Chimeric antigen receptor‐modified human regulatory T cells that constitutively express IL‐10 maintain their phenotype and are potently suppressive

Abstract Clinical trials of Treg therapy in transplantation are currently entering phases IIa and IIb, with the majority of these employing polyclonal Treg populations that harbor a broad specificity. Enhancing Treg specificity is possible with the use of chimeric antigen receptors (CARs), which can be customized to respond to a specific human leukocyte antigen (HLA). In this study, we build on our previous work in the development of HLA‐A2 CAR‐Tregs by further equipping cells with the constitutive expression of interleukin 10 (IL‐10) and an imaging reporter as additional payloads. Cells were engineered to express combinations of these domains and assessed for phenotype and function. Cells expressing the full construct maintained a stable phenotype after transduction, were specifically activated by HLA‐A2, and suppressed alloresponses potently. The addition of IL‐10 provided an additional advantage to suppressive capacity. This study therefore provides an important proof‐of‐principle for this cell engineering approach for next‐generation Treg therapy in transplantation.


Introduction
FOXP3 + regulatory T cells (Tregs) are a subset of CD4 + T cells that function to maintain self-tolerance and prevent inappropriate immune activation [1,2]. Tregs are currently under investigation as an adoptive cell-based therapy to prevent transplant rejection and for the treatment of autoimmune diseases [3,4]. Polyclonal Treg cellular therapy trials have shown promise in the prevention of graft-versus-host disease after allogeneic HSC transplantation [5,6], as well as in the maintenance of C-peptide levels in type 1 diabetes [7,8]. We have completed two phase I/II clinical trials, the ONE Study (NCT02129881) and ThRIL (NCT02166177), both assessing the safety and feasibility of adoptive transfer of polyclonal Tregs [9][10][11][12][13][14]. These trials have found Treg therapy to be safe, well tolerated, and with some early signs of efficacy [13,15]. However, evidence indicates that donor-specific Tregs are superior to polyclonal Tregs in pre-clinical models of transplantation [16][17][18], with donor-specificity achieved by culturing Tregs with allogeneic antigen-presenting cells (APCs) or by transduction of Tregs with T-cell receptors (TCR) specific for alloantigens [19].
Antigen specificity may also be conferred through the genetic engineering of Tregs to express chimeric antigen receptors (CARs). CARs are synthetic fusion proteins that comprise an extracellular antigen-targeting domain, hinge and transmembrane domains, one or more intracellular costimulatory domains, and a TCR-derived intracellular signaling domain [20][21][22]. CAR technology has the benefit of being customizable, from the target antigen to the signaling domains. Notably, CARs bypass major histocompatibility class (MHC) restriction. We and others have generated human Tregs expressing an HLA-A2-specific CAR and have shown that A2-CAR-Tregs are functionally superior in vitro and in vivo compared with polyclonal Tregs in a variety of humanized mouse models [23][24][25]. CAR engineering provides additional opportunities to produce cells that express other molecules as additional payloads. IL-10 is an anti-inflammatory cytokine produced by a wide range of both innate and adaptive immune cells that acts to limit inflammatory responses and prevent autoimmunity [26]. IL-10 signals through activation of the Jak kinase Stat transcription factor pathway, and although Stat3 is indispensable for this process, both Stat1 and Stat 5 have also been shown to be relevant to its signaling. IL-10 suppresses through the inhibition of proinflammatory cytokine secretion, as well as direct suppression of Th2 and Th17 cells and APCs [27,28]. Tregs can express IL-10 in vivo, but only after stimulation. The precise signals required for this are not yet clear but may be related to TGF-β signaling [29]. Interestingly, IL-10 can act in an autocrine fashion to further stimulate Tregs and enhance their function [30]. IL-10 has also been shown to promote the generation of suppressive regulatory-type 1 T cells (Tr1 cells) in vitro [31]. However, the effects of IL-10 are complex and its regulation key to the eventual outcomes [27]. For example, in specific situations IL-10 can promote the function of NK cells, CD8 + T cells, and B cells [32], resulting in enhanced anti-tumor immune responses in some experimental settings [33]. These pleiotropic effects may explain the failures in clinical trials of direct IL-10 administration in inflammatory bowel disease [32], and there is therefore an argument for a targeted IL-10 approach to control for any aberrant pro-inflammatory effects. In this context and to explore the potential role of IL-10 as an additional CAR-Treg payload, we built on our previously CAR-Treg work [24] to examine the impact of co-expression of IL-10 in HLA-A2 CAR Tregs.

