Lactic acid improves Treg manufacturing and in vivo function

Abstract

Adoptive cell therapy using regulatory T cells (Tregs) is a promising approach to suppress immune responses in autoimmunity and transplantation, but it is challenging to expand pure and optimally suppressive cells. Lactic acid (LA) is associated with enhanced Treg function in tumors so we hypothesized that it may be beneficial during Treg expansion. We found that addition of LA at day 3 post-stimulation onwards improved viability and purity, increased glycolysis upon re-stimulation, and led to superior suppressive function. In Tregs expressing chimeric antigen receptors (CARs) specific for HLA-A2, LA not only enhanced viability and purity but also significantly reduced tonic signaling-associated expression of exhaustion-associated markers (PD-1, TIM-3, LAG-3, TOX, and BLIMP-1). The effects of LA were not fully recapitulated by either pH-neutral lactate or low pH. In immunodeficient mouse models of chronic stimulation and xenogeneic graft-versus-host disease, LA-conditioned human Tregs demonstrated enhanced stability, reduced exhaustion marker expression, and improved efficacy. Thus, LA has a multimodal effect on human polyclonal and CAR Treg purity, viability, and function, representing a method to generate an optimal Treg product for cell therapy.

Graphical abstract

Addition of lactic acid to a human Treg manufacturing protocol enhanced viability, expansion, purity, and suppressive function. Lactic acid also reduced tonic CAR signaling, preventing expression of exhaustion makers. In immunodeficient mouse models, lactic acid-conditioned CAR Tregs maintained higher stability, expressed fewer exhaustion markers, and had greater therapeutic efficacy.

Introduction

Adoptive cell therapy using CD4⁺ regulatory T cells (Tregs) aims to harness the naturally immunosuppressive phenotype of Tregs for applications in organ transplantation, autoimmunity, and beyond. Autologous polyclonal Tregs have now been tested in multiple phase 1/2 clinical trials, revealing a strong safety profile and the need to further improve efficacy.^1,2 One approach to improve efficacy is genetic modification to introduce antigen receptors, including chimeric antigen receptors (CARs) against disease-relevant targets.³ Our group and others have shown that expression of a CAR specific for HLA-A2, a commonly mismatched antigen in transplantation, directs Treg homing into HLA-A2+ allografts and their draining lymph nodes, resulting in prolonged graft survival.^4,5,6,7 HLA-A2-specific CAR (A2-CAR) Tregs are now undergoing clinical testing in clinical trials of kidney and liver transplantation (NCT04817774 and NCT05234190).⁸

A major barrier to the clinical application of Tregs is efficient cell manufacturing. First, as a naturally rare cell population, Tregs must be expanded several hundred-fold in ex vivo culture. In a recent phase 1/2 trial of donor alloantigen-reactive Tregs in liver transplantation, 40% of patients could not be infused due to failure to manufacture the minimum dose.⁹ Second, because Tregs are hypoproliferative compared to conventional T cells (Tconvs), any contaminating Tconvs in the initial population can rapidly expand resulting in an undesirable increase in their proportion, leading to purity challenges in large-scale expansion. Third, the in vivo persistence of adoptively transferred Tregs is often low,^9,10 which may limit long-term efficacy. Thus, advancements in culture strategies are needed to enable large-scale production of a highly functional, pure, and long-lived Treg product for cell therapy.

Manipulating the metabolic pathways used by Tconvs for energy production has shown promise for enhancing adoptive cell therapy for cancer.¹¹ While less explored in Tregs, a growing body of evidence suggests that metabolic pathways also influence Treg phenotype and function.¹² Notably, Tregs appear to thrive in the lactic acid (LA)-rich, glycolytic tumor environment through their unique ability to take up and incorporate lactate into metabolic pathways, resulting in a proliferative and survival advantage over Tconv in high lactate conditions.^13,14,15 Consequently, Treg-specific knockout of the lactate transporter, MCT1, or inhibition of lactate production in tumors results in greater tumor control in mice.^14,15 Similarly, lactate enhances Treg mitochondrial function and sensitivity to transforming growth factor β (TGF-β)-induced FOXP3 expression by altering the glycosylation and lactylation of various intracellular proteins.^16,17 Here, our aim was to leverage the characteristic ability of Tregs to thrive in lactate-rich environments to improve the ex vivo expansion of human polyclonal and CAR Tregs.

Results

Lactic acid enhances polyclonal Treg viability and phenotype

To investigate the effect of LA on polyclonal Treg culture, naive human Tregs (CD4⁺CD45RO⁻CD45RA⁺CD127⁻CD25^hi) were sorted from the peripheral blood of healthy human donors (Figure S1). First, we tested the impact of adding 15 mM LA at the time of Treg stimulation (day 0) (Figure S2A), finding that the addition of LA at day 0 significantly impaired proliferation, with a >50% reduction in cell number compared to the control (Figure S2B). However, viability was not affected, indicating that LA did not cause cell death, and there was no effect on the proportion of cells expressing FOXP3 and Helios (Figure S2C). A small, but significant increase in the ability of LA-cultured Tregs to suppress CD8 T cell proliferation was observed (Figure S2D).

We hypothesized that addition of LA at the time of activation may impair the early pathways needed to stimulate Treg proliferation. Thus, we next tested addition of 15 mM LA 3 days after naive Treg stimulation (Figure 1A). Although culture in LA resulted in a small delay in Treg proliferation, LA-cultured Tregs exhibited significantly higher viability throughout 15 days of culture (Figure 1B). Importantly, Tregs cultured in control media underwent a contraction phase after 9 days, characterized by reduced cell numbers and low viability, which was prevented by culture in LA. LA also significantly and consistently improved Treg purity as exhibited by the increased proportion of FOXP3^posHelios^pos cells and reduction in contaminating FOXP3^negHelios^neg Tconvs (Figure 1C). The expression of other functional Treg markers at rest and in the absence of additional stimulation was only marginally affected by LA, with a small reduction in CTLA-4 expression and increase in CD39 expression but no impact on CD25, LAP, or GARP (Figures 1D and S3A).

Figure 1.

