Genetic Influence on Peripheral Blood T Lymphocyte Levels

MA Hall1, KR Ahmadi2, P Norman3, H Snieder2, AJ MacGregor2, RW Vaughan3, TD Spector2 and JS Lanchbury1 1Molecular Immunogenetics Unit, Department of Rheumatology, Division of Medicine, 5th Floor Thomas Guy House, Guy’s, King’s and St Thomas’ Hospitals School of Medicine, King’s College, Guy’s Hospital Campus, London SE1 9RT, UK; 2Twin Research and Genetic Epidemiology Unit, St Thomas’ Hospital, London SE1, UK; 3South Thames Tissue Typing, 3rd Floor, New Guy’s House, Guy’s Hospital, London SE1 9RT, UK

T lymphocytes are a major component of the adaptive immune system. CD4 positive T cell subpopulations regulate B cell and macrophage effector function while CD8 positive T cells are largely responsible for anti-viral cytotoxic activity. The degree of natural variation in the levels and ratios of the various T cell subpopulations is a possible risk factor for the development of autommune disease, infectious disease and cancer. There is some evidence from studies of inbred strains of mice and humans which suggests that variation in T cell subpopulations is genetically influenced. However, family studies alone cannot distinguish between common environmental and shared genetic influences and provide less robust estimates of the heritability than twin studies. To comprehensively examine genetic influences on a selection of important T cell phenotypes, we investigated variation in levels of total lymphocytes, CD3+, CD4+, CD8+, CD3+CD4+, CD3+CD8+ lymphocytes and in CD4:CD8 ratio as a proportion of lymphocytes and of T cells using the classical twin model approach. Healthy female twin pairs were sampled from the St. Thomas’ UK Adult Twin Registry. A maximum of 103 monozygotic (MZ) and 186 dizygotic (DZ) twins aged 18–80 years participated in the study. Whole blood samples were analysed for T cell subsets by flow cytometry. The relative genetic contribution to these phenotypes was estimated using a variance components model-fitting approach. Heritability estimates were calculated of 65% for CD4:CD8 T cell and lymphocyte ratios, around 50% for absolute lymphocyte, CD3+ and CD4+ counts, and 56% for CD8+ numbers. Unique (rather than shared) familial environment explains the remainder of the variance. Genetic factors have a major influence on the variation in peripheral T cell subset numbers. Polymorphism dictating such variation should be taken into account when assessing risk factors for T cell immune-mediated disease with a genetic background. Genes and Immunity (2000) 1, 423–427.

