs. S2A and S2B). Control-MSCs (engineered with a sham or nonexpression lentiviral vector) were used as control. Antitumor activity of different combinations was assessed by comparing the tumor burden fold change 10 days after treatment. Extensive in vivo screening of over 20 payloads (cytokines, chemokines, and immune activators) with different mechanisms of action tested either alone or in combinations in both syngeneic immunocompetent IP tumor models (CT26-ip and B16F10-ip). Although multiple combinations achieved significant tumor burden reduction in the CT26-ip model that contains more T-cell infiltrates (Fig. 1D; refs. 36, 37), the immune refractory B16F10-ip model, revealed that the combination of IL12 and IL21 resulted in the most robust antitumor response (Fig. 1E; Supplementary Fig. S3). IL12 and IL21 are potent immune-stimulatory cytokines with multimodal immune modulatory mechanisms of action and have previously been shown to demonstrate antitumor activities in multiple nonclinical models of cancer (IL12; refs. 39–42), IL21; refs. 43–46) and in the clinic (17–21). This combination was selected as the candidate, SENTI-101.

To optimize the expression of IL12 and IL21 in human BM-MSCs, we evaluated a library of commonly used promoters to drive the expression of an EGFP reporter in our lentiviral transfer vector backbone. hBM-MSCs were transduced with a lentiviral vectors encoding different promoters driving a fluorescent reporter protein, EGFP. To select the best promoters from this panel, transduction efficiency was evaluated using reporter fluorescence intensity as measured from flow cytometry (Fig. 2A and B; Supplementary Fig. S4A). In parallel, constructs with IL12 and IL21 were generated using spleen focus-forming virus (SFFV) or elongation factor-1 alpha (EF1a) promoters and cytokine production was compared (Fig. 2C). In both instances, SFFV promoter outperformed other promoters such as EF1a or EFS in driving robust and durable expression of the expression cassette for multiple passages and was selected for our construct. Furthermore, we tested different orientations as well as the presence of different signal sequence peptides to enhance the secretion of these cytokines (Fig. 2D; Supplementary Fig. S4B). Protein function and signaling was confirmed for each cytokine using flow-based and cell-based reporter assays to detect induction of STAT phosphorylation and subsequent activation of downstream signals. All assays confirmed that our engineered MSCs secrete functional IL12 and IL21 cytokines (Fig. 2E and F; Supplementary Fig. S5). Although there was a modest increase in IL21 protein secretion by using IL8 signal sequence (construct SB00971), the difference was not sufficient to warrant such modification, and thus we selected the construct with the naïve codon-optimized sequences of IL12p70 single chain and IL21 linked with a T2A self-cleaving peptide (construct SB00880) as our lead candidate for manufacturing and development of SENTI-101 [Fig. 2G; Supplementary Fig. S6; Supplementary Table S1, United States Patent Application 2020017109; United States Patent Publication No. 2020/0206271; United States Patent No. 10,993,967 (pending) 47].

To characterize the pharmacokinetic (PK) and pharmacodynamic (PD) parameters of SENTI-101, we used a murine surrogate, mSENTI-101 (murine BM-MSCs engineered to express the equivalent dose of murine IL12 and IL21; Supplementary Fig. S7A). We delivered mSENTI-101 via intraperitoneal injection into syngeneic immunocompetent tumor-bearing mice (CT26-ip and B16F10-ip). A bolus dose of recombinant cytokines was injected in a separate group of mice. The dose of recombinant cytokines was four times the daily dose produced by mSENTI-101 in vitro to account for the sustained production over approximately 4 days by the engineered MSCs in vivo. PF and serum were collected at multiple time points, and cytokine concentrations of IL12 and IL21 as well as cytokines and chemokines that were predicted to be downstream of IL12 and IL21 were determined via Luminex. IL12 cytokine concentration in locoregional and systemic compartments was measured and the PKs of recombinant cytokines and Senti-101 were nonlinearly fit with a weighted (1/y) single-phase exponential decay using GraphPad Prism. The same nonlinear regression was applied to both treatment arms. (Fig. 3A; Supplementary Fig. S7B). SENTI-101 administration resulted in a more durable and localized presence of IL12 compared with administration of recombinant cytokines. Half-life of IL12 in the mSENTI-101 group was more than four times longer compared with the recombinant cytokines (10.39 hours for mice treated with mSENTI-101 vs. 2.3 hours for mice treated with recombinant cytokines; Supplementary Fig. S7B). Importantly, the ratio between PF and serum levels of IL12 was 10-fold lower in mice treated with SENTI-101 compared with the treatment with recombinant cytokines (Fig. 3B). The limited systemic exposure of IL12 and IL21 observed with SENTI-101 treatment offers an important advantage over recombinant cytokines, reducing the potential risk for systemic immunotoxicities that has been previously observed with IL12 (18, 23).

