Soluble PD-L1-secreting umbilical cord-derived mesenchymal stromal cells transiently attenuate established xenogeneic T cell activation in vivo Authors: Ryosuke Uchibori, Ken Ohmine, Keiya Ozawa Published: 2026 ABSTRACT Immune overactivation damages tissues in T cell-mediated disorders, and cell-based therapies are being developed to suppress T cell responses. Mesenchymal stromal cells (MSCs) modulate immunity; however, their suppressive effects depend on inflammatory cytokine priming. Umbilical cord-derived MSCs are attractive for non-invasive collection and expansion, but their effects diminish after intravenous infusion. PD-1/PD-L1 signaling lowers T cell activation; therefore, soluble PD-L1 (sPD-L1)-secreting MSCs may provide contact-independent checkpoint engagement and we tested whether sPD-L1-MSCs can modulate an established human T-cell response in vivo. Using a model in which human peripheral blood mononuclear cells were transferred into immunodeficient NOG mice, umbilical cord- or bone marrow-derived MSCs did not restrain established T cell activation, whereas sPD-L1-MSCs transiently blunted activation during dosing without reducing T cell numbers. In vitro, a neutralizing anti-PD-L1 antibody partially relieved this effect at early time points, consistent with PD-1/PD-L1 signaling involvement. Biodistribution imaging and ELISA showed predominant pulmonary trapping and short-lived sPD-L1 exposure after intravenous MSC administration. This limited and transient exposure may contribute to the variability in MSC-based immunomodulation. Therefore, in this late-intervention xenogeneic model, the timing, location, and magnitude of exposure to MSC-derived mediators strongly constrain efficacy, and checkpoint delivery from engineered MSCs can provide transient immunomodulatory activity.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-026-56402-y. Introduction Mesenchymal stromal cells (MSCs) have been widely investigated as a cellular therapy platform that combines tissue repair capacity with immunoregulatory activity. Among MSC sources, umbilical cord-derived MSCs (UC-MSCs) offer advantages over bone marrow-derived MSCs, including noninvasive procurement, greater primary expansion capacity, and lower immunogenicity, making them attractive for clinical translation1–3. However, MSC immunosuppression critically depends on the cell source, manufacturing conditions, and exposure to inflammatory cytokines, which contributes to variability and uncertainty in therapeutic efficacy4,5. Therefore, MSCs with tunable immunosuppressive functions that can be reproduced across batches are required. The PD-1/PD-L1 checkpoint lowers T cell activation via SHP2 recruitment to dampen PI3K-AKT and RAS-ERK signaling6–8. Although MSCs can upregulate PD-L1 in inflammatory milieus9–12, the magnitude and duration of expression depend on exposure to cytokines such as IFN-γ and TNF-α, creating batch-to-batch variability and complicating dose standardization. Together with uncertainty regarding the necessity for cell–cell contact and the effective range of action across MSC-secreted mediators, current MSC approaches make precise and timely control of T cell suppression difficult13–17. Previous studies have shown that PD-L1-engineered MSCs can suppress T cell function18,19. Here, we evaluated an approach to render UC-MSCs contact-independent and dose-programmable by engineering them to secrete sPD-L1. This strategy addresses two practical limitations of MSC therapy, namely cytokine-dependent variability and short in vivo persistence due to first-pass pulmonary sequestration20,21, by placing PD-1 engagement under the exogenous control of dosage and schedule rather than inflammation-driven PD-L1 induction. Because the effect can be neutralized with anti-PD-1/PD-L1 antibodies, pharmacologic control remains inherently reversible, supporting safety management. If validated in more disease-relevant settings, this platform may support MSC-based medicines for T cell-driven inflammatory conditions in which transient and controllable immunomodulation is desirable. In this study, we used three reporter systems with distinct purposes: iELuc-T cells as an activation-dependent reporter of the T-cell activation state, ELuc-T cells as a constitutive reporter reflecting the overall T-cell number, and ffLuc as a constitutive reporter to track the biodistribution of gene-modified MSCs in vivo. Results Unmodified UC-MSCs did not suppress established T-cell activation in vivo in either late- or earlier-intervention cohorts The T cell activation inhibitory effect of MSCs was evaluated in vivo using a mouse model. To monitor T cell activation, we used an expression cassette (iELuc) in which luciferase expression is driven by a T cell activation-inducible promoter (see Supplementary Fig. S1A online). A retroviral vector carrying this expression cassette was used to modify PBMCs from healthy donors genetically, creating genetically-modified T cells (iELuc-T cells) (see Supplementary Fig. S1B). In vitro, stimulation of iELuc-T cells with a T cell activation reagent (ImmunoCult™ Human CD3/CD28 T Cell Activator) significantly increased ELuc expression compared with non-activated T cells (see Supplementary Fig. S1C). Furthermore, after transfer into immunodeficient NOG mice, iELuc-T cells produced increased activation-dependent reporter output indicating progressive activation through xenogeneic recognition of murine tissues, as reflected by the increased activation-dependent reporter output (Fig. S1D, E). We first examined whether unmodified intravenously administered UC-MSCs could suppress an already established xenogeneic T-cell response in vivo. In a late-intervention cohort, treatment was initiated 21 days after the transfer of 2 × 106 iELuc-T cells, when activation-dependent bioluminescence was already strongly increasing (Fig. 1A–C). UC-MSCs were administered thrice, and no suppression of T cell activation was observed. Two mice in the UC-MSC-treated group died during the study, potentially reflecting infusion-associated events. We then examined a second, earlier-intervention cohort in which UC-MSC administration was initiated 14 days after iELuc-T-cell transfer, when activation was present but less advanced than that in cohort 1 (Fig. 1D–F). In this cohort, UC-MSCs were administered four times, but no suppression of T-cell activation was observed. No deaths occurred in the UC-MSC-treated group in cohort 2.Fig. 1Unmodified UC-MSCs did not suppress established T-cell activation in vivo in either late- or earlier-intervention cohorts. (A) Schematic timeline of cohort 1 (late-intervention cohort). NOG mice received 2 × 106 iELuc-T cells, and UC-MSCs (1 × 106 per dose, i.v.) were administered beginning 21 days later. MSCs were infused three times (D0, D2, and D4). (B) Representative longitudinal in vivo bioluminescence imaging of mice in cohort 1 after UC-MSC administration (n = 5 per group). (C) Quantification of whole-body photon flux in cohort 1 (n = 5 per group). Individual values are shown. Statistical analysis was performed using a two-way ANOVA followed