Keywords: T cells; genetics; immune system; heritibility

Introduction ulatory activities. Studies of the impact of genetic variation on the functioning of the immune system have up The human immune response is a key physiological sys- to now largely concentrated on qualitative issues of tem upon which mammalian survival depends and response and non-response. Experimental manipulation would be expected to be a crucial target for the action of of the murine genome using ‘knock out’ and transgenic natural selection. T cells are a fundamental component of technologies has enabled insight to be gained in systems the adaptive arm of this response and are involved in which are driven to phenotypic extremes. The effects of defence against pathogens including bacteria, viruses, CD4 and CD8 phenotypes are examples of the former2,3 1 fungi, protozoa, and multicellular parasites. CD4 posi- while useful insights into T cell development have been tive T cell subpopulations regulate B cell and macro- gained from rearranged T cell receptor transgenics as phage effector function while CD8 positive T cells are examples of the latter.4 responsible for anti-viral cytotoxic and some immunoreg- In contrast, the existence of genetically determined natural variation in levels of T cell subpopulations in mice and humans has received little attention. Studies in Correspondence: Dr MA Hall, Molecular Immunogenetics Unit, Depart- mice first showed that T cell subset representation was ment of Rheumatology, Division of Medicine, 5th Floor Thomas Guy 5 House, Guy’s, King’s and St Thomas’ Hospitals Schools of Medicine, strain dependent and therefore under genetic control. 6 King’s College, Guy’s Hospital Campus, London SE1 9RT, UK. E-mail: Amadori et al more recently studied the genetics of CD4 margaret.a.hallȰkcl.ac.uk and CD8 T cell ratios in healthy human nuclear families We are grateful to the Special Trustees of St Thomas’ Hospital and and concluded that CD4:CD8 T cell ratio was under gen- Gemini Genomics Limited for support in initiating this research. etic control. The degree of natural variation in the levels The Twin Research Unit also receives support from the Arthritis and ratios of the various T cell subpopulations may con- Research Campaign, the Wellcome Trust and the Chronic Disease Research Foundation. The experiments described in this paper com- tribute to part of the genetic risk for the development ply with the laws of the United Kingdom. of autommune disease, infectious disease and cancer. To Received and revised 17 June 2000; accepted 21 June 2000 explore this hypothesis we have measured a number of Genetic influence on peripheral blood T lymphocyte levels MA Hall et al 424 fundamental T cell parameters in humans in a large bution to the total variance of the T cell (1%, P Ͻ 0.001) cohort of monozygotic (MZ) and dizygotic (DZ) twins. and lymphocyte (2%, P Ͻ 0.001) CD4:CD8 ratios. From a genetic perspective, MZ twins share all their genes while DZ twins are similar to ordinary siblings in Discussion that they share on average 50% of their segregating genes. However, unlike sibs, use of DZ twins has the vir- In this study, we investigated the extent to which genetic tue of removing any confounding effects of age and con- and environmental influences control the absolute levels trols for differences in pre- and postnatal circumstances and cellular ratios of various T lymphocyte subpopula- of gestation and rearing. In this study we establish the tions using the classical twin model approach. The study degree of genetic and environmental control over a num- population was entirely female and the age range ber of peripheral T cell parameters using the classical covered the adult period from 18 to 80 years. Our data twin model approach. indicate that quantitative variation in absolute levels of peripheral blood T lymphocyte levels is under significant Results genetic control. Heritability, the proportion of total variance due to additive genetic factors, ranged from 45%, The age range of the twins under study was 18–80 years. for the CD3+ T cells, to 65%, for the CD4:CD8 T cell ratio. The mean age of the MZ twins was 47.7 ± 15.6 years and The variance due to unique environmental differences, for the DZ twins it was 50.0 ± 13.0 years. Table 1 summar- ranged from 33%, for CD4:CD8 lymphocyte ratios, to 52% ises the number of complete twin pairs (N), the mean for lymphocyte and CD3+ T cell levels. Age was shown values (±s.d.), and the intra-class correlations (ICC) of the to have a small but significant effect on absolute levels various immune markers for MZ and DZ twins. of CD3+, CD8+ and CD3+ CD8+ T cells, as well as the The mean values and standard deviations for the MZ CD4:CD8 T cell and lymphocyte ratios and explained up and DZ twin populations were comparable, although to 4% of the observed variance. some means were slightly higher in the MZ group. On This is the first classical twin study of an adult popu- testing for the differences in the means between the MZ lation which investigates the relative influences of genes and DZ group we showed that although the differences and environment on the absolute levels of peripheral were significant for lymphocytes, CD3+, CD4+ and blood T cells. Previous studies have either examined CD3+CD4+, this did not significantly influence the herita- small nuclear families6 or twins at age 12.7 Amadori et al6 bility estimates. studied subjects across a wide age range while Evans et For both the lymphocyte and CD3+ T cell populations, al7 reported a twin study of broadly similar T cell pheno- the intraclass correlations suggest a genetic effect as the types and sample size. Our results agree with these stud- ICC for the MZ twins is greater than the DZ intraclass ies in showing that levels of T cell subpopulations and correlations. ratios are under significant genetic control. Interestingly, The results of the univariate model fitting are shown our heritability estimates are somewhat lower for equiv- in Table 2. The AE model (ascribing variance due to addi- alent T cell phenotypes than those of Evans et al7 obtained tive genetic and unique environment) represented the from a mixed sex group of 12-year-old twins. A possible best-fitting and most parsimonious model for lympho- explanation is that genetic control relaxes with increasing cyte, CD3+, CD8+, CD3+CD4+ and CD3+CD8+ populations age and the influence of unique environment becomes (C and D could be dropped without any significant more important with age through the cumulative change in ␹2). Age was shown to contribute significantly exposure to a range of environmnetal and self antigens. to the total variance of CD3+ (2%, P Ͻ 0.00001), In agreement with the family study, age had a small CD3+CD8+ (3%) cells and CD8+ (4%, P Ͻ 0.00001). The but significant influence on not only the T cell (age vari- absolute numbers of CD3+ and CD3+CD8+ lymphocytes ance ෂ1%) and T lymphocyte (ෂ2%) ratios but also on all decreased with age. Univariate analysis for CD3+ (ෂ2%), CD3+CD8+ T cells (ෂ3%), as well as CD8+ CD3+CD4+:CD3+CD8+ T cell and CD4+:CD8+ lymphocyte lymphocytes (ෂ4%). Our findings broadly reflect the ratios also suggests AE to be the best fitting model. Test- gross changes in T cell levels which have been observed ing the effect of age showed a small but significant contri- in cross-sectional and longitudinal studies in mice and

