T Cell Repertoire Abnormality in Immunodeficiency Patients with DNA Repair and Methylation Defects

Genomic DNA from 19 IEI patients with evidence of DNA repair defects and radiosensitivity and 14 age-matched healthy donors was extracted from peripheral blood, and 18 patients underwent whole-exome sequencing (WES). All patients and normal individuals were used for TCRβ repertoire sequencing. Approvals were obtained from the human research ethics committees at the Institutional Review Boards of the Karolinska Institutet, the Tehran University of Medical Sciences, and the Institute of Research in Biomedicine of BGI-Shenzhen. Informed consent for the performed tests was obtained from all patients and/or their parents. All patients were diagnosed as IEI based on the updated clinical diagnostic criteria of the European Society for Immunodeficiencies (ESID) [29, 30].

Agilent SureSelect Human All Exon V5 (50 Mb, Agilent Technologies, Santa Clara, CA) were used for whole-exome capture and Illumina HiSeq2000, and 90 bases paired-end sequencing (Illumina) was performed for exon library sequencing. The candidate gene screening pipeline of WES data was performed as previously described [31]. Burrows-Wheeler Aligner (BWA), Genome Analysis Toolkit (GATK), and Variant Effect Predictor (VEP) were carried out for reads alignment to the GRCh37/hg19 human reference genome, variants calling, and annotation, respectively. Public databases including the 1000 Genomes Project (KG), Exome Variant Server (ESP), Exome Aggregation Consortium (ExAC), and The Genome Aggregation Database (gnomAD) were used to identify and remove polymorphisms. Copy number variants (CNV) were detected by ExomeDepth.

Multiplex primers and two complete sequencing adapter primers described before [3, 32] were used to amplify the CDR3 sequences of TCRβ genes for equal amounts of DNA samples (1.2 µg) from all TCR subjects. AMPure XP (Beckman, A63882) was used to clean the first 15 cycles amplicons and purified PCR products were then amplified for a second round (25 cycles) with a pair of communal primers. Libraries were loaded onto a BGIseq500 (BGI-Shenzhen, Shenzhen, China) and underwent 200 bp single end to perform sequencing. A mean of 5,971,100 (range 1,345,371–20,526,738) total reads and 5,159,138 (range 824,569–17,542,667) clean reads were obtained for the sequenced samples.

TCR sequencing data were analyzed using iMonitor [3]. The final alignment result was obtained according to the optimal score after two rounds of Basic Local Alignment Search Tool (BLAST) [3], while the International ImMunoGenneTics database (IMGT, www.​imgt.​org) was used as a reference. In-frame and out-of-frame determinations; V, D, and J segments usage calculation; deletion/insertion nucleotide; and amino acid sequence determination at the rearrangement were performed with default parameters as described previously [3]. The rearrangement frame was labeled as “out-of-frame” if the sequence contains a stop codon or has a length of non-multiple of three (frameshift); otherwise, it was labeled as “in-frame”. Evenness is best expressed as a partial order, and this post-structure can adequately be illustrated by Shannon–Wiener’s diversity (or entropy) index (H’) and the Gini evenness coefficient (G’) measured by considering three basic requirements: permutation invariance, scale invariance, and the transfer principle.

In total 20,814 pathology-associated CDR3 records of humans (Table S1) were collected from a manually curated database of T and B cell receptors targeting known antigens (TBAdb) (https://​db.​cngb.​org/​pird/​tbadb/​) and online databases McPAS-TCR (http://​friedmanlab.​weizmann.​ac.​il/​McPAS-TCR/​), where all records were derived from a manual literature search. We used all CDR3β sequences as a reference to perform alignment for all studied patients, and identical CDR3 sequences that are absolutely aligned to the reference were then calculated and annotated as specific for various pathogens.

