Thymic selection of the T cell receptor repertoire is biased toward autoimmunity in females
- Immunology, Immunopathology, Immunotherapy i3 Lab, INSERM UMRS-959, Sorbonne Université, France
- Biotherapy (CIC-BTi) and Inflammation-Immunolopathology-Biotherapy Department (i2B), Assistance Publique -Hôpitaux de Paris (AP-HP), France
- Institut Universitaire de France, France
Figures
Figure 1 with 2 supplements
Schematic overview of the generation of the thymic T cell receptor (TCR) dataset and the analytical pipeline.
Top panel: Generation of the thymic TCR dataset. From deceased human thymuses of males and females, we isolated key T-cell subtypes through cell sorting. These subtypes included double-positive (DP) cells (CD3+CD4+CD8+), single-positive (SP) CD8 + cells (CD3+CD4 CD8+), SP CD4 + cells (CD3+CD4+CD8-) that were further separated into T effector (Teff, CD3+CD4+CD8-CD25-) and Treg (CD3+CD4+CD8 CD25+) cells. TCR libraries were generated from the RNA of each cell population using rapid amplification of cDNA ends by PCR (5'RACE PCR). Following sequencing, data preprocessing involved quality sequencing checks, contig alignment, and quality control. The final dataset comprised DP samples (n=20; male-to-female ratio of 1:1), SP CD8+samples (n=21; 1.1:1), SP CD4+samples (n=6; 1:5), SP CD4 Teff samples (n=16; 1.67:1), and SP CD4 Treg samples (n=14; 1.33:1). Males are depicted in violet and females in orange. Bottom panel: Analytical pipeline. We compared the TCR repertoires of males and females across various dimensions. We evaluated general aspects of the TCR repertoire were evaluated, including diversity, gene usage, CDR3aa length distribution, and aa usage within the CDR3 region. Additionally, we analyzed the probability of sequence generation and the TCR repertoire structure based on CDR3aa sequence similarity. We identified differentially expressed TRB CDR3aa motifs between sexes and analyzed TRB CDR3aa sequence specificity. Figure 1 was created with BioRender.com.
Figure 1—figure supplement 1
Age distribution of male and female donors is comparable across thymic subsets.
Age of individual donors plotted separately for females (orange) and males (violet) for each thymic cell subset: double-positive (DP) thymocytes (n=20), CD8 single-positive (CD8) (n=21), CD4 effector single-positive (CD4 Teff) (n=22), and CD4 regulatory T-cell single-positive (CD4 Treg) (n=14). For each subset, a two-sample Kolmogorov–Smirnov (K–S) test was performed; the D statistic and corresponding p-value are displayed below each panel. In all cases, no significant difference was detected, indicating balanced age distributions between sexes for every subset.
Figure 1—figure supplement 2
Rank–frequency distributions of thymic TRA and TRB clonotypes.
Log–log rank–frequency plots of unique TRA (left) and TRB (right) clonotypes are shown for each donor and thymic subset (double positive [DP] – n=20, CD8 – n=21, CD4 Teff – n=22, CD4 Treg – n=14, from top to bottom). For each sample, clonotypes are ranked by decreasing abundance (x-axis, log scale), and their corresponding relative frequencies are plotted (y-axis, log scale). Curves display broadly comparable shapes across donors, with no obvious systematic differences between males and females, indicating similar clonotype abundance distributions and sampling depth across sexes.
Figure 2 with 4 supplements
Comparable overall T cell receptor (TCR) gene usage between males and females.
(A) Principal component analysis (PCA) derived from the distribution of TRAV (left) and TRBV (right) gene usage frequencies across sex groups (males vs females), showing results for double-positive (DP; n=20), CD8 (n=21), CD4 Teff (n=22), and CD4 Treg (n=14) cells (displayed from top to bottom). Each point on the graph represents an individual. Ellipses indicate 95% confidence intervals. (B) Heatmap showing the Jensen-Shannon Divergence (JSD) score between samples, derived from the distributional usage of TRAV-TRAJ (left) and TRBV-TRBJ (right) gene associations in DP (n=20), CD8 (n=21), CD4 Teff (n=22), and CD4 Treg (n=14) cells (displayed from top to bottom). Hierarchical clustering was performed using the Euclidean distance and the complete linkage method. Males are shown in violet and females in orange.
