Identical twins discordant for multiple sclerosis have a shift in their T-cell receptor repertoires

Abstract

CD4 T-cells have an important role in the autoimmune response in multiple sclerosis (MS). We investigate the possibility that a shift occurs in the T-cell receptor (TR) repertoire of identical twins discordant for MS. We compare the CDR3 spectratype distributions of 24 different TR V beta (TRBV) segments in naïve CD4 T-cells from discordant MS twins and from healthy identical twins. We also compare the CDR3 spectratype distributions in unrelated healthy pairs, formed by combining members of different healthy twins, with the CDR3 spectratype distributions in unrelated pairs of MS patients and in unrelated pairs of their apparently healthy cotwins, formed by combining members of different discordant twins. We use the correlation coefficient (r-value) as a measure of similarity of CDR3 spectratypes in each pair, and we test for the significance of the difference between r-values from the different pairs. We observe that the r-value for the CDR3 spectratype distributions among discordant twins differs significantly from the corresponding r-value for the healthy twins for two TRBV segments. Further, the r-values, for both the unrelated MS patient pairs and the unrelated pairs of their apparently healthy cotwins, differ significantly from the r-values for healthy unrelated pairs of individuals. We conclude that both the MS patients and their apparently healthy cotwins have shifts in their CDR3 repertoires. Because we study naïve CD4 T-cells, we postulate that CDR3 repertoire shifts precede MS and predispose to MS, but are unlikely to be sufficient to cause MS.

INTRODUCTION

Multiple sclerosis (MS) is a complex, possibly heterogeneous disease of the central nervous system (CNS) of uncertain cause [1]. An antimyelin autoimmune response occurs in MS patients and in its animal models. CD4 T-cells have a central role in initiating this autoreactivity [2,3]. Analysis of the CD4 T-cell response suggests that T-cell receptor (TR) usage is oligoclonal and restricted in some individual MS patients, but is probably diverse between patients [4]. In one study, members of healthy identical twins and members of twins concordant for MS selected similar TR segments to recognize both a myelin antigen and tetanus toxoid. By contrast, the members of discordant twins selected different TR segments to recognize the same antigens [5]. However, the investigators did not investigate if the altered TR segment usage was due to external factors or the disease process in MS [5]. Also, they did not investigate, if the altered TR segment usage precedes the onset of MS [6].

We reported previously that the correlation coefficient (r-value), obtained as a measure of the overall similarity of the TR J beta (TRBJ) segments among identical twins discordant for MS, differs significantly from the corresponding r-value for healthy identical twins. Thus, we concluded that the discordant twins have a shift in the overall diversity of their TRBJ segments, i.e. of their TRBJ repertoires [7]. However, in that study, we did not explore whether this shift occurred only in the MS patients, or in both MS patients and their apparently healthy cotwins. Further, TRBJ segment usage varies with the expression of different TRBV segments [8]. Consequently, the TRBJ repertoire shifts could be secondary to altered TRBV segment usage by the MS patients of the discordant MS twins. A further issue is that the diversity of TRB chains is concentrated in the complementarity-determining regions 3 (CDR3 regions), which are formed by combination of V, Diversity (D) and J segments, and by germline-independent nucleotide additions and deletions at the joins between V, D and J segments [9]. In order to address these concerns, we concentrate on the CDR3 repertoire of naïve CD4 T-cells in the present study. These T-cells do not normally respond to a recall antigen, i.e. the majority have not been previously antigen-stimulated [10,11]. Thus, the chronic immune stimulation in MS should not affect naïve CD4 cells and any shift in the CDR3 repertoires, if observed, could precede the onset of MS. In an attempt to control for possible genetic influences on the CDR3 repertoire, we study healthy identical twins and identical twins discordant for MS.

An issue we had was how to determine whether the discordant twins and the healthy twins have significantly different CDR3 repertoires. To address this issue, we first determine the extent of similarity of naïve CD4 T-cell CDR3 repertoires among the healthy identical twins, by correlation analysis [12], and use this similarity or r-value as our reference point of normality. We then determine the extent of similarity of naïve CD4 T-cell CDR3 repertoires from the discordant twins, and test whether this r-value differs significantly from the r-value observed in the healthy twin pairs. Further, we test whether only MS patients, or both MS patients and their apparently healthy cotwins, have shifts in their CDR3 repertoires when compared with healthy subjects.