Consistent with the results obtained in the CT26-ip model, mSENTI-101 resulted in a peritoneal localized and sustained production of IL12 and IL21 with a longer half-life when compared with recombinant cytokines in B16F10-ip models (Supplementary Figs. S7C and S7D). Of note, there were differences in the kinetics and cytokine systemic exposure between CT26-ip and B16F10-ip models, which can be due to different baseline immune landscapes present in the tumors, as well as different mouse backgrounds (Balb/c vs. C57Bl/6; refs. 37, 38, 48). Consistent with increased cytokine half-life, mSENTI-101 treatment resulted in increased antitumor activity and proportion of tumor-free survival in tumor-bearing mice compared with recombinant cytokines (Fig. 3C; Supplementary Fig. S7E).

IL12 and IL21 have pleiotropic and non-overlapping roles, but combination of both results in activation of T and NK cells, which lead to IFNγ production (39, 49). We quantified the levels of IFNγ as well as other IFN-induced chemokines such as IP10, as part of our PD panel. We observed a significant and durable increase in multiple immune-stimulatory factors (IFNγ, IP10, sCD25) as a result of the treatment with mSENTI-101, indicative of a robust antitumor immune response (Fig. 3D). Our panel also included other inflammatory cytokines typically involved in immune-mediated toxicities and cytokine release syndrome (CRS) such as TNFα or IL6. Although in mice treated with mSENTI-101, we observed relative increases in such cytokines in the early time points (3 hours), these increases were minor when compared with baseline (Supplementary Fig. S7D).

Recombinant IL12 and IL21 have each been evaluated for the treatment of cancer due to its immune stimulatory functions, which can result in enhanced antitumor immunity (39, 43, 44). Importantly, several strategies to safely and locally deliver IL12 have been tested in clinical studies for cancer (NCT04006119, NCT03393884). To test for antitumor activity of our lead candidate, we used the murine surrogate (mSENTI-101) in both CT26-ip and B16F10-ip models. Immune competent mice with established syngeneic tumors were treated on day 7 posttumor implant with IP-delivered engineered MSCs (Supplementary Fig. S8A). Treatment with mSENTI-101 resulted in a significant and dose-dependent reduction in tumor burden as measured by bioluminescence imaging (Fig. 4A and B). This included the complete eradication of over 90% of the tumors in both models 10 days after treatment (Fig. 4C and D; Supplementary Figs. S8B and S8C). Importantly, in the CT26-ip tumor model that develops ascites, treatment with mSENTI-101 significantly reduced the presence of malignant ascites (Fig. 4E), which is an important feature of peritoneal carcinomatosis and advanced ovarian cancer (2, 3). None of the doses of SENTI-101 resulted in any toxicities or decreased body weight in the mice (Supplementary Fig. S9).

To confirm that this antitumor activity was mediated by the immune system, we used depletion antibodies to selectively decrease different immune components that are known to mediate antitumor immunity: namely CD4, CD8, and NK cells (50). Mice were treated with depleting antibodies weekly prior to tumor implantation and throughout the duration of treatment (Supplementary Fig. S10A). Flow cytometry was performed to assess the reduction of the targeted cells (Supplementary Figs. S10B and S10C). Once immune depletion was confirmed, mice were implanted with peritoneal tumors and treated with intraperitoneally-injected MSCs 7 days thereafter. mSENTI-101-driven antitumor activity was significantly reduced when different immune subsets were depleted (Fig. 4F and G), demonstrating that SENTI-101 antitumor activity does indeed rely on the endogenous immune system, and specifically requires T cells and NK cells at a minimum.