Lactic acid enhances Treg expansion and purity

Naive human Tregs (CD4⁺CD45RO⁻CD45RA⁺CD127⁻CD25^hi) were stimulated using anti-CD3/CD28 Dynabeads and cultured for 9–15 days. At day 3, media was supplemented with 15 mM LA for the remainder of culture. (A) Schematic of Treg culture protocol. (B) Fold expansion and viability over time (n = 18). (C) Expression of FOXP3 and Helios at day 9 (n = 13). Representative figure shown on left. (D) Expression of CD25, CTLA-4, LAP, GARP, and CD39 at day 9 (n = 10). Connected lines indicate individual donor pairs. (E and F) Treg media was supplemented with 15 mM LA, 15 mM sodium lactate (SL), or HCl (pH 6.7) at day 3. (E) Fold expansion and viability over time (n = 9). (F) Expression of FOXP3 and Helios at day 9 (n = 12). Statistical analysis was carried out by two-way ANOVA with Sidak’s multiple comparisons test (B and E), paired t test (C, left and D), Wilcoxon matched-pairs signed rank test (C, right), or Friedman test with Dunn’s multiple comparison’s test (F). Colored asterisks indicate significant differences between control and LA (blue), SL (green), or HCl (purple). Data are represented as mean ± SEM. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001.

To determine the effect of culturing in LA on the cytokine profile of Tregs, cells at day 9 were re-stimulated polyclonally in the absence of LA and cytokine production was measured by cytokine concentration in supernatant and intracellular cytokine content. Tregs cultured in either control or LA conditions produced similarly low levels of all measured cytokines, indicating that there was no notable gain of effector function (Figures S3B and S3C). Little to no IL-17A/F was produced by Tregs cultured in either condition. Interestingly, although Tregs secreted similar levels of interferon (IFN)-γ and tumor necrosis factor (TNF)-α into supernatant (Figure S3B), fewer than 5% of cells were IFN-γ⁺ whereas approximately 60% were TNF-α⁺ (Figure S3C). This is consistent with a previous observation that Tregs predominantly produce a membrane-bound, rather than secreted, form of TNF-α, which primarily binds to tumor necrosis factor receptor 2 (TNFR2).¹⁸ Overall, the cytokine profile of Tregs is not affected by culture in LA.

LA is composed of a lactate ion as well as an acidic (H+) component. To delineate the contribution of lactate versus acidity to the effects of LA, we cultured Tregs in the presence of 15 mM LA, 15 mM pH-neutral sodium lactate, or pH-matched HCl (pH 6.7). We observed that both LA and HCl (pH 6.7) enhanced viability and expansion, although effects with HCl (pH 6.7) were weaker (Figure 1E). However, only LA significantly enhanced Treg purity, as exhibited by an increased proportion of FOXP3^posHelios^pos cells and reduction in FOXP3^negHelios^neg cells (Figure 1F). pH-neutral sodium lactate had no impact on any measures.

We also investigated the effect of LA concentration on Treg culture. Treg viability was improved even at a low LA concentration (5 mM) but further enhanced in a dose-dependent manner (Figure S3D). Although 20 mM LA resulted in the best preservation of long-term viability, it also strongly impaired proliferation, indicating that excessively high LA concentrations can be harmful to the manufacturing process. LA also had a dose-dependent effect on Treg purity, with an increased reduction in FOXP3^negHelios^neg cells at higher LA concentrations (Figure S3E). Considering the need to balance impaired proliferation with improved purity and viability, we continued with 15 mM LA as the optimal concentration for Treg culture.

Many clinical trials utilize bulk Tregs (CD4⁺CD127⁻CD25^hi) rather than naive Tregs (CD4⁺CD127⁻CD25^hiCD45RO⁻CD45RA⁺), as they are more numerous and easier to isolate. Therefore, we investigated the effect of culture in 15 mM LA, added at day 3, on sorted bulk Tregs, which were stimulated using anti-CD3/CD28 beads and cultured for up to 15 days. LA-conditioning had a similar impact on bulk Treg expansion and viability, with an initial lag phase in proliferation that recovered by day 11, as well as an improvement in viability (Figure S4A). LA also significantly enhanced the purity of Tregs, exhibited by a reduction in contaminating FOXP3^negHelios^neg Tconvs (Figure S4B). This improvement in purity was particularly evident after 15 days of culture for donors that lost a significant proportion of FOXP3-expressing cells (Figure S4C). Although a slight reduction in CTLA-4 expression was observed, similar to naive Tregs, there was no major impact on other phenotypic markers (Figures S4D and 1D). We also tested the suppressive function of bulk Tregs, finding that LA had a small, but significant effect on enhancing suppression of CD4 and CD8 T cells (Figure S4E), as well as CD80/86 on B cells (Figure S4F). In summary, LA-conditioning has a similar beneficial effect on naive and bulk Treg populations.

To ask if LA enhances Treg purity solely via an inhibitory effect on contaminating Tconvs, we cultured naive human CD4⁺CD25⁻ T cells in LA and tested effects on expansion, viability, and FOXP3 expression. LA did not significantly affect proliferation but did lead to a moderate enhancement in viability (Figure S5A). We further investigated the phenotype of Tconvs, observing an increase in the proportion of FOXP3-expressing cells (Figure S5B). Because FOXP3 expression can be induced upon activation, we tested the expression of the activation marker CD69, finding that it was highly upregulated at day 9 (Figure S5C), indicating that FOXP3 expression may be activation-induced at day 9. We, therefore, also tested FOXP3 expression at day 15, a time when the cells were no longer activated, as evidence by low CD69 expression (Figure S5C). At day 15, cells cultured in LA still exhibited a significantly higher level of FOXP3 expression (Figure S5D), suggesting that the LA-mediated increase in FOXP3 expression is stable and not activation-induced. Thus, these data suggest that some of the beneficial effect of LA on Tregs may be via promotion of Treg induction from Tconv.

In summary, addition of 15 mM LA to Treg culture at day 3 post-stimulation, but not immediately at the time of activation, results in enhanced purity and viability. Thus, we continued to explore the effect of LA addition at day 3, termed “LA-conditioning” (Figure 1A) on other aspects of Treg function and phenotype.

Treg suppressive function is enhanced by LA-conditioning

To test the effect of 15 mM LA on Treg function, control or LA-conditioned Tregs were polyclonally stimulated with anti-CD3/CD28 beads for 24 h in the absence of LA. Both control and LA-conditioned Tregs upregulated ICOS, CTLA-4, GARP, LAP, and CD39 expression to a similar extent, suggesting that overall activation capacity was unaffected (Figures 2A and S6A). Although PD-1 expression has been previously reported to increase on Tregs following stimulation in the presence of LA,¹⁵ we observed that LA-conditioned Tregs consistently expressed lower levels of PD-1, as well as 4-1BB, when re-stimulated in absence of LA (Figure 2A).

Figure 2.

