Design, generation and expression of MOG CAR construct in human and mouse Tregs
To engineer an anti-MOG CAR, a library of fully human anti-MOG scFvs was screened and the optimal candidate that is cross-reactive to both human and mouse MOG was cloned into a lentiviral vector to generate a MOG-CAR. The CAR contained a hinge domain, a transmembrane domain, and CD28-CD3ζ signaling domain. (Fig. 1a). Additionally, affinity of these human and mouse scFv was determined by measuring extracellular domain of CAR binding mouse or human MOG protein coated beads. In this assay, it was shown that the scFv has slightly less affinity for mouse MOG protein than to human MOG (Fig. 1a).
To generate human MOG-CAR Tregs, CD4+ CD45RA+ CD127lo CD25+naïve Tregs were sorted from peripheral blood of healthy donors and expanded prior to lentiviral transduction with a CAR construct. To generate mouse MOG-CAR Tregs, mouse CD4 + CD25 + CD44-CD62L + naïve Tregs were sorted from the spleen of C57BL/6 mice aiming to mimic the human process as much possible. After transduction, cell surface expression of MOG-CAR was confirmed with staining with protein-L for human construct and NGFR (nerve growth factor receptor) for mouse construct which bound to the scFv. CAR expression monitoring showed that transduction efficiencies ranged on average at 45.9 +/- 18.9% (mean+/-SD) for human construct and 62.5+/- 3.9% for mouse construct (Fig. 1b).
Moreover, human MOG-CAR Tregs maintained regulatory T cell phenotype following expansion, with high expression of CD4, CD25, CD45RA, CTLA-4, FOXP3, HELIOS. (Fig. 1C). For mouse Tregs, expression of FOXP3 remained on average over 80% positive after in vitro expansion. (Fig. 1d).
Altogether, MOG-CAR expression did not alter the phenotype of human and murine Tregs following transduction and expansion, fulfilling the basic requirement of a cell therapy product.
Human and mouse MOG CAR Tregs exhibit high signal-to-noise ratio and are specifically activated by MOG antigen in vitro
Both human and mouse MOG-CAR Tregs were analyzed after restimulation. Human and mouse MOG-CAR Tregs displayed negligible tonic signaling with basal activation level (% CD69) similar to untransduced polyclonal Tregs (mean+/- SD; 6.9+/-2.3% CD69 + for human MOG-CAR Tregs vs 5.7+/-2.4% for human polyclonal Tregs and 8.8 +/- 13.5% CD69 + for mouse MOG-CAR Tregs vs 5.6 +/- 6.4% untransduced mouse Tregs) indicating that transduced Tregs may not be prone to acquire an exhausted phenotype due to constitutive signaling through the CAR as previously reported[24]. Recombinant human MOG protein-coated plates were used to assess CAR-specific activation of MOG-CAR Tregs. Following overnight exposure to rMOG-coated plates, MOG-CAR Tregs significantly upregulated CD69 to a level similar or higher to TCR stimulation using anti-CD3/CD28 beads (Fig. 2a and b).
MOG protein represents only 0.05% of myelin sheath proteins in the CNS. To assess if MOG-CAR Tregs could be activated with physiological levels of MOG, an ex-vivo co-culture system was set-up. In this assay, MOG-CAR human and mouse Tregs are co-cultured overnight with spinal cord from C57Bl/6 mice. In this culture system, both human and mouse MOG-CAR Tregs significantly upregulated CD69 at a similar level to TCR stimulation using CD3/CD28 beads indicating that the CAR can activate Tregs upon target engagement as effectively as through their TCR (Fig. 2a and b).
Next, we analyzed whether the scFv we selected to engineer MOG-CAR was able to recognize the MOG protein expressed in human tissue using immunochemistry. We demonstrated that the scFv binds to human MOG expressed in the brain and spinal cord but not to heart tissue used as a negative control indicating that the scFv contained within our MOG CAR is highly specific to MOG (Fig. 2C).
