Treatment of Uveitis by In Situ Administration of Ex Vivo–Activated Polyclonal Regulatory T Cells

This article is featured in In This Issue, p.1981

CD4⁺CD25⁺Foxp3⁺ regulatory T (Treg) cells play a major role in preventing autoimmune diseases. Profound Treg cell deficiency is associated with a severe autoimmune syndrome leading to death early in life in humans and mice (1, 2), and partial Treg deficiency has been reported in several human autoimmune diseases (3). In mouse models of organ-specific autoimmune diseases, it was shown that Treg cells specific for Ags expressed by the target tissue are much more efficient than polyclonal Treg (poly-Treg) cells to suppress the disease. For example, injection of small numbers of pancreatic islet Ag–specific Treg cells could efficiently control autoimmune diabetes, whereas poly-Treg cells had no beneficial effect (4–6). This is likely due to the preferential activation of Ag-specific Treg cells in the target organ or its draining lymph nodes, turning on their suppressive activity, inducing bystander inhibition of pathogenic cells (7–10). Treg cell therapy is thus a promising strategy for the treatment of chronic inflammatory and autoimmune diseases (11). However, Treg cells specific for given autoantigens are difficult to produce in conditions that can be translated to the clinic because of technological limitations. Also, autoantigens involved in many autoimmune diseases, such as in uveitis tested in this study, are unknown, impeding therapeutic application of autoantigen-specific Treg.

A therapeutic alternative would consist in injecting preactivated poly-Treg cells directly into the site of the autoimmune attack, postulating that their suppressive function, turned on in vitro, would be maintained after injection. This approach can be evaluated in autoimmune uveitis. Cells would be injected directly into the vitreous, where they remain confined because of the structure of the eye. In this work, we tested this cell therapy approach in uveitis in mice. We showed that poly-Treg cells were able to suppress uveitis but only when they were preactivated and administered in the vitreous. The clinical improvement was associated with decreased reactive oxygen species (ROS) produced by infiltrating immune cells.

Only female mice were used throughout the study. Six- to eight-week-old female BALB/c mice were obtained from Charles River Laboratories. The TCR-hemagglutinin (HA) transgenic BALB/c mice expressed a TCR recognizing I-Ed–restricted HA epitope 110–120 (SFERFEIFPKE) (12), and the CD90.1 BALB/c were bred in our animal facility under specific pathogen-free conditions and were manipulated according to the European Union guidelines. The animal protocol was approved by the Comité d’ethique en expérimentation animal Charles Darwin N°5.

First, HA was transduced in the retina using an adeno-associated AAV2/5 viral vector expressing HA under the control of the CMV promoter. The vector was produced by transient transfection in 293 cells with the pGG2-HAWT plasmid and with accessory plasmids providing adenoviral helper genes and adeno-associated viral rep2 and cap5 genes. The viral particles were purified by double cesium chloride gradient centrifugation and titrated by quantitative PCR. Subretinal injection of this AAV vector was performed under direct retinoscopy using a binocular microscope (Wild M3B; Leica) in mice anesthetized by ketamine (50 mg/kg) and xylazine (10 mg/kg), as described preciously (13). Briefly, a 33-gauge needle (Hamilton) was brought into focus between the retina and retinal pigment epithelium and 2 μl viral supernatant (± 3 × 10⁸ viral genome/ml) was injected in the two eyes, visualized by retinal detachment. One month later, mice were injected in the retro-orbital sinus with in vitro–activated 2 × 10⁶ CD25⁻ HA-specific cells prepared as followed: cells from the spleen and peripheral lymph nodes of TCR-HA mice were incubated with a biotin-labeled anti-CD25 mAb (7D4; BD Biosciences), followed by anti-biotin magnetic microbeads (Miltenyi Biotec). The CD25⁻ cell fraction was then obtained using LS columns or an autoMACS pro separator (Miltenyi Biotec) and CD25⁻ cells (1 × 10⁶/ml) were activated by syngenic dendritic cells (0.2 × 10⁵/ml, purified from the spleen using CD11c Miltenyi magnetic microbeads) pulsed with 10 μg/ml HA_111–119 peptide (PolyPeptide group) in complete culture medium (RPMI 1640 medium/10% FCS) supplemented with 10 ng/ml GM-CSF (R&D systems) for 4 d. After removal of dead cells using a Ficoll gradient, 2 × 10⁶ CD25⁻ cells were injected in the retro-orbital sinus for uveitis induction. Because HA was expressed in both eyes, mice developed bilateral uveitis, which allowed reducing number of mice used in the protocol for ethical reasons. Indeed, clinical scores of uveitis were the average of the clinical scores of the two eyes, increasing their significance, and flow cytometry analyses that required pooling several eyes necessitated the utilization of reduced mouse numbers.