MATERIALS AND METHODS

The study subjects

We studied four healthy identical twin pairs (aged 42, 44, 48 and 52 years), with no known medical history, and four identical twin pairs (aged 32, 43, 52 and 53 years) discordant for MS by standard clinical criteria [13]. The patients from the discordant twins had MS for 7, 21, 21 and 11 years, respectively. The 52 years old MS patient had secondary progressive MS. The other patients had relapsing-remitting (RR) MS.

Cell preparation

Naïve human CD4 T-cells coexpress CD45RA and CD62L and are negative for CD45RO (CD45RO^–) [10,11,14]. Co-expression of CD62L and CD45RA enhances detection of truly naïve CD4 T-cells [14,15]. We isolated CD4⁺CD45RA⁺RO^– T-cells from peripheral blood samples on CD4⁺CD45RO^– columns (R & D Systems, Minneapolis, MN, USA), and then positively selected for CD62L expression using fluorochrome-conjugated anti-CD62L antibodies. The purity of naïve CD4 T-cells was greater than 95% in all experiments using a FACScan and CellQuest software (Becton Dickinson, Mountain View, CA, USA).

RNA isolation, first strand cDNA synthesis and PCR amplification of cDNA

We isolated total RNA from naïve CD4 T-cells using Trizol (Gibco BRL, Rockville, MD, USA) and used 2–5 µg of total RNA for oligo d(T) primed first strand cDNA synthesis as previously reported [7]. For PCR amplifications, we used the set of 24 different TRBV segment primers and the TRB constant (TRBC) region primer labelled with [γ-³²P] ATP described by Gorski et al. [16]. Here, we follow the internationally accepted nomenclature of TRB genes [17]. We prepared a PCR master-mix, and then added aliquots to individual reaction tubes so that each tube contained 1·25 U Taq polymerase, 250 µm dNTPs, 1·5 mm MgCl₂ and 0·165 µm of the radiolabelled TRBC primer and from 0·25 to 1 µl of cDNA. We then added a TRBV-specific primer (0·165 µm) to each tube. After 5 min denaturation at 94°C, we performed PCR for 30 cycles on a PTC-100 Programmable Thermal Cycler (MJ Research Inc., Boston, MA, USA). Each cycle consisted of 30 s at 94°C, 30 s at 58°C and 30 s at 72°C, with a final 7-min extension at 72°C. PCR products were run on 8% acrylamide/7 m urea sequencing gels. We then used autoradiography to identify the radiolabelled bands that constitute each CDR3 repertoire or spectratype. By doubling or halving the amount of cDNA, we confirmed that PCR conditions were nonsaturating. We scanned each dried denaturing gel, and captured the counts in each CDR3 spectratype band using the Cyclone Storage Phosphor System^TM, and OptiQuant^TM Image Analysis Software (Packard Instrument Company, Meriden, CT, USA).

Comparison of spectratypes and data analysis

Previously, we reported a statistical approach to test whether discordant twins have significant shifts in their TRBJ repertoires when compared with a group of healthy twins [18]. We modified this method for our current study, and our analysis of TRBV20-1 spectratypes illustrates the method. Briefly, after PCR amplification and CDR3 spectratyping, we had 11 TRBV20-1 spectratype bands in the various twins (Fig. 1a,b). In each twin, the sum of the percentage contribution of the 11 spectratype bands to the total TRBV20-1 CDR3 repertoire is 100. Therefore, once the percentage expression of 10 spectratype bands is determined, the percentage expression of the 11th band is known, i.e. 10 of the 11 bands have independent levels of expression. Thus, to achieve independence of the percentage expression of spectratype bands, we decided to drop the contribution of the first spectratype band to the total TRBV20-1 CDR3 repertoire. We did this for each twin. In order to have sufficient observations for the purpose of comparison, we calculated an r-value for the combined results of percentage expression of the TRBV20-1 spectratype bands of all healthy twins. That is, we had four pairs of 10 different percentage expressions, i.e. data points, for the healthy twins. We then used a standard normal Z-Score using Fisher's z transformation to test for the significance of the difference between these r-values (details of the equations are provided in [7]). We used the same method to test whether discordant twins have significant shifts of their CDR3 repertoires for the other TRBV segments.

Fig. 1.

TRBV20-1 CDR3 spectratype distributions of naïve CD4 T-cells from (a) four healthy identical twin pairs (the r-value for each twin pair varied between 0·982 and 0·998) and (b) four discordant MS twin pairs (the r-value for each twin pair varied between 0·972 and 0·997).