Human MOG-CAR Tregs mediate target-antigen-specific immunosuppression and immunomodulation
To evaluate the function of human MOG-CAR Tregs, their ability to inhibit the proliferation of CD4 responder T cells (Tconv) were examined in a suppression assay. MOG-CAR Tregs were left unstimulated (NS) or were stimulated through their CAR or TCR prior to co-culture with CD4 Tconv. MOG-CAR Tregs significantly inhibited the proliferation of Tconv upon pre-stimulation with either MOG-coated plate or anti-CD3/CD28 beads, demonstrating an antigen-specific immunosuppression (Fig. 3a).
To demonstrate that human MOG-CAR Tregs possess the same primary functions as polyclonal Tregs in vitro, we performed additional assays to evaluate their immune-modulatory activity. Since CD8 T cells have been described to play an important role in the pathology of MS, the capacity of MOG-CAR Tregs to inhibit the proliferation of CD8 T conv cells was assessed in an in vitro suppression assay similar to the assay with CD4 T conv. As for CD4 T conv., MOG-CAR Tregs significantly inhibited the proliferation of CD8 Tconv upon pre-stimulation with either MOG-coated plate or anti-CD3/CD28 beads (Fig. 3b)
Given that CAR-Tregs-mediated DC suppression in vitro appears to be a superior predictor of in vivo Treg function [25], the ability of Tregs to suppress APCs has been suggested as an important functional criterion for an effective CAR-Treg therapeutic product [26]. Hence, we tested whether MOG-CAR Treg inhibited DC maturation and function in vitro. Monocyte-derived immature DC (imDC) were stimulated with LPS (DC alone) to induce their maturation or co-cultured with unstimulated (NS) or MOG pre-stimulated MOG-CAR-Tregs (MOG) in the presence of LPS. As previously described, the cell surface expression of costimulatory molecules CD80, CD86, CD40 and antigen presentation molecule HLA-DR was higher on mDCs compared to imDC. Moreover, it was demonstrated that Treg-treated DC displayed a less mature phenotype with decreased cell surface expression of these markers[27, 28] and increased expression of DC-SIGN. Consistent with these studies [29, 30], we show that DCs treated with MOG-CAR Tregs stimulated through their CAR inhibited DC maturation and increased level of DC-SIGN molecule on DCs (Fig. 3c).
Overall, these results show that the expression of MOG-CAR on human Tregs can suppress both T cells and antigen presenting cells upon their activation through the CAR.
MOG-CAR Tregs engineered from MS patient blood are functionally active in vitro
We aim to develop an autologous MOG-CAR Treg cell therapy to treat MS patients. There are conflicting reports on Treg in blood of MS patients with data indicating that the number of Tregs may be reduced or functionally impaired [5]. Therefore, we analyzed naive CD4+CD45RA+ CD127lo CD25hi Treg subset, regarding their phenotype and functionality in 5 patients with remitting-relapsing MS (RRMS) compared to healthy donors. We found that the proportion of naïve Tregs among CD4 + T cells in RRMS patients is similar to that found in healthy donors (mean+/- SD; 1.17%+/-0.56 and 1.14%+/-0.57 respectively) (Fig. 4a). The number of CD4+CD45RA+CD127loCD25hi Treg was also not significantly different in RRMS patients compared to healthy donors (mean +/- SD; 1.07 +/- 0.28 x 106 and 1.43 +/- 0.93 x106 respectively; not shown). After stimulation in vitro, the fold of expansion of RRMS naïve Tregs is not statistically reduced compared to healthy donors with an average fold expansion of 60.6+/-42 for RRMS Tregs versus 223 +/-343 for healthy donors (mean+/-SD) (Fig. 4b). Transduction efficiency of the MOG-CAR construct was similar in RRMS Tregs compared to healthy donors (mean+/-SD; 35.5%+/- 13.2 and 39.7% +/-16.3 respectively) (Fig. 4c). Moreover, after transduction and expansion in vitro, Tregs from RRMS patients compared to Tregs from healthy donors express the same level of FOXP3 (97.3+/-1.6 and 96.2 % /- 2.5% FOXP3 + mean+/-SD respectively) and Helios (mean +/- SD; 75.4 +/-7.9 and 69.9 +/-15 % elios + respectively) and have high level of FOXP3 hypomethylation (100% +/-0 FOXP3 TSDR hypomethylation for RRMS vs 87.7% +/- 14.0% for healthy donors; mean +/-SD) indicating that they are phenotypically identical and stable (Fig. 4d).