Experimentally induced uveoretinitis scores were evaluated by clinical examination (slit lamp biomicroscopy) after pupil dilatation with tropicamide eye drop (Mydriaticum, Pharma Stulln). Ocular inflammation was graded as follows: grade 0, absence of apparent inflammation; grade 0.5, mild inflammation revealed by enlargement of iris capillaries without vitritis and protein exudation; grade 1, mild protein exudation in the aqueous humor and vitritis; grade 1.5, moderate protein exudation in the aqueous humor and reduced dilatation of the pupil without synechiae; grade 2, iris synechiae; grade 2.5, more than three synechiae but expandable pupil; and grade 3, fibrin clumps in the anterior chamber. Each grade includes the criteria of the lower grades.

Cells from spleen and peripheral lymph nodes of BALB/c mice, CD90.1 BALB/c or TCR-HA transgenic mice were incubated with a biotin-labeled anti-CD25 mAb (7D4; BD Biosciences), followed by anti-biotin magnetic microbeads (Miltenyi Biotec). The CD25⁺ cell preparations were then separated using LS columns or an autoMACS Pro separator (Miltenyi Biotec). This enriched freshly prepared Treg cell population contained 70% of CD4⁺CD25^highFoxp3^highcells. The other cells were composed of B cells (63%), myeloid cells (18%), CD8 and CD4 conventional T cells (16%) and dendritic cells (2%). For some experiments, Treg cells were further enriched by flow cytometry by sorting CD4⁺CD25^highCD62L^high cells after labeling cells with anti-CD4 mAb (RM4.5, FITC), L-selectin (MEL-14, PE) mAb, and streptavidin-PE-Cyanine 5, given highly purified Treg cells (99% Foxp3⁺ cells). Then, cells were activated at 10⁶ cell per ml during 18h in the presence of syngeneic splenic dendritic cells (0.2 × 10⁵/ml) in complete culture medium supplemented with 1 μg/ml anti-CD3 mAb (2C11; BD Biosciences), 1 μg/ml anti-CD28 mAb (37.51; BD Biosciences), 10 ng/ml GM-CSF, and 10 ng/ml mouse IL-2 (R&D Systems). Cells were injected in the blood by tail vein administration (2 × 10⁶ cells in 100 μl) or the vitreous (2 × 10⁵ cells in 2 μl) under anesthesia using isoflurane 4% or ketamine and xylazine, respectively. Control mice received the PBS diluent. Enriched Treg cells (70% pure) were used in Figs. 1, 2, 3, and 6. Highly purified Treg cells (99% pure) were used in Fig. 5. Enriched and highly purified Treg cells were both used in Fig. 4 and Supplemental Figs. 1–3. In these latter figures, data were pooled because similar results were obtained with both types of Treg cells.

Cells of lymphoid tissues were analyzed after mechanical dissociation. Cell analyses of the eyes were performed from the whole organ or from its choroid or retina fractions, mechanically separated in PBS. Then, tissues were digested using 100 μg/ml liberase (Roche) and 83 μg/ml DNAse-I (Roche) at 37°C for 30, 20, and 15 min for the whole eye, the choroid, and the retina, respectively. For the whole eye, cells were further mechanically dissociated and filtered on 70-μm cell strainer.

