Additional file 2 of Genetic and environmental determinants of human TCR repertoire diversity

Additional file 2: Fig. S1. Cohort assembly and filtering. Flowchart depicting the studies in which TCR sequencing and HLA genotyping were performed, and steps used to select individuals for analysis. Fig. S2. Association of HLA-I and II polymorphism with TCR repertoire Shannon entropy in CMV- individuals. a Association of HLA-I polymorphism with increased Shannon entropy in CMV- individuals; HLA-I P = 0.008, estimate = 0.33; age P = 0.0005, estimate = -0.02. P-values are from a linear model incorporating the number of unique HLA-I alleles and age. b Association of full HLA-I heterozygosity (6 unique HLA-I alleles) with increased Shannon entropy; full HLA-I heterozygosity P = 0.01, estimate = 0.46; age P = 0.0005, estimate = -0.02. P-values are from a linear model incorporating a binary variable encoding full HLA-I heterozygosity, and age as a continuous variable. c No association between HLA-II polymorphism and Shannon entropy; HLA-II P = 0.24, estimate = 0.1; age P = 0.002, estimate = -0.02. P-values are from a linear model incorporating number of unique HLA-II alleles and age. d No association between full HLA-II heterozygosity (10 unique HLA-II alleles) and Shannon entropy; full HLA-II heterozygosity P = 0.65, estimate = -0.10; age P = 0.002, estimate = -0.02. P-values are from a linear model incorporating a binary variable encoding full HLA-II heterozygosity, and age as a continuous variable. Fig. S3. Neither HLA-I nor HLA-II polymorphism is associated with TCR repertoire diversity in CMV+ individuals. a No association between HLA-I polymorphism and number of unique CDR3s; HLA-I P = 0.70., estimate = -3318.9; age P = 0.41, estimate = -378.5. P-values are from a linear model incorporating the number of unique HLA-I alleles and age. b No association between HLA-I polymorphism and Shannon entropy; HLA-I P = 0.80, estimate = -0.04; age P = 0.03, estimate = -0.02. P-values are from a linear model incorporating the number of unique HLA-I alleles and age. c No association between HLA-II polymorphism and number of unique CDR3s; HLA-II P = 0.45, estimate = 3918.7; age P = 0.37, estimate = -414.3. P-values are from a linear model incorporating the number of unique HLA-II alleles and age. d No association between HLA-II polymorphism and Shannon entropy; HLA-II P = 0.70, estimate = 0.04; age P = 0.03, estimate = -0.02. P-values are from a linear model incorporating the number of unique HLA-II alleles and age. Fig. S4. Age and HLA-I polymorphism independently affect TCR repertoire Shannon entropy in CMV- individuals. a Association between age and Shannon entropy in CMV+ individuals; age P = 0.03, estimate = -0.02; HLA-I P = 0.80, estimate = -0.04. P-values are from a linear model incorporating age and number of unique HLA-I alleles. b Association between age and Shannon entropy in CMV- individuals; age P = 0.0005, estimate = -0.02; HLA-I P = 0.008, estimate = 0.33. P-values are from a linear model incorporating age and number of unique HLAI alleles. c AIC analysis of three linear models with Shannon entropy as the dependent variable, and either age alone, number of unique HLA-I alleles alone, or both as the independent variables. All models were fit in CMV- individuals. Data show that the best model that explains the observed TCR repertoire diversity across these individuals is the one with both age and number of unique HLA-I alleles (AIC = 575.11). d AIC analysis of three linear models with Shannon entropy as the dependent variable, and either age alone, number of unique HLA-I alleles alone, or both as the independent variables. All models were fit in CMV+ individuals. Data show that number of unique HLA-I alleles adds no effect beyond the effect of age alone. Fig. S5. Mean HLA evolutionary divergence is associated with increased TCR repertoire diversity in CMV- individuals. a Association of high mean HED (Mean HED >= median) with increased Shannon entropy in CMV- individuals; high mean HED P = 0.03, estimate = 0.37; age P = 0.001, estimate = -0.02. P-values are from a linear model incorporating corresponds a binary variable encoding high mean HED, and age as a continuous variable. b AIC analysis of three linear models with number of unique CDR3s as the dependent variable, and either age alone, high mean HED alone, or both as the independent variables. All models were fit in CMV- individuals. Data show that the best model that explains the observed TCR repertoire diversity across these individuals is the one with both age and high mean HED (AIC = 577.6). c No association of high mean HED (Mean HED >= median) with Shannon entropy in CMV+individuals; high mean HED P = 0.87, estimate = -0.04; age P = 0.03, estimate = -0.02. P-values are from a linear model incorporating a binary variable encoding high mean HED, and age as a continuous variable. d AIC analysis of three linear models with Shannon entropy as the dependent variable, and either age alone, high mean HED alone, or both as the independent variables. All models were fit in CMV+ individuals. Data show that high mean HED adds no effect beyond the effect of age alone. Fig. S6. Association of HLA-I diversity with TCR repertoire diversity measured using the normalized Shannon-Wiener index in CMV- individuals. a Association of HLA-I polymorphism with normalized Shannon-Wiener index in CMV- individuals; HLA-I P = 0.07, estimate = 0.006; age P = 0.0002, estimate -0.0007. P-values are from a linear model incorporating number of unique HLA-I alleles and age. b AIC analysis of three linear models with TCR normalized Shannon-Wiener index as the dependent variable, and either age alone, number of unique HLA-I alleles alone, or both as the independent variables. Data show that the best model that explains the observed TCR repertoire diversity across these individuals is the one with both age and number of unique HLA-I alleles (AIC = -727.0241). c Association of high mean HED (Mean HED >= median) with increased normalized Shannon-Wiener index in CMV- individuals; high mean HED P = 0.01, estimate = 0.01; age P = 0.01, estimate = -0.0006. P-values are from a linear model incorporating corresponds a binary variable encoding high mean HED, and age as a continuous variable. d AIC analysis of three linear models with TCR normalized Shannon-Wiener index as the dependent variable, and either age alone, high mean HED alone, or both as the independent variables. Data show that the best model that explains the observed TCR repertoire diversity across these individuals is the one with both age and high mean HED (AIC = -729.9617). Fig. S7. Association of sex with TCR repertoire diversity in CMV- individuals. a Linear model testing the association of sex with number of unique CDR3s in CMV- individuals. b Linear model testing the association of sex with Shannon entropy in CMV- individuals.