LA-conditioning enhances Treg suppressive function

(A) Control or LA-conditioned human Tregs were stimulated for 24 h using anti-CD3/CD28 beads. Expression of ICOS, CTLA-4, PD-1, and 4-1BB was determined by flow cytometry (n = 6). Representative figures show unstimulated (dotted line) and stimulated (solid line) Tregs. Connected lines indicate paired donors. (B and C) Control or LA-conditioned Tregs were co-cultured with PBMCs and stimulated using anti-CD3/CD28 beads for 72 h. (B) Suppression of CD4 (left) and CD8 (right) T cell proliferation was determined by dye dilution (n = 18). (C) Expression of CD80 and CD86 was determined on B cells (n = 17). Representative figures show indicated cells in PBMC alone (black), control Treg (gray), and LA Treg (blue) conditions. Statistical analysis was carried out by paired T test (A) or two-way ANOVA with uncorrected Fisher’s LSD test (B and C). Data are represented as mean ± SEM. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001.

We also explored the impact of LA on the suppressive function of Tregs. Peripheral blood mononuclear cells (PBMCs) were stimulated and cultured in the presence of control or LA-conditioned Tregs for 72 h (Figures 2B and 2C). To avoid a possible impact of LA on PBMCs, LA was washed out from Tregs prior to co-culture. We observed that LA-conditioned Tregs reduced CD4⁺ and CD8⁺ T cell proliferation to a greater extent than control Tregs (Figure 2B). Furthermore, LA-conditioned Tregs had an increased ability to downregulated CD80 and CD86 on B cells (Figure 2C). These effects were observed at all concentrations of LA up to 15 mM, whereas the improved suppressive function was abolished with 20 mM LA (Figures S3F and S3G), further supporting earlier observations of the detrimental effects of high LA.

Culture in LA alters Treg glycolysis

Previous studies have demonstrated that LA can alter T cell and Treg metabolism by inhibiting glycolysis and promoting oxidative phosphorylation.^17,19 We first explored the effect of LA on glucose consumption and lactate production by Tregs. Tregs were highly glycolytic following stimulation and produced large amounts of lactate (Figures 3A and 3B). LA significantly reduced the uptake of glucose by Tregs during the initial 4 days after addition of LA (Figure 3A). Lactate production was similarly reduced, suggesting dampened glycolysis or reduced lactate export (Figure 3B). However, upon removal from an LA-containing environment to carry out a Seahorse assay, LA-conditioned Tregs exhibited a higher basal extracellular acidification rate, whereas basal oxygen consumption rate was unchanged (Figures 3C and 3D). Moreover, when Tregs were re-stimulated for 48 h in an LA-free media, LA-conditioned Tregs consumed significantly more glucose and produced more lactate (Figure 3E). Together, these data suggest that glycolysis may be impaired in the presence of LA but that a rebound effect occurs after the removal of LA.

Figure 3.

LA alters Treg metabolism

Naive human Tregs were stimulated using anti-CD3/CD28 and cultured in control or LA-containing media for 9 days. (A and B) Glucose (A) and lactate concentration (B) was determined in media at 2-day intervals between D3-5, D5-7, and D7-9 (n = 3). Dotted line indicates concentration of glucose (A) or supplemented LA (15 mM) (B) in media at the beginning of culture. Net lactate production was calculated by subtracting 15 mM from the LA condition. (C and D) Oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were measured by Seahorse assay at day 9. (C) OCR over time (n = 3). (D) Basal OCR and ECAR (n = 3). (E) Control or LA-conditioned Tregs were re-stimulated at day 9 using anti-CD3/CD28 beads for 48 h in the absence of LA. Glucose and lactate concentration was determined in media (n = 3). Statistical analysis was carried out by two-way ANOVA with Sidak’s multiple comparisons test (A) or paired t test (D and E). Data are represented as mean ± SEM. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗∗p < 0.0001.

LA reduces CAR tonic signaling in Tregs

CAR expression can modify T cell and Treg metabolism in a co-stimulatory domain-dependent manner.¹¹ Thus, we next explored the effect of LA on HLA-A2-specific CAR Tregs, comparing effects on four different types of CAR Tregs encoding different co-stimulatory domains: a 1^st generation CAR with no co-stimulatory domain (A2.ζ), a CD28-containing CAR (A2.28ζ), a TNFR2-containing CAR (A2.TNFRζ), or a 4-1BB-containing CAR (A2.BBζ) (Figure 4A).²⁰ These were compared to control Tregs expressing the transduction marker, NGFR, but no CAR. Human Tregs were stimulated with anti-CD3/28-beads with lentiviral transduction of CARs carried out after 2 days (Figure 4A). To prevent an effect of LA on transduction, 15 mM LA was added to media 3 days post-transduction (day 5 post-stimulation) and cells were cultured for an additional 7 days.

Figure 4.

LA reduces CAR tonic signaling on Tregs

Naive human Tregs were stimulated using anti-CD3/CD28 dynabeads, lentivirally transduced with HLA-A2-specific CARs with no co-stimulatory domain (A2.ζ) or with a CD28 (A2.28ζ), TNFR2 (A2.TNFRζ), or 4-1BB (A2.BBζ) domain. LA was supplemented to media from day 5 onwards. Cells were analyzed at day 12. (A) Schematic of CAR Treg transduction and expansion protocol. (B) CAR⁺ cell yield and viability (n = 8). (C) Cell-surface CAR expression, detected via the extracellular Myc tag (n = 5). (D) Proportion of FOXP3 and Helios expressing cells (n = 9). (E) Expression of FOXP3, CTLA-4, and LAP (n = 9). (F) Expression of Helios on FOXP3^pos cells relative to control-treated NGFR cells. Statistical analysis was carried out by two-way ANOVA with Fisher’s LSD test (B–E). Data are represented as mean ± SEM. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001.

Expression of 2^nd-generation CARs (A2.28ζ, A2.TNFRζ, and A2.BBζ) significantly increased yield over that of A2.ζ, suggesting the second-generation construct may mediate antigen-independent activation via a process known as tonic signaling.^21,22 Consistent with this possibility, whereas the A2.ζ-CAR Treg yield was unaffected by 15 mM LA, it significantly reduced the yield of A2.28ζ-, A2.TNFRζ-, and A2.BBζ-CAR Tregs (Figure 4B). As observed with polyclonal Tregs (Figure 1B), viability was enhanced by culture in LA, particularly for Tregs expressing 2^nd generation CARs (Figure 4B). Expression of the CAR itself was only minimally impacted by LA (Figure 4C). Culture in LA significantly and consistently reduced the proportion of contaminating FOXP3^negHelios^neg cells for all CARs investigated (Figure 4D).