Figure 2—figure supplement 1
Some differential V and J gene usage in T cell receptor (TCR) repertoires of double-positive (DP) cells between males and females.
(A–D) Frequencies of TRAV (A), TRAJ (B), TRBV (C), and TRBJ (D) gene usage between males and females (n = 10 per group). Each dot represents one donor. Horizontal red lines indicate the mean for each sex group. Statistical comparisons between groups were performed using two-sided Wilcoxon test. Asterisks indicate significant differences between males and females based on the test p-value (*: p<0.05, **: p<0.01). (E–F) Principal component analysis (PCA) derived from the distribution of the frequency of usage of TRAV-TRAJ (E) and TRBV-TRBJ (F) gene associations across donors. Each point represents an individual. Ellipses show 95% confidence intervals. Males are depicted in violet (n=10) and females in orange (n=10).
Figure 2—figure supplement 2
Some differential V and J gene usage in T cell receptor (TCR) repertoires of CD8 cells between males and females.
(A–D) Frequencies of TRAV (A), TRAJ (B), TRBV (C), and TRBJ (D) gene usage between males (n = 11) and females (n = 10). Each dot represents one donor. Horizontal red lines indicate the mean for each sex group. Statistical comparisons between groups were performed using two-sided Wilcoxon test. Asterisks indicate significant differences between males and females based on the test p-value (*: p<0.05, **: p<0.01). (E–F) Principal component analysis (PCA) derived from the distribution of the frequency of usage of TRAV-TRAJ (E) and TRBV-TRBJ (F) gene associations across donors. Each point represents an individual. Ellipses show 95% confidence intervals. Males are depicted in violet (n=11) and females in orange (n=10).
Figure 2—figure supplement 3
Some differential V and J gene usage in T cell receptor (TCR) repertoires of CD4 Teff cells between males and females.
(A–D) Frequencies of TRAV (A), TRAJ (B), TRBV (C), and TRBJ (D) gene usage between males and females (n = 11 per group). Each dot represents one donor. Horizontal red lines indicate the mean for each sex group. Statistical comparisons between groups were performed using two-sided Wilcoxon test. Asterisks indicate significant differences between males and females based on the test p-value (*: p<0.05, **: p<0.01). (E–F) Principal component analysis (PCA) derived from the distribution of the frequency of usage of TRAV-TRAJ (E) and TRBV-TRBJ (F) gene associations across donors. Each point represents an individual. Ellipses show 95% confidence intervals. Males are depicted in violet (n=11) and females in orange (n=11).
Figure 2—figure supplement 4
Some differential V and J gene usage in T cell receptor (TCR) repertoires of CD4 Treg cells between males and females.
(A–D) Frequencies of TRAV (A), TRAJ (B), TRBV (C), and TRBJ (D) gene usage between males (n = 8) and females (n = 6). Each dot represents one donor. Horizontal red lines indicate the mean for each sex group. Statistical comparisons between groups were performed using two-sided Wilcoxon test. Asterisks indicate significant differences between males and females based on the test p-value (*: p<0.05, **: p<0.01). (E–F) Principal component analysis (PCA) derived from the distribution of the frequency of usage of TRAV-TRAJ (E) and TRBV-TRBJ (F) gene associations across donors. Each point represents an individual. Ellipses show 95% confidence intervals. Males are depicted in violet (n=8) and females in orange (n=6).
Figure 3 with 1 supplement
Comparable thymic T cell receptor (TCR) repertoire diversity between males and females.
Boxplots display Shannon (A), Simpson (B), and Berger-Parker (C) index values for TRA (left) and TRB (right), across thymic T cell subtypes, displayed from top to bottom, in double-positive (DP; n=20), CD8 (n=21), CD4 Teff (n=22), and CD4 Treg (n=14). Each point on the graph represents the median value from 50 rarefactions per sample. Statistical analysis (two-sided Wilcoxon test) showed no significant sex bias in TCR repertoire diversity (p>0.05). Males are shown in violet and females in orange.
Figure 3—figure supplement 1
Minimal differences in diversity profile of CD8 and CD4 Treg thymic T cell receptor (TCR) repertoire between males and females.