Statistical analysis and data visualization was performed using R (R version 3.3.2). To account for the different sequencing depths among samples, normalization by a total number of reads was performed. Metrics calculated from normalized one million reads of data were used to describe the CDR3 sequences of each sample, including Shannon’s H index for measurement of repertoire diversity, Gini coefficiency of clonality, frequency of top 100 CDR3 clones, Pielou’s evenness index of V-J pairing, frequency of in-frame and out-of-frame rearrangements, nucleotide composition of the CDR3 sequences, and V gene usage. Differences in these metrics between groups were investigated, and their statistical significance was tested by two-sided Wilcoxon Rank Sum Test. In the V gene usage comparison, Bonferroni correction was used for multiple tests correction. Length distribution of the CDR3 sequences and insertion/deletion sequences were further examined to investigate the patterns in each group during pre-selection and post-selection of TCRβ rearrangement, where two-sided bootstrap Kolmogorov–Smirnov test was used for testing the distribution differences and one-sided Wilcoxon Rank Sum Test was used to test for the significance of length difference between each group. PCA analysis and visualization were performed using the built-in R function and function from additional packages ggplot2 downloaded from CRAN (https://​cran.​r-project.​org/​). Adjustments of all statistical tests have been made to multiple testing corrections; p values ≤ 0.05 or FDRs ≤ 0.1 were taken to be significant.

After performing WES, 19 IEI patients that were found to have mutations in DNA repair and methylation genes were recruited into this study. The pathogenicity of all disease attributable gene variants was re-evaluated using the updated guideline for interpretation of molecular sequencing by the American College of Medical Genetics and Genomics (ACMG) criteria [33, 34], considering the allele frequency in the population database, computational data, immunological data, and clinical phenotyping. The causative genetic variations later were confirmed by Sanger sequencing and segregation validation in each affected pedigree (Table 1). The patients consisted of 13 cases with a clinical diagnosis of atypical combined immunodeficiency (CID) and 6 patients with AT [35]. According to the molecular diagnosis, we subdivided our patients into four groups including atypical T⁺ SCID (n = 3), AT (n = 6), ICF1 (n = 6), and ICF2 (n = 4). We classified patients with a hypomorphic mutation in NHEJ components (RAG1 hypomorphic with 48% residual activity [36], DCLRE1C- OMIM #602450) and a patient with mutation in JAK3 (OMIM #600802) [37] as atypical SCID, due to incompatibility with the standard diagnostic criteria of the European Society for Immunodeficiencies (ESID) and recent phenotypic classification of the International Union of Immunological Societies (IUIS) [38]. The median age at the time of the study was not significantly different between the four groups (9.2, 10.6, 9.1, and 13.2 years, respectively).

The clinical and immunological phenotypes of the patients (8 male, 11 female) are described in Table 2 and Fig. S1, where the AT patients presented a more severe phenotype than the other patients, even as compared to the atypical SCID group. In our AT patients, low counts of lymphocytes, absolute T and B cell numbers, and total CD4⁺ and CD8⁺ T cell numbers were observed, with a mean value of 1,663 lymphocytes, 1,295 T cells, 135 B cells, 723 CD4⁺ cells, and 512 CD8⁺ cells per microliter (µL), respectively (Table 2 and Fig. S1A-E), which is consistent with previous studies [25, 39]. The lymphocyte counts in AT patients was significantly reduced as compared with the other groups (p = 0.0238, Lymphocyte _AT vs Lymphocyte _{Atypical SCID}, p = 0.0043, Lymphocyte _AT vs Lymphocyte _ICF1 and 0.0260 CD8⁺ cells _AT vs CD8⁺ cell _ICF1) (Table 2 and Fig. S1A,and E). Correlation analysis shows that the T cell count, as expected, is positively correlated with the concentration of lymphocytes, CD4⁺ and CD8⁺ cells in all patients (Fig. S1G-I).

It is generally thought that TCR repertoires are dynamically changed and that diversity declines with aging [40‐42] and thus, age-matched controls have usually been used in immune repertoire analyses [39, 43]. To investigate the V(D)J recombination defect and analyze the TCRβ repertoire in patients with different pathogenic backgrounds, we compared their repertoire diversity and complexity using next-generation sequencing in four groups of patients together with 14 age-matched controls.

TCRβ rearrangements were detected in all patients. To evaluate the diversity and clonality of repertoire, we introduced some economic indexes, such as Shannon’s H index [3] and Gini coefficient, to measure repertoire diversity (diversity of CDR3, the V/J gene usage, and V-J pairing) and to evaluate repertoire homogeneity, respectively. Of note, immune repertoires in patients with atypical SCID showed a restricted diversity reflecting the hypomorphic nature of the identified variants in our patients reported previously [35].