We also investigated the intensity of FOXP3 expression in Tregs. Although the overall proportion of FOXP3-expressing cells was higher in LA-treated CAR Tregs (Figure 4D), Tregs expressing CARs, particularly the A2.28ζ- and A2.TNFRζ-CARs, displayed a higher amount of FOXP3 expression compared to control NGFR-Tregs (Figure 4E). This was consistently reduced by culture in LA. This reduction in FOXP3 MFI was not observed in control NGFR-expressing Tregs, indicating a specific effect of CAR expression. In contrast, the expression of Helios, a marker of Treg stability,²³ was not reduced by LA (Figure 4F), suggesting that reduced FOXP3 in LA-treated CAR Tregs is not indicative of destabilization. Since FOXP3 expression can be upregulated by activation,²⁴ we also measured expression of two other Treg-relevant proteins that are known to be upregulated by activation: CTLA-4 and LAP. Expression of both CTLA-4 and LAP was increased in CAR-expressing cells and most notably for the A2.28ζ- and A2.TNFRζ-CARs (Figure 4E), suggesting an enhanced degree of activation at baseline. As with FOXP3, both CTLA-4 and LAP were reduced by culture in LA.

LA reduces CAR tonic signaling-induced expression of exhaustion-associated markers

To further explore the possibility that LA improved CAR-mediated tonic signaling, we quantified the effect of LA on the expression of three proteins associated with CAR tonic signaling-induced T cell exhaustion—PD-1, TIM-3, and LAG-3—after 12 days of culture (Figures 5A and S7A–S7C). We found that Tregs transduced with 2^nd generation CARs (A2.28ζ, A2.TNFRζ, and A2.BBζ) all had an increased proportion of cells expressing one or more exhaustion-associated markers. Notably, LA significantly reduced the proportion of A2.28ζ- and A2.BBζ-CAR Tregs expressing 2 or 3 exhaustion-associated markers and increased the proportion of Tregs expressing no markers (Figures 5A and S7A–S7C). The expression of TOX and BLIMP-1, transcription factors associated with Treg and CD4⁺ T cell exhaustion,^22,25 was also heightened relative to NGFR-Tregs, and reduced by culture in LA (Figure 5B).

Figure 5.

LA reduces CAR Treg exhaustion marker expression

Naive human Tregs were transduced with an HLA-A2-specific CAR (A2.ζ, A2.28ζ, A2.TNFRζ, or A2.BBζ) and cultured in control or LA-containing media. Cells were analyzed at day 12. (A) The proportion of CAR Tregs expressing 0–3 exhaustion markers (PD-1, LAG-3, and TIM-3) (n = 10). Asterisks indicate significant differences in proportion of cells between control and LA. (B) The expression of TOX and MFI of BLIMP-1 relative to NGFR-Tregs (n = 4–8). (C–E) CAR Tregs were co-cultured with HLA-A2⁺ dendritic cells (DCs) for 96 h. (C) Schematic of DC suppression assay (made with BioRender.com). (D) Expression of CD80 and CD86 on DCs after co-culture n = 6). (E) Expression of Ki67 and ICOS on CAR Tregs after co-culture (n = 5–6). Statistical analysis was carried out by two-way ANOVA with Sidak’s multiple comparisons test (A) or two-way ANOVA with Fisher’s LSD test (B, D, and E). Bars represent mean with individual donors connected by lines. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

LA also increased the expression of CCR7, a lymph node-homing marker highly expressed on less differentiated cells, in A2.28ζ- and A2.BBζ-CAR Tregs but not in NGFR Tregs, which maintained higher expression even after 12 days of culture (Figure S7D). Importantly, neither sodium lactate nor low pH fully recapitulated the effects of LA on tonic CAR signaling or exhaustion in Tregs (Figures S8A and S8B).

A major suppressive mechanism of Tregs is the downregulation of co-stimulatory molecules, CD80 and CD86, on antigen-presenting cells.² Thus, we cultured control or LA-conditioned A2-CAR Tregs with A2+ monocyte-derived dendritic cells (DCs) for 72 h and measured the expression of CD80/86 on DCs (Figures 5C and 5D). All CAR-Tregs were highly effective at downregulating CD80 and CD86 expression and no effect of LA was observed. However, LA-conditioning significantly increased CAR-stimulated expression of Ki67 on A2.ζ- and A2.28ζ-CAR Tregs as well as ICOS on A2.28ζ-CAR Tregs (Figure 5E), suggesting stronger antigen-induced activation. Expression of LAP, 4-1BB, and CD39 were not affected by LA-conditioning (Figure S7E).

In summary, LA reduces expression of exhaustion-associated markers—PD-1, TIM-3, LAG-3, TOX, and BLIMP-1—on CAR-Tregs, likely due to a reduction in tonic signaling. This is associated with enhanced CAR-stimulated activation upon culture with A2+ DCs.

LA-conditioning increases CAR Treg stability and function in vivo

We next explored the stability, persistence, and function of LA-conditioned CAR Tregs using two humanized mouse models. To first test for effects on Treg stability and persistence with chronic antigen stimulation, we injected human autologous PBMCs and either control or LA-conditioned A2.28ζ-CAR Tregs into HLA-A2-transgenic non-obese diabetic SCID gamma (NSG) mice (Figure 6A). In this model, A2.28ζ-CAR Tregs receive strong and persistent stimulation by HLA-A2 expressed throughout the mouse but do not develop xenogeneic graft-versus-host disease (xenoGVHD) during the experimental timeline. Mice were bled weekly to characterize circulating Treg phenotype and numbers. No overall difference in the engraftment and persistence of control versus LA-conditioned A2.28ζ-CAR Tregs was seen (Figure 6B) and PBMC engraftment did not differ between groups (Figure S9A). However, cells conditioned with LA maintained significantly higher proportions of FOXP3^posHelios^pos cells, with fewer FOXP3^negHelios^neg cells over time compared to control A2.28ζ-CAR Tregs (Figures 6C and S9B). Notably, the expression of Ki67 on FOXP3^negHelios^neg CAR-expressing cells was also significantly reduced in cells that had undergone LA-conditioning (Figure 6D), indicating a long-term impact on phenotype up to 4 weeks in vivo.

Figure 6.

LA-conditioning enhances CAR Treg purity and reduces expression of exhaustion markers under chronic stimulation in vivo

LA-conditioned or control A2.28ζ-CAR Tregs were injected into HLA-A2⁺ NSG mice with autologous PBMCs at a 1:1 ratio. CAR Treg engraftment and phenotype was tracked weekly in blood. (A) Schematic of chronic stimulation model. (B) Absolute count of CAR⁺ cells in blood over time (n = 7). (C) Proportion CAR⁺ cells expressing FOXP3 and Helios in blood (n = 4–7). (D) Proportion of CAR⁺FOXP3^negHelios^neg cells expressing Ki67 in blood (n = 3–7). (E) Proportion of CAR⁺ cells in blood, liver, lung, and spleen at day 28 expressing 0–3 exhaustion markers (LAG-3, TIM-3, and PD-1). Statistical analysis was carried out by two-way ANOVA with Fisher’s LSD test (B–E). Schematic made using Biorender.com (A). Data are represented as mean ± SEM. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗∗p < 0.0001.