Diversity profile with Rényi diversity index values (α ranging from 0 to infinity) for all cell subsets (double-positive [DP] – n=20, SP CD8 – n=21, SP CD4 Teff – n=22, and SP CD4 Treg – n=14) and for both TRA (left) and TRB (right). Clonotypes of each sample were rarefied fifty times to their effective diversity number [i.e. ] and the median of their Rényi values was used for analysis. Dotted lines and points show the mean value for each sex group, while the shaded area represents the SD for each sex group. The overall shape of the curves was compared statistically using the Kolmogorov-Smirnov test, with the D value and associated p-value indicated.
Figure 4 with 2 supplements
CDR3aa length and amino acid composition of TRB CDR3s in males and females.
(A) Distribution of TRB CDR3 amino acid (CDR3aa) length usage in double-positive (DP; n=20), CD8 (n=21), CD4 Teff (n=22), and Treg (n=14) single-positive (SP) thymic cells, displayed from top to bottom. Asterisks indicate significant differences between males and females based on the two-sided Wilcoxon test p-value (*: p<0.05, **: p<0.01). (B) The data on amino acid (aa) usage between males and females are presented as the log2 fold change of the median per-donor usage in females over males for each aa in the p108 to p114 CDR3aa region for TRB. A line at log2 fold change = 0 is indicative of the direction of the difference in usage frequency. The bars are color-coded according to the hydropathy class of the aa as defined by the Kyte-Doolittle-based IMGT classification: neutral aa by gray, hydrophilic aa by blue-green, and hydrophobic aa by gold. Asterisks indicate statistical differences of usage between males and females based on the two-sided Wilcoxon test p-value (*: p<0.05, **: p<0.01). (C) Position-specific usage of hydrophobic aa (excluding alanine, due to its weak hydrophobicity) at IMGT positions p109 and p110 in TRB across thymic T cell subtypes, including DP (n=20), CD8 (n=21), CD4 Teff (n=22), and CD4 Treg (n=14). For each donor, the values represent the proportion of unique TRB CDR3aa sequences carrying a hydrophobic amino acid at the indicated position. Asterisks indicate significant sex differences based on the two-sided Wilcoxon test p-value (*: p<0.05), with a significant increase in hydrophobic usage at p109 in female CD8 SP cells. Males are depicted in violet and females in orange.
Figure 4—figure supplement 1
Comparable TRA CDR3aa length distribution between males and females.
Distribution of CDR3aa length usage for TRA in all thymic T cell subsets, including double-positive (DP; n=20), CD8 (n=21), CD4 Teff (n=22), and CD4 Treg (n=14). For each donor, the values represent the proportion of unique TRB CDR3aa sequences carrying a hydrophobic amino acid at the indicated position. Asterisks indicate significant differences between males and females based on the two-sided Wilcoxon test p-value (*: p<0.05, **: p<0.01). Males are depicted in violet and females in orange.
Figure 4—figure supplement 2
Some differences in TRA CDR3aa usage between males and females.
The data on amino acid (aa) usage between males and females are presented as the log2 fold change of the median per-donor usage in females over males for each aa in the p108 to p114 CDR3aa region for TRA, across thymic T cell subsets: double-positive (DP; n=20), CD8 single-positive (SP; n=21), CD4 Teff SP (n=22), and CD4 Treg SP (n=14). A line at log2 fold change = 0 is indicative of the direction of the difference in usage frequency. The bars are color-coded according to the hydropathy class of the aa as defined by the Kyte-Doolittle-based IMGT classification: neutral aa by gray, hydrophilic aa by blue-green, and hydrophobic aa by gold. Asterisks indicate statistical differences of usage between males and females based on the two-sided Wilcoxon test p-value (*: p<0.05, **: p<0.01).
Figure 5
Probabilities of generation of thymic T cell receptors (TCRs) in males and females.
The figure shows the distribution of log10 probability of generation (Pgen) values of sequences between males and females for TRA and TRB in double-positive (DP) and CD8 single-positive (SP) thymic T cells. A V(D)J recombination model was created using 100,000 non-productive random sequences derived from the nonproductive sequences of all individuals for each TCR chain, both for DP and CD8 cells (Marcou et al., 2018). These models were then used to estimate the Pgen values of sequences for each individual (Sethna et al., 2019). The overall distribution comparison between males and females was tested using two-sided Kolmogorov-Smirnov tests, with the D value and associated p-value indicated in each panel. Males are depicted in violet and females in orange.