We created three additional Groups. Each Group consisted of subgroups of two pairs of individuals, in order to permit comparisons between subgroups having similar numbers of CDR3 spectratype bands. Group 1 consisted of all possible subgroups of two pairs of healthy unrelated individuals; each pair was formed by combining members from two different healthy identical twins. Group 2 consisted of all possible subgroups of two pairs of apparently healthy unrelated individuals, formed by combining apparently healthy members of the discordant twins. Group 3 consisted of all possible subgroups of two pairs of unrelated MS patients, formed by combining MS patients from the discordant twins. For each subgroup, we pooled the CDR3 spectratype distribution information and calculated an r-value in the same way described earlier. When comparing group 1 with Group 2, we compared the r-value of each subgroup of Group 1 with the r-value of each subgroup of Group 2. We used a similar procedure when comparing Group 1 with Group 3. The pairs of individuals in each group had from 18 to 22 pairs of spectratype bands.The standard normal Z-test, using Fisher's z transformation, requires 25 or more data points, i.e. 25 or more pairs of spectratype bands. Consequently, when testing for significant differences between two subgroups, we used an alternative to Fisher's z-transformation, which is applicable to sample sizes of 10 or greater [12].

RESULTS

Selection of TRBV segments for detailed analysis

Initially, we followed the approach to spectratyping described by Gorski et al. [16], i.e. we visually inspected the autoradiographs obtained for all 24 TRBV segments from all twins. TRBV21-1, and as reported previously TRBV30 [16], consistently gave multiple out-of-frame bands and TRBV12-3 consistently failed to give spectratype bands, i.e. these TRBV spectratypes were uninterpretable. Seventeen of the remaining 21 TRBV spectratypes showed no difference between healthy and discordant twins and had a Gaussian distribution of band intensities on the autoradiographs. In contrast, TRBV5-5, BV7-8, BV27 and BV24-1 spectratypes showed some differences between healthy and discordant twins. In order to investigate the significance of these differences, we decided to repeat the spectratyping, and performed duplicate PCR amplifications and CDR3 spectratyping of these four TRBV segments on cDNA from both healthy and discordant twins. For comparison purposes, we similarly performed duplicate PCR amplification-CDR3 spectratyping on TRBV20-1, which was one of the TRBVs that had uniform Gaussian distributions of spectratypes in all twins. We then compared the duplicate analyses from each twin member using correlation analysis. We arbitrarily decided to consider the findings as reproducible and worthy of further analysis, when the duplicate spectratype results gave r-values of 0·9 or higher. The duplicate analyses of all healthy and discordant twin members gave r-values of 0·9 or higher for TRBV20-1 and BV5-5 spectratypes. Similarly, the duplicate analyses of all members of the four discordant twin pairs and of three of the healthy twin pairs gave comparable r-values for TRBV7-8: we only had sufficient cDNA from three of the four healthy twin pairs to complete this analysis. These findings imply reproducibility of the percentage expressions of the CDR3 spectratype bands for these three TRBVs. In contrast, the r-values for duplicate TRBV27 and TRBV24-1 spectratypes were consistently less than 0·8 in members of the healthy and discordant twins, likely because these particular TRBV segments are typically used at low frequencies in most individuals [for example, see 19]. Thus, in all subsequent studies, we only analysed the TRBV20-1, BV5-5 and BV7-8 spectratypes from the healthy and discordant twins and calculated the final percentage expression of each spectratype band by averaging the duplicate results of percentage expressions.

CDR3 spectratype diversity in naïve CD4 T-cells

In order to investigate whether genetic factors influence the CDR3 repertoire, we compared the TRBV20-1, BV5-5 and BV7-8 CDR3 spectratype distributions among the healthy identical twins with those among the unrelated healthy pairs in Group 1 (see Materials and methods). In all possible comparisons, the Z-scores for TRBV20-1 CDR3 spectratype distributions did not differ significantly between these groups. TRBV5-5 spectratype distributions did not differ significantly between these groups in 83% of the possible comparisons. Similarly, TRBV7-8 spectratype distributions did not differ significantly between groups in 79% of the possible comparisons. Thus, genetic factors have little or no demonstrable influence on the CDR3 repertoires of these three TRBV segments in healthy subjects.

