Research ArticleMultiple Sclerosis

GDP-l-fucose synthase is a CD4+ T cell–specific autoantigen in DRB3*02:02 patients with multiple sclerosis

See allHide authors and affiliations

Science Translational Medicine  10 Oct 2018:
Vol. 10, Issue 462, eaat4301
DOI: 10.1126/scitranslmed.aat4301

Move over myelin

Although it is well established that autoreactive lymphocytes induce demyelination in multiple sclerosis, the exact antigenic targets that initiate disease are undefined. Planas et al. studied CD4+ T cells from the cerebrospinal fluid of patients with multiple sclerosis. One CD4+ T cell clone was reactive to the human enzyme GDP-l-fucose synthase; T cells from other patients were then identified, as well as myelin-reactive cells. Intriguingly, some of the GDP-l-fucose synthase–reactive cells could also be stimulated by a bacterial version of the enzyme. These tantalizing results identify a new autoantigen and suggest that one possible trigger of disease could be cross-reactivity to microbiota-derived peptides.

Abstract

Multiple sclerosis is an immune-mediated autoimmune disease of the central nervous system that develops in genetically susceptible individuals and likely requires environmental triggers. The autoantigens and molecular mimics triggering the autoimmune response in multiple sclerosis remain incompletely understood. By using a brain-infiltrating CD4+ T cell clone that is clonally expanded in multiple sclerosis brain lesions and a systematic approach for the identification of its target antigens, positional scanning peptide libraries in combination with biometrical analysis, we have identified guanosine diphosphate (GDP)–l-fucose synthase as an autoantigen that is recognized by cerebrospinal fluid–infiltrating CD4+ T cells from HLA-DRB3*–positive patients. Significant associations were found between reactivity to GDP-l-fucose synthase peptides and DRB3*02:02 expression, along with reactivity against an immunodominant myelin basic protein peptide. These results, coupled with the cross-recognition of homologous peptides from gut microbiota, suggest a possible role of this antigen as an inducer or driver of pathogenic autoimmune responses in multiple sclerosis.

INTRODUCTION

Multiple sclerosis (MS) is considered a demyelinating autoimmune disease of the central nervous system (CNS) (14). The inflammatory infiltrate in demyelinating brain lesions, the intrathecal production of oligoclonal immunoglobulin G (IgG), the genetic trait consisting of multiple immune-related genes (5, 6), the positive effect of immunotherapies targeting B and T cells, and the similarities with the animal model experimental autoimmune encephalomyelitis, all support that MS is an immune-mediated disease. The dominance of CD8+ T cells in MS brain lesions and the predisposition and protection conferred by the HLA-A*03:01 and HLA-A*0201 alleles, respectively (6, 7), as well as evidence in experimental animal models (8), hint at a relevant role of these cells in MS pathogenesis. However, the fact that the HLA-DR15 haplotype is by far the strongest genetic risk factor associated with MS and that immunization with myelin components induces relapsing or chronic demyelinating disease models, which can be transferred by myelin-specific CD4+ T cells to naive animals, underscores that autoreactive CD4+ T cells play a central role in MS pathogenesis (1). Although myelin protein/peptides are considered relevant autoantigens in MS (9), the full spectrum of target antigen(s) driving the immune response in this disease has yet to be defined.

The etiology of MS involves both a complex genetic background with more than 100 quantitative trait loci (5, 6) and several environmental risk factors including vitamin D, smoking (10), viral infections (11), and gut microbiota (12). Several studies have reported distinct fecal microbial community profiles in patients with MS compared to healthy controls (1316), which might affect autoimmune responses by regulating permeability of the blood-brain barrier (BBB), by modulating the maturation, activation, and function of immune cells (12, 1517) or by controlling the maturation and function of microglia (18) and astroglia (19). To be able to cross the BBB and infiltrate the brain, autoreactive T cells first have to be activated outside the CNS. Cross-reactivity between autoantigens and peptides from pathogens or gut microbiota, i.e., molecular mimicry, has been considered a putative trigger of the autoimmune reaction in MS (12).

The identification of autoantigens and molecular mimics targeted by pathogenic CD4+ T cells in MS may improve our understanding of disease mechanisms, help in diagnostic classification of patients with MS, and enable the development of antigen-specific immunotherapies (20). In the past, the search for candidate autoantigens in MS has concentrated on myelin components based on the fact that demyelination is a hallmark of MS brain lesions. Furthermore, immunological research in patients with MS focused on studying myelin-specific T cells from the peripheral blood because the brain and spinal cord are rarely available for immunological studies. Autoimmune diseases such as primary biliary cirrhosis can be exquisitely organ or tissue specific, although the T cell autoreactivity is directed against ubiquitous autoantigens such as pyruvate dehydrogenase (21). Thus, the focus on myelin-specific CD4+ T cells and on cells circulating in the peripheral blood has had limitations in the search for candidate autoantigens and molecular mimics in MS. Nonmyelin target antigens should also be considered in MS. To overcome these limitations, we focused on a putatively disease-relevant T cell clone (TCC) 21.1, a CD4+ TCC that was identified as clonally expanded in active pattern II demyelinating brain lesions from a patient with secondary progressive MS (SPMS) using deep T cell receptor (TCR) β chain sequencing and that was isolated from autologous cerebrospinal fluid (CSF)–infiltrating cells as previously described (22). The release of T helper 2 (TH2) cytokines and its ability to help B cells (22) support a pathogenic role of this TCC in pattern II demyelination that is mediated by antibodies and complement (23). Using positional scanning peptide libraries (24, 25) and biometrical analysis (26, 27) that allow the systematic identification of stimulatory peptides, we have identified guanosine diphosphate (GDP)–l-fucose synthase as the main specificity of TCC21.1. Reactivity against this protein and against myelin and potential molecular mimics was then examined more broadly in CSF-infiltrating CD4+ T cells from patients with MS and led to the identification of a putative target autoantigen in MS.