To further characterize the immune response induced by SENTI-101, we profiled the immune composition of tumors, PF, peritoneal lymph nodes, and circulating PBMCs from treated mice. mSENTI-101 was used to treat mice with established CT26-ip or B16F10-ip tumors and tissues were collected at 72 hours posttreatment and immune system composition was analyzed via multiplexed-IHC and flow cytometry. Treatment with mSENTI-101 resulted in increased T- and B-cell infiltration in peritoneal tumors (Fig. 5A and B; Supplementary Figs. S11A–S11C). These T cells and B cells were localized in well-organized clusters, indicative of tertiary lymphoid structures that were particularly evident in the immune “cold” model B16F10-ip (Supplementary Figs. S11B and S11C). These tertiary lymphod structures have been associated with improved prognosis as well as response to checkpoint inhibitors in multiple tumor types, including ovarian cancer (51–55). This was accompanied by a significant increase in activated CD8⁺ T cell subsets in the PF (Fig. 5C; Supplementary Fig. S12A). T cells in the PF of mice treated with mSENTI-101 expressed two or more cytokines (Fig. 5D; Supplementary Fig. S12B), suggestive of a stronger T-cell function, which has been shown to correlate with improved antitumor functions (56, 57). Consistent with an increase in IFNγ secretion, T cells in the peritoneal fluid expressed exhaustion markers such as Lag3 and PD1 (Supplementary Fig. S12C). Concomitantly, we observed a significant increase in antigen-presenting dendritic cells in the peritoneal tumor-draining lymph nodes (Fig. 5E; Supplementary Fig. S12D). Antigen-presenting dendritic cells (DC1 and DC2) are critical for durable and sustained antitumor activity and play a role in initiating the cancer immunity cycle (58–61). We also observed a reduction in myeloid derived suppressive cells (MDSC) in the blood of mice treated with mSENTI-101 (Supplementary Fig. S12E). Consistent with the localized cytokine production, the extent of the immune activation was much greater in the PF and tumors compared compared to circulation, with T-cell increases of 4- to 5-fold in tumor and PF of mice treated with mSENTI-101 compared with control-treated mice, whereas the increase in T cells was much lower in the periphery (<1.5-fold in treated mice compared with control; Fig. 5F; Supplementary Fig. S12F). These data support that treatment with SENTI-101 induces a potent but localized antitumor immune response, thus reducing the risk for systemic immune-mediated toxicities.

An important hallmark of a successful antitumor immunotherapeutic is the achievement of durable responses and tumor-free survival. To this end, we conducted long-term survival studies using both CT26-ip and B16F10-ip tumor models treated with mSENTI-101. In both cases, treatment with mSENTI-101 resulted in long-term tumor-free survival (73% in CT26-ip and 60% in B16F10-ip tumor models; Fig. 6A and B; Supplementary Fig. S13). Of note, the B16F10 model is highly immune-suppressed and refractory to multiple immune-oncology drugs, including anti-PD1/PD-L1 antibodies (37, 38). Treatment with mSENTI-101 significantly outperformed anti-PD1 antibodies and increased the response rate to such treatment from 27% to 85% of mice being tumor-free after 100 days (Fig. 6B). As observed by flow cytometry, treatment with SENTI-101 resulted in the engagement of multiple compartments of the cancer immunity cycle, which should ultimately lead to the acquisition of antitumor immune memory. We quantified the presence of CD8 T-cell clones that recognize the tumor antigen H2a using a tetramer staining against the CT26 Tetramer. We observed a significant increase in the presence of tetramer positive T cells in the PF of mice 14 days after treatment with mSENTI-101 (Fig. 6C). Concomitantly, T-cell memory subsets were significantly increased (Fig. 6D). Antitumor immune memory can be confirmed by a re-challenge of treated and tumor-free mice with new implantation of tumor cells (Supplementary Fig. S13C). All mice that were confirmed tumor-free for more than 90 days after treatment with mSENTI-101 were rechallenged with new tumor cells implanted subcutaneously in the flank. With the CT26 model, all the survivor mice treated with mSENTI-101 were able to reject the newly implanted tumors, demonstrating a successful achievement of antitumor immune memory (Fig. 6E). With the B16F10 model, 11 of 13 of the surviving mice after treatment with mSENTI-101 (85%) were able to reject the newly implanted tumor cells and, in all the cases, there was a significant delay in tumor growth, suggestive of a partial or initial antitumor immune response (Fig. 6F) in the 2 remaining mice. Overall, these results confirm that SENTI-101 treatment results in a robust and durable antitumor immune response and immune memory in syngeneic models of peritoneal cancer.