Since we observed an impact of LA on expression of exhaustion-associated markers in vitro, we also explored the expression of TIM-3, LAG-3, and PD-1 on CAR Tregs in blood and tissues after 4 weeks. Both LAG-3 and PD-1 expression increased over time on CAR Tregs, whereas TIM-3 expression peaked early followed by a progressive decrease (Figure S9C). Overall, no differences in exhaustion-associated marker expression were observed up to 3 weeks in blood. However, 4 weeks post-injection, we observed that the proportion of CAR⁺ cells expressing exhaustion-associated markers in blood, lung, liver, and bone marrow was significantly reduced by LA-conditioning (Figures 6E and S9C).

To explore the impact of LA-conditioning on CAR Treg efficacy, we tested the ability of a suboptimal dose of A2.28ζ-CAR Tregs to delay disease in a xenoGVHD model. We injected a ratio of 8 HLA-A2+ PBMCs to 1 A2.28ζ-CAR Treg, a dose which was not expected to significantly delay xenoGVHD in pre-conditioned NSG mice (Figure 7A). Onset of xenoGVHD symptoms was tracked over time in addition to weekly phenotyping of circulating CD45⁺ cells in blood. Consistent with the enhanced suppressive activity of LA-conditioned Tregs in vitro (Figure 2), LA-conditioned Tregs significantly prolonged survival compared to mice receiving either PBMCs alone or control A2.28ζ-CAR Tregs (Figure 7B). In contrast, at this low ratio, control A2.28ζ-CAR Tregs were not able to control disease onset. Similarly, LA-conditioned A2.28ζ-CAR Tregs were superior to control A2.28ζ-CAR Tregs in reducing xenoGVHD symptom scores and preventing body weight loss (Figures S10A and S10B). No significant differences in PBMC engraftment were observed between any groups (Figure S10C).

Figure 7.

LA-conditioning enhances CAR Treg efficacy in vivo

NSG mice were pre-conditioned with busulfan prior to injection of A2⁺ PBMCs with a sub-optimal ratio of LA-conditioned or control A2.28ζ-CAR Tregs (8:1 PBMC:CAR Treg). Mice were bled weekly and xenoGVHD scoring carried out at 1–3 day intervals. (A) Schematic of xenogeneic graft-versus-host disease (xenoGVHD) model. (B) Proportion of surviving mice over time (n = 8–9). (C) Absolute count of CAR⁺ cells in blood over time (n = 8–9). (D) Proportion of CAR⁺ cells expressing FOXP3 and Helios in blood (n = 8–9). (E) Correlation between the proportion of FOXP3^negHelios^neg cells at day 7 and survival. Statistical analysis was carried out by a Log rank (Mantel-Cox) test (B), two-way ANOVA with Fisher’s LSD test (C), or unpaired t test (D). Schematic made using Biorender.com (A). Data are represented as mean ± SEM. ∗p < 0.05, ∗∗p < 0.01.

The engraftment and phenotype of CAR Tregs was also quantified in the xenoGVHD model. CAR Tregs were abundant 7 days post-injection but rapidly disappeared from circulation (Figure 7C). However, at day 7, LA-conditioned CAR Tregs exhibited superior purity, evidenced by a significantly lower proportion of FOXP3^negHelios^neg cells and a trend toward increased FOXP3^posHelios^pos cells (Figure 7D). Notably, the proportion of FOXP3^negHelios^neg cells present in blood at day 7 had a significant negative correlation with the overall survival of mice (Figure 7E), suggesting that Treg purity is a critical factor in therapeutic efficacy. In contrast, no significant correlation was observed between the number of circulating CAR⁺ cells at day 7 and survival (Figure S10D).

In summary, LA-conditioning enhances CAR Treg purity in vivo, even when observed long-term. This enhancement in purity is correlated with increased efficacy in a humanized model of xenoGVHD.

Discussion

Achieving high purity and yield are major barriers in the ex vivo expansion of Tregs and CAR Tregs for adoptive cell therapy. Here, we show that culture in LA, added at day 3 post-stimulation, significantly enhanced the purity and suppressive function of polyclonal and HLA-A2-specific CAR Tregs both in vitro and in vivo. Moreover, LA-conditioning significantly reduced CAR-mediated tonic signaling, resulting in lower expression of T cell exhaustion-associated markers. Remarkably, the LA-enhanced purity was maintained long-term following in vivo injection and was correlated with better efficacy in controlling xenoGVHD. Thus, LA-conditioning by addition of 15 mM LA 3 days post-stimulation is a method to improve Treg and CAR Treg culture for the purpose of adoptive cell therapy.

LA-conditioning led to an overall improvement in phenotypic purity, as seen by a significant reduction in contaminating FOXP3^negHelios^neg cells and an increase in FOXP3^posHelios^pos cells. A number of mechanisms driving this enhancement in purity are possible. We found that culture in LA led to an increase in FOXP3 expression in naive Tconv, suggesting that LA could help support Treg induction. LA, in a pH-dependent manner has been previously shown to enhance TGF-β-mediated FOXP3 induction in Tconvs.²⁶ Although we did not add exogenous TGF-β into the culture system, Tconvs may produce endogenous TGF-β that could promote this pathway. LA may also promote Treg purity by lineage stabilization. Although we did not investigate demethylation of the Treg-specific demethylated region (TSDR), an epigenetic marker of Treg stability, we observed an increase in FOXP3^posHelios^pos cells. Since Helios expression is correlated with TSDR demethylation,^20,23 LA likely also enriches for pre-existing, lineage-stable Tregs.

A number of strategies have been tested to enhance Treg purity and stability during expansion and following in vivo infusion. One such strategy is rapamycin, an inhibitor of mechanistic target of rapamycin (mTOR).^27,28 Several mechanisms have been proposed for the effect of rapamycin, including increased Treg stability,²⁹ reduced proliferation of Tconv,²⁷ and increased induction of Tregs from Tconvs.^30,31 We observed that LA has a similar effect in increasing Treg purity, suggesting that these may act on similar pathways. Rapamycin is known to promote Treg induction by inhibiting the antagonist effect of mTOR on intracellular SMAD3/4-mediated TGF-β signaling.³² Similarly, both low pH and lactate have been reported to enhance intracellular TGF-β signaling.^16,26 Notably, although low pH can activate latent TGF-β, it has been shown that pH values obtained by addition of LA (∼pH 6.7–6.8) are not sufficient to mediate activation, so effects are likely mediated by an intracellular mechanism rather than increased TGF-β availability.²⁶ In addition to the beneficial effect on purity, we observed several effects of LA not previously observed with rapamycin, including improved viability and expansion as well as reduction of tonic CAR signaling and exhaustion. Overall, LA likely improves Treg proliferation and purity by several parallel mechanisms.