by Šídák’s multiple-comparisons test. No statistically significant group differences were detected at any time point. (D) Schematic timeline of cohort 2 (earlier-intervention cohort). NOG mice received 2 × 106 iELuc-T cells, and UC-MSCs (1 × 106 per dose, i.v.) were administered beginning 14 days later. MSCs were infused four times (D0, D4, D7, and D11). (E) Representative longitudinal in vivo bioluminescence imaging of mice in cohort 2 after UC-MSC administration (n = 5 per group). (F) Quantification of whole-body photon flux in cohort 2 (n = 5 per group). Individual values are shown. Statistical analysis was performed using a two-way ANOVA followed by Šídák’s multiple-comparisons test. No statistically significant group differences were detected at any time point. Unmodified UC-MSCs did not suppress established T-cell activation in vivo in either late- or earlier-intervention cohorts. (A) Schematic timeline of cohort 1 (late-intervention cohort). NOG mice received 2 × 106 iELuc-T cells, and UC-MSCs (1 × 106 per dose, i.v.) were administered beginning 21 days later. MSCs were infused three times (D0, D2, and D4). (B) Representative longitudinal in vivo bioluminescence imaging of mice in cohort 1 after UC-MSC administration (n = 5 per group). (C) Quantification of whole-body photon flux in cohort 1 (n = 5 per group). Individual values are shown. Statistical analysis was performed using a two-way ANOVA followed by Šídák’s multiple-comparisons test. No statistically significant group differences were detected at any time point. (D) Schematic timeline of cohort 2 (earlier-intervention cohort). NOG mice received 2 × 106 iELuc-T cells, and UC-MSCs (1 × 106 per dose, i.v.) were administered beginning 14 days later. MSCs were infused four times (D0, D4, D7, and D11). (E) Representative longitudinal in vivo bioluminescence imaging of mice in cohort 2 after UC-MSC administration (n = 5 per group). (F) Quantification of whole-body photon flux in cohort 2 (n = 5 per group). Individual values are shown. Statistical analysis was performed using a two-way ANOVA followed by Šídák’s multiple-comparisons test. No statistically significant group differences were detected at any time point. We next examined whether the lack of effect observed with UC-MSCs was related to their tissue source. To address this, we evaluated bone marrow (BM)-MSCs alongside UC-MSCs. Each MSC type was seeded at 2.0 × 106 cells per T-225 flask and cultured for 10 days with medium changes until sub-confluence. No difference in proliferation was observed between the two MSC types (Fig. 2A). MSC administration was initiated 14 days after transfer of iELuc-T cells into NOG mice (Fig. 2B), and MSCs were administered four times on day (D)0, D4, D7, and D11. Compared with the control group, neither UC-MSCs nor BM-MSCs significantly suppressed activation-dependent bioluminescence (Fig. 2C, D). These results indicate that, under these conditions, the lack of T-cell suppression was not explained by MSC tissue source.Fig. 2BM-MSCs with no in vivo suppression of established T cell activation. (A) Fold expansion of UC-MSCs and BM-MSCs from seeding to harvest. MSCs at passage 4 were seeded at 2 × 10⁶ cells per T-225 flask and cultured for 10 days (with medium changes on D4 and D7; n = 13 per group, mean ± SD). A two-tailed unpaired t-test showed no statistically significant difference between the groups (P = 0.5918). (B) Schematic timeline of the experiment. UC-MSCs or BM-MSCs (1 × 10⁶ per dose, i.v.) were infused into NOG mice that had received 2 × 10⁶ iELuc-T cells 14 days prior. MSCs were infused four times (D0, D4, D7, and D11). (C) Representative longitudinal in vivo bioluminescence imaging of NOG mice after each MSC infusion (n = 5 per group). (D) Quantification of whole-body photon flux over time (n = 5 per group). Statistical analysis was performed using a two-way ANOVA followed by Šídák’s multiple-comparisons test. No statistically significant group differences were detected at any time point. BM-MSCs with no in vivo suppression of established T cell activation. (A) Fold expansion of UC-MSCs and BM-MSCs from seeding to harvest. MSCs at passage 4 were seeded at 2 × 10⁶ cells per T-225 flask and cultured for 10 days (with medium changes on D4 and D7; n = 13 per group, mean ± SD). A two-tailed unpaired t-test showed no statistically significant difference between the groups (P = 0.5918). (B) Schematic timeline of the experiment. UC-MSCs or BM-MSCs (1 × 10⁶ per dose, i.v.) were infused into NOG mice that had received 2 × 10⁶ iELuc-T cells 14 days prior. MSCs were infused four times (D0, D4, D7, and D11). (C) Representative longitudinal in vivo bioluminescence imaging of NOG mice after each MSC infusion (n = 5 per group). (D) Quantification of whole-body photon flux over time (n = 5 per group). Statistical analysis was performed using a two-way ANOVA followed by Šídák’s multiple-comparisons test. No statistically significant group differences were detected at any time point. Lentiviral vector-mediated gene modification of UC-MSCs Simply administering MSCs is insufficient to suppress already activated T cells; therefore, we attempted to create genetically-modified MSCs with enhanced T cell suppression capacity. To achieve this, we established a gene delivery method using lentiviral vectors. First, 2 × 106 UC-MSCs were seeded into T-225 flasks and cultured for 3 days (see Supplementary Fig. S2A). To achieve efficient gene delivery, we employed the ViraDuctin™ Lentivirus Transduction System. A complex of chemical polymers and ZsGreen1-expressing lentiviral vectors was added to the culture medium, and the flasks were incubated for 24 h. The amount of viral vector used for gene delivery was 1.0 × 109 copies per T-225 flask. The next day (day 4, D4), the medium was completely aspirated, and the flasks were washed. The medium was then replenished, and culturing was continued until D10, when the cells were harvested. Compared with the non-transduced control group (Mock), the transduced MSCs showed a slightly lower proliferation rate (10.4 ± 1.1-fold vs 7.2 ± 0.8-fold) (see Supplementary Fig. S2B). Fluorescent microscopy revealed ZsGreen1 fluorescence under excitation light (see Supplementary Fig. S2C), and the proportion of ZsGreen1-positive cells was also high (0 ± 0% vs 90.8 ± 6.9%) (see Supplementary Fig. S2D, E). Biodistribution, circulating sPD-L1 exposure, and transient modulation of established T-cell activation by gene-modified UC-MSCs We evaluated the biodistribution and circulating exposure associated with intravenously administered gene-modified MSCs. After establishing the general feasibility of lentiviral transduction with a ZsGreen1 reporter construct, we characterized UC-MSCs transduced with the sPD-L1-containing constructs used in subsequent functional experiments. Firefly luciferase (ffLuc) expression lentivirus vectors and lentivirus vectors co-expressing ffLuc and soluble PD-L1 (sPD-L1) were constructed and used to genetically transduce UC-MSCs, following the procedure described above (see Supplementary Fig. S2A). The cell proliferation rate (Fig. 3A) was