Thymic-Specific Serine Protease Limits Central Tolerance and Exacerbates Experimental Autoimmune Encephalomyelitis

Multiple sclerosis (MS) is a chronic inflammatory demyelinating disease of the CNS that is thought to be driven by CD4 T cells. This notion is supported by genome-wide association studies showing a strong association of MS susceptibility with HLA-DRB1 and with nucleotide polymorphisms within regulators of T cell activation, proliferation, and effector T cell differentiation (1–3). Experimental autoimmune encephalomyelitis (EAE), an animal model of MS, can be induced by immunization with myelin proteins or with immunodominant epitopes of myelin proteins (4). In several strains of mice, including C57BL/6 (B6) and NOD mice, EAE is induced by immunization with the immunodominant epitope of myelin oligodendrocyte glycoprotein (MOG)35–55, which triggers the expansion and differentiation of CD4 T cells into distinct effector T cell subsets, including the highly pathogenic Th1 and Th17 subsets (5).

The strong immune response induced by MOG_35–55 immunization indicates that central and peripheral T cell tolerance to MOG_35–55 is incomplete. This is common for proteins with tissue-specific expression, like MOG, which is expressed by oligodendrocytes in the CNS (6). Tolerance to self is primarily established in the thymus by deletion or deviation from the regulatory lineage of developing thymocytes. In the thymus, medullary thymic epithelial cells (mTECs) have the unique capacity to express a vast array of peripheral tissue-restricted Ags, resulting, at least in part, from expression of the transcription factor AIRE (7, 8). These self-antigens may be directly presented by mTECs or transferred to and presented by thymic-resident dendritic cells (DCs) for CD4 T cell deletion (9–13) or regulatory T cell (Treg) generation (9, 14–18). An exhaustive analysis of the relative contribution of mTECs and DCs to the deletion of self-reactive CD4 T cells or Treg generation further showed that both stromal APCs contributed equally to these processes but impacted distinct T cell repertoires (19). In agreement with promiscuous expression of tissue-specific Ags by mTECs, MOG transcripts are found in mTECs but not in thymic DCs or cortical thymic epithelial cells (cTECs) (7). However, this expression has only a moderate impact on central tolerance to MOG, suggesting that the immunodominant MOG_35–55 peptide is not abundantly presented by thymic APCs (20).

Generation of the MHC/peptide complexes recognized by CD4 T cells relies on the sequential proteolysis of endosomal proteins by different Ag-processing enzymes, such as the cathepsin (Cat-) family of cysteine proteases, the aspartyl proteases Cat-D and Cat-E, or the asparaginyl endopeptidase (21). Ag-processing enzymes may also impair generation of some MHC class II/peptide complexes, as described for Cat-S and asparaginyl endopeptidase (22, 23). Thymus-specific serine protease (TSSP) is another remarkable example of such “destructive” proteases. Initial studies showed that TSSP is specifically expressed by cTECs and regulates positive selection of some I-A^b–restricted CD4 T cells (24). However, more recent transcriptomic studies suggested that mTECs, including AIRE^Pos mTECs, may also express low levels of TSSP mRNA (25–27). In addition, we have shown that TSSP is expressed by thymic DCs but not peripheral DCs (28). In NOD mice, expression of TSSP by thymic stromal cells limits the presentation of several islet Ags by unknown mechanisms and, thus, limits central tolerance to these self-antigens (28–30). Hence, in NOD mice, lack of TSSP expression by thymic DCs or thymic epithelial cells (TECs) enhances thymic negative selection, as demonstrated by the increased deletion of several islet-reactive CD4 T cells, and prevents diabetes development (28–30).

We show in this article that, in NOD mice, lack of TSSP expression by thymic DCs also favors the intrathymic deletion of MOG_35–55-reactive CD4 T cells, thereby reducing EAE severity. Remarkably, although NOD and B6 thymic DCs express comparable levels of TSSP, lack of TSSP expression had no impact on disease severity in B6 mice expressing the I-A^b class II haplotype, indicating that the effect of TSSP on Ag presentation is constrained by the class II haplotype. When analyzing TSSP expression by human thymic DCs, we found that some individuals show high expression levels. whereas others show a 10-fold lower level of expression. Altogether, these results suggest that the level of TSSP expression by thymic DCs may contribute to MS severity.