Figure 6
Comparable T cell receptor (TCR) repertoire network structure based on CDR3 amino acid sequence similarity between males and females.
For each sample, 100 random subsamplings were performed on the minimum number of CDR3aa per cell subtype. Two CDR3aa are linked if they have a Levenshtein distance of one. (A) TRA Network structure of subsampled TCR repertoire of a male subject for double-positive (DP), CD8, CD4 Teff, and CD4 Treg single-positive (SP; from left to right). Each point on the graph represents a CDR3aa. (B–C) Comparison of the proportion of linked sequences (B) and network density (C) between male and female samples, in DP (n=20), CD8 (n=21), CD4 Teff (n=22), and CD4 Treg (n=14). Each point on the graph represents the median value from 100 subsampling iterations for each sample. Statistical analysis using the two-sided Wilcoxon test revealed no significant sex differences for these two metrics (p>0.05). Males are depicted in violet and females in orange.
Figure 7 with 5 supplements
Thymic TRB T cell receptor (TCR) sex-associated motifs.
Different structural motifs found differentially expressed between males and females in our dataset. We distinguish local motifs as distinct aa sequences, and global motifs as motif regions with one variable aa position maintaining a BLOSUM62 score of ≥0. (A) Number of male and female associated motifs by cell subset. (B) Euler diagram illustrating the distribution and overlap between all sex-associated motifs. The numbers indicate the number of motifs in overlap zones. (C–D) Validation of these sex-associated TRB CDR3aa motifs. The following heatmap illustrates the usage of all the TRB CDR3aa motifs in the external thymic pediatric dataset (Heikkilä et al., 2021; Mattila et al., 2023) (C) and those of TRB CD8 in the peripheral dataset (D). Sex and total CDR3aa number are depicted by sample. Males are depicted in violet and females in orange, then local motifs in blue-green and global motifs in magenta.
Figure 7—figure supplement 1
Double-positive (DP) TRB thymic sex-associated motifs in our dataset.
Different structural motifs in the CDR3 amino acid (CDR3aa) region found differentially expressed between males and females within our dataset. Local motifs refer to distinct amino acid sequences, whereas global motifs represent motif regions with a single variable amino acid position maintaining a BLOSUM62 score ≥0. The heatmap showcases the differential usage of TRB CDR3aa motifs between males and females in DP cells. Hierarchical clustering reveals clear segregation of individuals by sex, with most CDR3aa motifs being almost exclusively expressed in one sex while absent in the other. Sample attributes are visualized as follows: (1) Sex: males in violet, females in orange; (2) Age class: children (<18 years) in green, young adults (18–50 years) in blue, older adults (>50 years) in red; (3) sequencing platform: HiSeq 2500 samples in persimmon, NovaSeq 6000 samples in purple; (4) the total CDR3aa number. Motifs overexpressed in males are shown in violet, those overexpressed in females in orange; local motifs are indicated in blue-green and global motifs in magenta.
Figure 7—figure supplement 2
CD8 TRB thymic sex-associated motifs in our dataset.
Different structural motifs in the CDR3 amino acid (CDR3aa) region found differentially expressed between males and females within our dataset. Local motifs refer to distinct amino acid sequences, whereas global motifs represent motif regions with a single variable amino acid position maintaining a BLOSUM62 score ≥0. The heatmap showcases the differential usage of TRB CDR3aa motifs between males and females in CD8 cells. Hierarchical clustering reveals clear segregation of individuals by sex, with most CDR3aa motifs being almost exclusively expressed in one sex while absent in the other. Sample attributes are visualized as follows: (1) Sex: males in violet, females in orange; (2) Age class: children (<18 years) in green, young adults (18–50 years) in blue, older adults (>50 years) in red; (3) sequencing platform: HiSeq 2500 samples in persimmon, NovaSeq 6000 samples in purple; (4) the total CDR3aa number. Motifs overexpressed in males are shown in violet, those overexpressed in females in orange; local motifs are indicated in blue-green and global motifs in magenta.
Figure 7—figure supplement 3
CD4 Teff TRB thymic sex-associated motifs in our dataset.