For intracellular cytokine staining, cells were restimulated with 1 μg/ml PMA (Sigma-Aldrich) and 0.5 μg/ml ionomicyn (Sigma-Aldrich) for 4 h (2 h in Fig. 5B) in the presence of GolgiPlug (BD Biosciences). For detecting IL-10, LPS (10 ng/ml; Sigma-Aldrich) was added to the culture. Cells were preincubated with 2.4G2 before staining. After cell surface staining, intracellular staining was performed using either the CytoFix/CytoPerm kit (BD Biosciences) or the fixation/permeabilization kit (eBioscience) after cell fixation in 2% paraformaldehyde (Fig. 5B). The following mAbs from BD Biosciences, BioLegend, or eBioscience were used for flow cytometry analyses: PE-labeled anti–IL-5 (TRFK5), –IFN-γ (XMΓ1.2), –IL-10 (JES5-16E3), and –glucocorticoid-induced TNFR-related protein (GITR; DTA-1); PerCP-labeled anti-CD90.1 (OX7); allophycocyanin-labeled anti-CD25 (7D4), -CTLA4 (UC10-AF10-11), -CD11c (HL3), -ICOS (C398-4A), -Foxp3 (FJK-16S), -IL17a (ebio1787), and –TNF-α (MP6-XT22); Alexa 700–labeled anti-CD4 (RM4.5), -CD8 (53-6.7), and -CD11b (M70); PE-Cyanin 7–labeled anti-CD11c (HL3); PE-CF594–labeled anti-CD45 (30-F11); brilliant violet 650–labeled anti-CD11b (M1/70); brilliant violet 510–labeled anti-CD4 (RM4-5); allophycocyanin-eFluor 780–labeled anti-CD19 (OX-7); and Alexa eFluor 700–labeled anti-CD8 (53-6-7) and eFluor 450–labeled anti-Foxp3 (FJ-16s). For detection of ROS, cells were preincubated with 10 μM 2′,7′-dichlorodihydrofluorescein diacetate (H₂DCFDA; Molecular Probes) for 30 min at 37°C, washed in PBS and stimulated with LPS (10 ng/ml) for 1 h in RPMI 1640 medium before cell surface staining and flow cytometry analyses. Cells were acquired on an LSRII cytometer (BD Biosciences) and analyzed using Flow Jo.

N-Acetylcysteine was administered at day 3 by i.p. injection (100 mg/kg [pH 7.4]) and once per day at days 3, 4, and 5 with one drop per eye (0.1% [pH 7.4]). Blocking anti–IL-10R mAb (200 μg per injection, 1B121C4) was injected i.p. at days 2, 4, and 6, and blocking anti-TNF mAb (500 μg/injection, XT3; BioXCell) was injected i.p. at days 3, 5, and 7. Control mice received the PBS diluent.

After dissection of the eyes, aqueous humor and vitreous (3–5μl/mice) was collected by puncture using a 23-gauge needle under a binocular microscope (Wild M3B; Leica) and stored at −80°C until testing. Myeloperoxidase was measured by ELISA kits (Abcam), according to the manufacturer’s procedures.

Eyes were fixed in Bouin liquid, paraffin embedded, sectioned (7 μm) through the papillary-optic nerve plane, and stained with H&E. Lesions of uveitis were graded as described previously (14): grade 0.5, mild inflammatory cell infiltration; no damage; grade 1, infiltration; retinal folds and focal retinal detachments; few small granulomas in choroid and retina; perivascularitis; grade 2, moderate infiltration; retinal folds and detachments; focal photoreceptor cell damage; small- to medium-sized granulomas; perivascularitis and vascularitis; grade 3, medium to heavy infiltration; extensive retinal folding with detachments; moderate photoreceptor cell damage; medium-sized granulomatous lesions; subretinal neovascularization; and grade 4, heavy infiltration; diffuse retinal detachment with serous exudate and subretinal bleeding; extensive photoreceptor cell damage; large granulomatous lesions; subretinal neovascularization.

Mice were anesthetized with ketamine and xylazine, and their pupils were dilated with tropicamide eye drop (Mydriaticum; Pharma Stulln, Stulln, Germany) before image acquisition using a scanning-laser ophthalmoscope. Before angiography, mice were injected in the peritonea with 100 μl 5% fluocyne (sodium fluorescein; Serb-labo, Paris, France). A glass coverslip was directly applied on the cornea with a drop of methyl-cellulose (0.5 g Gellaser; Alcon, Rueill Malmaison Cedex, France), and the signal was analyzed using the 488-nm laser. For spectral domain optical coherence tomography, we used the HRA2⁺ apparatus (Heidelberg Engineering, Heidelberg, Germany) supplemented with a 35-diopter lens fixed directly to the outlet of the device with tracking eye movement and real-time averaging scans. Resulting data were exported as 24-bit color bitmap files and processed in Adobe Photoshop.

Statistical data were calculated using the two-tailed unpaired Mann–Whitney test. The p values (*p < 0.05, **p < 0.01, and ***p < 0.001) were calculated with GraphPad Prism.