Tssp-deficient (Tssp°) B6 and NOD (NOD Tssp°), NOD wild-type (WT), NODscid, Tssp° NODscid, NOD-Tcra–deficient (NOD Cα°), and Tssp° NOD Cα° mice have been described previously (24, 28). For generation of bone marrow (BM) chimeras, 10–12-wk-old NOD Cα° and Tssp° NOD Cα° mice were irradiated (9.5 Gy) the day before reconstitution with 8 × 10⁶ T cell–depleted NOD WT BM cells mixed at a 3:1 ratio with NODscid or Tssp° NODscid BM cells, as previously described (30). This ratio is sufficient to induce deletion of islet-reactive CD4 T cells by TSSP-deficient APCs (28–31).

All experiments involving animals were performed in accordance with national and European regulations and INSERM institutional guidelines. Protocols were approved by the “Midi Pyrénées” ethical committee for animal studies.

For active EAE, mice were immunized s.c. in the tail base with 100 μg of HPLC-purified (>95% purity) MOG_35–55 peptide (GeneCust) emulsified in CFA (Sigma) and supplemented with 100 μg of Mycobacterium tuberculosis H37RA (Difco Laboratories). Pertussis toxin (List Biological Laboratories) was given i.v. at day 0 (200 ng) and at day 2 (400 ng) after immunization.

In some experiments, mice were adoptively transferred a day before EAE induction with 1 × 10⁶ T cells purified, by negative selection, from the lymph nodes (LNs) and spleen of mice.

To examine the encephalitogenic potential of conventional CD4 T cells, CD4 T cells were purified from the spleen of NOD WT or NOD Tssp° mice by negative selection and subsequently depleted of CD25⁺ T cells using PE–anti-CD25 and anti-PE MicroBeads, as previously described (31). In parallel, CD8 T cells were purified from the spleen by negative selection. NOD Cα° mice were adoptively transferred with 4 × 10⁵ CD8 T cells, together with 6 × 10⁵ conventional CD4 T cells corresponding to the normal CD4/CD8 ratio. EAE was induced the next day, as described above.

To analyze the suppressive activity of CD4 Tregs, CD4 T cells were purified from the spleen of NOD Tssp° and NOD WT mice by negative selection, and CD25⁺ T cells were subsequently isolated using PE–anti-CD25 and anti-PE MicroBeads, as previously described (28). NOD Cα° mice were adoptively transferred with 4 × 10⁵ NOD WT CD8 T cells, 5.1 × 10⁵ NOD WT CD4 T cells, and 9 × 10⁴ CD4⁺CD25⁺ Tregs from NOD WT or NOD Tssp° mice, corresponding to the normal ratio of conventional CD4 T cells and Tregs. EAE was induced the next day, as described above.

Disease severity was scored daily on a five-point scale: 1, tail atony; 2, hind limb weakness; 3, hind limb paralysis; 4, quadriplegia; and 5, moribund or death.

Mice were immunized with 100 μg of MOG_35–55, as described above. Nine days later, CD4 T cells were isolated from draining LNs (dLNs) by negative selection, as previously described (31). For ELISA, 5 × 10⁵ purified CD4 T cells were stimulated in the presence of MOG_35–55 peptide, along with 5 × 10⁵ irradiated (25 Gy) NOD splenocytes. Supernatants were collected 3 d later, and cytokine production was measured using the adapted DuoSet ELISA System (R&D Systems) and TMB Substrate Reagent Set (BD Biosciences), according to the manufacturers’ instructions. For ELISPOT assays, triplicates of 2.5 × 10⁵ purified CD4 T cells were stimulated in the presence of MOG_35–55 peptide, along with 5 × 10⁵ irradiated (25 Gy) T cell–depleted NOD splenocytes for 24 h. IFN-γ and IL-17 production was measured using mouse ELISPOT Ready-SET-Go! kits (eBioscience) and Multiscreen-IP Sterile Plates (Millipore), according to the manufacturers’ instructions. Spots were monitored using an ImmunoSpot S6 Micro Analyzer (CTL). The number of spots per 10⁶ CD4 T cells was calculated after subtraction of the background values (stimulation in the absence of peptide), which was less than two spots per 10⁶ CD4 T cells.