Table 1 Number of complete pairs (n), means, standard deviation (s.d.) and intraclass correlations (ICC) of various immune phenotypes for MZ and DZ twins. The mean and standard deviations of immune phenotypes are presented as absolute values ×106

Immune phenotype MZ DZ Mean (s.d.) Mean (s.d.)

n Absolute ICC n Absolute ICC

Age (range 18–80 yrs) — 47.7 (15.6) — — 50.0 (13.0) — Lymphocyte count 92 1.79 (1.07) 0.48 183 1.70 (0.96) 0.17 CD3+ 92 1.30 (0.73) 0.47 182 1.21 (0.64) 0.21 CD4+ 89 0.86 (0.39) 0.53 182 0.77 (0.37) 0.22 CD8+ 89 0.45 (0.21) 0.53 182 0.42 (0.19) 0.38 CD4:CD8 T-cells 97 2.32 (1.02) 0.62 183 2.14 (0.85) 0.43 CD4+CD3+ 78 0.89 (0.45) 0.52 150 0.80 (0.38) 0.23 CD8+CD3+ 78 0.38 (0.18) 0.48 150 0.37 (0.15) 0.39 CD4:CD8 Lymphocytes 103 1.88 (0.88) 0.62 186 1.79 (0.76) 0.45

Genes and Immunity Genetic inﬂuence on peripheral blood T lymphocyte levels MA Hall et al 425 Table 2 Estimates of additive genetic (h2), unique environmental (e2) and age (age2) variance components. Values in brackets represent the 95% conﬁdence interval

Immune phenotype h2 e2 age2

Lymphocyte count 0.48 (0.3–0.63) 0.52 (0.37–0.69) — CD3+ 0.45 (0.28–0.59) 0.52 (0.37–0.69) 0.02 (0.001–0.05) CD4+ 0.56 (0.41–0.68) 0.43 (0.31–0.58) — CD8+ 0.57 (0.43–0.67) 0.39 (0.29–0.52) 0.04 (0.008–0.08) CD4:CD8 lymphocytes 0.65 (0.55–0.73) 0.33 (0.25–0.43) 0.02 (0.0006–0.05) CD4+CD3+ 0.49 (0.32–0.63) 0.50 (0.36–0.67) — CD8+CD3+ 0.56 (0.41–0.67) 0.40 (0.29–0.55) 0.03 (0.008–0.08) CD4:CD8 T cells 0.64 (0.53–0.73) 0.34 (0.26–0.45) 0.01 (0.0006–0.04)

