SLE patients and their relatives were recruited in several hospitals throughout Portugal in collaboration with the Portuguese Lupus Association. A total of 148 SLE patients (136 females and 12 males) were enrolled in the present study. 76 of the SLE patients were part of 54 families, from which 166 relatives (98 females and 68 males) were also included in the study. All patients met the revised 1997 American College of Rheumatology criteria for SLE [23], independently of their disease activity at recruitment. The SLE Disease Activity Index (SLEDAI) [24] was assessed for each patient at the time of collection (SLEDAI range: 0–41). Irreversible damage caused by the disease was calculated using the Systemic Lupus Erythematosus International Collaborating Clinics (SLICC) damage index. Information on medication from each patient was collected at recruitment. Patients and their relatives completed a 10 question screening questionnaire [25], used here as an indication of autoimmune disease manifestations in each individual (Table 1). Control individuals (N = 117, 58 females and 59 males) were blood donors recruited from several regional blood centers. The population sample characteristics and the numbers used in each of the subsequent analyses are described in Additional file 1. Approval for this study was obtained from the Portuguese Ethics Committee for Life Sciences Research, and collection of all biological samples was performed after written informed consent from all participating individuals.

PBMC were isolated using Vacutainer CPT™ tubes (Becton Dickinson). For the CD4CD25CD45RO staining 1 × 10⁶ cells were fixed in 2% PFA, washed twice in PBS 2% FCS and subsequently incubated for 20 min at 4°C with the optimal dilution of each conjugated anti-human mAb (CD4-FITC, CD25-Cy-chrome, CD31-PE and CD45RO-APC, Becton Dickinson). Cells were washed again and fluorescence intensity staining was analyzed using a FACScan flow cytometer (FACScalibur™ and CellQuest™ software, Becton Dickinson). CD4⁺CD25⁺ T cells were defined, for all the individuals, as the population of CD4 positive T cells whose CD25 expression exceeded the level of CD25 positivity seen in the CD4 negative T cells. The CD4-FITC, CD25-APC, FOXP3-PE staining was performed according to eBioscience manufacture protocol.

Reacti-Bind™ EIA plates (Pierce) were coated overnight at 4°C with 1 mg/mL purified dsDNA (Sigma-Aldrich). The plates were blocked with a solution of 1% gelatin for 5 h at 4°C and incubated overnight at 4°C with test plasma (in triplicates) at a total protein concentration was 0.5 mg/mL. Plates were then incubated overnight at 4°C with 50 μL of alkaline phosphatase-conjugated anti-human IgG (Sigma-Aldrich) and bound IgG was revealed using 1 mg/mL p-nitrophenyl phosphate (Sigma-Aldrich). Absorbance at 405 nm was measured in an ELISA plate reader (BioRad). Quantification of each reactivity was obtained by fitting ODs to a standard curve present in duplicates on each plate, and consisting of 12 serial dilutions of respective positive control sera. In the readout, one unit was defined as the equivalent of the highest concentration used for the respective positive control serum, which was previously chosen to be below the saturation level. Fractions of one unit were consequently defined by the dilution factor of this standard, at which it yielded an OD equivalent to that obtained for a respective sample [26].

Genomic DNA was isolated from PBMC by standard methods. The polymorphic markers tested were selected for their putative role in gene expression or function. Within the CTLA4 gene, we genotyped a functional SNP involving a missense A>G transition at +49 exon 1 (rs231775) and a 3'UTR microsatelite marker proposed to influence mRNA stability and turnover [27]. In the TGFβ gene, we genotyped a C>T transition in the promoter (rs1800469) and a G>C SNP in exon 1 (rs1800471), both known to be associated with the serum concentration of TGF-β [28,29]. Within the FOXP3 gene two microsatelite markers were genotyped (one in intron 0, which has been described as having a promoter/enhancer function and other in intron 5) and a -1128 G>A SNP (rs12843496) [30]. The IL2 gene, -384 G<T (rs2069762) and +114 G>T (rs3087209) in the promoter and in exon 1, respectively, were genotyped as previously described [30,31]. Within the CD25 gene, we genotyped a (CA)n repeat 10 kb upstream of the gene (D10S189) and two SNPs, rs2274037 (a A>G transition in exon 4) and rs1570538 (a C>T transition in the 3'UTR region of CD25) as previously described [32]. The D10S189 microsatelite marker was genotyped using the Applied Biosystems fluorescence based automated sequencer. Semi-automated fragment sizing was performed using GENESCAN^® 3.1.