At 15 and 34 d post-EAE induction, mice were euthanized and perfused with PBS. Brain and spinal cord were harvested and digested with Collagenase D (2 mg/ml), TLCK (10 μg/ml), and DNase I (20 μg/ml) (Roche), followed by Percoll gradient centrifugation. The leukocyte infiltrates were characterized using Abs against CD45 (A20), CD4 (RMA4-5), CD8 (53-6.7), CD19 (1D3), TCRβ (H57-597), and TO-PRO-3 as viability dye. Samples were analyzed using a BD LSRFortessa cytometer (BD Biosciences) and FlowJo (TreeStar). Alternatively, the different CNS APCs were stained with Abs against CD45 (A20), CD19 (1D3), Thy1 (53-2.1), CD14 (Sa2-8), CD11b (M1/70), CD11c (N418), and CMHII (RT1B) and FACS sorted using a FACSAria II High Speed Cell Sorter (BD Biosciences).

I-A^g7–MOG_42–55, I-A^b–MOG_39–50, I-A^g7–CLIP, and I-A^b–CLIP tetramers were obtained from the NIH Tetramer Core Facility. Single-cell suspensions from spleens or dLNs were incubated for 2 h at room temperature with the optimal concentrations of allophycocyanin- and PE-conjugated tetramers. For some experiments, tetramer-positive cells were enriched using anti-PE and anti-APC magnetic beads, as previously described (6).

Human thymus tissue from children undergoing cardiac surgery was obtained from the Cardiology Department of the Purpan Children’s Hospital in Toulouse. The study received ministry approval (DC-2014-2088), and the participants’ legal representatives did not oppose the use of the tissue for research purposes. Thymic tissues were digested with Liberase (Roche) and DNase (Sigma). T cells and B cells were depleted using anti-CD3 and anti-CD19 magnetic beads (Miltenyi Biotec). The pre-enriched DC preparation was stained with anti-CD3–FITC, anti-CD19–FITC, anti-CD14–VioBlue, anti-BDCA1 PE-Vio770, anti–HLA-DR–PerCP–Cy5.5, anti-BDCA3–BV510, and anti-CD11c–PE Abs and propidium iodide, and the DC1 subset was FACS sorted.

Mouse DCs were positively selected from digested thymic tissue using CD11c MicroBeads (Miltenyi Biotec).

RNA was extracted using an RNA XS Isolation Kit (Macherey-Nagel), and cDNA was synthetized using SuperScript II enzyme and random primers (Invitrogen), according to the manufacturers’ instructions. Quantitative PCR was performed with a LightCycler 480 Real-Time PCR System (Roche), and results were analyzed with LightCycler 480 Software, Version 1.5. Primer sequences are provided in Supplemental Table I. The cycling threshold value of the endogenous housekeeping control genes was subtracted from the cycling threshold value of Tssp to generate the change in cycling threshold (ΔC_T).

Clinical scores were compared by two-way ANOVA with the Bonferroni correction. Otherwise, data were compared using the Mann–Whitney U test.

In agreement with previous studies (32–34), we found that NOD WT mice develop mild relapsing-remitting EAE following MOG_35–55 immunization (Fig. 1A). The first clinical signs were observed by day 13 postimmunization, and 100% of the mice had developed disease by day 19 (Fig. 1). Remission was observed at day 34, which was followed by a relapse of enhanced severity starting at day 41. Similarly, NOD Tssp° mice developed disease following MOG_35–55 immunization starting at day 13 postimmunization (Fig. 1A). However, in NOD Tssp° mice, disease severity was milder overall. Although all NOD Tssp° mice eventually developed disease, only 77% of the mice showed clinical signs at day 19 postimmunization (Fig. 1B), and most of the mice underwent remissions and mild relapses without late worsening (Fig. 1A). Accordingly, the maximal and cumulative disease scores were significantly lower in NOD Tssp° mice compared with NOD WT mice (Fig. 1B).