We assessed the therapeutic potential of Treg cells in uveitis in a mouse model that we have described previously (13). Briefly, the disease was induced by transferring effector T (Teff) cells specific for the HA model Ag (HA-Teff cells) into mice that stably expressed HA in the retina. Clinical signs of uveitis appeared as early as 3 d later, persisted for several weeks, followed by a slow remission over a 3-mo period. In this model, we previously showed that i.v. injection of HA-specific Treg (HA-Treg) cells controlled uveitis (13). In this study, we analyzed the therapeutic efficacy of these HA-Treg cells after in situ injection (2 × 10⁵ cells injected in the vitreous) versus systemic injection (2 × 10⁶ cells injected in the blood). We also compared the therapeutic effect of freshly purified versus preactivated Treg cells that were stimulated in vitro for 24 h before injection to turn on their suppressive function. Because the preparation of Treg cells that can be injected in patients is mostly done by sorting CD25^high expressing cells using magnetic beads, we purified mouse Treg cells using the same procedure, generating 70% of Foxp3⁺ cells. HA-Treg cells were injected at days 3–4 (meaning 3–4 d after HA-Teff cell injection), a time when mice had already severe uveitis. Clinical scores, assessed as depicted in Fig. 1A, were measured at day 12. Control mice that received PBS in the vitreous showed severe signs of uveitis with an average clinical score of 1.6. In situ administration of HA-Treg cells reduced clinical signs by >50%, even when they were not preactivated before injection (Fig. 1B). This therapeutic effect was similar or even better compared with the one obtained after systemic injection of 10 times more HA-Treg cells. We then assessed the kinetics of clinical scores after in situ injection of preactivated HA-Treg cells at days 3–4. The disease was suppressed as soon as day 8 and until day 19, when the experiment was stopped (Fig. 1C). Then, we tested whether these mice were protected from a new wave of pathogenic Teff cells by reinjecting HA-Teff cells at days 19–21. Compared with controls, mice initially injected with HA-Treg cells in the vitreous (at days 3–4) maintained low levels of clinical scores after readministration of pathogenic HA-Teff cells (Fig. 1D). Taken together, we showed that in situ administration of Treg cells specific for a retinal Ag provided long-term protection from uveitis.

We then tested the efficacy of autologous poly-Treg cells, an approach more relevant to the clinical context. Compared with controls that developed severe uveitis, poly-Treg cells injected at days 3–4 suppressed uveitis when they were preactivated prior to transfer and injected in situ (Fig. 2A). In contrast, no or minimal disease amelioration was observed when freshly purified poly-Treg cells were injected in situ or when 10 times more cells, even preactivated, were injected in the blood. The beneficial effect of in situ injection of preactivated poly-Treg cells on uveitis was very reproducible because it was observed on >20 independent experiments (data not shown).

In our uveitis model, mice developed a pan-uveitis that severely damaged the retina (13), which we studied by histology at days 12–14. PBS-injected controls showed extensive infiltrates and damage of the retina with focal destruction of the photoreceptor cell layer, associated with subretinal exudates (Fig. 2B, 2C). Poly-Treg cells injected at days 3–4 led to significant reduction of these lesions, showing a retinal structure relatively similar to the one of healthy unmanipulated mice. This beneficial effect of poly-Treg cells was observed only when cells were activated prior to transfer and injected in situ (Fig. 2B), confirming the clinical data. Similar findings were observed on live mice by optical coherence tomography and fluorescein angiography of the retina (Fig. 2C). In PBS controls, optical coherence tomography revealed dome-shaped retina, subretinal detachment, and vitreous opacities because of cell infiltration. Fluorescein angiography emphasized vasculitis, identified by enlarged and irregular vessels, and vascular leakage characterized by focal hyperfluorescence. All these signs of uveitis were abolished, or much reduced, in mice injected in situ with preactivated poly-Treg cells (Fig. 2C).

Because of its clinical relevance and therapeutic effect on uveitis, we have focused the rest of the study on the effect of in situ injection of preactivated poly-Treg cells. When injected at days 3–4, disease scores were reduced as soon as day 8 and up to day 19, when the experiment was stopped (Fig. 2D). Even when Treg cells were injected at the peak of disease (day 6), they were able to control the disease although with slightly less efficacy. Some mice, initially injected in situ with preactivated poly-Treg, were rechallenged at days 19–21 by reinjecting pathogenic HA-Teff cells, and the disease score was assessed 8–10 d later. Mice were not or minimally protected from uveitis by this challenge, suggesting that the effect of Treg cells was only transient (Fig. 2E). Altogether, these experiments showed that poly-Treg cells can efficiently suppress pan-uveitis but only if they were preactivated and injected in situ.