Although the effect of LA on Tregs has been studied by a number of groups,^14,15,16 few have specifically delineated the contribution of the lactate and acid components to observed effects. Recently, lactate in the form of sodium lactate was shown to alter the N-glycosylation of mitochondrial proteins related to oxidative phosphorylation, resulting in enhanced Treg function in a model of xenoGVHD.¹⁷ TGF-β-mediated FOXP3 induction is also enhanced by sodium lactate,¹³ although a more recent study found the reverse: Treg induction was enhanced in a low pH environment independent of lactate.²⁶ We found that sodium lactate had no effect on Treg phenotype, whereas only some effects of LA were recapitulated by low pH, namely enhanced expansion and viability, as well as a moderate impact on reducing exhaustion. This suggests that co-administration of lactate and acid is important for the functional effects of LA on Tregs. The contradictory findings across studies can possibly be reconciled by considering the proton-dependence of monocarboxylate transporters (MCTs), responsible for shuttling lactate with H⁺ ions across the cell membrane. Lactate transport by MCTs occurs bi-directionally along the concentration gradient of lactate/H⁺, with highest enzymatic activity occurring in the absence of a pH gradient.³³ Thus, considering our observation that human Tregs are highly glycolytic upon stimulation and produce large amounts of LA, acidification of the extracellular media could elevate intracellular lactate even without addition of exogenous lactate. Indeed, in the presence of HCl, CD8⁺ T cells exhibit intracellular elevation of both endogenously produced and exogenously supplemented lactate.³⁴ Thus, even without exogenous lactate, low pH may drive lactate-dependent effects in Tregs. The synergy between low pH and lactate in elevating intracellular lactate may explain the requirement for LA over lactate or low pH alone to enhance Treg function in our study.

Treg phenotype and function are closely linked with metabolic characteristics.¹² We found that LA regulates glucose metabolism as exhibited by a reduction in glucose uptake and lactate export in the presence of LA. Paradoxically, upon removal of exogenous LA, Tregs exhibited an elevated rate of glycolysis and secreted significantly more lactate. Inhibition of glycolysis by LA or low pH has been previously described for CD8⁺ and CD4⁺ T cells¹⁹; our data demonstrate the effects on glucose uptake in Tregs. Interestingly, acidity-mediated inhibition of T cell glycolysis has been shown to be rapidly reversible upon normalization of extracellular pH,¹⁹ aligning with our observations of reduced glucose consumption and lactate production in the presence of LA but enhancement of ECAR in the context of a pH-neutral Seahorse assay. The enhancement of glycolysis upon removal of LA suggests that Tregs may develop compensatory mechanisms under prolonged LA-rich conditions that lead to enhanced glycolysis via altered expression of transporters or glycolytic enzymes. Indeed, such effects have been shown for CD8⁺ T cells cultured transiently in glucose-restricted conditions, resulting in upregulated glucose transporter Glut1 expression, enhanced ECAR upon glucose re-exposure, and improved cytotoxic function in vivo.³⁵ The increased rate of glycolysis observed following LA-conditioning may contribute to the increased suppressive function observed in our study. Recent findings have shown that increased glycolysis by Tregs cultured in IL-7 and IL-15 is associated with improved control of xenoGVHD.³⁶ Furthermore, TNFR2-driven glycolysis is critical for supporting Treg cell identity and suppressive function, whereas inhibition of glycolysis resulted in loss of FOXP3 expression and poor suppressive function.³⁷ In contrast, excessive glycolysis can also promote Treg instability,³⁸ indicating the requirement for a tightly controlled level of glycolysis. We did not observe any instability associated with the increased glycolysis, suggesting that the LA-induced enhancement in glycolysis does not overwhelm the glycolytic balance in Tregs but may instead contribute to their suppressive function.

A benefit of LA-conditioning is inhibition of tonic CAR signaling in Tregs. We observed that this was associated with reduced expression of exhaustion-associated markers, including PD-1, TIM-3, LAG-3, BLIMP-1, and TOX,²⁵ as well as increased proliferation upon antigen encounter. Although these features are suggestive of exhaustion in CD8⁺ T cells and CD4⁺ Tconvs, the characteristics of exhaustion in Tregs and its functional impact on Tregs are not well understood. Tonic-signaling CAR expression in Tregs has previously been associated with many features of conventional exhaustion, including PD-1, TIM-3, LAG-3, and TOX expression, as well as increased chromatin accessibility and altered DNA methylation.²² Accordingly, tonic-signaling CAR Tregs are less suppressive and exhibit reduced proliferation, together supporting the existence of an exhaustion-like phenotype in these cells.²² Similarly, following repeated stimulation for expansion, Tregs acquired DNA methylation alterations, including hypomethylation of promoters of exhaustion-related genes.³⁹ In tumors, high expression of the conventional exhaustion marker, PD-1, on Tregs is associated with IFN-γ secretion, reduced suppression, and an exhaustion-like gene signature.⁴⁰ However, many markers of conventional exhaustion may also be functionally important for Tregs. For example, LAG-3 is crucial for controlling Treg metabolism to maintain function in an inflammatory context,⁴¹ and LAG-3⁺ Tregs are highly suppressive.⁴² Likewise, TIM-3 expression on Tregs is associated with an effector-like and highly suppressive phenotype.^43,44 Thus, further work is needed to differentiate between true exhaustion, gain of Treg effector function, and other T cell states, such as senescence. However, it is clear that dysfunction arising from tonic CAR signaling or chronic stimulation is detrimental to Treg function but can be reduced by LA-conditioning.