RESULTS

The use of peptide mixtures to identify the specificity of a brain-infiltrating TCC

TCC21.1 is a CD4+ TCC isolated from CSF-infiltrating cells and clonally expanded into two active white matter demyelinating MS lesions from a patient with SPMS (1154SA) with pattern II demyelination (22). TCC21.1 releases TH2 cytokines and helps proliferation and antibody production by autologous B cells upon unspecific activation. TCRα and TCRβ chain sequences of TCC21.1 are TRAV12-2, TRAJ42*01, CDR3α CAVNEGSQGNLIF and TRBV11-2*01, TRBJ1-2*01, TRBD1*01, and CDR3β CASSGRGPSYGYT. To study the peptides recognized by this TCC of unknown specificity, we first identified the human leukocyte antigen (HLA) class II engaged by TCC21.1 (28). Because patient 1154SA was homozygous for the DR15 haplotype, we tested TCC21.1 initially only with the mixtures with amino acids defined at position 5, presented by Epstein-Barr virus (EBV)–immortalized bare lymphocyte syndrome (BLS) B cell lines (BLS-BCLs) transfected with different single autologous HLA-DR/DQ molecules (DRA*01:01/DRB5*01:01, DRA*01:01/DRB1*15:01, and DQA1*01:02/DQB1*06:02). As readout of T cell activation, we used granulocyte-macrophage colony-stimulating factor (GM-CSF) production (29), since TCC21.1 releases this cytokine after polyclonal stimulation with anti-CD3 and phorbol 12-myristate 13-acetate (PMA) (22). Only mixtures presented by BLS-BCL expressing DRB1*15:01 class II molecules were stimulatory (Fig. 1A). We then tested TCC21.1 with the complete decapeptide positional scanning library (200 mixtures) presented by BLS-BCL expressing DRB1*15:01 class II molecules. GM-CSF release in response to the complete library is shown in Fig. 1B (black histograms). Next, we generated a biometrical analysis scoring matrix by assigning numerical values to the stimulatory potential of each of the 20 defined amino acids in each of the 10 positions of the decapeptide library as previously described (26). Here, the values were calculated as the log base 10 of the median GM-CSF secretion of three independent experiments in the presence of mixtures, minus the secretion in the absence of mixtures (Fig. 1C). On the basis of a model of independent contribution of individual amino acids to peptide antigen recognition, the predicted stimulatory score of a given peptide is the sum of the matrix values of each amino acid contained in the peptide at each position (26). This scoring approach was applied to rank all natural overlapping 10-mer peptides of the protein sequences within the UniProt human protein database (30). On the basis of these predicted values, we synthesized and tested the 50 predicted human natural peptides with highest scores (table S1). Unexpectedly, none of the peptides was clearly stimulatory (Fig. 1D), demonstrating that the predictive capacity of the above approach was lower for this TCC than for others in previous studies (27).

Fig. 1 Single- and dual-defined decapeptide positional scanning mixtures in combination with biometric analysis to identify the specifity of TCC21.1.

(A) GM-CSF production by TCC21.1 in response to 20 combinatorial peptide mixtures with amino acids defined at position 5 and presented by BLS-BCL expressing only DRB1*15:01, DRB5*01:01, or DQB1*06:02 class II molecules. x axis, single-letter amino acid code; y axis, cytokine production. (B) GM-CSF (black histograms) and interleukin-10 (IL-10; white histograms) production by TCC21.1 in response to a complete decapeptide positional scanning library (200 mixtures) presented by BLS-BCL expressing only DRB1*15:01. (C) Score matrix designed with the log10 median of GM-CSF production of three independent experiments. Dark gray cells correspond to values ≥2, and light gray cells correspond to values between 1.5 and 2. Bold borders show mixtures selected for dual-defined mixtures, and continuous and dotted borders were based on GM-CSF and IL-10 results, respectively. (D) GM-CSF production by TCC21.1 in response to the 50 peptides with highest scores predicted using the GM-CSF–based score matrix. (E) GM-CSF production by TCC21.1 in response to 22 dual-defined mixtures. In green, mixtures with defined amino acids of the TCR motif in frame 1 and in blue in frame 2. Stimulatory responses of mixtures shown with darker colors (frame 1/2-HM) were integrated into the original matrix using the HM model. (F) TCR motif and dual-defined mixture activity values selected on the basis of the HM model and incorporated into the original matrix, frame 1-HM (green), and frame 2-HM (blue). (G) GM-CSF production by TCC21.1 in response to the 50 peptides with higher scores predicted using harmonic boost frame 1 and frame 2 score matrices. Complete decapeptide library, dual-defined mixtures, and individual decapeptides were presented by BLS-BCL expressing DRB1*15:01. Mixtures and individual decapeptides were tested at 200 and 5 μg/ml, respectively, for 72 hours. Histograms show means ± SEM and dot plots mean of three independent experiments. Cytokine released is expressed as pg/ml released by TCC21.1 in response to stimuli minus pg/ml released in the absence of stimulus (negative control).

To further explore this unusual discrepancy between the mixture activities and the lower than expected activity of the peptides described above, we first noted that TCC21.1 also released high amounts of IL-10 after polyclonal stimulation with anti-CD3 and PMA (22). TCC21.1 was then tested with the complete decapeptide positional scanning library using IL-10 production as readout (Fig. 1B, white histograms). A similar but not identical response pattern was obtained with both readouts. The most stimulatory mixtures resulting in GM-CSF or IL-10 production are shown in Fig. 1C. These mixtures have identical defined amino acids in consecutive positions, suggesting a possible recognition of one amino acid motif in multiple frames. To confirm this observation, a set of 22 dual-defined mixtures (positional scanning mixtures with two defined positions) was designed, synthesized, and tested (Fig. 1E); the results confirmed the presence of a unique recognition motif (LHSXFEV) with different flanking residues (Fig. 1F). To apply this recognition motif in the two frames to the original GM-CSF–derived matrix and perform biometrical analysis, we used the harmonic mean (HM) model (31) to integrate the stimulatory responses of some of the dual defined mixtures (Fig. 1E, frame 1/2-HM mixtures) into the original matrix (fig. S1). Using the new “harmonic boost” frame 1 and frame 2 matrices, all natural overlapping 10-mer peptides in the protein sequences within the UniProt human database were scored and ranked as before. The 50 predicted natural peptides with highest scores for both matrices were synthesized and tested for GM-CSF release (table S1). These two matrices allowed us the identification of three clearly stimulatory peptides (Fig. 1G). The two most stimulatory peptides, NVLHSAFEVG and DNVLHSAFEV predicted with harmonic boost frame 1 and frame 2, respectively, belong to GDP-l-fucose synthase encoded by the TSTA3 gene and overlap by nine amino acids. The third stimulatory peptide (KLLLHSGVEN) was predicted with harmonic boost frame 2 and belongs to a transmembrane protein (FLJ37396).

