Autoreactive memory CD4+ and CD8+ T cells specific for myelin antigen of the peripheral nerves are present in GBS patients.
To explore the autoreactive T cell immunity in GBS patients, we collected peripheral blood samples and clinical data from a total of 16 patients with AIDP, both at acute phase (range 0 - 36 days from disease onset, 12.7 ± 9.9 (mean ± SD)), and/or at follow-up visits during the recovery stage (range 135-509 days from disease onset, 244.6 ± 115.7 (mean ± SD)) (Extended Data Table 1). When available, we also obtained cerebrospinal fluid (CSF, n = 3) and nerve biopsy samples (n = 1) at disease onset.
We used an experimental approach that combines in vitro screening assays, generation of single T cell clones and TCR Vb sequencing, as described in Extended Data Figure 1 27. In vitro stimulation assay was performed on total memory CD4+ and CD8+ T cell populations from the blood of 15 GBS patients at different disease stages (Extended Data Table 2). We included 10 GBS patients sampled before the COVID-19 pandemic outbreak (non-COVID-19 GBS), in which the potential trigger was either unknown (n = 6) or associated with Varicella-zoster virus (VZV) (n = 2) or Cytomegalovirus (CMV) (n = 2) in the 2-3 weeks before disease onset as well as 5 GBS patients with SARS-CoV-2 infection as a preceding trigger (post-COVID-19 GBS, range 6-20 days post-infection, 11 ± 5.4 (mean ± SD)) (Extended Data Table 1 and 2). As controls, we obtained blood samples from healthy donors (HD, n = 15). Briefly, T cell populations were FACS sorted, labelled with carboxyfluorescein succinimidyl ester (CFSE) and co-cultured with autologous monocytes pulsed or not with peptide pools of selected myelin antigens of the peripheral nerves (PNS-myelin antigens), namely P0, P2 and PMP22, or positive control antigens (Flu vaccine for CD4+ T cells or EBV/CMV for CD8+ T cells). Notably, self-reactive memory CD4+ T cells targeting one or more PNS-myelin antigens were identified in 12 out of 15 GBS patients at different disease stages, whereas they were absent in HD (except for 1 out of 15), as indicated by one representative experiment (Figure 1a) and summary data from the different donors (Figure 1b). The autoreactive response was directed against 1 or 2 PNS-myelin antigens in 10 out of 15 of GBS patients, whereas only 2 GBS patients showed broad autoreactivity against all 3 PNS-antigens (Figure 1b). The P2 antigen was the immunodominant target being identified in 10 out of 15 patients at different disease stages, whereas P0 or PMP22 antigens were recognized only in 6 out of 15 (Extended Data Figure 2a). Interestingly, autoreactive CD4+ T cells were detected in all non-COVID-19 GBS patients (10 out of 10), but only in a minority of post-COVID-19 GBS patients (2 out of 5) (Extended Data Figure 2b, c). In non-COVID-19 GBS, P2 (p = 0.019) and, to a less extent, P0 (p = 0.055) were the main self-antigens targeted during the acute disease phase, whereas the autoreactive CD4+ T response was significantly increased against all three PNS-myelin antigens tested, namely P0 (p = 0.019), P2 (p = 0.008) and PMP22 (p = 0.004) during the recovery disease phase (Figure 1c). Conversely, CD4+ T cells from post-COVID-19 GBS patients (n = 5) showed high background proliferation in negative control cultures (No antigen), and despite PNS-myelin-reactive T cells being identified in 2 out of 5 patients (Extended Data Figure 2b, c), the response was not significant (Figure 1c).
We also performed in vitro screenings of memory CD8+ T cells from 11 GBS patients and 11 HD (Extended Data Table 2) showing that, contrarily to the memory CD4+ T cell subset, self-reactive CD8+ T cells could be detected only in a minority of patients (5 out of 11), but also in 2 out of 11 HD (Extended Data Figure 3 a-c and Extended Data Table 2). In line with what we observed in memory CD4+ T cells, the autoreactive memory CD8+ T cell response was mostly found in non-COVID-19 GBS patients (4 out of 7) compared to post-COVID-19 GBS patients (1 out of 4) (Extended Data Figure 3c).
Collectively, these data indicate that autoreactive memory CD4+ T cells and rare memory CD8+ T cells recognizing P2, P0 and PMP22 myelin antigens of the peripheral nerves are present in the blood of most GBS patients at the disease onset and recovery phases, but they are uncommon in healthy individuals.
Characterization of self-reactive T cell clones.