Results and discussion
Generation of A2-CAR-Tregs co-expressing IL-10 We engineered four expression cassettes containing (i) both IL-10 and the HLA-A2-CAR ("IL10-A2-CAR"; open reading frame [ORF] size 4.85 kb, provirus size 8.67 kb), (ii) CAR alone ("A2-CAR"), (iii) IL-10 alone ("IL10-poly"), and (iv) neither cassette ("Poly") ( Fig. 1A, Supporting Information Tables S1-S5). Notably, all constructs contained the established radionuclide-fluorescence reporter NIS-TRFP [34,35] intended to streamline CAR-Treg production and comparison (through its red fluorescence aiding flow cytometry/FACS) and enable future in vivo CAR-Treg tracking. It is noteworthy that this fusion reporter has already been shown to not impact negatively on T cells [36]. Human Tregs were isolated and expanded in vitro as previously described (Supporting Information Fig. S1 and B; [9]). Transduction efficiencies were evaluated by NIS-TRFP expression (Supporting Information Fig.  S1C) before transduced cells were FACS sorted on day 10 and further expanded until day 20 (Fig. 1B, Supporting Information Fig.  S1D). All further analyses refer to cells that were expanded for 10 days, FACS-sorted, and expanded for further 10 days before indicated assays, which also involved gating on NIS-TRFP-positive and thus successfully transduced cells after expansion. In transduced Tregs, we also found the radionuclide reporter to be expressed and functional (Fig. 1C, Supporting Information Fig. S2). Notably, Treg lines transduced with an expression cassette encoding for IL-10 secreted this cytokine at high levels ( Fig. 1D). Moreover, those Treg lines transduced with vectors encoding for the A2-CAR were confirmed to express the CAR by positive staining with an HLA-A2 dextramer (Fig. 1E).

Characterization of A2-CAR-Tregs co-expressing IL-10
We next evaluated whether the process of Treg engineering impacted on their ability to expand or changed their phenotypes. Transduced and untransduced Tregs from the same batches were expanded in parallel and their phenotypes compared. We found no significant differences between different Treg types, neither in their expansion properties ( Fig. 2A) nor in their phenotypes ( Fig. 2B; Supporting Information Fig. S3). Notably, the expression of homing molecules relevant for the migration and function of CD4 + CD25 + CD127 lo Tregs including CD62L, CCR4, CCR9, CCR10, CLA, and β7 also remained unaffected by engineering with IL-10 in the presence or absence of the A2-CAR (Fig. 2C). Next, we assessed whether IL-10 co-expression would impact Treg activation. We co-cultured each transduced Treg population with one of two irradiated B-lymphoblastic cell lines (B-LCLs) that expressed either HLA-A2 + or HLA-A2 − together with the same HLA-DR haplotype (DR11), and subsequently analyzed CD69 upregulation in the Tregs as a measure of activation. As expected, both IL10-A2-CAR-Tregs and A2-CAR-Tregs upregulated CD69 after co-culture with the A2 + B-LCLs (49.9 ± 4.6% and and after FACS sorting (days 10-20; right) and compared to mock-treated untransduced cells (gray). Geometric means and SD of n = 6 different Treg batches (different donors, one donor per experiment) are shown. No significant differences were found between Treg types neither before (p = 0.9840) nor after FACS sorting (p = 0.5445; both by one-way ANOVA); however, expansion slowed across Treg types after FACS (p = 0.0340; two-way ANOVA with Tukey's multiple comparison correction). (B) Different Treg types (colors as in (A)) were analyzed for expression of indicated markers. Cells were first gated on CD4 + CD25 + . (C) Phenotypic marker analysis as in (B) but based on gating on CD4 + CD25 + CD127 lo cells (cf. CD127 marker in (B)). For both (B) and (C), cells were analyzed on day 20. Each individual data point belongs to one Treg batch/donor (symbol shapes identify donors). Error bars are mean ± SD from n = 4 different donors (one donor per experiment); no significant differences in marker expression between Treg types were found by one-way ANOVA (one test per marker with Tukey's multiple comparison correction).
54.9 ± 5.9%, respectively) but not after co-culture with the A2 − B-LCLs (Fig. 3A, Supporting Information Fig. S4). Importantly, Tregs lacking the A2-CAR did not show significant CD69 upregulation after co-culture with A2 + B-LCLs. Moreover, A2-CAR-expressing Tregs produced IL-4 and TNF-α when co-cultured with A2 + but not A2 − B-LCLs, whereas Tregs lacking the A2-CAR did not produce these cytokines when stimulated in this manner ( Fig. 3B and C). Co-expression of IL-10 did not impact on IL-4 or TNF-α production. Both were higher in A2-CAR-Tregs than in Tregs without the CAR; however, this phenomenon did not appear to be detrimental to Treg function, while its significance is yet to be determined [23]. Due to the constitutive expression of the IL-10 payload, secreted IL-10 levels were comparable in all conditions between the IL-10 co-expressing Treg types. Notably, A2-CAR Tregs also produced IL-10 upon stimulation with A2 + B-LCLs, although this was at much lower levels compared to IL-10expressing Treg types (Fig. 3D). The various engineered Tregs produced very little IL-17A on activation with either A2 + or A2 − B-LCLs with no significant differences between conditions, and generally at levels very similar to untransduced Tregs (Supporting Information Fig. S5A). Notably, at day 20 only a small percentage of our engineered Tregs express PD-1 (Supporting Information Fig. S5B), which could have been indicative of exhaustion if expressed in larger number of cells.
Taken together, these data demonstrate that the co-expression of additional payloads such as IL-10 and the imaging reporter did not impact on expansion capacity, Treg-specific phenotype, and expected activation of the correspondingly engineered Tregs.