Total RNA was isolated from 5 × 10⁶ PBMC using the RNeasy Mini Kit (QIAGEN). The first strand cDNA was generated by reverse transcription with oligo dT primer (Invitrogen). The CTLA-4, TGF-β, FOXP3, IL-2 and CD25 mRNA levels were quantified with the LightCycler (Roche Molecular Biochemicals) in 33 patients and 24 controls, which were representative of the total population sample. 50 ng of first-stranded cDNA were amplified and real-time fluorimetric intensity of SYBR green I was monitored. RT-PCR reactions consisted in an initial denaturation step at 95°C for 10 minutes followed by 45 cycles of denaturation at 95°C for 15 seconds, 56°C for 5 seconds and 72°C for 23 seconds. The individual samples were standardized by the amount of HPRT RNA. This RT-PCR reaction consisted on an initial denaturation step at 95°C for 10 minutes followed by 45 cycles of denaturation at 95°C for 15 seconds, 60°C for 5 seconds and 72°C for 23 seconds. Additionally, CTLA-4 and CD25 mRNA expression levels were normalized by the proportion of CD4⁺ cells, given that these molecules are mainly expressed by CD4⁺ T cells.

Nonparametric analysis of variance was performed using the Mann-Whitney U and Kruskal-Wallis tests to compare the distributions of T cell frequencies in patients and controls. Results were considered significant at the 0.05 level. Heritability estimation was performed by the SOLAR software [33]. In this test the likelihood of the sporadic model of no inheritance is compared with the likelihood of a polygenic model of inheritance. The sporadic model constitutes our null hypothesis in which it is assumed that there is not a genetic component underlying the frequencies of Treg cells. The polygenic model constitutes our testing hypothesis, where the quantitative trait is considered to be determined by a large number of genetic loci acting independently and/or additively. The models are then compared to find the hypothesis that is better supported by the data. Hardy-Weinberg equilibrium (HWE) was analyzed by the HWE exact test. Linkage disequilibrium (LD) between two polymorphic markers was quantified as D' using the HAPLOXT program implemented in the GOLD software [34].

The previous observation that a CD4⁺CD25⁺ T cell subpopulation expressing the CD45RO activation marker exhibits 75% higher suppressive activity [35] suggests that the majority of Treg activity is restricted to CD45RO⁺ cells. We therefore evaluated the frequency of both CD4⁺CD25⁺ and CD4⁺CD25⁺CD45RO⁺ T cell populations in total lymphocyte PMBC samples of SLE patients, their relatives and healthy controls and expressed it as a percentage of total CD4⁺ cells (Figure 1A). CD4⁺CD25⁺ and CD4⁺CD25⁺CD45RO⁺ T cell frequencies were significantly decreased in patients compared to controls (Mann-Whitney U test, z = -6.339, P < 0.00001 and z = -6.161, P < 0.00001, respectively). CD4⁺CD25⁺ and CD4⁺CD25⁺CD45RO⁺ T cell distribution in relatives was intermediate between that of patients and controls (Figure 1B). As the frequency distributions of either cell population were identical, the results obtained for the CD4⁺CD25⁺CD45RO⁺ cell subset was considered for the remaining of this study.

SLE affects primarily women and consequently only about 10% of our patients are male, while our control population is balanced in terms of gender. We therefore tested whether this gender bias was not a confounding factor in our interpretation. CD4⁺CD25⁺CD45RO⁺ T cell frequency was significantly lower in healthy females than in males (Mann-Whitney U test: z = 4.121, P < 0.001). The CD4⁺CD25⁺CD45RO⁺ T cell frequency was also lower in females than in males in the patient and relative groups, although the difference did not reach statistical significance (Figure 1C). Separate analysis within each gender group shows that the CD4⁺CD25⁺CD45RO⁺ T cell frequency is always significantly lower in patients (Mann-Whitney U test, Males: z = -2.504, P = 0.012; females: z = -3.864, P < 0.001).

CD25 is a marker for both Treg and activated cells. To rule out the possibility that this T cell population is mainly constituted by activated cells we measured the frequency of the CD4⁺CD45RO⁺ T cells, since CD45RO expression is associated with early activation events. No difference was found in the frequency of CD4⁺CD45RO⁺ cells between patients and controls (Mann-Whitney U test: z = -1.724, P = 0.089), indicating that the CD25⁺ population was largely a regulatory population.

To exclude a global thymic output defect, we evaluated the percentage of recent thymic emigrants (RTE) inside the CD4⁺ population using the phenotypic markers CD45RO^-CD31⁺ [36]. As expected, CD45RO^-CD31⁺ cell frequency decreased with age, but there were no differences between patients and controls in absolute frequency or in the kinetics of decay (Figure 2A). Moreover, the frequency of RTE did not correlate with the percentages of CD4⁺CD25⁺CD45RO⁺ cells (ρ = 0.10, P = 0.094). However, the CD31⁺ cells inside CD4⁺CD25⁺ cells decreased faster with age in patients than in controls (Figure 2B), suggesting a defect specifically in thymic Treg cell production in SLE patients. No significant difference was found in the frequency of CD4⁺CD45RO^-CD31⁺ or CD31⁺ cells inside CD4⁺CD25⁺ between males and females (data not shown).