GDP-l-fucose synthase as the main autoantigen for a brain-infiltrating TCC

To identify autoantigens recognized by TCC21.1 in the two autologous brain lesions in which it was clonally expanded (22), the harmonic boost frame 1 and frame 2 matrices were then used to score and rank, according to their stimulatory score, all natural overlapping 10-mer peptides in the protein sequences within a brain-protein subdatabase created with RNA sequencing–based transcriptome data from these two lesions (GSE60943). Of the 40 predicted natural brain peptides with highest scores for the frame 1 matrix, 38 peptides were already predicted from the unbiased UniProt human database. For the frame 2 matrix, the top 40 peptides were previously predicted (table S1). The two new peptides for frame 1 were synthesized and tested. NVLHSAFEVG [GDP-l-fucose synthase(96–105)] and DNVLHSAFEV [GDP-l-fucose synthase(95–104)] were the only peptides from proteins transcribed in autologous brain lesions found to stimulate TCC21.1 (table S1).

Transcript abundance, expressed as reads per kilobase of exon model per million mapped reads, of GDP-l-fucose synthase in the two autologous brain lesions in which TCC21.1 was clonally expanded [LI and LIII; (22)] is shown in Table 1. Transcript values for other brain-specific genes are also shown as quality control of the samples and as reference for genes expressed at high [myelin basic protein (MBP) and myelin proteolipid protein 1 (PLP1)] and medium [myelin-associated glycoprotein (MAG), myelin-associated oligodendrocyte basic protein (MOBP), and oligodendrocyte myelin glycoprotein (OMG)] abundance.

Table 1 GDP-l-fucose synthase transcripts and peptides identified in brain tissue.

RPKM, reads per kilobase of exon model per million mapped reads; PSMs, peptide spectrum matches. Proteome data correspond to white matter (WM) and gray matter (GM) from patients with MS (MS; n = 15) and controls without MS (non-MS; n = 9). In bold are GDP-l-fucose synthase values.

View this table:

Seventeen GDP-l-fucose synthase peptides were identified in white and gray matter brain tissue from other patients with MS and controls without MS by proteomic analysis. The peptide sequences are listed in Table 1, and the peptide spectrum matches in the different samples. The percentage of the GDP-l-fucose synthase amino acid sequence covered by identified peptides was 56%. Analysis of other brain-specific proteins as reference is also included (Table 1 and table S3).

Characterization of GDP-l-fucose synthase recognition by a brain-infiltrating TCC

As mentioned above, the two GDP-l-fucose synthase peptides recognized by TCC21.1 overlap by nine amino acids. We synthesized and tested on DRB1*15:01-expressing BLS-BCL nine additional 10-mer peptides overlapping by nine amino acids and identified an additional peptide, VLHSAFEVGA(97–106), which induced GM-CSF release by TCC21.1 (Fig. 2A). Peptide NVLHSAFEVG(96–105) was the most stimulatory with a median effective concentration of 0.2 μg/ml (Fig. 2B) and could be presented by DRB1*15:01 and DQB1*0602 molecules (Fig. 2C). As we demonstrated previously, the cross-reactivity and cross-restriction of TCC21.1 might facilitate its activation (32, 33). The common amino acids of the three stimulatory GDP-l-fucose synthase peptides are VLHSAFEV(97–104). Single-alanine (A) substitutions of these three overlapping peptides revealed that the substitution of H99 and S100 abrogated the T cell responses, suggesting that they are the primary amino acid involved in TCR or HLA interaction. The substitution of V97, L98, F102, or E103 for A in the most active peptide NVLHSAFEVG(96–105) resulted in peptides with some stimulatory activity, in contrast to the two other overlapping peptides in which their substitution resulted in nonstimulatory peptides. Thus, the residues V97, L98, F102, and E103 appear to be secondary TCR or HLA contacts (Fig. 2D).

Fig. 2 Characterization of TCC21.1 response to GDP-l-fucose synthase peptides.

(A) GM-CSF production by TCC21.1 in response to 11 decapeptides overlapping by nine amino acids presented by BLS-BCL expressing DRA*01:01/DRB1*15:01 and tested at 0.5 and 5 μg/ml. (B and C) GM-CSF production by TCC21.1 in response to GDP-l-fucose synthase peptide (96–105) presented by BLS-BCL expressing DRA*01:01/DRB1*15:01, DRA*01:01/DRB5*01:01, or DQA1*01:02/DQB1*06:02 class II molecules. (D) GM-CSF production by TCC21.1 in response to 13 A scan peptides presented by BLS-BCL expressing DRA*01:01/DRB1*15:01. Alanine substitution is shown in gray. (E and F) Proliferative responses (white bars) and production of cytokines (black bars) by TCC21.1 in response to GDP-l-fucose synthase peptides presented by autologous peripheral blood mononuclear cells (PBMCs) or autologous BCL and nonspecific stimulus (anti-CD3 and PMA; anti-CD3 and anti-CD28; and anti-CD3, anti-CD28, and anti-CD2 beads). (F) Proliferative responses (white bars) and GM-CSF and IL-3 production (black bars) by TCC21.1 and other TCCs from other patients with MS with different functional phenotypes in response to the corresponding specific peptides presented by autologous PBMCs (for TCC21.1, also autologous BCL) and to nonspecific stimuli (anti-CD3 and PMA; anti-CD3 and anti-CD28; and anti-CD3, anti-CD28, and anti-CD2 beads for TCC21.1; and only anti-CD3, anti-CD28, and anti-CD2 beads for the other TCCs). Cytokine release is expressed as pg/ml released by the TCC in response to stimuli minus pg/ml released in the absence of stimulus (negative control). All results show the means ± SEM. Dotted lines show a putative threshold for positivity: stimulatory index (SI) = 2 in proliferation histograms and the amount of interferon-γ (IFN-γ), IL-4, IL-17, IL-9, and IL-22 released by T cells with TH1, TH2, TH17, TH9, or TH22 functional phenotype, respectively. (G) Representative flow cytometry analysis of intracellular IL-4, IFN-γ, GM-CSF, and IL-3 production by TCC21.1 in response to GDP-l-fucose synthase peptide (96–105) and PMA/ionomycin (Io). Numbers represent the percentage of positive cells. (H) Representative flow cytometry analysis of the mean fluorescence intensity of CD28, CCR4, CCR6, and CRTH2 on the surface of resting TCC21.1 (black lines) in comparison to the isotype control staining (gray shaded). Flow cytometry analysis of TCC21.1 was performed three times after independent expansions.