To further dissect the autoreactive T cell response in GBS patients, we generated PNS-myelin reactive single T cell clones (n = 1004) from CFSElowCD25highICOS+ proliferating T cell fractions from in vitro screenings, which were further characterized for their TCR Vb sequences, HLA restriction and targeted epitopes, as previously described (Supplementary Table 1) 27. We obtained a total of 949 CD4+ T cell clones from 13 patients with specificity against P0 (n = 290 from 10 patients), P2 (n = 504 from 11 patients) or PMP22 (n = 155 from 7 patients) (Figure 2a). We also obtained 55 CD8+ T cell clones from 5 patients targeting P0 (n = 8), P2 (n = 14) or PMP22 (n = 33) antigens (Extended Data Figure 3e). The TCR clonotype composition of the autoreactive T cell clones was determined by performing TCR Vb Sanger sequencing, which led to the identification of TCR Vβ sequences from 682 CD4+ T cell clones encompassing 41 P0-, 81 P2- and 27 PMP22-reactive unique clonotypes (Figure 2a) as well as from 41 CD8+ T cell clones including 4 P0-, 7 P2- and 6 PMP22-reactive unique clonotypes (Extended Data Figure 3f). Both CD4+ and CD8+ autoreactive T cell clones showed a polyclonal TCR repertoire, including a broad spectrum of TCR Vb genes even in the same individual (Figure 2b and Extended Data Figure 3f and 4a). Interestingly, in 4 GBS patients we were able to isolate P0- and P2-reactive sister T cell clones carrying the same TCR Vb clonotypes within the acute and recovery matched blood samples of the same GBS patient (Figure 2c and Supplementary Table 1).
We next determined the TCRB complementarity-determining region 3 (CDR3b) length of PNS-myelin reactive CD4+ T cell clones (n = 148) in comparison to the ones of SARS-CoV-2-specific T cell clones isolated from post-COVID-19 GBS patients (n = 92) or previously characterized microbe-reactive CFSElow fractions from HD (Supplementary Table 2) 28,29. The analysis was also performed on CDR3b sequences of total memory CD4+ T cells obtained ex vivo from PBMCs of GBS patients as well as from publicly available datasets from HD or patients with other autoimmune disorders 27-29 (Supplementary Table 2). Notably, the CDR3b lengths of PNS-myelin reactive T cell clones were shorter than the ones of virus-specific or total memory CD4+ T cell counterparts, further supporting the self-reactive nature of their TCRb repertoire, as previously reported 30,31 (Extended Data Figure 4b).
We next determined the HLA restriction of 97 P0-, 190 P2-, and 83 PMP22-specific CD4+ T cell clones from 13 patients accounting respectively for 27, 60 and 19 identified clonotypes by measuring the inhibition of peptide-induced T cell proliferation by blocking antibodies. This analysis revealed that the majority of PNS-myelin reactive T cell clonotypes derived from different donors of different HLA haplotypes (Supplementary Table 3) were HLA-DR-restricted (85.8%, n = 91) with a minority being HLA-DP- (7.5%, n = 8) or HLA-DQ-restricted (8.5%, n = 9) (Figure 2d, e and Supplementary Table 1). Finally, we successfully mapped the epitope specificity of 61 P0-reactive clones (comprising 25 clonotypes) from 9 patients, 123 P2-reactive clones (comprising 47 clonotypes) from 9 patients as well as 34 PMP22-specific clones (comprising 20 clonotypes) from 5 patients. Self-reactive cells recognized multiple sites that collectively spanned across the whole length of P0, P2 or PMP22 sequences (Figure 2f and Supplementary Table 1), however certain epitopes emerged as immunodominant encompassing specific regions targeted by several clonotypes in different GBS patients. Notably, the 191-205 amino acid region in P0 was recognized by 6 clonotypes derived from 4 out of 9 of GBS patients (44%) as well as the 1-15 amino acid region in P2 was targeted by 17 clonotypes derived from 5 out of 9 patients (55.5%). Similarly, 8 clonotypes isolated from 5 out of 5 GBS patients targeted the region comprising amino acids 81-100 in PMP22.
Overall, our data show that PNS-myelin reactive T cells in GBS patients are mostly HLA-DR restricted, have a polyclonal TCR repertoire, short CDR3b lengths, and recognize multiple epitopes of the self-antigens with some immunodominant regions targeted across patients.
TCR repertoire analysis identifies persistent and public autoreactive T cells in GBS patients.