Given the major involvement of FOXP3 in the differentiation and function of Treg, we analyzed the frequency of FOXP3⁺ cells in a subset of 19 patients and 10 healthy individuals. While the frequency of FOXP3⁺ cells in CD4⁺PBMC was not significantly different between patients and controls (Figure 3A), the frequency of CD25⁺ in the CD4⁺FOXP3⁺ cell subset was significantly reduced in patients when compared to healthy controls (Mann-Whitney U test, z = 2.457, P = 0.014) (Figure 3B). In contrast, FOXP3^-CD25⁺ activated cells were significantly increased in SLE patients (Mann-Whitney U test, z = -2.377, P = 0.017) (Figure 3C). This diminished expression of CD25 in patients' FOXP3⁺ cells indicates that the overall reduction in CD4⁺CD25⁺ cells in SLE patients likely results specifically from a defect in the ability to convert FOXP3⁺CD25^- into FOXP3⁺CD25⁺ cells.

Autoantibody production against dsDNA is a hallmark of SLE. We therefore analyzed the plasma immunoreactivity profiles against dsDNA from SLE patients, relatives and healthy individuals. Overall, there is a negative correlation between Treg frequency and anti-dsDNA in the total population (ρ = -0.19, P = 0.0026) (Figure 4A), reinforcing the notion that perturbations in the system that maintains immunologic unresponsiveness to self-constituents lead to the loss of peripheral tolerance and to SLE. This negative correlation between anti-dsDNA autoantibodies and the frequency of CD4⁺CD25⁺CD45RO⁺ T cells was also found in our patient (ρ = -0.37, P = 0.005) and control groups (ρ = -0.30, P = 0.022) (Figure 4A). Surprisingly, there is a positive correlation between the frequency of CD4⁺CD25⁺CD45RO⁺ T cells and anti-dsDNA antibodies in the relatives' group (ρ = 0.29, P = 0.0004) (Figure 4A). No difference was found between males and females.

The CD4⁺CD25⁺CD45RO⁺ T cell frequency was inversely correlated with disease activity measured by the SLEDAI index (ρ = -0.502, P = 0.002) (Figure 4B). However, no correlation was found with the SLICC damage index (ρ = 0.071, P = 0.658). There was no significant difference between the frequency of CD4⁺CD25⁺CD45RO⁺ T cells in patients untreated or treated with corticosteroids (Kruskall-Wallis Test: z = 0.07, P = 0.94), antimalarial agents (Kruskall-Wallis Test: z = 1.41, P = 0.158), or both drugs (Kruskall-Wallis Test: z = -1.32, P = 0.132). Noteworthy, no correlation was found between total lymphocyte counts and the frequency of CD4⁺CD25⁺CD45RO⁺ T cell (ρ = 0.021, P = 0.758), excluding the possibility that lymphopenia led to this phenotype. There was also no correlation with age (ρ = 0.153, P = 0.100).

In two SLE patients variations in CD4⁺CD25⁺CD45RO⁺ T cell frequency as a function of disease onset and progression and with treatment were evaluated. A 33-year-old Caucasian woman was initially enrolled in this study as a healthy relative of an SLE patient. At the time of recruitment, the frequency of CD4⁺CD25⁺CD45RO⁺ T cells in this subject was 4.60% (Figure 5A). One year later, this woman presented classical SLE clinical symptoms, lymphopenia, positive anti-dsDNA antibody, positive lupic anticoagulant and a SLEDAI of 10. Upon establishment of the diagnosis of SLE, a new measurement of Treg frequency showed a decrease to 2.88% (Figure 5B), suggesting the involvement of CD4⁺CD25⁺CD45RO⁺ T cell frequency in the onset of symptoms. In a second case a 67-year-old female SLE patient, diagnosed three years before, developed an SLE flare with symptoms of acute isquemic stroke, hemolytic anemia, impaired renal function, antinuclear, anti-dsDNA and anticardiolipin antibodies and hipocomplementemia. The SLEDAI score was 42. A regimen of low-dose corticosteroids and antimalarials had previously provided a good control of disease. At this point the CD4⁺CD25⁺CD45RO⁺ T cell frequency was 4.22% (Figure 5C). After anticoagulant and intravenous corticosteroid therapies, the CD4⁺CD25⁺CD45RO⁺ T cell frequency was 4.30% (Figure 5D). In addition to anticoagulants and intravenous corticosteroids therapies, IVIg therapy administration was deemed necessary. Upon IVIg treatment, the patient's CD4⁺CD25⁺CD45RO⁺ T cell frequency increased to 7.12% (Figure 5E) and she showed progressive clinical and laboratorial improvement, and the SLEDAI score decreased to 12 after three weeks of treatment. Taken together these results support the idea that decreased CD25⁺ T cell frequency is associated with disease activity.