To explore the mechanism of action of preactivated poly-Treg cells in uveitis, we analyzed their persistence and activation in the injected eyes using a congenic marker (Fig. 3A). When injected in situ at day 3, numbers of donor CD4⁺ T cells progressively declined from days 5 to 20 (Fig. 3B). Among them, the percentage of Foxp3⁺ cells also progressively declined from 62% at day 5, which was similar to their initial proportion, to 32% at day 20 (Fig. 3C). This may be explained by the loss of Foxp3 expression in an inflamed environment, as described previously (15).

We then assessed whether Treg cell administration induced an increase of whole Treg cell numbers at day 5 in the eye. PBS controls had already a significant amount of recipient Treg cells in both the retina and the choroid (Fig. 3D), which represented 20–25% of the recipient CD4⁺ T cells. In Treg cell–treated mice, whole Treg cell numbers, composed of recipient and donor cells, were slightly but not significantly increased in the retina. Interestingly, donor Treg cells expressed higher levels of GITR but lower levels of ICOS and CTLA4, when compared with recipient Treg cells (Fig. 3E). Also, Treg cell injection did not modify reproducibly the expression of these activation markers on recipient Treg cells (data not shown). In conclusion, donor Treg cells persisted transiently in the eye and expressed different levels of activation markers compared with recipient Treg cells.

To further analyze the mechanism of uveitis control by injected Treg cells, we assessed numbers of infiltrating cells in the eye. In PBS controls, the numbers of immune cells per eye were in the range of 0.5–3 millions at day 5 or 8, with cells present in both the retina and choroid (Supplemental Fig. 1). Their numbers were <0.4 million in unmanipulated mice. Surprisingly, mice treated with Treg cells had similar numbers of whole immune infiltrating cells. We then determined the type of infiltrating cells in unfractioned eye or the retina or choroid. They were composed of CD4⁺ and CD8⁺ Teff cells and CD11b⁺ and CD11c⁺ myeloid cells (Supplemental Figs. 2, 3). The proportions of these cell subsets and DC subsets were similar in Treg cell–treated mice compared with controls (Supplemental Fig. 2 and data not shown).

We then analyzed whether Treg cells would exert their suppressive activity by inhibiting cytokine production by Teff cells, as shown in numerous studies (reviews in Refs. 16–18). In PBS controls, a significant fraction of CD4⁺ and CD8⁺ Teff cells produced IFN-γ and TNF, and low numbers of CD4⁺ Teff cells produced also IL-17 and IL-5 with variations between experiments (Fig. 4). This cytokine profile suggests that uveitis was a Th1- or Th17-mediated disease, as previously described in other models (19–23). As expected, Treg cells produced very low amount of these different cytokines (Fig. 4A). Again, quite unexpectedly, injection of therapeutic Treg cells did not reproducibly modify the production of these cytokines by Teff cells (Fig. 4). In conclusion, our model of uveitis was characterized by high inflammation in the eye, but surprisingly, Treg cell treatment did not reduce Teff or myeloid cell numbers or the ability of Teff cells to produce inflammatory cytokines.

One of the major suppressive mechanisms of Treg cells is IL-10 mediated (17), and IL-10 is a potent immunoregulatory cytokine in uveitis (22). We thus measured the capacity of eye infiltrating cells to produce this cytokine. In PBS controls, <1% of cells produced IL-10. This proportion was slightly but significantly increased in Treg cell–treated mice (Fig. 5A). This was confirmed using an improved protocol of IL-10 detection, which allowed detecting an increased proportion of cells producing the cytokine (Fig. 5B). These IL-10–producing cells were derived from the host and were composed of CD4 and CD8 T cells, as well as B cells, CD11b⁺ and CD11c⁺ cells. Treg cell injection did not modify the proportions of these cell subsets among IL-10–producing cells (Fig. 5B). To further assess the role of IL-10, we tested the effect of a blocking mAb. Quite remarkably, blocking IL-10 inhibited the beneficial effect of Treg cells in uveitis, suggesting that this cytokine played a role (Fig. 5C). We then analyzed factors that would be potentially pathogenic in uveitis and would be reduced by IL-10. TNF was a good candidate because anti-TNF drugs have therapeutic effects on uveitis in mice and patients (24, 25), and IL-10 reduced production of TNF by myeloid cells (26). We thus measured levels of TNF produced by CD11b⁺ cells. Compared with controls, we found a small but reproducible decrease of CD11b⁺ cells able to produce TNF in Treg cell–treated mice (Fig. 5D). However, blocking TNF using anti-TNF drugs had no effect on uveitis, suggesting that TNF was not pathogenic in our model of uveitis (Fig. 5C).