Next, we characterized the response of TCC21.1 to GDP-l-fucose synthase peptides presented by autologous irradiated PBMCs and an EBV-transformed autologous BCL. GDP-l-fucose synthase peptides presented by the two types of antigen-presenting cells (APCs) were able to induce proliferation, although PBMCs were more efficient (Fig. 2E). We also analyzed the functional phenotype of this response (Fig. 2E). TCC21.1 displayed a TH2 phenotype, releasing mainly TH2 cytokines and lower amounts of IL-22 and IL-10. Unexpectedly, when peptides were presented by autologous BCL instead of autologous PBMCs, they induced higher levels of IFN-γ. In addition, TCC21.1 also released GM-CSF and IL-3 in response to GDP-l-fucose synthase peptides (Fig. 2F). Intracellular cytokine staining confirmed the TH2 functional phenotype of TCC21.1 in response to GDP-l-fucose synthase(96–105) (Fig. 2G). More than 60% of TCC21.1 cells were IL-4+ after stimulation with GDP-l-fucose synthase(96–105), whereas only about 12% were IFN-γ+. About 60% of the cells were GM-CSF+, and 47.7% of these were also IL-4+. Further characterization of TCC21.1 demonstrated expression of CD28 and the chemokine receptor CRTH2 (Fig. 2H).

Recognition of GDP-l-fucose synthase by CSF-infiltrating CD4+ T cells from patient 1154SA

Sixty-two 15-mer peptides overlapping by 10 amino acids and covering the entire GDP-l-fucose synthase protein (table S2) were synthesized and tested for their ability to induce TCC21.1 proliferation when presented by autologous PBMCs. Seven immunodominant/encephalitogenic myelin peptides (34), CEF [(cytomegalovirus, EBV, and influenza virus) and tetanus toxoid] peptide pool, and control beads were tested in parallel (Fig. 3A and table S2). TCC21.1 only recognized two overlapping GDP-l-fucose synthase peptides, (91–105) and (96–110), containing the three stimulatory decapeptides [(95–104), (96–105), and (97–106)] that we identified previously. Bulk CSF-infiltrating CD4+ T cells from the same patient after short and long phytohemagglutinin (PHA) expansion (see Materials and Methods) were tested with the GDP-l-fucose synthase, myelin, and CEF peptides (Fig. 3A). Long, but not short, PHA–expanded CSF-infiltrating CD4+ T cells proliferated in response to three GDP-l-fucose synthase peptides, suggesting a low frequency of specific TCCs in the initial CSF infiltrate and a preferential expansion with enrichment of some TCCs, including TCC21.1, during the long PHA expansion. Proliferation of long PHA–expanded CSF-infiltrating CD4+ T cells to peptides (91–105) and (96–110) was mainly mediated by TCC21.1 because Vβ21+ (that included TCC21.1), but not Vβ21, CSF-infiltrating CD4+ T cells proliferated in response to this peptide (Fig. 3B). The third GDP-l-fucose synthase peptide recognized by long PHA–expanded CSF-infiltrating CD4+ T cells (191–205) was not recognized by TCC21.1 (Fig. 3A). We generated a T cell line (TCL) from these proliferating cells (TCL-39; Fig. 3C), and its TCR sequencing confirmed the existence of three TCCs other than TCC21.1. Although we have not been able to identify the individual TCC responsible for GDP-l-fucose-specific response, note that two of these TCCs were clonally expanded in active brain lesion III and ranked in positions 51 and 56 regarding their frequency [Fig. 3D and (22)]. Functional phenotype analysis of responses to peptides (91–105) and (191–205) showed a TH2 response for peptide (91–105), confirming the activation of TCC21.1, but a TH1/TH2 response for peptide (191–205) supporting activation of other TCCs (Fig. 3E). Bulk CSF-infiltrating CD4+ T cells did not recognize either the myelin peptides or the CEF peptide pool.

Fig. 3 Recognition of GDP-l-fucose synthase peptides by TCC21.1 and CSF-infiltrating CD4+ T cells from patient 1154SA.

(A) Proliferative responses expressed as SI of TCC21.1 and CSF-infiltrating CD4+ T cells after short and long PHA expansion to GDP-l-fucose synthase, myelin, or CEF peptides presented by autologous PBMCs, as well as to control anti-CD3, anti-CD28, and anti-CD2 beads. TCC21.1 graph shows the mean SI ± SEM, whereas, in the CSF graphs, each dot represents one well. Positive peptides are shown in black. (B) Proliferation of long-expanded CSF-infiltrating CD4+ T cells in response to GDP-l-fucose synthase peptide (91–105), CEF pool, or PHA using BLS-BCL DRA*01:01/DRB1*15:01 as APCs and assessed by 5-ethynyl-2′-deoxyuridine (EdU) incorporation. Numbers represent the percentage of EdU+ cells. (C) Proliferative responses expressed as stimulation SIs of positive wells to GDP-l-fucose synthase peptide (191–205) for generation of TCL-39. (D) TCRBV sequencing of TCCs present in TCL-39. (E) Production of cytokines of long expanded CSF-infiltrating CD4+ T cells to GDP-l-fucose synthase peptides (91–105) and (191–205) presented by autologous PBMCs and to control beads. Cytokine release is expressed as pg/ml released by stimulated cells minus pg/ml released in the absence of stimulus (negative control). Results show the means ± SEM.