Recent successes of immunotherapies in certain cancer indications have revolutionized the treatment paradigm in oncology. Unfortunately, most solid tumors, among them ovarian cancer and peritoneal carcinomatosis, remain unresponsive to these advances (13). Thus, there remains an unmet need for novel therapeutics that can effectively modify and activate the immune system to fight cancer.

Allogeneic BM-MSCs have been extensively used in the clinic, primarily for treatment of graft-versus-host diseases; however, the use of gene-engineered MSCs for the treatment of cancer is more limited (NCT02530047, NCT02008539; ref. 24). Despite remarkable success of engineered cell-based immune therapeutics, such as CAR-T cells that have revolutionized the treatment paradigm for liquid tumors, some important challenges remain. Autologous engineered cell-based therapeutics require a long and expensive manufacturing process. Furthermore, the immunosuppressive tumor microenvironment present in solid tumors is a barrier to efficacious treatments. MSCs offer several advantages over autologous T-cell therapeutics: (i) MSCs are naturally immune privileged, allowing for allogeneic off-the-shelf products; (ii) MSCs are easy to manufacture and have extensive expansion potential; (iii) MSCs innately localize and home into solid tumors, thus reducing the risk for systemic toxicities; and (iv) MSCs are not inhibited by the suppressive tumor microenvironment.

Here we present a novel cell-based therapeutic, SENTI-101, consisting of allogeneic MSCs as a modality to deliver two potent cytokines (IL12 and IL21) into the tumor microenvironment, to elicit durable immune-mediated antitumor responses and induce immune memory. The combination of IL12 and IL21 was chosen among more than 20 immune-stimulatory factors evaluated on the basis of in vivo antitumor efficacy. These two immunomodulatory cytokines (IL12 and IL21) engage both adaptive and innate immune responses against the tumor, have complementary and multimodal mechanisms of action, (39–41, 43, 44, 46) and have been shown to induce antitumor activity in multiple nonclinical models of cancer (40, 42, 43, 62). IL12 is a covalently linked heterodimer composed of two chains, p40 and p35, connected by a disulfide bridge, which is mainly secreted by subclasses of antigen-presenting cells (39). Its membrane receptor consists of two chains, IL12Rβ1 and IL12Rβ2, and is present on immune cell populations such as NK cells, CD4⁺ and CD8⁺ T cells. Upon binding to its receptor, IL12 promotes the secretion of Th1 cytokines (e.g., IFNγ) and the subsequent induction of an antitumor immune response. IL21 is a four-α-helical bundle cytokine mainly secreted by activated CD4⁺ T cells and iNKT cells. Its heterodimeric receptor (IL21R) comprises a distinct α-chain and a common cytokine receptor γ chain shared with the IL2 family of cytokines (46). IL21R has a pleiotropic function and is expressed on B cells, iNKT cells, CD4 and CD8 T cells as well as myeloid cells and is known to drive B-cell maturation and proliferation. In addition, IL21 can induce antitumor activities, from promoting NK function to supporting T-cell memory phenotype and persistence (44–46).