In agreement with the lower disease severity, NOD Tssp° mice showed limited hematopoietic cell infiltration in the CNS at day 15 compared with NOD WT mice; this correlates with a 5–10-fold reduction in the total number of infiltrating B cells, CD4 T cells, and CD8 T cells (Fig. 1C). However, at a later time point, when all NOD Tssp° mice had developed disease, hematopoietic cell infiltration into the CNS of NOD Tssp° mice was similar to that of NOD WT mice (Supplemental Fig. 1).

Collectively, the data show that, in NOD mice, lack of TSSP expression is associated with reduced neuroinflammation and EAE severity following MOG_35–55 immunization.

Restimulation of encephalitogenic T cells in the CNS is critical for EAE onset (35–37). Given that TSSP is a protease of the class II presentation pathway, it may contribute to Ag presentation by different CNS APCs, thus explaining the reduced EAE severity in NOD Tssp° mice. We have shown that, although TSSP is expressed by thymic DCs, it is not expressed by peripheral DCs, even upon TLR stimulation (28). However, we found that the inflammatory environment of the CNS associated with EAE induced TSSP expression by CNS-infiltrating or resident APCs (Fig. 2A). Indeed, microglia and macrophages, but not DCs, isolated from the CNS of NOD WT mice at day 15 post-EAE induction expressed significant levels of TSSP mRNA. To further examine the relative contribution of TSSP expression by CNS APCs to disease severity, we adoptively transferred NOD WT T cells into T cell–deficient NOD Cα° or Tssp° NOD Cα° mice and induced EAE by MOG_35–55 immunization. We found that disease initiation, progression, and severity were comparable, regardless of whether or not the host expressed TSSP (Fig. 2B, 2C). Hence, TSSP expression by CNS APCs is not contributing to EAE severity. In addition, the results further demonstrate that TSSP expression by peripheral DCs or B cells is not mandatory for full-blown EAE.

Therefore, the reduced EAE severity of NOD Tssp° mice is not due to a defect in the presentation of myelin Ags by CNS APCs.

Th17 effector CD4 T cells have a critical role in EAE development through the production of IL-17 and GM-CSF, whereas IFN-γ is thought to facilitate lymphocyte infiltration in the CNS (38–43). Therefore, we determined whether TSSP deficiency may reduce the magnitude of the encephalitogenic T cell response. For this experiment, CD4 T cells were isolated from the dLNs of NOD WT and NOD Tssp° mice 9 d after MOG_35–55 immunization and were restimulated in vitro with MOG_35–55 peptide. The production of IL-17, GM-CSF, and IFN-γ was assessed by ELISA. We found that CD4 T cells isolated from immunized NOD Tssp° mice produced significantly lower levels of IL-17 and GM-CSF compared with those from NOD WT mice (Fig. 3A, 3B). IFN-γ production by MOG_35–55-reactive CD4 T cells isolated from NOD Tssp° mice was also reduced compared with those from NOD WT mice, but this did not reach statistical significance (Fig. 3C). We further assessed whether this corresponded to a reduced frequency of IL-17–producing and IFN-γ–producing MOG-reactive CD4 T cells by ELISPOT assay. We found that the frequency of IL-17–producing and IFN-γ–producing MOG-reactive CD4 T cells was significantly reduced in immunized NOD Tssp° mice compared with NOD WT mice (Fig. 3D, 3E).