slightly lower in the ffLuc and ffLuc/sPD-L1 groups than in the Mock cells, consistent with the ZsGreen1 experiments. ELISA confirmed robust sPD-L1 secretion from ffLuc/sPD-L1-MSC supernatants (Fig. 3B). Next, to monitor the in vivo dynamics of tail vein-injected UC-MSCs, ffLuc-expressing or ffLuc/sPD-L1-transduced UC-MSCs (1 × 106 cells) were administered 14 days after the infusion of activated T cells (non-luminescent) into NOG mice, followed by time-lapse in vivo imaging (Fig. 3C). MSCs were administered four times at D0, D4, D7, and D11. Both showed marked pulmonary sequestration immediately after intravenous administration, with minimal redistribution beyond the lungs and signals declining within days (Fig. 3D, E). Peripheral blood ELISA analysis detected circulating sPD-L1 during the MSC administration period, although the measured concentrations varied across time points and fell below the detection limit after dosing had ceased (Fig. 3F).Fig. 3Biodistribution and circulating sPD-L1 exposure after administration of gene-modified MSCs. (A) Fold expansion of non-transduced (mock), ffLuc-transduced, and ffLuc/sPD-L1-transduced UC-MSCs from seeding to harvest. UC-MSCs at passage 4 were seeded at 2 × 10⁶ cells per T-225 flask and cultured for 3 days. UC-MSCs were transduced with ffLuc or ffLuc/sPD-L1-expressing lentiviral vector (1.0 × 109 copies per flask) with transduction enhancer on D3. Transduced cells were harvested on D10 (n = 13 per group, mean ± SD). Statistical analysis was performed using a one-way ANOVA followed by Tukey’s multiple-comparisons test (***p < 0.001, ****p < 0.0001). (B) PD-L1 secretion in the cell culture supernatant. Soluble PD-L1 (sPD-L1) levels were measured in the conditioned medium from cells using a human PD-L1 sandwich ELISA kit (n = 13 per group, mean ± SD). Statistical analysis was performed using a one-way ANOVA followed by Tukey’s multiple-comparisons test (****p < 0.0001). (C) Schematic timeline of the experiment. Gene-modified UC-MSCs (1 × 10⁶ per dose, i.v.) were infused into NOG mice that had been injected with 2 × 106 T cells 14 days prior. MSCs were infused four times (D0, D4, D7 and D11). Blood samples were collected to measure PD-L1 levels (D2, D9, and D15). (D) Representative longitudinal in vivo bioluminescence from gene-modified MSCs of NOG mice (n = 10 per group). (E) Quantification of the pulmonary photon flux (n = 10 per group, mean ± SD). The arrows indicate the time point of MSC administration. Statistical analysis was performed using a two-way ANOVA followed by Šídák’s multiple-comparisons test. No statistically significant group differences were detected at any time point. (F) Measurement of sPD-L1 levels in the blood. sPD-L1 was measured in the blood using a human PD-L1 sandwich ELISA kit (n = 3 per group at each time point, mean ± SD). Statistical analysis was performed using a two-way ANOVA followed by Šídák’s multiple-comparisons test (**** p < 0.0001). Biodistribution and circulating sPD-L1 exposure after administration of gene-modified MSCs. (A) Fold expansion of non-transduced (mock), ffLuc-transduced, and ffLuc/sPD-L1-transduced UC-MSCs from seeding to harvest. UC-MSCs at passage 4 were seeded at 2 × 10⁶ cells per T-225 flask and cultured for 3 days. UC-MSCs were transduced with ffLuc or ffLuc/sPD-L1-expressing lentiviral vector (1.0 × 109 copies per flask) with transduction enhancer on D3. Transduced cells were harvested on D10 (n = 13 per group, mean ± SD). Statistical analysis was performed using a one-way ANOVA followed by Tukey’s multiple-comparisons test (***p < 0.001, ****p < 0.0001). (B) PD-L1 secretion in the cell culture supernatant. Soluble PD-L1 (sPD-L1) levels were measured in the conditioned medium from cells using a human PD-L1 sandwich ELISA kit (n = 13 per group, mean ± SD). Statistical analysis was performed using a one-way ANOVA followed by Tukey’s multiple-comparisons test (****p < 0.0001). (C) Schematic timeline of the experiment. Gene-modified UC-MSCs (1 × 10⁶ per dose, i.v.) were infused into NOG mice that had been injected with 2 × 106 T cells 14 days prior. MSCs were infused four times (D0, D4, D7 and D11). Blood samples were collected to measure PD-L1 levels (D2, D9, and D15). (D) Representative longitudinal in vivo bioluminescence from gene-modified MSCs of NOG mice (n = 10 per group). (E) Quantification of the pulmonary photon flux (n = 10 per group, mean ± SD). The arrows indicate the time point of MSC administration. Statistical analysis was performed using a two-way ANOVA followed by Šídák’s multiple-comparisons test. No statistically significant group differences were detected at any time point. (F) Measurement of sPD-L1 levels in the blood. sPD-L1 was measured in the blood using a human PD-L1 sandwich ELISA kit (n = 3 per group at each time point, mean ± SD). Statistical analysis was performed using a two-way ANOVA followed by Šídák’s multiple-comparisons test (**** p < 0.0001). Next, we tested whether UC-MSCs engineered to secrete sPD-L1 could attenuate the activation of an established xenogeneic T-cell response in vivo. We constructed an sPD-L1 lentiviral vector and genetically transduced UC-MSCs using the procedure described previously (see Supplementary Fig. S2A). As previously observed, the proliferation rate of sPD-L1-transduced MSCs was slightly lower than that of non-transduced UC-MSCs (Mock) (Fig. 4A). Total RNA was extracted from cells harvested on D10, and gene expression analysis by RT-qPCR confirmed the amplification of the PD-L1 gene in sPD-L1-transduced UC-MSCs (Fig. 4B). Furthermore, ELISA analysis of the culture supernatant confirmed sPD-L1 secretion (Fig. 4C). NOG mice infused with 2 × 106 iELuc-T cells were administered either non-genetically modified UC-MSCs (Mock) or sPD-L1-expressing UC-MSCs at 1 × 106 cells, and in vivo imaging was performed over time (Fig. 4D). MSCs were administered four times at D0, D4, D7, and D11. The increase in bioluminescence intensity was significantly suppressed in the sPD-L1 group at D4, D7, and D9 (Fig. 4E, F). ELISA analysis of peripheral blood detected sPD-L1 levels throughout the MSC administration period (Fig. 4G). However, no statistically significant differences were observed among the three groups in terms of body weight changes or survival analysis (Fig. 4H, I). To confirm that the suppression of luminescence intensity on D7 was not due to suppression of T cell number, similar observations were performed in NOG mice infused with ELuc-expressing T cells (Fig. 4J). In this constitutive-expression model (ELuc-T cells), luminescence correlates with T cell number. Time-course analysis of in vivo imaging revealed no statistically significant differences in luminescence intensity among the three groups at any time point during the observation period (Fig. 4K, L). Therefore, the suppression of the increase in luminescence intensity observed in the sPD-L1 group at D7 is consistent with the inhibition of T cell activation rather than a reduction in T cell numbers.Fig. 4Transient delay in T cell activation progression with sPD-L1-expressing UC-MSCs. (A) Fold expansion