Beyond modulation of exhaustion-associated markers, a secondary effect of reducing CAR tonic signaling using LA was a reduction in the intensity of FOXP3 expression. Despite this decreased FOXP3 expression level, the proportion FOXP3-expressing cells was increased by LA. FOXP3 is well-recognized to be regulated by activation,²⁴ but the functional impact of activation-inducible FOXP3 on Tregs is not well understood. Virus-driven expression systems have shown that stable FOXP3 expression, rather than fluctuating, activation-inducible expression, is critical for establishing suppressive function in human FOXP3-transduced CD4⁺ T cells.⁴⁵ Considering that LA-treated CAR⁺ cells stably express FOXP3 as well as Helios, a known marker of Treg stability that correlates with greater demethylation of the TSDR,^20,23 the reduced intensity of FOXP3 is likely not indicative of destabilization. Nevertheless, we cannot rule out that a reduction in the total amount of FOXP3 protein does not influence other aspects of Treg phenotype and function.

CAR tonic signaling can be modulated by a variety of engineering strategies, including altering charge density on the extracellular domain, increasing single-chain variable fragment stability, and modulating expression level.^46,47,48 Such engineering approaches are associated with reduced exhaustion during expansion and greater efficacy in vivo. However, antigen-independent activation has also been associated with enhanced CAR T cell efficacy in vivo by maintaining survival in the absence of ligand.^48,49 As a result, reversible inhibition of tonic CAR signaling may be advantageous over receptor engineering approaches that strongly and permanently diminish tonic signaling. This has previously been achieved through in vitro use of dasatinib, a tyrosine kinase inhibitor that reduces T cell signaling, or by increasing the ionic strength of culture media using a pH-neutral salt, resulting in enhanced in vivo function in cancer models.^48,50 Although either of these approaches could likely reduce tonic signaling in CAR Tregs, the dual function of LA in enhancing Treg viability, expansion and purity, as well as reducing tonic signaling offers a significant advantage.

The use of LA to enhance Treg culture presents a highly translatable methodology to improve scaled-up clinical expansion protocols as it is both low-cost and straightforward. Although there is an initial lag in cell proliferation in the presence of LA, current clinical trials use expansion protocols ranging from 9 days to 15 days, with potential inclusion of second or third stimulations generally after >9 days.^{28,51,52,53,54,55} By reducing the contraction phase following expansion, LA is likely highly beneficial for these longer expansion protocols. Moreover, LA promotes the generation of consistently pure (high FOXP3⁺ proportion) cell products, which is required to meet release criteria for use in the clinic. Depending on the precise methodology for expansion, certain questions will require further investigation, such as how LA may synergize with other agents commonly used in Treg expansion, such as rapamycin. Furthermore, although we showed that LA-conditioning has a similar impact on bulk Tregs and naive Tregs, how LA impacts individual Treg subsets, including effector/central memory and naive-like memory Tregs,⁵⁶ requires further study. These Treg subsets have been shown to display unique phenotypes, stability, and suppressive function, influencing the efficacy of the Treg product.⁵⁶

Overall, our study demonstrates that pre-conditioning in LA, which is added at day 3 post-stimulation, enhances CAR Treg function in vivo by improving purity and reducing exhaustion. These effects are optimally driven by the synergistic function of lactate and low pH, rather than either component alone. Critically, enhancement of CAR Treg phenotype in vitro led to prolonged improvement in function in vivo, underscoring the importance of optimizing Treg phenotype during the expansion phase. Thus, LA represents a cost-effective, easy-to-establish method to improve Treg and CAR Treg function for adoptive cell therapy.

Materials and methods

PBMC and Treg isolation and expansion

Buffy coat products from healthy donors were obtained from the Canadian Blood Service with informed consent and ethical approval from the University of British Columbia Clinical Ethics Board and Canadian Blood Service Research Ethics Board (H18-02553). PBMCs were isolated from buffy coats by density gradient centrifugation with Lymphoprep Density Gradient Medium (STEMCELL Technologies) according to manufacturer protocol. For CD4⁺ T cell isolation, RosetteSep Human CD4⁺ T cell Enrichment Cocktail (STEMCELL Technologies) was used prior to density gradient centrifugation.

To isolate Tregs from CD4⁺ T cells, CD25⁺ cells were separated using CD25 MicroBeads II (Miltenyi Biotec). Naive Tregs (CD4⁺CD45RA⁺CD45RO⁻CD127⁻CD25hi) or bulk Tregs (CD4⁺CD127⁻CD25^hi) were sorted from the CD25⁺ fraction using a FACSAria Fusion (BD Biosciences) or MoFlo Astrios (Beckman Coulter) cell sorter. Naive Tconv (CD4⁺CD45RA⁺CD45RO⁻CD127⁺CD25⁻) cells were sorted from the CD25⁻ fraction.

Sorted Tregs or Tconv were stimulated with anti-CD3/CD28 Human T-Activator Dynabeads (Thermo Fisher Scientific) at a 4:1 bead-to-cell ratio and maintained in ImmunoCult-XF media (STEMCELL Technologies) supplemented with 1% penicillin-streptomycin (Gibco) and 1,000 IU/mL recombinant human IL-2 (Proleukin, Prometheus Laboratories Inc.) for Tregs and 100 international units (IU)/mL IL-2 for Tconv. Media was supplemented with L-(+)-lactic acid (Sigma), sodium L-lactate (Sigma), or HCl (Sigma) as indicated. Media was fully refreshed at intervals of 2–3 days. Assays were carried out at day 9 or day 12 as specified in Immunocult-XF media, unless otherwise described. Cell counts and viability were assessed using a CellDrop FL cell automated cell counter (DeNovix) with acridine orange and propidium iodide dye (Nexcelom).

Flow cytometry

Cells were washed and stained in fluorescence-activated cell sorting (FACS) buffer (PBS + 1% bovine serum [Hyclone]) with fluorescent antibodies against surface markers and fixable viability dye eF780 (Thermo Fisher Scientific) for 40 min at 4°C (Table S1). Cells were fixed and permeabilized using eBioscience FOXP3/Transcription Factor Staining Buffer Set (Invitrogen) prior to staining for intracellular markers (Table S1). Data were acquired using a BD LSR Fortessa or BD FACSymphony A1 or A5 (BD Life Sciences). Data were analyzed using Flowjo V10 Software (BD Life Sciences).

PBMC suppression assay

All suppression assays were carried out in Immunocult-XF with 1% penicillin-streptomycin and 5% human serum (NorthBio). Tregs and PBMCs were labeled with CPD eFluor670 or CPD eFluor450 (Thermo Fisher Scientific), respectively. 1 × 10⁵ PBMCs were cultured with serially diluted Tregs and stimulated with anti-CD3/CD28 Human T-Activator Dynabeads for 72 h. PBMCs and Tregs were stained with fluorescent antibodies (Table S1) and analyzed by flow cytometry. Division index of responder CD4⁺ and CD8⁺ T cells was determined using the FlowJo Proliferation tool. T cell suppression was calculated as previously described relative to the PBMC only condition.⁵⁷ CD80 and CD86 expression on CD19⁺CD3⁻ B cells were measured by flow cytometry and quantified relative to PBMC alone.