The above results suggest that lack of TSSP expression may reinforce central tolerance to MOG_35–55 and, thereby, reduce EAE severity. To directly test this possibility, we first assessed the pathogenicity of NOD WT T cells that had developed in a thymic environment in which only thymic APCs lack TSSP expression. Therefore, we generated mixed BM chimeras by reconstituting NOD Cα° (WT host) mice with a mix of NOD WT BM cells and Tssp° NODScid BM cells (knockout [KO] BM) as a source of TSSP-deficient DCs or, as control, NODScid BM cells (WT BM, WT control chimeras). To mimic NOD Tssp° mice, a second control group was generated by reconstituting Tssp° NOD Cα° mice (KO host) with a mix of NOD WT BM cells and Tssp° NODScid BM cells (KO control chimeras). Eight weeks later, T cells were isolated from these different mixed BM chimeras, and their encephalitogenic potential was examined upon adoptive transfer into NOD Cα° mice and MOG_35–55 immunization. As expected, the transfer of T cells isolated from WT control chimeras (BM→host, WT→WT) induced chronic EAE in 100% of the mice starting at day 12 postimmunization (Fig. 4A). Transfer of T cells from KO control chimeras (KO→KO) also resulted in EAE in 100% of the mice starting at day 13 postimmunization and with a maximal score comparable to that of WT control chimeras (Fig. 4). However, in the latter case, host mice showed relapsing-remitting disease, resulting in a significant reduction in the cumulative disease score. A similar disease course was observed when the adoptively transferred T cells developed in a thymic environment in which only the DC lacked TSSP expression (KO→WT, Fig. 4). Indeed, in this case, host mice also showed a relapsing-remitting disease with similar day of onset, maximal score, and cumulative score as that observed when T cells originated from KO control chimeras but significantly different from that observed when T cells originated from WT control chimeras. Hence, lack of TSSP expression by DCs reinforces central tolerance to MOG_35–55.

Given that TSSP-deficient TECs also impact the generation of the CD4 T cell repertoire, we next determined, using the same approach, whether TSSP-deficient TECs may also impact the development of encephalitogenic CD4 T cells. Therefore, we reconstituted Tssp° NOD Cα° (KO host) mice with a mix of NOD WT BM cells and NODScid BM cells (WT BM, WT→KO chimeras). We found that T cells isolated from these mixed BM chimeras induced a disease course similar to that observed for T cells isolated from KO→KO or KO→WT chimeras (Fig. 4). Indeed, in this case too, host mice showed relapsing disease and reduced cumulative score.

Given that thymic stromal cells can induce deletion of MOG-reactive conventional CD4 T cells or favor the development of Tregs, we assessed the representation and functional properties of the conventional and regulatory CD4 T cell compartments of NOD WT and NOD Tssp° mice. We first assessed the number of MOG_35–55-reactive CD4 T cells in naive NOD Tssp° mice and NOD WT mice by tetramer staining. As shown in Fig. 5A, the number of MOG_35–55-reactive CD4 T cells was significantly reduced in NOD Tssp° mice compared with NOD WT mice. A similar reduction was observed in immunized NOD Tssp° mice, suggesting that the clonal expansion of the remaining MOG-reactive CD4 T cells was not significantly impaired.

We next examined the pathogenicity of conventional CD4 T cells. We isolated CD4 T cells from NOD WT mice or NOD Tssp° mice, depleted them of CD25⁺ T cells to remove the vast majority of Tregs, and transferred them into NOD Cα° mice. EAE was induced by MOG_35–55 immunization the following day. One hundred percent of the recipient mice reconstituted with NOD WT CD4 T cells developed chronic EAE starting at days 8–10 postimmunization (Fig. 5B). Although the disease incidence, day of onset, and maximal score were similar when CD4 T cells were of NOD Tssp° origin, cumulative disease severity was significantly reduced (Fig. 5B, 5C). Furthermore, the transfer of NOD Tssp° conventional CD4 T cells induced a relapsing-remitting disease (Fig. 5B). We next compared the functional properties of NOD WT and NOD Tssp° CD4⁺CD25⁺ Tregs in regulating the encephalitogenicity of NOD WT conventional CD4 T cells. We found that NOD WT and NOD Tssp° Tregs were equally efficient at reducing EAE severity induced by WT effector CD4 T cells (Supplemental Fig. 2).

Collectively, the results indicate that lack of TSSP expression by thymic stromal cells favors the deletion of MOG_35–55-reactive CD4 T cells and significantly limits the encephalitogenic potential of conventional CD4 T cells.