of non-transduced (mock) and sPD-L1-transduced UC-MSCs from seeding to harvest. UC-MSCs at passage 4 were seeded at 2 × 10⁶ cells per T-225 flask and cultured for 3 days. UC-MSCs were transduced with sPD-L1-expressing lentiviral vector (1.0 × 109 copies per flask) with transduction enhancer on D3. Transduced cells were harvested on D10 (n = 31 per group, mean ± SD). Statistical analysis was performed using a two-tailed unpaired t-test (****p < 0.0001). (B) Measurement of PD-L1 mRNA at harvest using RT-qPCR. Expression data were normalized to GAPDH (n = 8 per group, mean ± SD). Statistical analysis was performed using a two-tailed unpaired t-test (****p < 0.0001). (C) Assessment of PD-L1 secretion in the cell culture supernatant at harvest by human PD-L1 sandwich ELISA (n = 31 per group, mean ± SD). Statistical analysis was performed using a two-tailed unpaired t-test (****p < 0.0001). The dotted line in the graph indicates the lower limit of the detection sensitivity. (D) Experimental timeline for validating the suppressive effect of sPD-L1-expressing MSCs on T cell activation progression using activation-dependent reporter T cells (iELuc-T cells). (E) Representative longitudinal in vivo bioluminescence from the iELuc-T cells of NOG mice after each MSC infusion (n = 10 per group). (F) Quantification of the whole-body photon flux (n = 10 per group, mean ± SD). Statistical analysis was performed using a two-way ANOVA followed by Šídák’s multiple-comparisons test (*p < 0.05, **p < 0.01, ***p < 0.001). (G) Circulating sPD-L1 concentrations at the indicated time points (n = 3 per group at each time point, mean ± SD). Statistical analysis was performed using a two-way ANOVA followed by Šídák’s multiple-comparisons test (****p < 0.0001). The dotted line in the graph indicates the lower limit of the detection sensitivity. (H) Body weight changes in mice in the in vivo animal experiment. Individual values are shown (n = 10 per group). Statistical analysis was performed using a two-way ANOVA followed by Šídák's multiple-comparisons test. No statistically significant group differences were detected at any time point. (I) Kaplan–Meier curve showing the survival of mice in each group (n = 10 per group). Survival curves were compared using the log-rank test. No statistically significant difference was detected between the groups. (J) Experimental timeline for verifying the effect of sPD-L1-expressing MSCs on in vivo T cell number using a constitutive reporter proportional to the cell count (ELuc-T cells). (K) Representative longitudinal in vivo bioluminescence from ELuc-T cells in NOG mice after each MSC infusion (n = 5 per group). (L) Quantification of the whole-body photon flux (n = 5 per group, mean ± SD). A two-way ANOVA followed by Šídák’s multiple-comparisons test detected no statistically significant group differences at any time point. Transient delay in T cell activation progression with sPD-L1-expressing UC-MSCs. (A) Fold expansion of non-transduced (mock) and sPD-L1-transduced UC-MSCs from seeding to harvest. UC-MSCs at passage 4 were seeded at 2 × 10⁶ cells per T-225 flask and cultured for 3 days. UC-MSCs were transduced with sPD-L1-expressing lentiviral vector (1.0 × 109 copies per flask) with transduction enhancer on D3. Transduced cells were harvested on D10 (n = 31 per group, mean ± SD). Statistical analysis was performed using a two-tailed unpaired t-test (****p < 0.0001). (B) Measurement of PD-L1 mRNA at harvest using RT-qPCR. Expression data were normalized to GAPDH (n = 8 per group, mean ± SD). Statistical analysis was performed using a two-tailed unpaired t-test (****p < 0.0001). (C) Assessment of PD-L1 secretion in the cell culture supernatant at harvest by human PD-L1 sandwich ELISA (n = 31 per group, mean ± SD). Statistical analysis was performed using a two-tailed unpaired t-test (****p < 0.0001). The dotted line in the graph indicates the lower limit of the detection sensitivity. (D) Experimental timeline for validating the suppressive effect of sPD-L1-expressing MSCs on T cell activation progression using activation-dependent reporter T cells (iELuc-T cells). (E) Representative longitudinal in vivo bioluminescence from the iELuc-T cells of NOG mice after each MSC infusion (n = 10 per group). (F) Quantification of the whole-body photon flux (n = 10 per group, mean ± SD). Statistical analysis was performed using a two-way ANOVA followed by Šídák’s multiple-comparisons test (*p < 0.05, **p < 0.01, ***p < 0.001). (G) Circulating sPD-L1 concentrations at the indicated time points (n = 3 per group at each time point, mean ± SD). Statistical analysis was performed using a two-way ANOVA followed by Šídák’s multiple-comparisons test (****p < 0.0001). The dotted line in the graph indicates the lower limit of the detection sensitivity. (H) Body weight changes in mice in the in vivo animal experiment. Individual values are shown (n = 10 per group). Statistical analysis was performed using a two-way ANOVA followed by Šídák's multiple-comparisons test. No statistically significant group differences were detected at any time point. (I) Kaplan–Meier curve showing the survival of mice in each group (n = 10 per group). Survival curves were compared using the log-rank test. No statistically significant difference was detected between the groups. (J) Experimental timeline for verifying the effect of sPD-L1-expressing MSCs on in vivo T cell number using a constitutive reporter proportional to the cell count (ELuc-T cells). (K) Representative longitudinal in vivo bioluminescence from ELuc-T cells in NOG mice after each MSC infusion (n = 5 per group). (L) Quantification of the whole-body photon flux (n = 5 per group, mean ± SD). A two-way ANOVA followed by Šídák’s multiple-comparisons test detected no statistically significant group differences at any time point. Early in vitro attenuation of the sPD-L1-MSC-associated suppressive effect by an anti-PD-L1 antibody To examine whether the in vitro suppressive effect associated with sPD-L1-expressing UC-MSCs was consistent with involvement of PD-1/PD-L1 signaling, we performed a controlled in vitro co-culture experiment. This in vitro assay was designed to assess the effect of early exposure to MSC-derived sPD-L1 under controlled conditions and was not intended to fully reproduce the timing of the in vivo intervention. First, 2 × 104 iELuc-T cells were mixed with either genetically unmodified UC-MSCs (Mock) or sPD-L1-expressing MSCs at MSC:T cell ratios of 0.1:1, 0.2:1, and 1:1. Then, the mixture was seeded into a 96-well plate, and a T cell activation reagent was added after 12 h to induce T cell activation (Fig. 5A). Equivalent increases in luminescence intensity were observed in wells seeded with T cells alone and in wells co-cultured with Mock cells, indicating that the Mock group had no inhibitory effect on activation (Fig. 5B). In contrast, co-culture with sPD-L1-expressing UC-MSCs showed dose-dependent attenuation of luminescence enhancement, which was most apparent at earlier time points after stimulation. ELISA analysis of culture supernatants from the co-culture group