The TCRβ repertoire of the AT patients was highly restricted, not just concerning the repertoire diversity as shown by a low Shannon’s H index (Fig. 1A), but also for the evenness of clonality, represented by a higher Gini coefficient (Fig. 1B) when compared to healthy controls. The total frequency of the top 100 CDR3 clonotypes (Fig. 1C) was significantly higher in the AT patients, and a reduction of the total number of unique clones (p = 0.016) was also observed in the AT group as compared to controls (Table S2). A lower ratio of in-frame CDR3 rearrangement (Fig. 1D) and a high proportion of out-of-frame rearrangements due to frameshifts (Fig. 1E) in AT patients versus healthy controls and the other DNA repair and methylation deficient patient groups were also observed.

In contrast, no significant change of the TCRβ repertoire was observed when the ICF1 group was compared to controls, except that ICF1 patients showed a significantly lower percentage of out-of-frame TCR genes containing a stop codon within their CDR3 (p = 0.036 versus controls, Fig. 1F). Interestingly, when comparing the ICF2 to the ICF1 group (Wilcoxon Rank Sum Test, Fig. 1A-B, and G), we observed significant differences in clone diversity (Shannon’s H, unevenness (Gini skewing index of TCR), and Gini skewing index of V-J pairing). This phenomenon correlates with the reduced number of lymphocytes in the peripheral blood of the ICF2 patients (Table 1 and Fig. S1A and B). Most notably, the TCRβ repertoire of the ICF2 group was characterized by restriction of diversity (Fig. 1A–B) when compared with normal individuals or ICF1 patients, representing a higher percentage of the top 100 clonotypes (Fig. 1C) and a smaller number of unique clones (Table S2) in comparison with healthy controls.

TCRβ repertoire diversity and clonality are initially created by random V(D)J recombination, further enhanced by Insertion/Deletion (InDel) of nucleotides in the junctional regions between the V, D, and J segments and followed by selection according to the receptor fitness and response to self-antigens [44]. Previously, twin studies have suggested that V gene segment usage in TCRβ is predominantly affected by the germ-line locus factor of recombination machinery [45] and preferred usage of specific V genes that might influence clonotypic expansion in response to antigens.

To determine the role of different underlying pathogenic variants in individual V gene usage or V-J pairing, we first compared the spectrum of V gene usage in unique TCRβ clones. Marked reduction of TRBV19 usage in the AT group (p < 0.001, Fig. S2A) and TRBV7-3 usage in the ICF2 group (p < 0.001, Fig. S2B) in comparison to normal individuals was observed.

A considerable decrease of the repertoire diversity was also demonstrated in ICF2 patients when compared to normal individuals or ICF1 patients and to a lesser extent in the AT patients using Gini’s skewing index of V-J paring (Fig. 1G) and treemaps, a graphical representation of the diversity of V-J pairing (Fig. S3). The low diversity of V-J pairing was also observed in the atypical SCID patient group which might be related to the pathogenicity of the identified variants in NHEJ repair function presenting a late-onset atypical disease, despite normal lymphocyte subsets.

CDR3 length and composition usually play a crucial role in adaptive immune responses to a variety of non-self-antigens as well as an increased risk of recognizing self-antigens in autoimmune diseases [46, 47]. We analyzed the length of the productive CDR3 (in-frame rearrangements) in different IEI groups and the differences in distribution compared to controls. Substantial reduction of the CDR3 length was observed in both AT and atypical SCID patients versus controls. By contrast, the CDR3 length distribution was increased both in the ICF1 and ICF2 patients compared with the healthy individuals (Fig. 2A–D and Fig. S4A).

It is generally thought that TCR repertoire bias can be introduced during the VDJ gene recombination process in the thymus, which could be evaluated by the non-productive rearrangements (out-of-frame CDR3s), or after the positive/negative selection in the thymus and antigenic selection in the periphery [43]. Furthermore, pre-selection of TCR formation (out-of-frame rearrangements) together with T cell selection in the thymus both depend on genetic factors. In order to better understand whether the aberrant CDR3 length happens before or after the selection, we repeated the CDR3 length distribution analyses for out-of-frame rearrangements, and a very similar trend was observed (Fig. 2E–H and Fig. S4B). It, therefore, implies that the pre-selection rearrangement gave rise to the aberrant CDR3 length.