Recognition of GDP-l-fucose synthase by CSF-infiltrating CD4+ T cells from patients with clinically isolated syndrome/MS

To find out how frequently specific recognition of GDP-l-fucose synthase occurs in CSF-infiltrating CD4+ T cells in patients with MS, we developed a new protocol to expand fresh CSF-infiltrating CD4+ T cells to high numbers in a single round to minimize deviations from the original T cell repertoire (see Materials and Methods). Eight patients with clinically isolated syndrome (CIS) with a CNS demyelinating event isolated in time and presence of oligoclonal bands in the CSF compatible with the future development of MS and 23 patients with MS were analyzed. Resting PHA-expanded CSF-infiltrating CD4+ T cells from these 31 patients with CIS/MS were tested in quadruplicate with the 62 overlapping GDP-l-fucose synthase peptides presented by autologous-irradiated PBMCs, as well as with the seven myelin peptides, CEF peptide pool, and control beads. All SIs (except control beads) were pooled, and SI values less than one were treated as having unit value. Cluster k-means analysis was performed to determine the optimal cutoff to differentiate responsive SI values from nonresponsive in this patient population. K-means clustering resulted in a cutoff value of 1.455 to differentiate positive from negative responses. A follow-up study of the negative control well SI values using Monte Carlo bootstrapping indicated that the 1.455 cutoff showed a 0.26% false-positive rate when applied to simulated medians of quadruplicate negative values (table S4). Subsequently, for each patient, all peptides having a median SI (of the quadruplicate wells) greater than 1.455 were identified as positive responses (table S5A).

Next, patient scores were constructed by calculating the sum of each responsive peptide median SI, weighted by the number of total patients that responded to that peptide. In this way, both the SI values themselves for each peptide and the relative immunogenicity of each peptide were factored in. Three-cluster k-means analysis was performed on log base 10 patient scores and clearly grouped patients into three categories: “nonresponders,” “moderate responders,” and “high responders” (table S5A). Nineteen patients (61.3%) were therefore characterized as nonresponders to GDP-l-fucose synthase (fig. S2A), six (19.35%) as moderate, and six (19.35%) as high responders (Fig. 4A). No significant differences between the three groups were found for CEF responses or for the positive or negative controls (fig. S2B). Immunodominant peptides were defined as peptides able to induce positive responses in at least three patients. Fourteen GDP-l-fucose synthase peptides achieving this criterion are shown in Fig. 4B. Functional analysis of the strongest responses to some immunodominant peptides revealed a TH1 phenotype with mainly production of IFN-γ (Fig. 4C).

Fig. 4 Recognition of GDP-l-fucose synthase peptides by CSF-infiltrating CD4+ T cells from moderate- and high-responder patients with CIS/MS.

(A) Proliferative responses expressed as SIs of CSF-infiltrating CD4+ T cells (single round of PHA expansion) to GDP-l-fucose synthase or CEF peptides presented by autologous PBMCs, as well as to control anti-CD3, anti-CD28, and anti-CD2 beads. Each dot represents one well, and median SIs are shown as solid lines. Positive peptides (black dots and red line) are peptides with median of all SIs > 1.455 (dotted red line). All peptides have been tested in four wells, except for peptides (156–170) and/or (161-175) in 748UR, 1129RE, 1005ME, 1173DI, 718MA, and 776JE, which were tested also for bacterial peptides by seeding eight wells. (B) Number of patients with CIS/MS with CSF-infiltrating CD4+ T cells that responded to GDP-l-fucose synthase peptides. Immunodominant peptides that were positive in at least three patients are shown in red. Solid red indicates high responder, and light red indicates moderate responder patients. (C) Production of cytokines of CSF-infiltrating CD4+ T cells from three high-responder patients to GDP-l-fucose synthase immunodominant peptides presented by autologous PBMCs.

Association of T cell responses to immunodominant GDP-l-fucose synthase peptides with reactivity against MBP(83–99)

To judge better the reactivity of CSF-infiltrating CD4+ T cells in patients with CIS/MS against GDP-l-fucose synthase peptides, we compared it with reactivity against the seven immunodominant/encephalitogenic myelin peptides, which had previously been shown to be the targets of high-avidity T cells in MS (Fig. 5A) (34). Four of 6 (67%) of the high-responder patients to GDP-l-fucose synthase also responded to myelin peptides, whereas only 1 of 6 (17%) moderate responders and 2 of 19 (11%) nonresponders did (Fig. 5A). The association between GDP-l-fucose synthase and myelin response is significant (P = 0.0138, Fisher’s exact test). When comparing the recognition of immunodominant GDP-l-fucose synthase peptides and specific myelin peptides (Fig. 5B), we found a significant association between recognition of peptide MBP(83–99) and three immunodominant GDP-l-fucose synthase peptides [(115); P = 0.0122; 156–170, P = 0.0122; and 256–270, P = 0.0481; Fisher’s exact tests with Bonferroni-Holm correction; Fig. 5B].

Fig. 5 Recognition of myelin peptides by CSF-infiltrating CD4+ T cells from patients with CIS/MS.

(A) Proliferative responses expressed as SIs of CSF-infiltrating CD4+ T cells (single round of PHA expansion) to myelin peptides presented by autologous PBMCs. Each dot represents one well, and median SIs are shown as solid lines. Positive peptides (black wells and red line) are peptides with median of all SIs > 1.455 (dotted red line). (B) Checkerboard graph illustrating GDP-l-fucose synthase immunodominant (red) and myelin (blue) peptides recognized by each patient with CIS/MS. Solid colors are GDP-l-fucose synthase and myelin peptides. The number of patients that responded (Y, yes) and that did not respond (N, not) to the peptides and the P values for Fisher’s exact tests with Bonferroni-Holm correction are shown.