We next tested whether the healthy identical twins have significantly different CDR3 spectratype distributions from the discordant twins. The TRBV20-1 spectratype distributions among the four healthy twins (four pairs of 10 independent data points) (Fig. 1a) did not differ significantly from those among the four discordant twins (four pairs of independent data points) (Fig. 1b). In contrast, the TRBV5-5 spectratype distributions among the four healthy twins (four pairs of 10 independent data points) (Fig. 2a) differed significantly from those among the four discordant twins (four pairs of 10 independent data points) (Fig. 2b). All values of Z-score varied between 2·88 and 3·22 and their P-values were less than 0·01. The TRBV7-8 spectratype distributions among three healthy twins (three pairs of 8 independent data points) (Fig. 3a) also differed significantly from those among four discordant twins (four pairs of 9 independent data points) (Fig. 3b). All values of Z-score varied between 2·1 and 2·6, with their P-values being less than 0·02. In summary, the TRBV5-5 and the TRBV7-8 spectratypes differed significantly between the healthy and discordant twins for all possible comparisons.

Fig. 2.

TRBV5-5 CDR3 spectratype distributions of naïve CD4 T-cells from (a) four healthy identical twin pairs (the r-value for each twin pair varied between 0·947 and 0·995) and (b) four discordant MS twin pairs (the r-value for each twin pair varied from 0·882 to 0·993).

Fig. 3.

TRBV7-8 CDR3 spectratype distributions of naïve CD4 T-cells from (a) three healthy identical twin pairs (the r-value for each twin pair was 0·997) and (b) four discordant MS twin pairs (the r-value for each twin pair varied from 0·975 to 0·979). (c) Image of sequencing gel of duplicate PCR amplifications and TRBV7-8 spectratypes from two discordant twins run in parallel. Lanes 1–4 show the duplicate CDR3 spectratypes from each member of one discordant pair, and lanes 5–8 show duplicate CDR3 spectratypes from each member of a second discordant pair.

One discordant twin pair had remarkably different TRBV7-8 spectratype distributions from the other discordant twins (Fig. 3b). Figure 3c shows the image of the sequencing gel from this twin pair and from a second discordant twin pair in which the spectratypes were run in parallel. The MS patient from the twin pair with the different TRBV7-8 spectratype distributions had RR MS, as did two of the three other patients, i.e. the MS subtype does not explain the TRB7-8 spectratype patterns in this pair of discordant twins. We questioned whether this discordant twin pair was responsible for the significant differences between the discordant and healthy twins for TRBV7-8 spectratypes. After removal of this twin pair from the analysis, however, 87% of possible comparisons of TRBV7-8 spectratype distributions showed significant differences between healthy and discordant twins. The significant values of Z-score varied between 2·05 and 2·55, and P-values were less than 0·04.

CDR3 diversity in the groups created from the healthy and discordant twins

The TRBV20-1 spectratype distributions did not differ significantly between any of the three groups. In contrast, for 99% of possible comparisons, the TRBV7-8 spectratype distributions of Group 1 differed from those of Group 2; the significant values of Z-score varied between 2·14 and 3·89, and their P-values were less than 0·03. The TRBV7-8 spectratype distributions of Group 1 differed from those of Group 3 in all possible comparisons; and their values of Z-score varied between 2·35 and 4·24, with their P-values being less than 0·02. The TRBV5-5 spectratype distributions of Group 1 differed from those of Group 2 in 73% of comparisons; the statistically significant values of Z-score varied between 2·43 and 3·86, and their P-values were less than 0·015. The TRBV5-5 spectratype distributions of Group 1 differed significantly from those of Group 3 in 93% of possible comparisons; the significant values of Z-score varied between 2·11 and 4·36, and corresponding P-values were less than 0·03. Thus, for most or for all comparisons the TRBV5-5 and BV7-8 CDR3 repertoires of unrelated healthy pairs differed significantly from those of the unrelated MS patient pairs and from those of the pairs of their apparently healthy cotwins.

DISCUSSION

Our major finding was that the naïve CD4 T-cell CDR3 spectratype distributions of TRBV5-5 and TRBV7-8 among the healthy identical twins differed significantly from those among the discordant twins. That is, the deviations from the usual Gaussian distribution of CDR3 spectratype band intensities [16,20] represent statistically significant shifts of the CDR3 repertoires in the discordant twins. Further, either for most or for all comparisons, the TRBV5-5 and BV7-8 spectratype distributions in Group 1 (unrelated healthy pairs) differed significantly from those in Group 2 (unrelated apparently healthy members of the discordant twins) and from those in Group 3 (unrelated MS patient pairs). We conclude therefore that both the MS patients and the apparently healthy members of the discordant twins had CDR3 repertoire shifts for TRBV5-5 and TRBV7-8 segments.