Different structural motifs in the CDR3 amino acid (CDR3aa) region found differentially expressed between males and females within our dataset. Local motifs refer to distinct amino acid sequences, whereas global motifs represent motif regions with a single variable amino acid position maintaining a BLOSUM62 score ≥0. The heatmap showcases the differential usage of TRB CDR3aa motifs between males and females in CD4 Teff cells. Hierarchical clustering reveals clear segregation of individuals by sex, with most CDR3aa motifs being almost exclusively expressed in one sex while absent in the other. Sample attributes are visualized as follows: (1) Sex: males in violet, females in orange; (2) Age class: children (<18 years) in green, young adults (18–50 years) in blue, older adults (>50 years) in red; (3) sequencing platform: HiSeq 2500 samples in persimmon, NovaSeq 6000 samples in purple; (4) the total CDR3aa number. Motifs overexpressed in males are shown in violet, those overexpressed in females in orange; local motifs are indicated in blue-green and global motifs in magenta.
Figure 7—figure supplement 4
CD4 Treg TRB thymic sex-associated motifs in our dataset.
Different structural motifs in the CDR3 amino acid (CDR3aa) region found differentially expressed between males and females within our dataset. Local motifs refer to distinct amino acid sequences, whereas global motifs represent motif regions with a single variable amino acid position maintaining a BLOSUM62 score ≥0. The heatmap showcases the differential usage of TRB CDR3aa motifs between males and females in CD4 Treg cells. Hierarchical clustering reveals clear segregation of individuals by sex, with most CDR3aa motifs being almost exclusively expressed in one sex while absent in the other. Sample attributes are visualized as follows: (1) Sex: males in violet, females in orange; (2) Age class: children (<18 years) in green, young adults (18–50 years) in blue, older adults (>50 years) in red; (3) sequencing platform: HiSeq 2500 samples in persimmon, NovaSeq 6000 samples in purple; (4) the total CDR3aa number. Motifs overexpressed in males are shown in violet, those overexpressed in females in orange; local motifs are indicated in blue-green and global motifs in magenta.
Figure 7—figure supplement 5
TRB thymic sex-associated motifs are not differentially expressed between males and females in peripheral T cell receptor (TCR) repertoire.
Heatmap representing TRB CDR3aa motif usage differentially expressed between males and females in CD4 Teff (n=75; n_males = 38, n_females = 37) (A) and CD4 Treg (n=64; n_males = 29, n_females = 35) (B) cells in a peripheral TCR repertoire dataset. Sex and total CDR3aa number are depicted by samples. Males are depicted in violet and females in orange and local motif in blue-green and global motifs in magenta.
Figure 8 with 3 supplements
Sex-biased enrichment of T cell receptors (TCRs) specific for known antigens.
From a pooled and curated database, an exact match with this database infers the specificity of TRB CDR3aa of our thymic dataset. Many specificity groups are defined according to the nature of the antigen peptide targeted (bacteria, virus, autoimmune disease [AID], cancer, and self-peptide not associated with disease [human]). This analysis compares the distributions of the proportion of unique TRB CDR3aa sequences with a specific specificity (A) and their usage (B) between females and males across cell subtypes, using the log2 fold change of the median values (females over males), following each specificity group, in double-positive (DP; n=20), CD8 (n=21), CD4 Teff (n=22), and CD4 Treg (n=14) cells. These groups of specificity are additionally classified as microorganisms in the top section (bacteria in gold and virus in light blue) and self at the bottom (AID in magenta, cancer in red, human in blue-green). Polyspecific CDR3aa are defined here as CDR3aa capable of recognizing multiple antigens from different organisms (for no self-antigens) or from different specificity groups (e.g. categorized microorganisms, categorized self-antigens, allergens…). The proportion of polyspecific CDR3aa among antigen-specific sequences (C) and their usage (D) is compared between males and females, in double-positive (DP; n=20), CD8 (n=21), CD4 Teff (n=22), and CD4 Treg (n=14) cells. Asterisks indicate significant differences between males and females based on the two-sided Wilcoxon test p-value (*: p<0.05, **: p<0.01). Males are depicted in violet and females in orange.
Figure 8—figure supplement 1
Enrichment of cancer-associated, unmodified self-peptide specific, and polyspecific T cell receptor (TCR), in our thymic dataset compared to the initial curated database.