with sPD-L1-expressing UC-MSCs detected MSC dose-dependent sPD-L1 levels (Fig. 5C). For the blockade experiment, anti-PD-L1 antibody was added at 0 h and again at 12 h after the initiation of co-culture, and the activation reagent was added at 12 h (Fig. 5D). Under these conditions, antibody treatment partially relieved the suppressive effect associated with sPD-L1-MSCs at early time points after stimulation, whereas the difference became less evident at later time points (Fig. 5E).Fig. 5Partial relief of the early sPD-L1-MSC-associated suppressive effect by anti-PD-L1 antibody in vitro. (A) Experimental timeline for validating the suppressive effect of sPD-L1-expressing MSCs on T cell activation progression in vitro using activation-dependent reporter T cells (iELuc-T cells). Non-transduced (mock) or sPD-L1-transduced UC-MSCs were cocultured with iELuc-T cells at MSC:T cell (E:T) ratios of 0.1:1, 0.2:1, or 1:1; T cells alone served as a control. T cell activation reagents were added to the culture 12 h after the co-culture. (B) Dose-dependent attenuation of T-cell activation by sPD-L1-expressing UC-MSCs. Attenuation was most apparent at earlier time points after stimulation. The luciferase assay was performed at the indicated time points after the addition of T cell activation reagents to the culture (n = 4 per group, mean ± SD). Statistical analysis was performed using a one-way ANOVA followed by Holm-Šídák’s multiple-comparisons test (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001). (C) Assessment of sPD-L1 in the cell culture supernatant at the indicated time by human PD-L1 sandwich ELISA. Soluble PD-L1 (sPD-L1) levels were measured in the conditioned medium from cells using a human PD-L1 sandwich ELISA kit (n = 4 per group, mean ± SD). Statistical analysis was performed using a one-way ANOVA followed by Holm-Šídák’s multiple-comparisons test (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001). (D) Experimental timeline for assessing the early effects of an anti-PD-L1 antibody on the sPD-L1-MSC-associated suppressive phenotype. The anti-PD-L1 antibody (2 µg/well) was added at 0 and 12 h; T cell activation reagent was added at 12 h. (E) Partial relief of the early sPD-L1-MSC-associated suppressive effect by an anti-PD-L1 antibody at early time points after stimulation. The luciferase assay was performed at the indicated time points after the addition of T cell activation reagents to the culture (n = 4 per group, mean ± SD). Statistical analysis was performed using a one-way ANOVA followed by Holm-Šídák’s multiple-comparisons test (*p < 0.05; **p < 0.01). Partial relief of the early sPD-L1-MSC-associated suppressive effect by anti-PD-L1 antibody in vitro. (A) Experimental timeline for validating the suppressive effect of sPD-L1-expressing MSCs on T cell activation progression in vitro using activation-dependent reporter T cells (iELuc-T cells). Non-transduced (mock) or sPD-L1-transduced UC-MSCs were cocultured with iELuc-T cells at MSC:T cell (E:T) ratios of 0.1:1, 0.2:1, or 1:1; T cells alone served as a control. T cell activation reagents were added to the culture 12 h after the co-culture. (B) Dose-dependent attenuation of T-cell activation by sPD-L1-expressing UC-MSCs. Attenuation was most apparent at earlier time points after stimulation. The luciferase assay was performed at the indicated time points after the addition of T cell activation reagents to the culture (n = 4 per group, mean ± SD). Statistical analysis was performed using a one-way ANOVA followed by Holm-Šídák’s multiple-comparisons test (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001). (C) Assessment of sPD-L1 in the cell culture supernatant at the indicated time by human PD-L1 sandwich ELISA. Soluble PD-L1 (sPD-L1) levels were measured in the conditioned medium from cells using a human PD-L1 sandwich ELISA kit (n = 4 per group, mean ± SD). Statistical analysis was performed using a one-way ANOVA followed by Holm-Šídák’s multiple-comparisons test (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001). (D) Experimental timeline for assessing the early effects of an anti-PD-L1 antibody on the sPD-L1-MSC-associated suppressive phenotype. The anti-PD-L1 antibody (2 µg/well) was added at 0 and 12 h; T cell activation reagent was added at 12 h. (E) Partial relief of the early sPD-L1-MSC-associated suppressive effect by an anti-PD-L1 antibody at early time points after stimulation. The luciferase assay was performed at the indicated time points after the addition of T cell activation reagents to the culture (n = 4 per group, mean ± SD). Statistical analysis was performed using a one-way ANOVA followed by Holm-Šídák’s multiple-comparisons test (*p < 0.05; **p < 0.01). Discussion Extensive in vitro studies and multiple preclinical models have established that MSCs can suppress T cell activation, proliferation, and effector function via indoleamine 2,3-dioxygenase (IDO), prostaglandin E₂ (PGE₂), TGF-β, IL-10, nitric oxide, and contact-dependent mechanisms23–26. These immunosuppressive programs are licensed by pro-inflammatory cytokines, especially IFN-γ together with TNF-α or other co-factors, which markedly reshape the MSC secretome and upregulate immunoregulatory molecules including PD-L110,27–30. Recent studies have shown that PD-L1 levels on MSCs and MSC-derived extracellular vesicles (EVs) are key determinants of their capacity to restrain T cell cytokine production and that anti-PD-L1 antibodies can reverse MSC-mediated suppression29–32. The fact that unmodified MSCs in our model did not attenuate T cell activation is unlikely to reflect a fundamental inability of MSCs to modulate T cell responses but rather a mismatch between the timing and magnitude of MSC-derived inhibitory cues and the advanced activation state of effector T cells at the time of intervention. Once T cells progress beyond the early activation phase, autocrine IL-2 signaling through high-affinity IL-2 receptors (CD25) creates a self-reinforcing positive feedback loop that stabilizes effector differentiation and diminishes susceptibility to external inhibitory signals33,34. In addition, early after T cell transfer, activation-dependent iELuc signals in a subset of mice remained close to background levels as xenoreactive T cells continued to expand. Therefore, we restricted our intervention and quantitative analyses to time points at which bioluminescence had clearly increased, reasoning that early interventions based on low, near-threshold signals would be difficult to interpret and might not reliably predict subsequent T cell dynamics. In our late-intervention xenogeneic human T-cell activation model, MSC administration was initiated 14–21 days after T-cell transfer, when xenoreactive T cells were already highly expanded and activation-dependent luminescence was rapidly increasing. This represents a stringent intervention context rather than a direct surrogate for clinical allogeneic graft-versus-host disease (GVHD). Concurrently, intravenously infused MSCs undergo pronounced pulmonary first-pass sequestration, with limited redistribution beyond