humans. A study on mice showed an age-dependent with a polygenic component and random environmental depletion of absolute CD8+ cell counts.8 However, they effect. Taken together with the mouse and rat data, the also reported a reduction in CD4 levels with age, which prospect exists of identifying candidate genes responsible we did not find in our study of humans. The related mea- for the major proportion of additive genetic variation. sure, CD4:CD8 ratio, was also shown to drop steadily as Genes within the HLA, T cell receptor alpha and T cell a function of age.8 Shifts in the CD4:CD8 ratio were receptor beta loci are attractive targets. The conservation shown to be due to decreases in the numbers of CD8+ T of genetically controlled variation in peripheral T cell lev- cells which were less pronounced than the decreases of els and CD4:CD8 ratios in humans, mice and rats CD4+ T cells.6,8 Cross-sectional studies of human periph- strongly implies a role for natural selection in the mainte- eral blood T cell populations have shown similar results nance of such variation. The functional relevance of this to the mouse studies. Absolute values for most lympho- variation has yet to be established but may relate to dif- cyte subpopulations have been shown to differ substan- ferential ability to respond to infectious pathogens. This tially with age. Absolute numbers of lymphocytes, CD3+ variation may also represent a risk factor for the develop- T cells, CD4+ T cells and CD8+ T cells and CD4:CD8 ratios ment of immune-mediated diseases with a genetic all declined with age.6,9–12 component, including autoimmune diseases. It has been previously reported that the CD4:CD8 ratio in inbred mice, strains of rat, and humans is under genetic control. In the mouse the ratio was concluded to be Conclusions under the control of a single dominant gene and signifi- These results show that additive genetic factors have a cantly influenced by age.5 Another study, using common major influence on the control of peripheral T cell subset laboratory mouse strains, indicated that the distinct rela- numbers. Detection of polymorphism dictating such vari- tive sizes of the CD4 and CD8 compartments are determ- ation should be pursued and taken into account when ined by genetic variation in the process of thymic lineage assessing risk factors for T cell immune-mediated disease commitment rather than by TCR-mediated positive or with a genetic background. negative selection.13 Like mice, different rat strains have been shown to differ in their peripheral CD4:CD8 T cell ratios. A study by Damoiseaux et al14 used the Lewis Materials and methods (LEW) and Brown Norway (BN) rat strains, their respect- − × ive MHC congenic, as well as (LEW BN) F2 hybrids Study population to decipher control of the peripheral CD4:CD8 T cell A maximum of 103 monozygotic (MZ) and 186 dizygotic ratios. Their results revealed that the development of the (DZ) twins aged 18–80 years from the St Thomas’ UK distinct peripheral CD4:CD8 T cell ratio between both adult twin registry participated in the study. The twin strains originates in the thymus and is determined during pairs were ascertained from the general population selection on radio-resistant stromal cells. Furthermore, through national media campaign in the United King- the difference was strictly correlated with the MHC dom. Informed consent was obtained from all subjects haplotype and was a result of a reduction in the absolute although the participants were unaware of the specific number of CD8+ and not CD4+ T cells in BN rats.14 These hypothesis tested. Zygosity was determined by standard- results agree with those from the mouse in the sense that ised questionnaire15 and confirmed by a combination of the thymus plays a crucial role in the establishment of DNA analyses which included full genome scans, the ABI the peripheral CD4:CD8 T cell ratio. However, there is a FESTR kit and the United Kingdom Forensic Science Ser- clear discrepancy between these two species. The vices QUAD system. Due to incomplete lysis of red cells, CD4:CD8 ratio in the mouse develops independently of the number of subjects studied for each phenotype varies. MHC haplotype, whereas in the rat the MHC genes were shown to play a crucial role in determining the peripheral Flow cytometry CD4:CD8 T cell ratio. Fasting venous blood samples (5 ml) were drawn into In humans, mapping of loci responsible for quantitat- glass vacutainers anticoagulated with 50 ␮l sterile pre- ive variation in T lymphocyte populations and ratios has servative-free heparin and analysed within 2–4 h. yet to be undertaken. However, the segregation analysis Samples were taken between 10.00 and 11.30 am to mini- by Amadori et al6 suggested that variation in CD4: CD8 mise any circadian fluctuation in leukocyte population ratio was controlled by a major recessive locus together levels. Flow cytometric analysis was performed using a