Association of T cell response to GDP-l-fucose synthase with DRB3* alleles

Table 2 summarizes demographic and clinical characteristics of patients with CIS/MS and their designation as nonresponders, moderate responders, or high responders to GDP-l-fucose synthase peptides. No significant differences between groups could be identified regarding disease course or clinical presentation, gender, age, CSF IgG index, or number of CSF-infiltrating cells. HLA class II typing is also summarized in Table 2. All 6 of the high-responder patients are DRB3*02:02, whereas only 1 of 6 (17%) moderate responders and 4 of 19 (21%) nonresponders express this HLA class II molecule. This association is significant (P = 0.0006, Fisher’s exact test). Furthermore, when looking only at the difference between moderate responders and nonresponders, an additional significant association was observed: 3 of 6 (50%) moderate responders are DRB3*01:01, whereas only 1 of 19 (5%) nonresponders is DRB3*01:01 (P = 0.0312, Fisher’s exact test). The predicted binding affinities of positive peptides to DRB3*01:01 and DRB3*02:02 are summarized in table S5B.

Table 2

Demographic and clinical characteristics of patients with CIS/MS and HLA-DR/DQ typing. RRMS, relapsing remitting MS.

View this table:

Cross-recognition of human and bacterial GDP-l-fucose synthase peptides by CD4+ CSF-infiltrating T cells

GDP-l-fucose synthase is a cytosolic enzyme that converts GDP-4-keto-6-deoxy-d-mannose into GDP-l-fucose that is then used to fucosylate oligosaccharides (35) including sugars and glycoproteins on mucosal intestinal cells. Gut microbiota can use a mammalian-like pathway of fucosylation to colonize the mammalian intestine (36). GDP-l-fucose synthase is an evolutionarily conserved protein (37). To search for molecular mimics that might induce or perpetuate GDP-l-fucose synthase–specific T cells, we therefore compared the protein sequence of human GDP-l-fucose synthase with several bacterial GDP-l-fucose synthase sequences. We focused on bacterial species that have been reported to be altered in gut microbiota of patients with MS and for which GDP-l-fucose synthase sequences were available. We also included bacterial species present in gut microbiota without association with MS and also some pathogens not present in the gut (table S6). Sequence comparison of bacterial peptides to the four main immunodominant peptides identified above is shown in table S7A. The most immunodominant human GDP-l-fucose synthase peptide (161–175) showed the highest similarity with bacterial sequences, and particularly with those associated with MS, with which it shares at least 40% identity. Eight 15-mer bacterial peptides with at least 40% identity with the human (161–175) peptide and shared by different bacteria species were synthesized as 15- and 10-mer (overlapping in five amino acids) peptides (table S7B). Reactivity of CSF-infiltrating CD4+ T cells against these bacterial peptides was tested in eight patients with CIS/MS. Six of these patients were patients from our previous cohort from whom autologous PBMCs were available (748UR, 1005ME, 1129RE, 1173DI, 718MA, and 776JE). Five of them (748UR, 1129RE, 1173DI, 718MA, and 776JE) had a positive response to human peptide (161–175), whereas patient 1005ME did not respond to this peptide and was included as control (Fig. 4A). Two additional patients (1440AM and 1489HE), from whom autologous PBMCs were not sufficient to test all GDP-l-fucose synthase peptides and who were not included in the previous cohort, were also identified as high responders to GDP-l-fucose synthase peptides, including peptide (161–175), after testing with a selection of six immunodominant peptides (fig. S3A). Both patients were also DRB3* (fig. S3B). 1489HE was DRB3*02:02- and 1440AM-expressed DRB3*03:01 with high similarity to DRB3*02:02 (38). CD4+ CSF-infiltrating T cells from four of these patients (1489HE, 1173DI, 748UR, and 776JE) proliferated in response to different bacterial peptides (Fig. 6). Four patients showed reactivity against peptides 3 and 4, which show the highest degree of similarity (53.3%) with the human peptide (Fig. 6). National Center for Biotechnology Information (NCBI) Basic Local Alignment Search Tool analysis of these bacterial peptides (fig. S3C) revealed that peptide 3 was shared by bacteria from the genera Akkermansia and Prevotella, which both have been reported to be altered in patients with MS (14). Peptide 4 is also shared by different bacterial genera present in gut microbiota and belonging to the phylum Bacteroidetes, the second most abundant phylum, after Firmicutes, in gut microbiota.

Fig. 6 Recognition of bacterial GDP-l-fucose synthase peptides by CSF-infiltrating CD4+ T cells from patients with CIS/MS.

Proliferative responses expressed as SIs of CSF-infiltrating CD4+ T cells (single round of PHA expansion) to eight bacterial GDP-l-fucose synthase peptides, sharing with the human homologous peptide at least 6 of the 15 amino acids (blue), presented by autologous PBMCs. Each dot represents one well, and median SIs are shown as solid lines. Positive peptides (black wells and red line) are peptides with median of all SIs > 2 (dotted red line).

DISCUSSION

Assuming that MS is an immune-mediated autoimmune disease, it is crucial to elucidate the target antigens responsible for inducing T and B cell activation to understand MS pathogenesis. In addition, it is a prerequisite for developing antigen-specific tolerization strategies. Using positional scanning peptide libraries in combination with biometric analysis, we have identified GDP-l-fucose synthase as the main specificity of TCC21.1, a brain-infiltrating and clonally expanded CD4+ TCC from a patient with SPMS with pattern II demyelination (22). In pattern II lesions, demyelination is mediated by deposits of antibody and complement in addition to macrophages and T cells (23). In addition, to be clonally expanded in active MS lesions, TCC21.1 was able to release TH2 cytokines and help B cells after nonspecific activation, characteristics that strongly support its putative pathogenic role in pattern II demyelination. Here, we confirmed a clear TH2 functional phenotype of TCC21.1 in response to GDP-l-fucose synthase peptides presented by autologous PBMCs and the expression of the chemokine receptor CRTH2, a marker of human TH2 cells. Furthermore, in response to GDP-l-fucose synthase peptides, TCC21.1 also released GM-CSF, which is considered relevant in MS pathogenesis (39, 40), and IL-3, reported to promote TH2 responses (41, 42). Reactivity against GDP-l-fucose synthase peptides was also identified in bulk CSF-infiltrating CD4+ T cells from the same patient after long, but not short, PHA expansion. TCC21.1 and at least one other GDP-l-fucose synthase–specific CD4+ TCC were responsible for this response, further supporting a putative role of GDP-l-fucose synthase peptides as autoantigens. The inability to detect GDP-l-fucose synthase–specific CSF-infiltrating CD4+ T cells after short PHA expansion suggests a low frequency of these cells in the CSF. The fact that TCC21.1 and most likely also the TCC specific for peptide (191–205) were clonally expanded in the most active lesion III (22) suggests that these TCCs might be retained in the brain parenchyma.