Relatives of SLE patients presented intermediate CD4⁺CD25⁺CD45RO⁺ T cell frequencies between patients and controls (Figure 1B), suggesting that Treg frequency might represent a genetic trait. Any potential bias due to recruitment design was excluded based on the following observations: for each family all samples were collected and processed on the same day; there were no significant differences in Treg frequency between families collected in different days. Treg frequencies in genetically unrelated family members and in controls collected and processed on the same day are within the range of healthy controls analyzed during the whole period of the study.

A high CD4⁺CD25⁺CD45RO⁺ T cell frequency heritability, defined as the proportion of variation of this trait attributable to genetic factors, was estimated at 0.85 ± 0.09 (P = 1.05e-10) (Additional file 2). As for most complex traits, there was no clear segregation pattern of CD4⁺CD25⁺CD45RO⁺ T cell frequency with affection status in the families. However, because relatives often present SLE-related autoimmunity symptoms even when not meeting the full diagnostic criteria, we tested the association of CD4⁺CD25⁺CD45RO⁺ T cell frequency with the severity of these suboptimal autoimmune symptoms through a 10-item survey questionnaire administered to patients and relatives (Table 1). The number of positive answers was inversely correlated with the frequency of CD4⁺CD25⁺CD45RO⁺ T cells (Kruskall-Wallis test: H = 20.7, P = 0.014) (Figure 6), as expected if the extent of clinical manifestations is associated with the frequency of CD4⁺CD25⁺CD45RO⁺ T cells. The results therefore provide evidence that CD4⁺CD25⁺CD45RO⁺ T cells frequency constitutes a genetic trait associated with SLE manifestations, which can be used for genetic mapping.

The evidence for familiar transmission prompted us to test putative functional polymorphisms in the candidate genes CTLA4, TGFβ, FOXP3, IL2 and CD25 for association with CD4⁺CD25⁺CD45RO⁺ T cell frequency. All genotypic distributions were in HWE and all the polymorphic markers within the same gene were in LD. A significant association of the CTLA4 3'UTR microssatelite marker with CD4⁺CD25⁺CD45RO⁺ T cell frequency was found (Kruskal-Wallis test: H = 40.57, P = 0.009), with genotypes 104/104 and 106/106 associated with lower (2.64%) and higher (5.47%) frequency of CD4⁺CD25⁺CD45RO⁺ T cells, respectively (Table 2). This result is particularly relevant because, in our sample, allele 104 has previously been found to increase susceptibility to disease, while allele 106 showed a protective effect [27]. The TGFβ rs180047 SNP was also significantly associated with the frequency of CD4⁺CD25⁺CD45RO⁺ T cells (Kruskal-Wallis test: H = 8.099, P = 0.004). Heterozygosis was associated with lower frequencies of CD4⁺CD25⁺CD45RO⁺ T cells (Table 2). Because the C allele has very low frequency in the general population (6%), there were no homozygous individuals for this allele in our sample.

Both the FOXP3 rs12843496 SNP and the intron 0 marker, located in the promoter/enhancer region of the gene, were nominally associated with SLE (χ² = 6.14, P = 0.013 and χ² = 10.30, P = 0.036, respectively) (Table 3). The A allele of rs12843496 was more frequent in patients than in controls. The susceptibility allele 4 of the intron 0 microsatellite marker was more frequent in patients (χ² = 6.63, P=0.01) and a protective allele 6 more frequent in controls (χ² = 4.03, P = 0.04) (Table 3). No association was found with CD4⁺CD25⁺CD45RO⁺ T cell frequency (Table 2). There was no evidence for association of any other tested gene marker with CD4⁺CD25⁺CD45RO⁺ T cell frequency or with SLE (data not shown).

Taken together these results indicate that CTLA4 and TGFβ gene variants contribute to the determination of CD4⁺CD25⁺CD45RO⁺ T cell frequency in PBMC, independently of disease activity, while the involvement of FOXP3 gene variants in SLE is likely not mediated by CD4⁺CD25⁺CD45RO⁺ T cell frequency.

No correlation was found between CTLA-4, TGF-β, IL-2 or CD25 expression levels and the frequency of CD4⁺CD25⁺CD45RO⁺ T cells, disease status or marker genotypes (data not shown). There was a trend towards higher levels of FOXP3 mRNA in SLE patients, although not significant (Mann-Whitney U-test: z = -1.78, P = 0.075), and in individuals carrying the A allele of FOXP3 rs12843496 (Mann-Whitney U-test: z = 4.95, P = 0.026). These results suggest that variability in CD4⁺CD25⁺CD45RO⁺ T cell frequency associated with SLE is not influenced by altered expression of any of the tested genes.