(A) Representation of the specificity groups in the pooled and curated database. (B) Proportion of unique CDR3aa TRB in the specific TCR repertoire for each sample for autoimmune disease (AID), bacteria, cancer, human, and virus groups is compared between males and females in double-positive (DP; n=20), single-positive (SP) CD8 (n=21), SP CD4 Teff (n=22), and SP CD4 Treg (n=14). (C) Proportion of polyspecific CDR3aa in the specific repertoire of samples is compared between males and females in DP (n=20), SP CD8 (n=21), SP CD4 Teff (n=22), and SP CD4 Treg (n=14). Asterisks indicate significant differences males or females with the pooled and curated database based on the two-sided Wilcoxon test p-value (*: p<0.05, **: p<0.01). Males are depicted in violet and females in orange.
Figure 8—figure supplement 2
Enrichment of autoimmune disease (AID)-associated T cell receptor (TCR) and bacteria-targeted TCR in CD8 female cells.
CDR3aa usage of CDR3aa with a specificity in autoimmunity disease, bacteria, cancer, human, and virus groups, in double-positive (DP; n=20), single-positive (SP) CD8 (n=21), SP CD4 Teff (n=22), and SP CD4 Treg (n=14). Asterisks indicate significant differences between males and females based on the two-sided Wilcoxon test p-value (*: p<0.05, **: p<0.01). Males are depicted in violet and females in orange.
Figure 8—figure supplement 3
No association between age and thymic usage of CDR3 sequences specific to autoimmunity or bacterial antigens.
Linear regression models assessing the effect of donor age on the cumulative usage (%) of TRB CDR3aa sequences with known specificity to self-antigens associated with autoimmunity (AID, left panels) or bacterial antigens (right panels), in CD8 single-positive (SP; top, n=21) and CD4 Treg SP (bottom, n=14) thymic subsets. Each dot represents one donor. The shaded area indicates the 95% confidence interval around the regression line. Reported slope, 95% confidence interval, R², and p-values show no significant effect of age.
Tables
Author response table 1
| Disease | Antigens | Epitopes | TRB CDR3 sequences |
|---|---|---|---|
| Type 1 diabetes | 12 | 42 | 83 |
| Celiac disease | 2 | 6 | 102 |
Additional files
-
Supplementary file 1
Minimal number of CDR3aa by cell subset.
- https://cdn.elifesciences.org/articles/109041/elife-109041-supp1-v1.xlsx
-
Supplementary file 2
HLA typing of the study cohort.
HLA class I (A, B, C) and class II (DP, DQA, DQB, DR) alleles for each subject included in the analysis. Due to the clinical nature of the data, typing is provided at low resolution (first-field/serological equivalent). "Not collected" indicates loci that were not documented in the donor’s clinical file at the time of organ procurement.
- https://cdn.elifesciences.org/articles/109041/elife-109041-supp2-v1.xlsx
-
Supplementary file 3
Detailed sample-level overview of the thymic cohort.
For each donor and thymocyte subset, the table reports: donor ID, sex, age at sampling, thymocyte subset (DP, CD8 SP, CD4 Teff SP, CD4 Treg SP), number of cells sorted, and for each TCR chain (TRA and TRB), the total number of CDR3 amino acid sequences retained after preprocessing (quality filtering, length filtering, and MiXCR-based error correction) and the corresponding number of unique clonotypes. This table provides the metadata referred to in the Samples, Thymocyte isolation and RNA extraction, and Data processing sections, and complements the global cohort description shown in Figure 1.
- https://cdn.elifesciences.org/articles/109041/elife-109041-supp3-v1.xlsx
-
Supplementary file 4
Annotation of self-antigens and their associated disease contexts.
This table lists the self-antigens included in the curated TCR specificity database, together with their associated diseases, when applicable, and their assignment to the corresponding specificity class used in the analyses. MAA = Melanoma associated antigen, TAA = Tumor associated antigen, DT1=type 1 diabetes, CeD = celiac disease.
- https://cdn.elifesciences.org/articles/109041/elife-109041-supp4-v1.xlsx
-
MDAR checklist
- https://cdn.elifesciences.org/articles/109041/elife-109041-mdarchecklist1-v1.docx
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Thymic selection of the T cell receptor repertoire is biased toward autoimmunity in females
eLife 14:RP109041.
https://doi.org/10.7554/eLife.109041.4