The use of cytokines for cancer immunotherapy has been extensively applied. For example, high-dose IL2 is approved for the treatment of metastatic renal cell carcinoma and metastatic melanoma (63). New approaches are under intense investigations, including late-stage clinical studies using pegylated IL2 (NKT-241) for the treatment of metastatic solid tumors (64, 65). Specifically, for IL12 and IL21, there are over 33 active clinical trials listed on Clinicaltrials.gov at the time of this writing. To date, two published examples of IL12-secreting cell therapies (lymphocytes) have been tested in clinical trials (NCT02498912; ref. 20, 66). A phase 1/2 study was conducted to evaluate the effect of IL12 expression under the control of a nuclear factor of activated T cells (NFAT) inducible promoter in autologous Tumor Infiltrating Lymphocytes (TILs) (20). This previous clinical experience with IL12-engineered T cells has shown that IL12-secreting cell therapy is feasible. However, the success of these strategies as well as previous clinical experiences using recombinant cytokines (17) has been hampered by severe toxicities due to systemic and uncontrolled immune activation, as well as the lack of efficacy caused by the poor PKs and short half-lives of these compounds. Recently, GEN-1, an IL12-encoding plasmid has been used with remarkable success in combination with neoadjuvant chemotherapy in early-stage clinical trials for ovarian cancer (NCT02480374). By leveraging the tumor-homing capability of MSCs to locally produce IL12 and IL21 in the tumor environment, the risk of systemic exposure may be significantly reduced, while broadening the effective therapeutic window of these cytokines in the locoregional tumor area.

Although MSCs innate function is to suppress immune activation and have been involved in cancer initiation or promotion (67), the expression of two potent immune stimulatory cytokines, IL12 and IL21, results in an overall antitumor immune response. In all our preclinical studies, control MSCs (engineered with an empty vector) did not demonstrate tumor increase or acceleration, which could be explained by the fact that we treated established (∼100 mg) tumors in syngeneic hosts. The preclinical and mechanistic studies described here demonstrate that SENTI-101 is able to act on multiple steps of the cancer immunity cycle (58), engaging both innate and adaptive immune compartments. Mice treated with SENTI-101 showed significant increase in antigen-presenting cells in their lymph nodes, characterized by expression of markers CD103 and XCR1. These cells have been demonstrated to enable the differentiation of T-cell memory phenotypes and a subsequent sustained antitumor response despite tumor or pathogen rechallenge (58, 60, 61). By activating both the innate and adaptive immune responses, SENTI-101 increases dendritic cells in the tumor-draining lymph nodes, T-cell local activation in the peritoneal fluid, and results in increased tumor-specific T cells and T-cell memory subsets. Our mechanistic studies using immunocompetent tumor models corroborate that this multimodal engagement of the cancer immunity cycle results in durable antitumor responses and immune memory as evidenced by rejection of newly implanted tumors in mice that were cured after treatment with SENTI-101.

SENTI-101 outperformed the efficacy and response rates of checkpoint inhibitor anti-PD1 in both preclinical models. The proposed mechanism of action of SENTI-101, with increased abundance and expression of activation markers on T cells, B cells, and APCs, would be indicative of turning “cold” tumors hot.” Our results suggest that Senti-101 could synergize or improve the response rate in those cases in which checkpoint inhibitors are not efficacious, as is the case of ovarian cancer. Analysis of clinical samples from HGSOC reveals the importance of the tumor immune infiltrate as a prognostic factor. Immune correlates induced by SENTI-101 such as the chemokines CXCL9 and 10, IFNγ, dendritic cells, T cells, B cells, and tertiary lymphoid structures (TLS) have been correlated with better prognosis in patients with ovarian cancer (55, 68–70). Recently, multiple studies have shown the importance of the immune landscape and the presence of T cells and B cells organized in TLS in determining response to checkpoint inhibitors (51–54), further supporting the testing of SENTI-101 in solid tumors either as a single agent or in combination with checkpoint inhibitors. Importantly, SENTI-101 demonstrated antitumor activity in different solid tumor models with very distinct immune landscapes such as CT26 (immune rich or “hot”) and B16F10 (immune excluded or “cold”), which suggests a broad applicability for SENTI-101 in multiple solid tumors, where cytokines could be utilized to polarize and re-educate the tumor microenvironment resulting in robust immune-mediated tumor-cell killing.