We then tested the role of ROS because these molecules are pathogenic in uveitis (27, 28) and because one of the pleiotropic immunoregulatory effects of IL-10 is to reduce ROS production by myeloid cells (29–31). Administration of N-acetylcysteine, a ROS scavenger, led to a significant reduction of clinical signs of uveitis, compared with nontreated mice (Fig. 6A), showing that ROS were involved in the pathogenic process. We then quantified levels of ROS in the eyes by flow cytometry. In PBS controls, microglia and dendritic cells produced the highest level of ROS, followed by macrophages, whereas lymphocytes produced low levels, in agreement with the literature (32). Strikingly, Treg cell injection was associated with a decrease of ROS levels in all myeloid and lymphoid cell subsets analyzed (Fig. 6B, 6C). The reduction of ROS production in the eyes of mice injected with Treg cells was confirmed when measuring the level of myeloperoxidase, an enzyme that is induced and upregulated by ROS mainly in granulocytes. Indeed, the amounts of myeloperoxidase levels were significantly reduced in mice injected with Treg cells compared with controls (Fig. 6D). In conclusion, uveitis suppression by Treg cells was associated with strong reduction of ROS production.

In this work, we showed that intravitreal injection of preactivated autologous poly-Treg cells suppressed an established uveitis. To our knowledge, this is the first demonstration that an autoimmune disease can be treated by in situ administration of activated poly-Treg. This approach may be envisioned in other organ-specific autoimmune diseases provided they meet two conditions: 1) the disease process can be suppressed within the target organ. This would be the case when the pathophysiological progression is perpetuated in the target tissue, after its initiation in lymphoid tissues, as reported in several organ-specific autoimmune diseases (33–36). 2) The target organ is accessible for cell injection, as for example, in arthritis where cells could be injected in the inflamed joints.

Compared with poly-Treg cells, the superior capacity of Treg cells specific for target tissue-Ags has been well described in different organ-specific autoimmune diseases (4–6, 37, 38). Similar findings have been reported in uveitis (39–42). In this study, we confirmed this observation because HA-Treg cells, but not poly-Treg cells, suppressed uveitis when injected i.v. Thus, there are two therapeutic strategies: in situ injection of preactivated poly-Treg cells or systemic injection of Treg cells specific for target-tissue Ags. Both approaches have conceptual pros and cons. Besides their better suppressive activity after systemic administration, Ag-specific Treg cells have another advantage. Compared with poly-Treg cells, HA-Treg cells induced long-term protection from a new wave of pathogenic Teff cells, probably because injected cells were chronically reactivated by cognate Ags, perpetuating their suppressive activity. On the other side, there are several major limitations for the use of Ag-specific Treg cells in the clinic. 1) In many autoimmune diseases, dominant T cell epitopes involved in the physiopathology are not or poorly known. 2) At the moment, no one has generated autoantigen-specific Treg cells in conditions that can be translated to the clinic. 3) Because there is no cell surface marker fully specific to human Treg cells, the process to obtain Ag-specific Treg cells would likely generate contaminating Teff cells with the same autoantigen specificities, which are potentially highly pathogenic. 4) Treg cells can trans-differentiate in pathogenic Teff cells, such as Th17 or Th1, especially in inflammatory contexts (15, 43, 44). These latter two concerns (contaminants and trans-differentiation) are much less critical in a cell therapy using poly-Treg cells because of the very low frequency, among the polyclonal repertoire, of putative auto-reactive pathogenic Teff cells. Altogether, compared with poly-Treg, Ag-specific Treg cells have potentially superior capacity to suppress an autoimmune disease but major technical and safety issues limit their utilization in the clinic.