The aberrant CDR3 length profile from patients could be the result of deviations of an InDel (either N-nucleotides inserted by terminal deoxynucleotidyl transferase (TdT) or deletions at the ends of V, D, and J segments). Therefore, to further examine the role of DNA repair defects genes in junctional diversity, we analyzed the variation of deletions or insertions in cases and controls, by calculating the length distribution of 6 segments (3’V, 5’D, 3’D, 5’J deletions and V-D, D-J insertions) in out-of-frame and in-frame rearrangements, separately.

Shorter deletions at the end of the V, D, and J segments result in an increased V, D, and J gene length in AT patients (Fig. S4, C-E, Table S3, 4). At the same time, in these patients, fewer nucleotides (N region) added to the sites of V-D and D-J junctions were also observed in both out-of-frame (Fig. S5 and Fig. 3A) and in-frame (Fig. S6 and Fig. 3A) rearrangements, and the same observation precedent was reported in XLF deficiency patients [4]. However, a reduced insertion length in the AT patients plays a more significant role than reduction in deletion length and leads to a shorter mean CDR3 length (Fig. 3A–B). The average insertion size is usually around 2–5 bp per coding joint which is optimal for base pair recognition patterns [48, 49]. Here, we found a statistically significant decrease in AT patients for both out-of-frame (average 3.63nt decreased) and in-frame (around 2.55nt reduced) versus controls (Fig. 3A–B and Table S4).

The InDel length distribution was significantly different between ICF patients (ICF1 and ICF2) and healthy controls (Table S3). A longer V gene length in CDR3 region (Fig. S4C) contributes to the increased mean CDR3 length in both the ICF1 and ICF2 group (Fig. S4A-B ).

To characterize the level of sequence similarity of the TCRβ repertoires within and between different diseases, we also calculated the fraction of shared identical clonotypes between sample pairs. The calculation revealed that more unique and total clonotypes were shared within the AT group, and an opposite trend was observed in ICF syndrome patients (Fig. 3C–D). The observation is consistent with previous reports that the shared clonotypes are characterized by a small size of insertion and the subsequent shorter CDR3 length, which increases the likelihood of two TCRβ sequences being identical [45, 50, 51]. Although AT patients and controls also have a high fraction of shared identical clones, the ratio of shared clonotypes in controls was smaller than what was found between AT patients (Fig. 3D), which might be due to a limited number of unique clones existing in the AT patients compared to controls (Table S2).

Hydrophobicity of positions 6 and 7 amino acids of the 13-amino acid-long CDR-B3 promotes the development of self-reactive T cells [52], as do the aromatic residues in CDR3 sequences enriched by positively charged tyrosine(s) [53]. Notably, the percentage of tyrosine residues in the CDR3 is significantly higher in AT patients (Fig. 4A–B) and also shows an appreciable divergence of hydrophobic amino acid usage at positions 6 and 7 of total clone sequences in AT and ICF2 patients (Fig. 4C–F), suggesting that the peripheral population in these patients differs from healthy controls.

Moreover, previous studies have shown that TdT preferentially uses dGTP and dCTP as compared to dAMP or dTMP during generation of junctional diversity which results in a high G/C content of the TCR insertion regions [4, 48, 54]. Clear distinctions were observed for the average percentage of inserted nucleotides in the sequence which is consistent with this notion in IEI patients or normal individuals (Table S5). However, no significant differences were observed for the average GC content among the different patient groups and normal individuals (Fig. S7).

Principal components analysis (PCA) was performed in order to assess whether the characteristics of the TCRβ repertoire could distinguish patients from normal individuals. The characteristics in patients with ATM deficiency, ICF2 patients, and healthy donors were distinct based on six variables (Pielou’s evenness (TCR), Gini skewing index (TCR), Gini skewing index (V-J paring), the percentage of in-frame rearrangement, mean CDR3 length of out-of-frame rearrangement, and the percentage of tyrosines in unique clones) (Fig. 4G). On the other hand, the ICF1 group and the atypical SCID patients could not be differentiated from each other, and the ICF1 patients clustered together with the ICF2 group and the normal individuals (Fig. 4G–H).