We next studied the frequency of autoreactive T cells in the blood of GBS patients by high throughput TCR Vb sequencing. To this end, we compared the TCR Vb sequences of the well characterized PNS-myelin reactive T cell clones with the ones obtained from CD4+ memory T cells directly isolated ex vivo from PMBCs obtained from the same patients (n = 6), as previously described 27. The analysis was performed on acute and recovery matched blood samples from patients PT5, PT7 and PT12, only on the acute sample from PT1, or recovery samples from PT4 and PT9 (Extended Data Table 2). In PT1, PT4, PT5, PT7 and PT12 we identified clonotypes corresponding to those of P0-, P2- and PMP22-specific T cell clones from the same patients (Figure 3a). In line with the results obtained on single T cell clones (Figure 2c), the same autoreactive TCR Vb clonotypes were found shared in the acute and recovery samples of the same patient showing variable frequencies in each donor (Figure 3a). PNS-myelin reactive TCR Vb clonotypes were next cross-referenced against the TCRb repertoire of memory CD4+ T cells ex vivo from PBMCs of GBS patients (n = 9). This analysis led to the identification of 22 autoreactive TCRb sequences, comprising 8 P0-, 9 P2- and 5 PMP22-specific clonotypes, which were shared across different GBS patients with different HLA haplotypes (Figure 3b and Supplementary Table 1 and 3). Conversely, such autoreactive TCR Vb clonotypes were not detected in our previously published TCR Vb datasets of memory CD4+ T cells from 15 different HD 27-29 (Figure 3b). In particular, 2 P0-specific clonotypes were detected in 6 out of 9 GBS patients (66.7 %) as well as 2 P2- and one PMP22-specific clonotypes were identified in 55.5% of GBS patients (5 out of 9) and, lastly, 2 P0-, 7 P2- and 4 PMP22-specific clonotypes were found shared across 2 or 3 patients (Figure 3c). The cumulative frequency of the shared P0-specific clonotypes, ranging from 3.3 x 10−5 to 1.4 x 10−3 (median 7.3 x 10−5), was slightly higher compared to the ones of the shared P2-specific (range 2.8 x 10−5 - 1.7 x 10−4, median 4 x 10−5) and PMP22-specific ones (range 6.7 x 10−6 - 9.3 x 10−6, median 8.1 x 10−6) in GBS patients at the acute disease phase (Figure 3d).
To further characterize the TCR Vb repertoire in GBS patients, we performed an unbiased analysis by using the grouping of lymphocyte interactions by paratope hotspots (GLIPH2) algorithm, which has been previously shown to reliably group CDR3b sequences of common specificity from different donors based on local and global similarity 32,33. We applied GLIPH2 algorithm to the TCR Vb repertoire of memory CD4+ T cells from GBS patients (n = 9) and to published TCR Vb datasets of memory CD4+ T cells from HD (n = 9) 27-29, including our reference dataset of known PNS-myelin specific TCR Vb clonotypes obtained from autoreactive single T cell clones. The analysis identified a total of 7 TCRβ specificity clusters that included PNS-myelin specific T cell clonotypes based on global (n = 6) or local similarity (n = 1), each comprising at least 4 unique clonotypes from 2 or more individuals and exhibiting significant final score (P < 1 × 10−5 by GLIPH2) (Figure 3e and Supplementary Table 4). Interestingly, among the identified TCRβ specificity clusters, 6 were exclusively found in 2 or more GBS patients and comprised P2- (n = 5) or P0-specific clonotypes (n = 1), whereas one was found shared in patients (n = 8) and healthy controls (n = 9) and included clonotypes with previously associated reactivities against self- and viral antigens 34-37 (Figure 3f and Supplementary Table 4). Notably, 5 out of 7 TCRβ specificity groups comprised public autoreactive clonotypes in CD4+ memory T cells across GBS patients, as identified above by high throughput TCR Vb sequencing (Figure 3b and e). Each cluster comprised a variable number of clonotypes in each individual accounting for a cumulative frequency ranging between 1.96 x 10−5 and 3.37 x 10−4 ex vivo in the blood of GBS patients (Figure 3g and h). Cluster one did not show a specific enrichment in GBS patients compared to HD neither in terms of clonotype number nor in terms of cumulative frequency (Figure 3f and g).
Overall, these findings confirm the existence of expanded autoreactive memory CD4+ T cells in the blood of GBS patients at disease onset and recovery phase and identify both public and private TCR Vb clonotypes with shared similarities and specificities against PNS-myelin antigens across GBS patients.