Genes and Immunity Genetic influence on peripheral blood T lymphocyte levels MA Hall et al 426 Coulter EPICS-XL I set up for three-colour detection and greater similarity between monozygotic compared with controlled with XL system II software. The cytometer was dizygotic twins reflects genetic influences. A higher MZ calibrated daily using ImmunoCheck beads (Beckman than DZ intraclass correlation (ICC) provides a first Coulter, High Wycombe, UK) in order to maintain a half impression of the magnitude of genetic influence.17 peak coefficient of variation (HPCV) of less than 2 for Details of model fitting to twin data have been each channel. described elsewhere.18,19 In short, the technique is based Cells were stained directly and prepared from whole on the comparison of the covariances (or correlations) in blood aliquots using fluorescent-conjugated monoclonal monozygotic and dizygotic twin pairs and allows separ- antibodies and the ImmunoPrep reagent system ation of the observed phenotypic variance into additive (Beckman Coulter). All monoclonal antibodies used were (A) or dominant (D) genetic components and common directly conjugated mouse anti human and were: IgG1- (C) or unique (E) environmental components. E also FITC 679.1Mc7, IgG1-PE 679.1Mc7, IgG1-Cy5 679.1Mc7, contains measurement error. Dividing each of these CD3-Cy5 UCHT1 (IgG1), CD45-Cy5 J.33 (IgG1) (All components by the total variance yields the different Immunotech/Coulter, High Wycombe, UK) and CD4- standardised components of variance, for example the FITC SK3 (IgG1), CD8-PE SK1 (IgG1) (Becton Dickinson, (narrow sense) heritability (h2) can be defined as the pro- Oxford, UK) portion of the total variance attributable to additive gen- Populations were gated as follows. A 2D scatter plot etic variation (also termed a2). By incorporating age into was generated for each parameter combination required. the model, the influence of age on the phenotype can also Lymphocytes were gated according to their forward and be quantified.20 A path diagram of the twin model is side scatter properties and 5000 events recorded. Data for shown in Figure 1. the CD4:CD8 ratio of lymphocytes was recorded for this The significance of age and components A, C and D region. In order to obtain a ratio of CD4:CD8 for T cells, was assessed by testing the deterioration in model fit a further analysis gate was taken around cells observed after each component was dropped from the full model. as CD3+. Positive thresholds were defined by quadrant Standard hierarchic ␹2 tests were used to select the best regions which enclosed 99% of the negative control popu- fitting models.18 Log transformation was carried out on lation on each axis. FlowCount (Coulter) beads were used the phenotype data prior to analysis to obtain normal to estimate the concentration of lymphocytes in a given distribution. This was tested for using a Kolmorogov– sample and the absolute number for each lymphocyte Smirnov test for normality. subset was calculated from this. Statistical software: Data handling and preliminary analy- Analytical approach ses were done with STATA.21 Genetic modelling was carried out with Mx, a computer program specifically Quantitative genetic model fitting of twin data: Twin meth- designed for the analysis of twin and family data.18,22 odology makes use of the fact that MZ twins share ident- ical genotypes, whereas DZ twins share only 50% of their Power: Given the sample size of our study there is 80% genes. It is assumed that both types of twins share their power (P = 0.05) to detect genetic influences for traits common family environment to the same extent16 so any with heritabilities as low as 15%. However, a higher num-

Figure 1 Observed variables for twin 1 and 2 are shown in squares. Unmeasured (latent) variables are shown in circles. A single arrow indicates a direct influence of one variable on another, its value represented by a path coefficient. Double headed arrows indicate a correlation between the two variables. The path model above represents the influences of additive genetic (A), shared environment (C),

speciﬁc environment (E), and age effects on phenotypes (P) of pairs of twins. The genetic correlation (rg) between variable A in twin1 and twin2 is ﬁxed at 1.0 for MZ and 0.5 for DZ twins. Correlation between variable C (rc) is 1.0 for MZ and DZ twins. Squaring the factor loading yields the variance explained by the various components. a, c, e and s = regression of A, C, E and age on phenotype, v = standard deviation of age.