Recent studies in B6 mice showed that MOG is expressed by mTECs, which leads to some level of deletion of MOG-reactive CD4 T cells, although a substantial fraction of this repertoire escapes central tolerance (20). Whether this results from TSSP expression by thymic stromal cells, as observed in the NOD mouse, is questionable given the structural differences in the class II I-A^b and I-A^g7 molecules. Indeed, the highly conserved Proβ56 and Aspβ57 of the β-chain is substituted with Hisβ56 and Serβ57 in I-A^g7 molecules, which considerably modifies the P1 and P9 pockets of I-A^g7, thereby reducing the stability of the I-A^g7/peptide complexes (44). Therefore, formation of some I-A^g7/peptide complexes may be disfavored as the result of prolonged exposure of the peptides to the protease TSSP. In addition, although the same MOG_35–55 peptide induces EAE in B6 and NOD mice, the I-A^b core peptide corresponds to MOG_38–50, whereas the I-A^g7 core peptide corresponds to MOG_42–55, thus allowing for subtle effects of TSSP (45, 46). Therefore, it was critical to determine whether lack of TSSP expression also reinforced central tolerance and conferred disease protection in B6 mice expressing the I-A^b haplotype. We found that disease onset and severity were comparable in B6 WT and B6 Tssp° mice, as were the frequency of IFN-γ–producing and IL-17–producing MOG_35–55-reactive CD4 T cells and the number of MOG-reactive CD4 T cells (Fig. 6, Supplemental Fig. 3A). Importantly, TSSP expression levels in NOD and B6 thymic DCs are comparable (Supplemental Fig. 3B). Therefore, the protective effect of TSSP deficiency is dependent on the MHC class II haplotype of the host.

The above results suggested that the level of TSSP expression by thymic DCs may determine EAE severity when combined with a given MHC class II haplotype. In humans, Prss16, the gene coding for TSSP, is associated with a diabetes-susceptibility locus (47–49). In addition, several polymorphisms within the promoter region, some exons, and a 3′ untranslated region have been detected in healthy individuals, as well as a 15-bp deletion observed in 17% of healthy individuals (47). Whether these polymorphisms modify the level or pattern of expression of TSSP is not known. Therefore, we examined the expression level of TSSP mRNA in human thymic DCs isolated from thymi of children undergoing cardiac surgery. Given that the DC1 subset of thymic DCs has a critical role in CD4 T cell deletion in the thymus (19), we isolated this subset based on BDCA3 expression (Fig. 7A). In agreement with our previous study of mouse DCs (28), we found that human DC1s express low levels of TSSP mRNA (Fig. 7B, 7C). Interestingly, we found that the samples segregated into two groups with either high or low expression of TSSP mRNA, regardless of the age or sex of the donors. The different samples showed homogeneous expression of Gapdh or ubiquitin mRNA (Fig. 7D), indicating that the observed differences indeed reflect differences in TSSP mRNA expression levels.

We show in this article that, in NOD mice, the protease TSSP impairs negative selection of the MOG_35–55-reactive CD4 T cell repertoire and consequently increases EAE severity. Indeed, we found that NOD Tssp° mice develop less severe EAE than their WT counterparts. Reduced disease severity is associated with a reduction in the frequency of MOG_35–55-specific naive CD4 T cells and a similar reduction in IFN-γ–producing and IL-17–producing MOG_35–55-specific effector CD4 T cells, suggesting that lack of TSSP expression by thymic stromal cells induces the deletion of MOG_35–55-reactive CD4 T cells but does not induce deviation from the regulatory lineage or some form of anergy. These results suggest that, in the absence of TSSP, thymic DCs, and possibly mTECs, more efficiently present MOG_35–55 peptide or a mimotope complexed with I-A^g7, thus leading to a more profound deletion of the MOG_35–55-reactive CD4 T cell repertoire. We had shown previously that, likewise, lack of TSSP expression by thymic APCs favors the deletion of several islet-reactive CD4 T cells and, therefore, leads to complete protection from autoimmune diabetes (28–30). Hence, TSSP significantly curtailed the peptide repertoire presented by I-A^g7–expressing thymic DCs and, thereby, predisposed to autoimmunity.