PNS-myelin reactive T cells are found in the CSF and peripheral nerves of GBS patients.
To investigate whether autoreactive T cells could be found also in the proximity of tissue immunopathology, we obtained a sample of CSF from 3 GBS patients at disease onset (Extended Data Table 1). Intrathecal CD4+ T cells were enriched by in vitro polyclonal expansion according to an established protocol 27 and further characterized for their clonotype composition by high throughput TCR Vb sequencing, which led to detection of 500-4000 clonotypes in different samples (Figure 4a). We next searched for the presence of PNS-myelin-specific TCR Vb clonotypes revealing the existence of autoreactive T cells in the CSF of 2 out of 3 patients. Such clonotypes were amongst the lowest expanded, which may explain the lack of detection of PNS-myelin reactive T cells in CSF by in vitro stimulation assays (data not shown). We identified one P2-specific clonotype in the CSF of PT10 as well as several P0- (n = 4), P2- (n = 5) and 1 PMP22-specific clonotypes in the CSF of PT11. In PT11, 3 PNS-myelin specific clonotypes in the CSF were also identified in the blood of the same patient showing frequencies ranging between 8.7 x 10−5 and 8.7 x 10−4 (Figure 4b). Notably, 10 out of the 11 PNS-myelin specific TCR Vb clonotypes identified in the CSF were amongst the ones described as public and shared across 2 or more GBS patients in the blood (Figure 3b and Supplementary Table 1). Conversely, none of the autoreactive TCR Vb clonotypes isolated from GBS patients was detected in our published TCRb sequence dataset of intrathecal CD4+ cells from patients with narcolepsy 27.
Finally, we obtained a nerve biopsy from one GBS patient (PT16) at disease onset and tissue infiltrating CD4+ T cells were enriched by FACS-sorting after in vitro polyclonal expansion. The cells were further analysed for their antigen specificity by in vitro stimulation assay, which revealed the existence of proliferating CFSElow PNS-myelin reactive T cells against P0 (17.5%) and PMP22 (17%), and to a less extent P2 (4.5%) (Figure 4c). In parallel, to further investigate autoreactivity in CD4+ T cells from the blood of the same patient, central memory (CM) and effector memory (EM) CD4+ T cell populations were stimulated in vitro with a combination of P0, P2 and PMP22 antigens. PNS-myelin reactive CFSElow T cells were identified in both CM and EM cell subsets with a slight enrichment in the EM population (Figure 4d). Furthermore, we studied the relationship between autoreactive T cells in the blood and nerve tissue by determining the TCR Vβ clonotype composition of the PNS-myelin reactive CFSElow T cell fractions from in vitro stimulated cultures in the blood and nerve biopsy. We obtained 274 unique TCR Vβ sequences in PMP22-myelin reactive T cells from nerve biopsy as well as 241 and 310 unique clonotypes in PNS-myelin reactive EM and CM CD4+ T cells from the blood, respectively. This analysis led to the identification of a total of 10 PNS-myelin specific clonotypes shared in nerve tissue infiltrating T cells and blood. Notably, 7 clonotypes were found in the CFSElow fractions from CM and 2 clonotypes from EM as well as 1 clonotype was identified in both fractions (Figure 4e).
To further investigate the relationship between autoreactive TCR Vb repertoires in different body compartments, we applied GLIPH2 analysis on total CD4+ T cells expanded from the nerve biopsy and matched total memory CD4+ T cells repertoire ex vivo from the blood of the same patient (PT16) as well as on the TCR Vb repertoires of total CD4+ T cells expanded from the CSF of GBS patients in the acute disease phase (n = 3; PT10, PT11, PT12). The analysis included our reference dataset of known PNS-myelin specific TCR Vb clonotypes obtained from autoreactive single T cell clones from GBS patients. Interestingly, we identified one TCR Vβ specificity cluster comprising one PMP22-specific public clonotype (clonotype ID: CD4_131, Supplementary Table 1 and 4) and encompassing several TCR Vβ clonotypes grouped by local similarity in nerve-tissue infiltrating CD4+ T cells (n = 43, PT16) and in total memory CD4+ T cells in matched-blood from the same GBS patient (n = 1, PT16) as well as in total CD4+ T cells from the CSF of 3 different GBS patients (PT10, n = 1; PT11, n = 1; PT12, n = 1) (Figure 4f and Supplementary Table 4).
Altogether, these results describe the existence of PNS-myelin reactive T cells in the affected nerve tissue and in the CSF compartment, thus pointing to their potential involvement in GBS immunopathology.