the lungs, where viable cells are rarely detected after 24–48 h20,21,26. In our imaging experiments, ffLuc-labeled MSCs exhibited rapid pulmonary sequestration with limited redistribution, and dose escalation or tighter dosing intervals were constrained by infusion-related mortality, consistent with concerns that large, adhesive MSCs in small pulmonary vessels can precipitate microembolic and hemodynamic events during intravenous administration35. Absolute pulmonary photon flux values immediately after each infusion may also vary despite identical nominal cell numbers because the measured signal can be influenced by the injection efficiency, initial distribution within the pulmonary vasculature, and imaging-related variability. Therefore, longitudinal interpretation should focus on relative trends rather than direct comparisons of absolute photon flux values across administrations. Taken together, these factors likely resulted in insufficient integrated exposure of activated T cells to MSC-derived suppressive mediators, explaining why neither UC-MSCs nor BM-MSCs altered the trajectory of an established xenogeneic response. Given the predominant pulmonary sequestration observed after intravenous administration, the transient systemic effect observed in this study is more plausibly explained by exposure to circulating sPD-L1 released from lung-trapped MSCs than by the widespread redistribution of viable MSCs to lymphoid organs. However, this remains a working hypothesis because we did not directly link pulmonary MSC activity to systemic sPD-L1 exposure and downstream T-cell modulation in the present study. This interpretation also highlights a limitation of the intravenous route and supports evaluation of alternative delivery routes or homing-enhancing strategies in future studies. Accordingly, the present results should be interpreted within the constraints of a late-intervention xenogeneic human T-cell activation model and a bioluminescence-based readout, and they do not exclude the potential benefits of MSCs in settings with earlier intervention, lower inflammatory burden, organ-specific disease, or alternative delivery routes. These observations underscore the utility and limitations of xenogeneic human T-cell transfer models for evaluating MSC-based immunomodulation. Such models provide a sensitive in vivo readout of human T-cell activation driven by recognition of murine tissues and have therefore been widely used to investigate human T-cell biology and immunotherapeutic responses36–39. However, the same features that make these models experimentally sensitive—accelerated T-cell expansion, intense systemic inflammation, and dependence on a dominant effector population—also compress the therapeutic window for immunomodulators, which require time and repeated dosing to exert their effects. In our study, increasing the MSC dose or dosing frequency aggravated infusion-associated death, further narrowing the feasible dosing range for intravenous MSC therapy in mice. Thus, while our results indicate that unmodified MSCs have a limited capacity to modulate an already established xenogeneic human T-cell response under these stringent conditions, they do not exclude potential benefits in settings with earlier intervention, lower inflammatory burden, organ-specific disease, or alternative delivery routes. These considerations also support the use of complementary allogeneic or conditioning-based models to directly assess the translational relevance of this approach. Using sPD-L1-engineered MSCs, we obtained supportive evidence that checkpoint delivery from an MSC platform can transiently dampen T cell activation in vivo under these conditions. The PD-1/PD-L1 axis attenuates TCR and CD28 signaling via the recruitment of phosphatases to immunoreceptor complexes, leading to reduced PI3K/AKT and RAS/ERK pathway activity and diminished cytokine production29,30,40. Our finding that activation-dependent luminescence was suppressed at time points when sPD-L1 was detectable in circulation and that this effect was absent in ELuc-reporter mice, in which luminescence reflects T cell number, indicates that sPD-L1 reduced the activation state of T cells rather than causing cytotoxic depletion. Partial relief of the early suppressive effect by an anti-PD-L1 antibody in vitro is consistent with the involvement of PD-1/PD-L1 signaling; however, it should not be interpreted as definitive proof of the mechanism underlying the in vivo observations. This result provides supportive evidence from a controlled in vitro assay and is consistent with prior studies implicating PD-L1 as a contributor to MSC-associated T-cell suppression29,30,32. Because the in vitro co-culture assay exposed T cells to MSC-derived sPD-L1 before activation, it should be interpreted as a controlled assay of early checkpoint exposure rather than a direct mimic of the late-intervention sequence used in vivo. Accordingly, the in vitro findings should be taken as supportive evidence for biological plausibility, while their relevance to already activated T-cell responses in vivo remains to be defined. Moreover, studies using PD-L1-Fc fusion proteins or soluble PD-L1 variants have shown that PD-1 engagement by secreted PD-L1 can prolong graft survival and dampen effector T cell responses in vivo, confirming that secreted PD-L1, at sufficient concentration, can function as an agonist checkpoint ligand39,41–43. Thus, our data are consistent with the broader PD-1/PD-L1 literature but also highlight that MSC-delivered sPD-L1 is constrained by the short systemic persistence imposed by the intravenous route. Plasma sPD-L1 was measured in a limited number of animals at each time point; therefore, differences between post-dose measurements should be interpreted cautiously and may reflect both biological and technical variability. Accordingly, these measurements should not be interpreted as establishing reproducible pharmacokinetic behavior across dosing cycles. A direct comparison with recombinant soluble PD-L1 or PD-L1-Fc was not included in this study; therefore, the present data do not establish whether MSC-based delivery provides advantages over direct protein administration in terms of potency, tissue exposure, and pharmacokinetics. A possible advantage of MSC-based delivery over the direct administration of recombinant sPD-L1 is that if MSCs can accumulate or persist at inflamed sites, they may provide more localized checkpoint exposure than a freely circulating protein. However, in the present intravenous model, the pronounced pulmonary sequestration of MSCs likely limited our ability to directly examine this possibility. Accordingly, future comparisons with recombinant soluble PD-L1 should ideally be performed in experimental settings that retain a vascular delivery context while allowing a better assessment of tissue-directed distribution beyond the lungs. Therefore, any potential advantage of MSC-based delivery over direct protein administration should be regarded as hypothetical and remains to be established experimentally. We also did not assess additional mechanistic