IL-10 expression enhances the suppressive capacity of A2-CAR Tregs
To assess suppressive function, the various Treg types were cocultured with autologous pre-labeled Teff cells and stimulated by either A2 + B-LCLs or A2 − B-LCLs. The ability of Teffs to proliferate under these conditions was evident in all donors (Supporting Information Figs. S6 and S7). Despite the Treg cultures containing a small number of untransduced cells, both the Treg populations that expressed A2-CARs suppressed Teffs significantly more effectively when cultured with A2 + B-LCLs than cultured with A2 − B-LCLs (Supporting Information Fig. S8). Importantly, production of IL-10 by Tregs significantly increased the suppressive capacity of A2-CAR Tregs in the presence of A2 + B-LCLs (Fig. 3E [red>green] and F [blue>black]). Notably, the effect was not significant in the presence of A2 − B-LCLs or upon stimulation with anti-CD3/CD28 beads ( Fig. 3E and F: bottom panels). The effect seen with the A2 − BLCLs was likely due to stimulation via HLA-DR. Our findings were further supported by blocking the IL-10 receptor (IL-10R) through addition of a corresponding blocking antibody to the suppression assays (Supporting Information Fig.  S9). In the presence of anti-IL-10R, the suppressive capacity of IL10-A2-CAR Tregs returned to levels similar to A2-CAR Tregs, demonstrating specificity of the IL-10-dependent enhancement in suppressive capacity.

Concluding remarks
Our in vitro data demonstrate that it is feasible to engineer Tregs efficiently with large expression cassettes using lentiviral technology (our largest provirus length was 8.67 kb) despite approaching provirus lengths empirically associated with inefficient virus production [37,38]. We exploited this to transfer up to three different constructs as one ORF (CAR, cytokine, reporter) separated by 2A self-cleaving sequences [39] into primary human Tregs at high efficiency and with all components being functional (Fig. 1). Tregs remained unchanged in expansion characteristics and overall phenotype (Fig. 2). We further demonstrated that the co-expression of IL-10 as an additional payload enhanced human Treg suppressive capacity (Fig. 3, Supporting Information Fig.  S9). This was also evident in A2-CAR-Tregs, which were already significantly more suppressive than polyclonal Tregs if stimulated via their CARs by A2 + B-LCLs. It is noteworthy that suppression of A2 + B-LCLs was also somewhat higher than suppression with A2 − B-LCLs in Treg types without CAR expression, and we believe this is caused by a slightly stronger capacity of the latter B-LCLs to activate T cells (Supporting Information Fig. S6B: middle panels). Importantly, when the CAR was present, the difference between the curves was much larger than the differences observed between IL-10-poly and Poly Treg lines.
In the future, we envisage CAR Tregs to be produced with precision using a specific antibody-binding domain against a mismatched HLA haplotype of interest. CAR constructs that contain the most common MHC class I antigen-binding domains may be highly beneficial to ensure full population coverage. In this context, it will be important to explore how activation of CAR-Tregs through their endogenous TCR impacts on their function and efficacy, and whether TCR deletion needs to be considered (its feasibility has already been shown [40,41]). A likely benefit of the TCR-independent allospecificity mediated through CAR technology is the requirement for a reduced number of cells to be administrated. Nevertheless, it is paramount to engineer these Tregs for optimal efficacy. The way we have approached this, by coexpression of IL-10, is another step toward optimized CAR-Treg production.

CAR-Treg generation
Lentiviral particles were produced as previously described [43]. Virus titers were determined and Tregs transduced with the same multiplicity of infection (MOI) for all constructs. After 3 days of activation using anti-CD3/CD28 Dynabeads (1:1 bead/cell ratio), Tregs were transduced in retronectin-coated plates (50μg/μL; TakaraBio, France) with optimized viral titers. Transduced Tregs were cultured for an additional 7 days and then sorted on a FAC-SAriaIII (BD Biosciences, UK) using NIS-TRFP reporter fluorescence for positive selection.

Radiotracer uptake assay
Note that 3 × 10 6 indicated CAR T cells were transferred into Eppendorf tubes, washed with ice-cold HBSS (ThermoFisher), and resuspended in 1 mL growth medium. Also 50 kBq generatoreluted [ 99m Tc]TcO 4 − (supplied by local King's Health Partners' Radiopharmacy and used within two half-lives) were added and cells incubated for 30 min at 37°C. Subsequently, cells were pelleted, supernatants collected, and cells washed twice with 1 mL HBSS before being resuspended in growth medium for γ-counting (1282 Compugamma, LKB-Wallac). Radiotracer uptake was calculated according to the following equation, in which Cpm represents decay-corrected radioactivity counts per minute: %Tracer uptake = Cpm Cells Cpm Cells + Cpm [Supernatant] · 100 Statistical analyses were performed using Prism v8 software (GraphPad, CA, USA) with details added to figure legends and in the text.