Using BM chimeras, we showed that lack of TSSP expression by radioresistant stromal cells induced a marked reduction in the encephalitogenic potential of polyclonal CD4 T cells to a level comparable to TSSP-deficient DCs. Whether this results from defective positive selection or enhanced negative selection of the MOG-reactive CD4 T cell repertoire is unclear. Yet, such a marked effect on a polyclonal repertoire is more likely reflecting enhanced negative selection. We previously described similar findings when analyzing diabetogenic CD4 T cells and showed that TSSP-deficient TECs induced negative selection of some islet-reactive CD4 T cells (28). Although initial studies showed that TSSP expression was confined to cTECs, more recent studies suggested that mTECs may also express low levels of TSSP (24–27). Therefore, it is possible that, in this BM chimera’s combination, TSSP-deficient mTECs induced deletion of MOG-reactive CD4 T cells.

In contrast to NOD mice, TSSP deficiency does not improve central tolerance to MOG in B6 mice. The amino acid composition of the class II–binding groove determines the physical and structural properties of the different peptide-binding pockets and, therefore, impacts the stability of the MHC/class II peptide complexes and, thereby, exposure of peptide to endosomal proteases. The specific features of the I-A^g7 class II molecules contribute to the MHC/peptide complex’s instability and permit peptide binding in a different register (44, 50, 51). Thus, the increased peptide dwell time in the endosomal compartment of NOD DCs may increase its exposure to the protease TSSP. Alternatively, TSSP may selectively “destroy” the MOG_35–55 I-A^g7 core peptide but not the I-A^b core peptide. Although the function and enzymatic characteristics of TSSP are still unclear, TSSP belongs to the family of proline-specific dipeptidyl peptidases, a family of Xaa-Pro aminopeptidase (52, 53). The I-A^g7 core peptide MOG_42–55, but not the B6 core peptide, contains a Ser⁴²-Pro⁴³ dipeptide in its N-terminal end that may be trimmed by TSSP, thereby destabilizing the MHC/peptide complex or reducing TCR avidity (54). Further characterization of the enzymatic property of TSSP is clearly required to determine whether this may be the mechanism by which TSSP impacts central tolerance to MOG_35–55 or other self-Ags. These studies will also define the repertoire of self-peptide “destroyed” by TSSP. In that regard, the HLA DRB1*15:01 allele, which confers the highest risk to MS, also presents specific features in the binding groove that lead to distinctive peptide-binding characteristics, as reported for myelin-derived peptides for instance (55–57). Interestingly, the HLA DRB1*15:01 binding epitope of myelin basic protein (MBP), MBP_85–95 (ENPVVHFFKNIVTPR), contains a proline at position 3. Whether this confers higher susceptibility to proteolytic cleavage by TSSP and, thus, alters the loading of these MBP peptides or other myelin-derived peptides on HLA DRB1*15:01 remains to be determined.

Our results suggest that the level of TSSP expression may modify the risk conferred by some MHC class II haplotypes to MS. The gene encoding TSSP, Prss16, is located in the extended class I region on chromosome 6 in humans. Polymorphisms within the Prss16 promoter and intronic regions have been described in humans, and polymorphisms in the region encoding for TSSP are associated with type 1 diabetes susceptibility (47–49). Whether these polymorphisms modify the level or pattern of TSSP expression remains unknown. Our observation that the human population segregates into two groups with high or low levels of TSSP expression in thymic DCs points to some level of polymorphism in the transcription factors, promoter, and/or signaling pathway that control TSSP expression in thymic DCs. With regard to MS, no polymorphism within Prss16 has been linked to disease susceptibility, although our study suggests a possible role for TSSP in disease severity, which was not assessed in published genome-wide association studies (1, 2, 58). Therefore, deciphering the pathways that control TSSP expression is critical to assess the contribution of this protease to MS susceptibility.

We thank R.S. Liblau for critical reading of the manuscript, M. Alloulou for advice on tetramer staining, F. Auriol and A. Garnier for help in obtaining human thymuses, F.-E. L’Faqihi-Olive, V. Duplan-Eche, and A. L. Iscache for technical assistance at the flow cytometry facility of INSERM U1043, Toulouse, the personnel of the INSERM US006 Centre Régional d'Exploration Fonctionnelle et Ressources Expérimentales animal facility for expert animal care, and the National Institutes of Health Tetramer Core Facility for providing the different tetramers.

The authors have no financial conflicts of interest.