readouts, such as T-cell proliferation, circulating cytokines, CD8+ T-cell activation status, cytotoxicity markers, or regulatory T-cell frequencies. The current data therefore support transient attenuation of activation-dependent reporter output, while broader immunologic effects remain to be defined. Regarding other engineered MSC approaches, IL-10-overexpressing MSCs enhance anti-inflammatory cytokine exposure and demonstrate improved control of GVHD and other inflammatory diseases in preclinical models, including IL-10 mRNA-engineered MSCs that retain potency after cryopreservation28,44. FasL-expressing MSCs further exploit death receptor pathways to delete effector T cells and ameliorate acute GVHD in murine models, while preserving MSC identity45. Recently, CXCR4/IL-10-coexpressing MSCs have been shown to combine strengthened immunosuppression with enhanced homing to SDF-1α-rich inflamed tissues, producing superior outcomes in humanized xenogeneic-GVHD models compared with wild-type MSCs6. In comparison, our sPD-L1-MSC platform delivers a mechanistically clean, drug-like inhibitory signal that is intrinsically reversible by approved anti-PD-1/PD-L1 antibodies. However, the transient nature of its effect suggests that combining checkpoint agonism with strategies to prolong MSC residence time in target tissues or with alternative delivery routes that increase exposure at disease-relevant sites could substantially increase its therapeutic impact. Clinically, these findings should be interpreted against an evolving landscape in which MSC therapies for steroid-refractory acute GVHD (SR-aGVHD) have moved from early trials to regulatory approval and yet continue to show heterogeneous efficacy. In Japan, Temcell HS Inj (BM-MSC) was the first regenerative medical product approved for SR-aGVHD under the new regulatory framework, and post-marketing data showed meaningful but not universal responses, with efficacy influenced by disease stage and patient characteristics47. Recently, remestemcel-L (Ryoncil) was approved in the USA for pediatric SR-aGVHD, supported by a high D28 overall response rate and encouraging survival; however, confirmatory studies highlight variability across subgroups and the need for optimized dosing48,49. Reviews consistently highlight that manufacturing heterogeneity, variable in vivo licensing, and uncertainty about optimal schedules limit predictability and may explain why many MSC trials still in phases I/II. Our data are mechanistically concordant with this view: unmodified MSCs administered late and intravenously cannot reliably quell a mature IL-2-driven T cell response, whereas rationally engineered pharmacologically controllable MSCs, such as sPD-L1-MSCs, can transiently tune activation but only within the constraints of MSC biodistribution. Conclusions In this late-intervention xenogeneic human T-cell activation model, neither unmodified UC- nor BM-MSCs suppressed an established T cell response, whereas UC-MSCs engineered to secrete soluble PD-L1 produced a transient reduction in T cell activation. This effect was accompanied by detectable circulating sPD-L1 and was supported by early in vitro antibody-blocking data, consistent with the involvement of PD-1/PD-L1 signaling. Together with biodistribution data showing predominant pulmonary trapping and short in vivo persistence after intravenous infusion, these findings indicate that efficacy is tightly constrained by the timing, location, and magnitude of exposure to MSC-derived mediators. Further studies in more disease-relevant models and with optimized delivery strategies are required to assess the translational relevance of this approach to GVHD and related T cell-mediated disorders. Materials and methods Cell culture Umbilical cord-derived MSCs (UC-MSCs) were provided by The Institute of Medical Science, The University of Tokyo, Cord blood and cORD bank (IMSUT CORD) (Tokyo, Japan). Bone marrow-derived human MSCs (BM-MSCs) were purchased from Lonza (Walkersville, MD, USA). MSCs were cultured in MesenPRO RS™ medium (Gibco, MA, USA) supplemented with MesenPRO RS™ growth supplement, 100 U/mL penicillin, and 100 µg/mL streptomycin. UC-MSCs and BM-MSCs were obtained from established sources and expanded under standard adherent MSC culture conditions. Formal de novo flow cytometric characterization of MSC surface markers was not performed in the present study. HEK-293 T cells (RIKEN BRC, Ibaraki, Japan) were maintained in DMEM/F-12 (Gibco) with 10% fetal bovine serum (Gibco) and penicillin–streptomycin. All cell lines were tested and confirmed negative for mycoplasma contamination. Peripheral blood (30 mL) was obtained from three independent healthy donors after obtaining written informed consent [approval No. 22-050]. Peripheral blood mononuclear cells (PBMCs) were isolated using Vacutainer® CPT™ tubes (BD, Franklin Lakes, NJ, USA) and washed twice with CELLOTION™ (Takara Bio Inc., Shiga, Japan). Cells were cryopreserved in a medium composed of CP-1 (Kyokuto Pharmaceutical, Japan), RPMI-1640, and human serum albumin (Albuminar, CSL Behring, Marburg, Germany), and stored in liquid nitrogen until use. For culture, PBMCs were maintained in GT-T503 (Takara Bio Inc.) supplemented with 0.6% autologous plasma, 1 × Antibiotic–Antimycotic (Gibco), and 175 IU/mL IL-2 (Immunase, Shionogi, Japan). All cultures were maintained at 37 °C and 5% CO₂. Plasmid construction A self-inactivating (SIN) retroviral reporter plasmid (iELuc) containing four NFAT-responsive elements, the minimal IL-2 promoter, and an ELuc gene (Toyobo, Japan) was prepared as described previously22. An ELuc control retroviral plasmid (ELuc) was also generated. Lentiviral transfer plasmids expressing ZsGreen1, ffLuc, sPD-L1 (extracellular domain, aa 1–231), or ffLuc/sPD-L1 (ffLuc and sPD-L1 were fused with a P2A peptide) were constructed in pLVSIN (Takara Bio Inc.) under the human EF-1α promoter. ZsGreen1 (Takara Bio Inc.) and ffLuc (Promega, USA) cDNAs were PCR-amplified and cloned into pLVSIN using Infusion cloning (Takara Bio Inc.). The sPD-L1 cDNA was obtained by RT-PCR (PrimeScript™ RT-PCR Kit, Takara Bio Inc.) from total RNA (NucleoSpin® RNA; Macherey–Nagel) isolated from Panc-1 cells (JCRB, Japan). ffLuc/sPD-L1 cDNAs were generated by PCR and infusion cloning. Primer sequences are provided in Supplementary Table 1. Retroviral vector production Reporter retroviral vectors were produced by co-transfecting 293 T cells with the retroviral transfer plasmid pVSV-G (VSV-G envelope), and pGP (gag-pol) using the calcium phosphate method. At 7 h post-transfection, the medium was replaced with fresh medium containing 5 mM sodium butyrate (STEMCELL Technologies, Canada), and cultures were maintained at 32 °C and 5% CO₂. Supernatants were collected at 48 h, clarified by filtration through a 0.45-µm membrane, aliquoted, and stored at ˗80 °C. Viral RNA titers were determined with the Retrovirus Titer Set for Real-Time PCR (Takara Bio Inc.) according to the manufacturer’s instructions. Retroviral transduction of T cells PBMCs were activated on RetroNectin® (Takara)-coated plates pre-bound