The phenotypic heterogeneity of immunodeficiencies is well established. For instance, opportunistic infections are rarely reported in AT patients [55‐57] but have been identified in several ICF patients [12, 58]. Furthermore, AT patients also have an increased risk of developing autoimmune diseases, including immune thrombocytopenia (ITP), several forms of arthritis, and vitiligo [59, 60]. Moreover, AT patients are predisposed to lymphoid malignancies [61]. However, ICF patients rarely develop cancer [62].

To further investigate the association between the TCR repertoire abnormality and the heterogeneous symptoms/complications in AT and ICF patients, we performed a comprehensive analysis of the enrichment of literature which reported pathology-associated TCR (paTCR) clonotypes within each patient group (see “Materials and Methods” section).

The proportion of known paTCR CDR3s out of all unique CDR3s in a sample may represent the abundance of paTCRs in the pre-selected repertoire, and the proportion of paTCRs in total CDR3s includes the influence of the selection process. We found that the proportion of known pathogen, autoimmunity, and cancer-associated clonotypes was significantly increased among the unique clones in the AT group (Fig. 5A–D and Fig. S8 A-I). However, there were no significant differences in a fair number of inflammatory diseases associated with clonotypes between AT patients and controls; these diseases include aseptic meningitis, transverse myelitis, polyradiculitis, inflammatory cranial neuropathy, and muscular dystrophy. Besides, only a significant increment of multiple sclerosis (MS)-specific clonotypes (Fig. S8H) out of autoimmune disease in AT patients and melanoma-specific clonotypes out of cancer-associated clonotypes enriched in AT patients (Fig. S8I, L), suggesting that the proportion of pathology-associated clones in the AT patients were not undifferentiated. As for the high proportion of melanoma-specific clonotypes observed in AT group, other studies also showed that mutations in ATM were associated with the risk of melanoma [63], and the change in phospho-ATM expression was associated with melanoma progression [64].

We furthermore checked the correlation between the frequency of pathology-associated clonotypes with the TCR diversity Shannon index, but no significant association among them was found (Fig. S9, A-C), implying that changes in the proportion of known paTCR CDR3s were independent of global TCR diversity and may independently reflect an increased risk of AT-associated complications.

In contrast to AT patients, a low frequency of pathogen-specific clonotypes was recorded in the ICF1 patients (Fig. 5A). Interestingly, ICF1 and ICF2 groups both showed a lower proportion of common clonotypes, which were defined as clonotypes presented in 5 or more control samples in our study (Fig. 5A–D and Fig. S8, A-I).

To test the contribution of antigen selection to the over-representation of paTCRs, we also compared the proportion of paTCRs in the total CDR3s among each group (Fig. S8, K-M). The proportion of pathogen and cancer paTCRs in the total CDR3s in the AT patients was also significantly higher than in healthy controls (Fig. S8, K-L). However, less magnitude of enrichment was observed in the total CDR3s compared to the unique hits, for both cancer and autoimmunity associated clones in the AT patients, perhaps reflecting that none of the pediatric AT patients in this study had as yet developed cancer or autoimmune disease. The frequency of pathogen-specific TCRs among the total clones fluctuated widely (Fig. S8M), indicating that the infection status might be different in each individual. Taken together, we propose that the over-representation of paTCRs in the pre-selected TCR repertoire contributes to the discrepant symptoms of AT patients compared to the ICF patients.

We supposed that the change of paTCR proportion in patients may be related to the shortening of CDR3 length. In order to examine this notion, we compared the length distribution of known pathology-specific clonotypes with those unknown clonotypes in each group. The known pathology-associated TCR showed a dramatically higher proportion of short CDR3 length compared to other unknown clonotypes (Fig. 5E). Therefore, the enrichment of paTCR might occur in the process of abnormal VDJ recombination which produces TCR repertoire with reduced CDR3 length.