The identification of TCC21.1 specificity using positional scanning libraries was less efficient than in previous studies (25, 27) because of recognition by this TCC of peptides with identical binding/contact motif but variable flanking C- and N-terminal amino acid. Major histocompatibility complex class II molecules have open binding grooves and are able to present peptides with a highly variable length, rendering TCR binding very flexible (43). However, at present, it is unclear why the recognition of a TCR motif in different frames influenced the results for TCC21.1 more so than other TCCs. The number of stimulatory peptides identified for TCC21.1 was unusually low compared with previous studies (25, 27). The limited cross-reactivity of TCC21.1 might underlie the special features of this TCC recognition profile of the decapeptide positional scanning library.

Analysis of bulk CSF-infiltrating CD4+ T cells from other patients with CIS/MS after a single PHA expansion demonstrated reactivity against GDP-l-fucose synthase peptides in 40% of patients. Fourteen immunodominant peptides that somewhat unexpectedly induced a TH1 response, excluding a firm association between GDP-l-fucose synthase reactivity and pattern II demyelination, were identified. Although processing and presentation of the native GDP-l-fucose synthase have not been demonstrated, interestingly, a significant association between recognition of GDP-l-fucose synthase and myelin has been found between the immunodominant GDP-l-fucose synthase peptides (115), (156–170), and (256–270) and the myelin peptide MBP(83–99), and all patients reacting to both GDP-l-fucose synthase peptides and MBP(83–99) were high responders. The putative cross-recognition of these peptides by CSF-infiltrating CD4+ T cells might play a relevant role in MS pathogenesis because CD4+ T cells specific for MBP(83–99) are able not only to induce a demyelinating disease similar to MS in experimental animal models but also to exacerbate the human disease (4448). However, whether single TCCs are able to cross-recognize MBP(83–99) and GDP-l-fucose synthase peptides, whether the recognition of one of these peptides facilitates the release of the other and the novo activation of new autoreactive T cells, a phenomenon known as epitope spreading (49), or whether these responses codevelop from as yet unknown reasons are still open questions.

HLA typing of patients with CIS/MS demonstrated reactivity against GDP-l-fucose synthase in DRB3*-positive patients and particularly strong reactivity in those expressing DRB3*02:02. The HLA-DRB1*03 and the DRB3* haplotype most likely appeared via gene duplication of a common DRB1* gene in that locus, and in consequence, DRB3* has been considered a secondary redundant locus (50). However, the ability of DRB3* molecules to present a distinct pattern of peptides and the existence of DRB3*-restricted T cells in peripheral blood have been demonstrated (51). Here, we found that the GDP-l-fucose synthase peptide (5165), the only peptide stimulatory in all high-responder DRB3*02:02 patients, showed the highest predicted binding affinity to this class II molecule, suggesting DRB3*02:02-restricted T cell recognition of this peptide. The DRB3*02:02 allele has been associated with MS in Korean children (52) and also with neuromyelitis optica (NMO) (53). DRB3*02:02 was overrepresented in a cohort of patients with NMO, in whom aquaporin 4–specific T cells cross-recognized a peptide from Clostridium perfringens, a normal component of gut microbiota (54).

The cytosolic enzyme GDP-l-fucose synthase converts GDP-4-keto-6-deoxy-d-mannose into GDP-l-fucose, which is then used by fucosyltransferases to fucosylate all oligosaccharides (35). Synthesis of GDP-l-fucose from free l-fucose represents a minor route. In mammals, fucosylated glycans play important roles in many biological processes including blood transfusion reactions, host-microbe interactions, and cancer pathogenesis (55, 56). Fucosylated glycans are highly expressed in brain tissue, and quantitative glycan profiling of human CNS myelin demonstrated an unusually rich variety of Lewis-type antigen containing fucosylated glycans compared with other tissues (57). Brain-fucosylated glycans have been implicated in the molecular mechanisms that underlie neuronal development, survival, and function (55, 5860), as well as in modulating immune responses. The encephalitogenic MOG (myelin-oligodendrocyte glycoprotein) is decorated with fucosylated glycans that support recognition by dendritic cell (DC)–SIGN on microglia and DCs. This recognition results in a tolerogenic signal characterized by IL-10 secretion and decreased T cell proliferation, whereas reduced fucosylation results in immunogenic signals (57). Reduction of brain-fucosylated glycans as a consequence of T cell reactivity against GDP-l-fucose synthase might directly affect neurons and facilitate MS development by inducing a proinflammatory immune response. Fucosylated glycans are highly expressed also in gut tissue, where they play a crucial role in the host-microbe interaction (61). Fucosylation is less common in prokaryotic organisms, and it seems to be involved in adhesion, colonization, and regulation of the host immune response. In addition, a putative role of fucosylated glycans as mechanism to mimic the host and evade the immune system has been suggested in these organisms (56). Such a mechanism might be very beneficial for pathogens colonizing highly fucosylated tissues such as the gut. In this context, the putative cross-reactivity of human and gut microbial GDP-l-fucose synthase peptides may represent a new role of gut microbiota in MS pathogenesis. Supporting this hypothesis, we demonstrated that CSF-infiltrating CD4+ T cells from four patients with CIS/MS that were reactive against the human GDP-l-fucose synthase peptide (161–175) also recognized two gut microbial GDP-l-fucose synthase peptides from bacterial genera that have been associated with MS (1316, 6265) and that, in addition, showed the highest similarity with the human peptide. However, because demonstration of cross-reactivity at single TCC level is missing, the putative involvement of gut microbiota via this mechanism in MS pathogenesis is still an open question. Recently, molecular mimicry between a specific β-cell autoantigen and a Bacteroides integrase peptide has been demonstrated in mice. Diabetogenic CD8+ T cells specific of the pancreatic peptide were able to cross-recognize the microbiota-derived peptide and suppress colitis (66).