with anti-CD3 (OKT-3, Invitrogen) for 4 days in the presence of IL-2 (175 IU/mL). For transduction, retroviral supernatant was loaded onto RetroNectin-coated 6-well plates and centrifuged at 1960×g for 2 h at 32 °C. Activated T cells were added to the virus-preloaded plates and centrifuged (1000×g, 10 min, 32 °C), then incubated overnight at 37 °C and 5% CO₂. Cells were washed twice with CELLOTION™, resuspended in fresh medium with IL-2, and cultured for an additional 6 days. Lentiviral vector production Lentiviral vectors were produced by co-transfecting 293 T cells with the lentiviral transfer plasmid and packaging plasmids (pCMV-VSV-G, pCgpV for gag-pol, and pRSV-Rev; Cell Biolabs) using calcium phosphate. At 7 h, the medium was replaced with fresh medium containing 5 mM sodium butyrate, and cultures were maintained at 37 °C and 5% CO₂. Supernatants were harvested at 48 h, filtered through a 0.45-µm membrane, aliquoted, and stored at − 80 °C. Viral RNA titers were measured with the Lenti-X™ qRT-PCR Titration Kit (Takara Bio Inc.) according to the manufacturer’s instructions. Lentiviral transduction of MSCs Lentiviral transductions were performed using the ViraDuctin™ Lentivirus Transduction Kit (Cell Biolabs) following the manufacturer’s protocol. MSCs (2 × 106 cells per T-225 flask; passage 4) were seeded and cultured for 3 days at 37 °C and 5% CO₂. ViraDuctin-lentivirus complexes were prepared with the indicated lentiviral vectors (1 × 109 copies per flask), added to the cultures, and incubated overnight. The next day, the medium containing the complexes was removed, cells were washed once with medium, replenished with fresh medium, and cultured for 6 days before harvest. Animal experiments Six-week-old male NOG (NOD/SCID/IL-2Rγnull) mice were obtained from In-Vivo Science Inc. (Tokyo, Japan) and acclimatized for 13 days before the experiment. Mice were inoculated with 2 × 106 unmodified T cells or gene-modified T cells by tail vein injection. At 14 or 21 days after T cell transfer, when xenogeneic responses were established, mice were randomly assigned to receive 1 × 106 unmodified or gene-modified MSCs injected through the tail vein. The dose of 1 × 106 MSCs per injection and the repeated-dosing schedule were selected empirically to provide measurable transient exposure after establishment of the xenogeneic response while remaining within a feasible and tolerated range for repeated intravenous administration in this model. The intervention time points were selected to initiate treatment only after activation-dependent bioluminescence had clearly risen above background, thereby testing an established response rather than an early low-signal phase. Bioluminescence imaging was performed using an IVIS Imaging System (PerkinElmer, Inc., USA) with Living Image software (PerkinElmer) to acquire imaging datasets. The investigators were not blinded to group allocation during treatment; however, image acquisition and analysis were performed using identical exposure settings and predefined regions of interest. IVIS imaging was performed under isoflurane anesthesia (Zoetis, Tokyo, Japan; induction, 3%; maintenance, 1.5%). For tail vein injection of T cells or MSCs, mice were restrained in a mouse restrainer (Braintree Scientific, Inc., Braintree, MA, USA) and injected while awake. As predefined humane endpoints, mice exhibiting ≥ 20% loss of body weight compared to baseline (D0) or a marked decrease in their activity were euthanized. Euthanasia was performed by isoflurane overdose using a chamber-type isoflurane vaporizer system (VetEquip, Inc., Pleasanton, CA, USA); the mice were exposed to 5% isoflurane until respiratory arrest. Death was confirmed by the cessation of respiration and heartbeat, and cervical dislocation was performed as a secondary method of euthanasia. Gene expression analysis RNA was extracted at D10, which corresponded to the common harvest point of the transduction workflow and allowed sufficient time for post-transduction transgene expression while matching the schedule used for other characterization assays. Total RNA was extracted from MSCs using a NucleoSpin® RNA kit (Macherey–Nagel) and quantified with a NanoDrop™ spectrophotometer (Thermo Fisher Scientific). One-step RT-qPCR was performed with the One Step TB Green® PrimeScript™ RT-PCR Kit II (Takara Bio Inc.) on a real-time PCR system according to the manufacturer’s instructions. GAPDH served as the endogenous control, and relative PD-L1 mRNA expression was calculated by the 2-ΔΔCt method after confirming single-peak melt curves. No-template and no-RT controls were included. Primer sequences are provided in Supplementary Table 2. ELISA Soluble PD-L1 (sPD-L1) in conditioned medium and mouse plasma was quantified using a Human PD-L1 ELISA kit (Invitrogen) according to the manufacturer’s instructions. Conditioned media were collected at harvest from cultures of Mock, ffLuc-MSCs, ffLuc/sPD-L1-MSCs, and sPD-L1-MSCs. In co-culture experiments, supernatants were collected from culture wells at the indicated time points. For animal studies, heparinized blood was obtained from the inferior vena cava, centrifuged (1500×g, 15 min, 4 °C), and plasma was collected and stored at − 80 °C until assay, avoiding repeated freeze–thaw cycles. Standards and samples were assayed in duplicate, and concentrations were interpolated from the standard curve using appropriate sample dilutions to remain within the kit’s dynamic range. In vitro luciferase assay Non-transduced (Mock) or sPD-L1-transduced UC-MSCs (2 × 103, 4 × 103, or 2 × 104 cells) were co-cultured with iELuc-T cells (2 × 104 cells) in 96-well white plates at MSC:T cell (E:T) ratios of 0.1:1, 0.2:1, or 1:1. After 12 h of co-culture, a T cell activation reagent (ImmunoCult™ Human CD3/CD28 T Cell Activator 0.5 µL/well; STEMCELL Technologies) was added; T cells alone served as the control. For blockade experiments, anti-PD-L1 antibody (2 µg/well; BioLegend, USA) was added at 0 h and again at 12 h after co-culture initiation. Luciferase activity was quantified using the Bright-Glo® Luciferase Assay System (Promega) according to the manufacturer’s instructions with a microplate luminometer (Fluoroskan FL; Thermo Fisher Scientific). Statistical analysis Data are presented as the mean ± SD. Analyses were performed using GraphPad Prism 9 (GraphPad Software, USA). Biological replicates (independent mice or independent cell preparations/donors) were treated as n. Two-group comparisons used two-tailed unpaired or paired t-tests, as appropriate. Comparisons among two or more groups used one-way ANOVA with Tukey’s or Holm-Šídák’s multiple-comparisons test, as appropriate. Longitudinal outcomes were analyzed by two-way repeated-measures ANOVA followed by Šídák’s multiple-comparisons test. Survival was analyzed by Kaplan–Meier with the log-rank (Mantel-Cox) test. Multiplicity-adjusted p values are reported; statistical significance was set at p < 0.05 (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). No statistical method was used to predetermine sample size. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1