Functionality of RRMS MOG-CAR Tregs compared to healthy donors MOG-CAR Tregs was assessed in vitro after stimulation using recombinant MOG protein-coated plates or CD3/CD28 beads. The tonic signaling of MOG-CAR Tregs was similar between RRMS patient Tregs compared to Tregs from healthy patients (mean+/-SD; 16.3 +/- 17.3% CD69 + and 12.4 +/- 13.3% CD69 + respectively). Following overnight exposure to rMOG-coated plates or CD3/CD28 beads, MOG-CAR Tregs from RRMS patients significantly upregulated CD69 to a level similar to MOG-CAR Tregs of healthy donors (mean+/- SD; 60.8+/-29.6% CD69 + for RRMS vs 66.9 +/- 16.8% CD69 + for healthy using rMOG-coated plates) or higher to TCR stimulation using anti-CD3/CD28 beads (mean +/- SD; 55.7 +/- 18.6% CD69 + for RRMS vs 49.9 +/- 15.7% CD69 + for healthy using CD3/CD28 beads) (Fig. 4e). This data suggests that MOG-CAR Tregs from RRMS patients can be activated in vitro through the CAR as efficiently as MOG-CAR Tregs from healthy patients.
The capacity of MOG-CAR Tregs of RRMS patients compared to healthy donor MOG-CAR Tregs to suppress CD4 + Tconv proliferation was assessed in vitro. MOG-CAR Tregs were left unstimulated or pre-stimulated with rMOG coated plates or CD3/CD28 beads and co-cultured with CD4 T conv for 3 days. MOG-CAR Tregs from RRMS patients inhibited the proliferation of Tconv upon pre-stimulation with either MOG-coated plate or anti-CD3/CD28 beads as efficiently as MOG-CAR Tregs from healthy patients (mean+/- SD; 41.1 +/- 24.4% suppression for RRMS MOG-CAR Treg preactivated with rMOG vs 51.7 +/- 17.4% suppression for healthy at ratio Tconv:Treg 2:1) (Fig. 4f), indicating that MOG-CAR Tregs from RRMS patient can still exert immuno-suppressive function in vitro.
Altogether, this data suggests that MOG-CAR Tregs isolated from RRMS patients are functionally active in vitro.
Taken together, we showed that naïve Tregs isolated from the blood of RRMS patients are suitable to manufacture an autologous MOG-CAR Treg cell therapy product to treat MS patients.
MOG-CAR Tregs localize in the CNS of EAE mice
As there is no model of EAE in humanized mice to assess the migration and efficacy of human MOG-CAR Tregs, we developed a surrogate mouse MOG-CAR to evaluate their mode of action in vivo.
The migration and the localization of MOG-CAR Tregs was monitored in a mouse EAE model using bioluminescence imaging. SJL mice were immunized with PLP in CFA and mouse luciferase-expressing MOG-CAR Tregs were injected intravenously 7 days post-immunization. As control, these cells were also injected in non-immunized mice. Two days post-injection, scarcely any MOG-CAR Tregs could be detected in the brain and spinal cord of the animals (CNS) (data not shown). However, MOG-CAR Tregs started to be detected in the CNS of immunized animals four days post-injection but not in non-immunized animals and accumulated in the CNS after 7 days as shown in imaging and the quantification of the bioluminescence signals (Fig. 5a). Ex-vivo analysis of organs showed that MOG-CAR Tregs remained in the spleen and lymph nodes of non-immunized animals and did not migrate to the CNS compared to immunized animals ( Fig. 5b), indicating that MOG-CAR Tregs localize specifically in inflamed tissues.