In conclusion, using positional scanning peptide libraries and biometrical analysis, we identified GDP-l-fucose synthase peptides as the main specificity of a brain-infiltrating clonally expanded TCC most likely involved in MS pathogenesis. Peptides from this putative autoantigen are widely recognized by CSF-infiltrating CD4+ T cells from DRB3*-positive patients with MS. Although we do not demonstrate processing and presentation of the native protein, the significant correlation between myelin and GDP-l-fucose synthase reactivity in DRB3*02:02 patients, as well as the putative cross-recognition of homologous peptides from gut microbiota, supports a role of this antigen as potential inducer or driver of relevant CD4+ T cell immune responses in MS. Further studies are needed to confirm the existence of DRB3*02:02-restricted TCCs able to cross-recognize myelin, human, and bacterial GDP-l-fucose synthase peptides that might provide the basis for a role of molecular mimicry in the development of MS.

MATERIALS AND METHODS

Study design

This study was designed to identify autoantigens and molecular mimics targeted by pathogenic T cells in MS. Initially, we focused on a clonally expanded brain-infiltrating CD4+ TCC identified by high-throughput TCRβ chain sequencing in active brain lesions from a patient with SPMS and isolated from the autologous CSF. Using positional scanning peptide libraries in combination with biometrical analysis to search for stimulatory peptides in a systematic way, we identified GDP-l-fucose synthase. The response of the TCC to stimulatory peptides from GDP-l-fucose synthase has been characterized by proliferation, enzyme-linked immunosorbent assay, and flow cytometry. Using proliferation, we also investigated the response of PHA-expanded CSF-infiltrating CD4+ T cells from different patients with MS to human and bacterial GDP-l-fucose synthase peptides, as well as to myelin peptides. Demographic and clinical parameters and HLA typing of patients were also analyzed. The sample size was dictated by the rate of sample collection, 5 to 10 ml of CSF, and 50 to 80 ml of peripheral blood for isolation of PBMCs. Blinding was not used. Primary data are located in table S9.

Statistical analysis

Three-cluster k-means analysis was performed on patient scores to group patients into three categories. Associations between response levels of peptides, patients, and HLA status were all performed using Fisher’s exact test with Bonferroni-Holm correction applied as appropriate, with 5% significance.

SUPPLEMENTARY MATERIALS

www.sciencetranslationalmedicine.org/cgi/content/full/10/462/eaat4301/DC1

Materials and Methods

Fig. S1. Integration of stimulatory response from testing dual-defined mixtures into the original scoring matrix using the HM model.

Fig. S2. Response of CSF-infiltrating CD4+ T cells.

Fig. S3. New patients and phylogeny of bacterial species sharing GDP-l-fucose synthase peptides.

Table S1. Summary of human decapeptides predicted with the biometrical approach, synthesized, and tested for stimulatory capacity.

Table S2. GDP-l-fucose synthase, myelin, and CEF peptides.

Table S3. Peptides from brain proteins identified by proteomic analysis in brain tissue.

Table S4. Negative bootstrap Monte Carlo summary.

Table S5. Patient classification and HLA binding.

Table S6. Bacteria selected for GDP-l-fucose synthase comparison.

Table S7. Sequence identity between human and bacterial GDP-l-fucose synthase peptides.

Table S8. Demographic and clinical characteristics of patients with MS and controls without MS.

Table S9. Primary data.

References (6775)

REFERENCES AND NOTES

Acknowledgments: We thank G. Nepom and B. Kwok (University of Washington, Seattle) for BLS tranfectants and F. Sallusto (Institute for Research in Biomedicine, Bellinzona, Switzerland) for IL-2 hybridoma. We acknowledge A. Willing, the INIMS outpatient clinic and day hospital at the University Medical Center Hamburg-Eppendorf, and the clinical staff of nims, Neurology Clinic, University Hospital Zurich for clinical samples. Funding: This work was supported by the DFG Clinical Research Group, KFO 228/1, DFG Center Grant (SFB 841 and SNF: 310030_146945), European Research Council Advanced Grant (340733) (to R.M.), Clinical Research Priority Program MS (CRPPMS) of the University Zurich (to R.M. and M.S.), Swiss National Science Foundation (Sinergia UnmetMS) (to R.M. and M.S.), and the Swiss MS Society (R.M.). R.P. was supported by UZH FK-13-046. UK Multiple Sclerosis Tissue Bank was supported by the Multiple Sclerosis Society of Great Britain and Northern Ireland (registered charity no. 207495). This work was supported in part by Multiple Sclerosis National Research Institute (to R.S. and C.P.). Author contributions: M.S. designed and supervised the study and wrote the manuscript. R.M. participated in study design and reviewed the manuscript. R.P. performed the experiments with T cells. R.S. and C.P. provided the peptide libraries and participated in the design and analysis of the results. P.T.-O. and C.C. help in the expansion of the CSF-infiltrating cells. A.L. supervised the provision of clinical samples. W.F. and N.S.-W. performed the proteomic analysis of the brain. C.E. and H.E. discussed and analyzed data and gave conceptual advice. All authors discussed the results and commented on the manuscript. Competing interests: R.M. and M.S. are inventors on a patent application (no. EP18180326.3) submitted by University of Zurich that covers immunodominant proteins and fragments in MS. All other authors declare that they have no competing financial interests in the context of this work. Data and materials availability: All data associated with this study are present in the paper or the Supplementary Materials. The transcriptomic analysis data discussed in this paper has been deposited in NCBI’s Gene Expression Omnibus (GEO) and are accessible through GEO Series accession number GSE60943. Decapeptide combinatorial libraries are available from Torrey Pines Institute for Molecular Studies (TPIMS) under a material transfer agreement with the University of Zurich.
View Abstract

Navigate This Article