Glucose and lactate measurements

Media samples were collected from Treg cultures at indicated times and stored at −20°C until analysis. Glucose and lactate concentration was measured using a Cedex Bio HT Analyzer (Roche) using the Glucose Bio HT and Lactate Bio HT test kits, respectively.

Seahorse assay

Tregs were harvested and re-suspended in Seahorse media (unbuffered RPMI supplemented with 1 mM pyruvate and 4 mM L-glutamine adjusted to pH 7.4) and plated at 2 × 10⁵ live cells per well in triplicate in a Seahorse XF96 culture plate coated with poly-D-lysine (Sigma) and allowed to adhere for 45 min at 37°C without CO₂. A hydrated XFe96 cartridge was loaded with oligomycin (assay concentration 1 μM), carbonyl cyanide-p-(trifluoromethoxy)phenylhydrazone (FCCP) (assay concentration 1.5 μM), and antimycin A (assay concentration 1 μM) + rotenone (assay concentration 100 nM). ECAR (mPh/min) and OCR (pmol/min) were analyzed using the Seahorse XFe96 Analyzer (Agilent Technologies).

Lentivirus generation and transduction

All CAR constructs were previously generated and cloned into a pCCL lentivirus backbone containing an EF1α promoter to drive CAR expression and a minimal-cytomegalovirus promoter controlling expression of truncated NGFR as previously described.²⁰ Lentiviral particles were generated as previously described.⁴⁵ HLA-A2− naive human Tregs were stimulated as described previously and transduced with a multiplicity of infection of 5 viral particles to 1 cell. At day 9, NGFR+ cells were magnetically purified using an EasySep Human CD271 Positive Selection Kit II (STEMCELL Technologies).

CD14⁺ cell isolation, monocyte-derived DC differentiation, and DC suppression assay

CD14⁺ cells were isolated from HLA-A2+ PBMCs using an EasySep Human CD14 Positive Selection Kit II (STEMCELL Technologies). Mature monocyte-derived DCs were differentiated as previously described.⁵⁸ Briefly, CD14⁺ monocytes were cultured in Immunocult-XF with 1% penicillin-streptomycin, 50 ng/mL granulocyte-macrophage colony-stimulating factor (GM-CSF) and 100 ng/mL IL-4 (STEMCELL Technologies) for 5 days. To induce maturation, monocyte-derived DCs were cultured with 10 ng/mL IL-1β, 100 ng/mL IL-6, 50 ng/mL TNFα (STEMCELL Technologies), and 1 μg/mL prostaglandin E2 (Tocris) from day 5 and 50 ng/mL IFN-γ (eBioscience) from day 6.

To measure DC suppression, 1 × 10⁵ HLA-A2+ monocyte-derived DCs were co-cultured with CAR Tregs at a 1:1 ratio for 96 h. Expression of CD80 and CD86 on monocyte-derived DCs and expression of activation-induced markers on Tregs was determined by flow cytometry.

In vivo CAR Treg chronic stimulation model

All in vivo experiments were approved by the University of British Columbia Animal Care Committee (A22-0120). To chronically stimulate A2-CAR Tregs in vivo, female or male HLA-A2-transgenic NSG mice (13–29 weeks old were intravenously injected with 1–1.5 × 10⁶ control or LA-conditioned A2-CAR Tregs with an equal number of autologous HLA-A2- PBMCs. Mice were age-, weight-, and sex-matched across groups. Mice were monitored at intervals of 3–4 days and peripheral blood was collected from the saphenous vein weekly to monitor PBMC/Treg engraftment and phenotype. No signs of xenoGVHD symptoms were observed and mice were euthanized after 28 days. At the time of euthanasia, blood was collected by cardiac puncture and red blood cells lysed with ammonium chloride (STEMCELL Technologies) prior to flow cytometry analysis. Spleen, lung, and liver samples were dissociated through a 70 μM cell strainer and analyzed by flow cytometry.

xenoGVHD model

All in vivo experiments were approved by the University of British Columbia Animal Care Committee (A22-0120). To test CAR Treg function in vivo, female NSG mice (11–13 weeks old) were conditioned with 2 × 15 mg/kg busulfan on day −2 and −1 prior to cell injection. Pre-conditioned NSG mice were intravenously injected with 7 × 10⁶ HLA-A2+ PBMCs and 0.875 A2-CAR Tregs (8:1 PBMC-to-Treg) or 7 × 10⁶ HLA-A2+ PBMCs for the PBMC alone group. Mice were age- and weight-matched across groups. GVHD was scored at 1- to 3-day intervals according to weight, fur texture, posture, activity, and skin integrity with 0–3 points per category. Peripheral blood was collected from the saphenous vein weekly to monitor PBMC/Treg engraftment and phenotype. Mice were euthanized upon reaching a total score of 6 or a score of 3 in any one category. At the time of euthanasia, blood was collected by cardiac puncture and red blood cells lysed with ammonium chloride (STEMCELL Technologies) prior to flow cytometry analysis. Spleen, lung, and liver samples were dissociated through a 70 μM cell strainer and analyzed by flow cytometry.

Statistical analysis

Data were analyzed using Graphpad Prism V10. Data are presented as mean ± SEM, with individual donors indicated by points and connecting lines. Statistical significance was determined as described in figure legends. Statistically significant differences were defined as ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001.

Data and code availability

All data supporting study findings are available within the article and supplementary materials.

Acknowledgments

This work was supported by grants from Breakthrough T1D (formerly JDRF Canada) via the Breakthrough T1D Center of Excellence at UBC (grant key: 3-COE-2022-1103-M-B) and CIHR (DT4-179512). K.T. was supported by fellowships from the Canadian Institute for Health Research, Michael Smith Health Research BC, and Canucks for Kids Fund Childhood Diabetes Laboratories. M.K.L. is a Canada Research Chair in Engineered Immune Tolerance and receives a Scientist Salary Award from the BC Children’s Hospital Research Institute. We thank Dr. Lisa Xu for her assistance with FACS.

Author contributions

This study was conceptualized by K.T., R.I.K.G., and M.K.L. Experiments were performed by K.T., S.M., M.M., V.C.W.F., M.H., E.S.Y.L., A.-S.A., and R.V.G. Experimental design was carried out by K.T., D.A.B., R.I.K.G., and M.K.L. Analysis was performed by K.T. The manuscript was written by K.T. and critically evaluated by M.K.L.

Declaration of interests

K.T. and M.K.L. have patents pending related to the use of LA in T cell manufacturing.