A. GonzalezJunca reports other support from Senti Biosciences, Inc., during the conduct of the study; also has a patent for US10993967 issued and a patent for US20200206271 pending. F.D. Liu reports other support from Senti Biosciences, Inc. during the conduct of the study. A. Mullenix reports other support from Senti Biosciences, Inc. during the conduct of the study. C. Lee reports other support from Senti B Biosciences, Inc. during the conduct of the study. R.M. Gordley reports other support from Senti Biosciences, Inc., during the conduct of the study; also has a patent for US10993967 issued and a patent for US20200206271 pending. D.O. Frimannsson reports other support from Senti Biosciences, Inc. during the conduct of the study; also has a patent for US Patent Application Publication No. US20200206271 pending to SENTI BIOSCIENCE and US Patent No. US10993967 issued to SENTI BIOSCIENCE. O. Maller reports other support from Senti Biosciences, Inc. during the conduct of the study; other support from Adicet Bio outside the submitted work. D. Iyer reports other support from Senti Biosciences, Inc. during the conduct of the study. M. Tian reports other support from Biosciences, Inc. during the conduct of the study. R. Martinez reports other support from Senti Biosciences, Inc. during the conduct of the study. R. Savur reports other support from Senti Biosciences, Inc. during the conduct of the study. A. Perry-McNamara reports personal fees from Senti Biosciences, Inc. during the conduct of the study. D. Nguyen reports other support from Senti Biosciences, Inc. during the conduct of the study. N. Almudhfar reports other support from Senti Biosciences, Inc. during the conduct of the study. C. Blanco reports other support from Senti Biosciences, Inc. during the conduct of the study. C. Huynh reports other support from Senti Biosciences, Inc. during the conduct of the study. A. Nand reports other support from Senti Bioscience, Inc. during the conduct of the study. A. Magal reports other support from Senti Biosciences, Inc. during the conduct of the study. S. Mangalampalli reports other support from Senti Biosciences, Inc. during the conduct of the study. P.J. Lee reports other support from Biosciences, Inc. outside the submitted work; also has a patent for US20200206271 pending to Senti Biosciences, Inc. and US10993967 issued to Biosciences, Inc. T.K. Lu reports other support from Senti Biosciences, Inc. during the conduct of the study; also has a patent for US Patent Application Publication No. US20200206271 pending to Senti Biosciences, Inc. and a US Patent No. US10993967 issued to Senti Biosciences, Inc; and is a co-founder of Senti Biosciences Inc., Synlogic, Engine Biosciences, Tango Therapeutics, Corvium, BiomX, Eligo Biosciences, Bota.Bio, Avendesora, and NE47Bio; he also holds financial interests in nest.bio, Armata, IndieBio, CognitoHealth, Quark Biosciences, Personal Genomics, Thryve, Lexent Bio, MitoLab, Vulcan, Serotiny, Avendesora, Pulmobiotics, Provectus Algae, Invaio, NSG Biolabs. G. Lee reports other support from Senti Biosciences outside the submitted work. No disclosures were reported by the other authors.

A. Gonzalez-Junca: Conceptualization, formal analysis, supervision, investigation, methodology, writing–original draft, writing–review and editing. F.D. Liu: Conceptualization, formal analysis, investigation, methodology, writing–original draft. A.S. Nagaraja: Conceptualization, formal analysis, investigation, methodology. A. Mullenix: Formal analysis, investigation, methodology. C-T. Lee: Supervision, investigation, methodology. R.M. Gordley: Conceptualization, resources, supervision. D.O. Frimannsson: Conceptualization, resources. O. Maller: Conceptualization, methodology. B.S. Garrison: Conceptualization, methodology. D. Iyer: Supervision, methodology. A. Benabbas: Resources, methodology. T.A. Truong: Methodology. A. Quach: Methodology. M. Tian: Methodology. R. Martinez: Methodology. R. Savur: Data curation, methodology. A. Perry-McNamara: Methodology. D. Nguyen: Methodology. N. Almudhfar: Methodology. C. Blanco: Resources, methodology. C. Huynh: Resources. A. Nand: Resources. Y.-A. E. Lay: Resources, methodology. A. Magal: Resources, methodology. S. Mangalampalli: Resources, methodology. P.J. Lee: Conceptualization, supervision, funding acquisition. T.K. Lu: Conceptualization, supervision, funding acquisition. G. Lee: Conceptualization, supervision, funding acquisition, project administration, writing–review and editing.