Next, the migration of MOG-CAR Tregs to the CNS was assessed in a passive model of EAE. Mouse Tregs were isolated from C57BL/6 mice, transduced with clinical surrogate MOG-CAR construct or a non-signaling MOG-CAR (control MOG-CAR Tregs) and expanded in vitro for 5 days. One day after intra-peritoneal injection of pathogenic cells, MOG-CAR Tregs or control MOG-CAR Tregs were injected intravenously. Mice were sacrificed 10 days later and brain and spinal cord were analyzed by flow cytometry for the presence of CAR Tregs. In this model, 4-fold more MOG-CAR Tregs were found in the CNS compared to MOG-CAR control Tregs of diseased animals (mean +/- SD; 159.6 +/- 178 MOG-CAR Tregs versus 33.0+/- 24.9 MOG-CAR control Tregs) (Fig. 5C). Altogether, this data suggests that MOG-CAR Tregs specifically migrate to the CNS of EAE mice upon inflammation.
MOG-CAR Tregs delay the onset of EAE and reduce the activity of pathogenic cells in the CNS
Next, we assessed the therapeutic efficacy of mouse MOG-CAR Tregs in reducing EAE disease score in a model of passive EAE as described above. In this model, the injection of MOG-CAR Tregs delay the onset of EAE by 2 days compared to mice injected with MOG-CAR Control Tregs or saline. Moreover, MOG-CAR Tregs were shown to reduce the incidence of EAE more efficiently than MOG-CAR control Tregs with 100% and 93% of mice showing signs of EAE at day 9 in the saline and MOG-CAR control Tregs group respectively versus only 50% for mice treated with MOG-CAR Tregs. At day 11, only 64.3% of mice showed signs of EAE in the MOG-CAR Treg treated group (Fig. 6a). Moreover, MOG-CAR Tregs were more efficacious than MOG-CAR control Tregs at reducing disease activity score (54.5% reduction of disease score with MOG-CAR Tregs versus saline at day 11 compared to 0.09% reduction with MOG-CAR control Tregs) up to 11 days post-injection. Interestingly, the number of mice with no EAE score was increased when treated with MOG-CAR Tregs compared to MOG-CAR control Tregs or saline (Fig. 6b). When analyzing the activity of pathogenic cells isolated from the CNS of the mice at day 11, it was observed that the percentage of CD4 + cells producing IFN-g was significantly reduced in mice treated with MOG-CAR Tregs compared to saline or MOG-CAR control Tregs treated mice (mean+/-SD; 25.8% +/- 9.1 for MOG-CAR Tregs vs 43.16% +/- 7.99 for saline and 35.34% +/- 8.5 for MOG-CAR control), indicating that MOG-CAR Tregs can efficiently modulate pathogenic cell activity in the CNS. (Fig. 6c). Moreover, Histopathologic analysis of the spinal cord of the animals harvested at day 11 showed that in mice treated with MOG-CAR Tregs, the number of spinal cord sections with demyelinating leukomyelitis was reduced compared to untreated mice or treated with MOG-CAR control Tregs (11 spinal cord sections affected for untreated mice, 9 for MOG-CAR control Tregs and 7 for MOG-CAR Treg treated mice) with a mean score of 0.82 for MOG CAR-Tregs treated mice versus 2.0 for untreated and 1.4 for MOG-CAR control Tregs (Fig. 6d). Immunostaining of spinal cord for Iba-1, a marker of microglial activation, showed a significant reduction of the percentage of Iba-1 positive staining in mice treated with MOG-CAR Tregs compared to untreated or MOG-CAR control treated mice (mean+/-SD; 0.22+/-0.55 for MOG-CAR Tregs versus 0.67+/-0.79 for MOG-CAR control Tregs and 0.7 +/- 0.89 for untreated mice) (Fig. 6e).
Altogether, despite the limitation of the model, this data suggests that MOG-CAR Tregs are efficient at reducing inflammation in the CNS of EAE mice by modulating the activity of CD4 + pathogenic cells and reducing microglial activation.