Duplicate PCR amplifications and spectratypings establish that the spectratype distributions of TRBV20-1 did not differ significantly between the healthy and discordant twins or between the members of any other group. This finding implies that the TR gene region is intact in all RNA samples, i.e. that the CDR3 repertoire shifts for TRBV5-5 and TRBV7-8 do not reflect differences in RNA quality between samples. We considered several additional explanations for our findings. First, we questioned whether some of these twin pairs were actually concordant, or would become concordant in the near future. The updated Canadian MS twin study concluded that 70% of identical twins are discordant throughout life: the mean twin age in their study was 48 years [21]. Two of our discordant twins were 52 and 53 years, respectively. We therefore consider the apparently healthy members of these twins to be disease free, although we did not investigate these twin members by MRI. We cannot know whether the apparently healthy members of the other discordant twin pairs (aged 32 and 43 years) will subsequently develop MS, but note that the 43-year-old-patient has had MS for 21 years. A second possibility was that the chronic immune stimulation associated with MS caused the CDR3 repertoire shifts. This seems unlikely as we studied naïve CD4 T-cells, the large majority of which have not been previously stimulated by antigen [10,11] (see also below). A third possibility was that minor differences in twin ages between the healthy and discordant twins caused the spurious appearance of CDR3 repertoire shifts. However, age is not known to shift the naïve CD4 CDR3 repertoires, although exceptionally old individuals (100 years) can have oligoclonal expansions of naïve CD4 T-cells [22]. Moreover, we found no evidence that CDR3 repertoires shift with age. Fourth, a minority of CD45RO⁺ memory T-cells can revert to the naïve T-cell phenotype [11,23]. We think it unlikely that more memory T-cells were present among the naïve CD4⁺ T-cells from the discordant twins than from the healthy twins, as the simplest explanation of our data is that the same mechanism was responsible for the CDR3 repertoire shifts in both members of the discordant twins. Thus, if enhanced memory T-cell reversion had occurred equally in both MS patients and their cotwins, the CDR3 repertoires of the discordant twins should not have shown the decrease in similarity associated with CDR3 repertoire shifts. Also, another study showed a minimal response to recall antigen [24] in naïve CD4 T-cells from MS patients, i.e. little evidence for memory T-cells in this subset. Fifth, superantigens can stimulate T-cells particular TRBVs [25,26], but it is unlikely that superantigens caused the CDR3 repertoire shifts. This is because superantigen-stimulated T-cells typically have a Gaussian distribution of spectratype band intensities [20].

We found little or no evidence for genetic influences on CDR3 spectratype distributions in healthy subjects. This finding is not surprising, in spite of the influence of genetics on TRBV [8] and TRBJ [27] segment usage, because 90% of TR diversity is due to terminal deoxynucleotidyl transferase (TdT) activity with random, template-independent addition of nucleotides in the CDR3 region [28]. Others report that genetics, including HLA, influences the TR repertoire in rheumatoid arthritis [29,30]. The question arises therefore whether MS susceptibility genes, or HLA differences between the healthy and discordant twins, could explain our findings. Because HLA alleles are markedly heterogeneous [31], our twins likely had considerable HLA differences. We did not test for HLA, however, as we developed our method of twin comparisons partly to exclude the effects of HLA heterogeneity. Thus, for example, for TRBV5-5 we compared the percentage expressions of all spectratype bands of the members of each twin pair with one another. We then determined the extent to which the percentage expressions are similar between members of each twin pair forming a group of healthy twins, and calculated an r-value as a measure of this similarity. We similarly calculated an r-value as a measure of the similarity of the percentage expressions for TRBV5-5 spectratypes between members of each twin pair forming a group of discordant twins. We then used a z-test to test whether these two r-values differ significantly from one another. Thus, we used the healthy twins as a reference point of normality and tested whether the pairs of healthy twins have more similar TRBV5-5 spectratype distributions among themselves than did the pairs of discordant twins. By comparing similarities within a twin pair, and then within and between groups of twin pairs, each pair of which is HLA identical, the HLA type becomes irrelevant.

Study of patients with complete DiGeorge syndrome indicates that the thymus is likely the only source of naïve human T-cells [32]. Thus, the CDR3 repertoire shifts could have occurred during TR repertoire formation in the thymus and before the onset of MS. In keeping with the concept of a thymic abnormality in MS, a recent study of T-cell receptor excision circle content suggests that patients with RR MS have alterations in thymic export [33]. As postulated earlier [34], we propose that the CDR3 repertoire shifts predispose to MS but are insufficient to cause MS as both the apparently healthy members and the MS patients from the discordant MS twins had TCR repertoire shifts. Nevertheless, marked CDR3 repertoire shifts may be a causative factor in MS. Further studies are needed in this direction.