The eye is an immune privileged site with potent suppressive mechanisms, partly involving Treg cells that differentiate locally from naive conventional T cells (45–47). Tolerance to autoantigens expressed in the eye is broken during uveitis, even though the tissue contains substantial amounts of Treg cells, at least in mouse models [our work and the one of others (48)]. Their inability to control the disease may be due to their quantitative or functional defect in this inflammatory environment, as shown in mice (47), or in patients during the active phase of uveitis (49–52). In this study, uveitis control by injected Treg cells was probably not due to an increase of whole Treg cell numbers in the eye, which was minor, but rather to a better efficacy of donor compared with recipient Treg cells. Interestingly, recipient and donor Treg cells exhibited differential expression of the GITR, CTLA4, and ICOS activation markers, which was probably due to their distinct origin and type of activation (in vitro versus in vivo in the eye). Compared with recipient Treg cells, donor Treg cells expressed lower levels of CTLA4 and ICOS, molecules that have been described to be part of the suppressive mechanisms of Treg cells (16–18). These findings suggest that uveitis suppression by donor Treg cells was not dependent on CTLA or ICOS. Surprisingly, Treg cell treatment had no effect at the peak of disease on Th1 or Th17 Teff cells that are pathogenic in uveitis (19–23, 53). Rather, Treg cell–mediated uveitis control in mice was correlated with a decrease of ROS production in the eye, revealed by the level of myeloperoxidase and by ROS activity in infiltrating cells. ROS have been shown to be pathogenic in uveitis (27, 28) by damaging endothelium cells and the retinal pigment epithelium, increasing the blood-retinal barrier permeability (28, 54).

To our knowledge, our study is the first to show that Treg cells induce a decrease of ROS production by immune cells. This decrease was observed in myeloid cells, the main ROS producing cells, which include granulocytes revealed by myeloperoxidase activity, microglial cells, inflammatory monocytes, macrophages or dendritic cells. This effect may be part of the immunosuppressive arsenal of Treg cells revealing a new mechanism of Treg cell–mediated suppression. Indeed, ROS act as critical signaling molecules for T cell activation (55) and oxidative stress has some inflammatory impact (32). Decrease of ROS production induced by Treg cells may be IL-10 mediated because we observed an increase of IL-10 production in the eye of Treg cell–treated mice and it has been shown that IL-10 reduced ROS production (29–31). It is unknown whether the beneficial effect of Treg cell injection could be fully replaced by local IL-10 administration. If it would the case, one may envisage using a slow-release nanodelivery devise to administer IL-10 locally. Nanotechnology for ocular drug delivery has already being used in the clinic. However, it is unclear whether such administration would be possible for IL-10 because of high physical and chemical constraints (56).

Autoimmune uveitis represents one of the major causes of severe visual decline and blindness worldwide (57). These patients are usually treated with corticosteroids, immunosuppressive drugs, or biological agents. However, these treatments have important side effects and are sometimes not efficient (58). Alternative treatments have to be proposed when all current therapeutic options have failed. Treg cell therapy is able to control uveitis in mice, as shown in this study and in other studies (13, 39, 40, 42), and a Treg cell deficiency has been reported in patients (49–52). On the basis of our data, we have initiated a phase I/IIa clinical trial (http://ichgcp.net/clinical-trials-registry/NCT02494492) to test the feasibility and safety of intravitreous injection of autologous preactivated poly-Treg cells in patients suffering from severe bilateral noninfectious uveitis. Considering the size of the eye, Treg cells are injected without requirement for prior in vitro expansion. The safety is quantified by evaluating visual acuity and retinal thickness as a readout of inflammation. We will also quantify levels of ROS, cytokines, and chemokines in the aqueous humor before and after Treg cell injection.

In conclusion, we have validated a new concept in cell therapy, consisting of in situ administration of autologous preactivated poly-Treg cells to suppress an inflammatory disease. We also showed that Treg cell–mediated suppression was associated with a reduction of ROS produced by immune cells, revealing a new mechanism of Treg cell suppression. This work led us to initiate an ongoing clinical trial in uveitis.

We thank Maxime Ferrand (Genethon), Carole Elbim, Angéline Rouers, Romain Vallion, and Marie-Christine Naud for excellent technical support and the Vector Core of the University Hospital of Nantes supported by the Association Française contre les Myopathies for the production of the AAV.

The authors have no financial conflicts of interest.