Genes and Immunity Genetic influence on peripheral blood T lymphocyte levels MA Hall et al 427 ber of twins will be required to detect small non-additive human blood lymphocyte subpopulations. II. Varying kinetics genetic (D) and common family environmental (C) influ- of percentage and absolute count measurements. Clin Immunol ences. Immunopathol 1994; 70: 152–158. 10 Erkeller-Yuksel FM, Deneys V, Yuksel B et al. Age-related changes in human blood lymphocyte subpopulations. J Pediatr Acknowledgements 1992; 120: 216–222. 11 Erkeller-Yusel F, Hulstaart F, Hannet I, Isenberg D, Lydyard P. MAH is an Arthritis Research Campaign Postdoctoral Lymphocyte subsets in a large cohort of patients with systemic Research Fellow. AM is an Arthritis Research Campaign lupus erythematosus. Lupus 1993; 2: 227–231. Senior Fellow. HS is a fellow of the British Heart Foun- 12 Hannet I, Erkeller-Yuksel F, Lydyard P, Deneys V, DeBruyere dation. M. Developmental and maturational changes in human blood lymphocyte subpopulations [see comments]. Immunol Today 1992; 13: 215, 218. References 13 van Meerwijk JP, Bianchi T, Marguerat S, MacDonald HR. Thy- mic lineage commitment rather than selection causes genetic 1 Janeway CA, Travers P. Immunobiology. The Immune System in variations in size of CD4 and CD8 compartments. J Immunol Health and Disease. 2nd edn. Churchill Livingstone: London, 1998; 160: 3649–3654. 1996. 14 Damoiseaux JG, Cautain B, Bernard I et al. A dominant role for 2 Yeung RS, Penninger J, Mak TW. Genetically modified animals the thymus and MHC genes in determining the peripheral and immunodeficiency [published erratum appears in Curr Opin CD4/CD8 T cell ratio in the rat. J Immunol 1999; 163: 2983–2989. Immunol 1993; 5: 974]. Curr Opin Immunol 1993; 5: 585–594. 15 Goldsmith HH. A zygosity questionnaire for young twins: a 3 Rahemtulla A, Kundig TM, Narendran A et al. Class II major research note. Behav Genet 1991; 21: 257–269. histocompatibility complex-restricted T cell function in CD4- 16 Kyvik KO. Generalisability and assumptions of twin studies. In: deficient mice. Eur J Immunol 1994; 24: 2213–2218. Spector TD, Snieder H, MacGregor AJ, eds. Advances in Twin 4 Robey EA, Fowlkes BJ, Gordon JW et al. Thymic selection in and Sib-Pair Analysis. Greenwich Medical Media: London, 1999, CD8 transgenic mice supports an instructive model for commit- pp 67–77. ment to a CD4 or CD8 lineage. Cell 1991; 64: 99–107. 17 Falconer DS. Introduction to Quantitative Genetics. 3rd edn. Long- 5 Kraal G, Weissman IL, Butcher EC. Genetic control of T-cell sub- man: Harlow, England, 1989. set representation in inbred mice. Immunogenetics 1983; 18: 18 Neale MC, Cardon LR. Methodologies for Genetic Studies of Twins 585–592. and Families. Kluwer Academic Publishers: Dordrecht, 1992. 6 Amadori A, Zamarchi R, De Silvestro G et al. Genetic control 19 Spector TD, Snieder H, MacGregor AJ. Advances in Twin and Sib- of the CD4/CD8 T-cell ratio in humans. Nature Med 1995; 1: Pair Analysis. Greenwich Medical Media: London, 1999. 1279–1283. 20 Snieder H. Path analysis of age related disease traits. In: Spector 7 Evans DM, Frazer IH, Martin NG. Genetic and environmental TD, Snieder H, MacGregor AJ, eds. Advances in Twin and Sib-Pair causes of variation in basal levels of blood cells. Twin Res 1999; Analysis. Greenwich Medical Media: London, 1999, pp 119–130. 2: 250–257. 21 StataCorp. STATA Statistical Software: Release 5.0. StataCorp: Col- 8 Callahan JE, Kappler JW, Marrack P. Unexpected expansions of lege Station, Tex, 1997. CD8-bearing cells in old mice. J Immunol 1993; 151: 6657–6669. 22 Neale MC. Mx: Statistical Modelling. 2nd edn. Medical College 9 Hulstaert F, Hannet I, Deneys V et al. Age-related changes in of Virginia: Richmond, Va., 1997.

Genes and Immunity