The multi-tetramer approach allows stable detection of autoreactive T cells at low frequencies in healthy controls
To study citrulline-reactive T cell specificities with our multi-tetramer assay, we created a panel of HLA-DRB1*04:01 tetramers loaded with eight citrullinated RA-associated self peptides derived from four candidate autoantigens, namely α-enolase, vimentin, fibrinogen-b as well as CILP. Additionally, we employed two commonly used viral control epitopes from influenza matrix proteins [20 21]. Table 2 lists the sequence and position of the peptides as well as the tetramer fluorophores. As autoreactive T cells have been shown to be present at low frequencies in the circulation of RA patients and healthy individuals [5 6] we first determined the sensitivity of our multi-tetramer assay by applying it to PBMC from seven HLA-DRB1*04:01-positive healthy control subjects. Here, we unambiguously detected T cells specific for α-enolase, CILP and fibrinogen at frequencies ranging up to 10 per million CD4+ T cells (median frequency 4.7 and 1.7 per million CD4+ T cells, respectively; Figure 1a). As expected, the number of cells reactive to our positive influenza peptide controls was higher, typically in a range of 20 to 1000 per million CD4+ T cells (median frequency 61 per million CD4+ T cells). Repeat experiments performed on cells from the same time points and donors demonstrated on average comparable frequencies of tetramer-positive cells in the two experiments for each donor (Figure 1a).
Besides enumerating the tetramer-positive CD4+ T cells, we also determined their differentiation state by examining the surface expression of CD45RA and CCR7 (Figure 1b). As expected, T cells specific for influenza were predominantly of a memory phenotype and distributed between a Tcm, central memory (51%) and a Tem, effector memory phenotype (44%). Conversely, the majority of autoreactive T cells in these healthy subjects displayed a naïve phenotype, expressing CCR7 and CD45RA simultaneously (Figure 1b and c). Still, it should be noted that we also detected central memory type T cells in a subset of the samples, while effector memory T cells were consistently a minor phenotype.
Autoreactive T cells are found in most RA patients, even in the absence of concurrent disease activity
Next, in order to further validate our panel also in patient samples, we analysed a longitudinal cohort of 14 RA patients from which we obtained samples from repeat blood draws approximately 2-3 weeks apart and therefore could analyse intra-individual variance. The patients included in this cohort were recruited according to the following criteria: having ACPA-positive RA and at least one HLA-DRB1*04:01 allele. All patients had long disease duration (>5 years), overall no signs of active disease around the time of sampling and stable anti-rheumatic treatment according to standards (see Additional file 1: Table S1.1).
We detected frequencies between 1 and 35 tetramer-positive cells per million CD4+ cells of CILP/fibrinogen- and α-enolase-specific T cells in these RA patients (Figure 2a). These frequencies were slightly increased in patients compared to healthy controls (Figure 1a and 2a). Not all specificities were present in all patients, with α-enolase-specific T cells being detected in eight out of fourteen and CILP/fibrinogen-specific T cells in thirteen out of fourteen patients. Specificities within individual patients were reliably detected in the repeat blood draws in half of the individuals. Other patients showed citrulline-specific T cells only at one or two of the three time points, as indicated by single dots and dotted lines connecting the frequencies of the remaining time points in Figure 2a. In contrast, influenza-specific T cells were steadily found in all patients in each of the three repeats and always at 10-20 times higher frequency compared to autoreactive T cells (Figure 2a). Examining the overall distribution of the cells within the different memory and naïve states, we detected - similarly to healthy subjects - a high proportion of influenza-specific T cells in the central and effector memory compartment and very little amounts of naïve T cells (Figure 2b). Again, we found a broad distribution of the proportion of naïve citrulline-reactive T cells between different subjects. Within the memory subset, central memory type T cells were overrepresented among CILP/fibrinogen- compared to α-enolase-reactive T cells (Figure 2b). To a lower extent (<20%) we also detected citrulline-specific T cells with a terminal effector memory phenotype, such Temra cells were virtually absent amongst the influenza-specific T cells (Figure 2b).
Analysing the expression of chemokine receptors on the tetramer-positive cells revealed a higher percentage of CXCR3+ cells among α-enolase- and CILP/FGB-specific compared to influenza-specific T cells (Figure 2c). CXCR5 expressing cells on the other hand were mainly detected among CILP/fibrinogen- and influenza-specific T cells and rather less among cells recognizing α-enolase. We found little expression of both CD25 and CCR6 on the influenza-specific as well as the general CD4 population, while such phenotypes were detected in α-enolase- and CILP/fibrinogen-specific T cells (Figure 2c). Of note, among influenza- and citrulline-specific cells as well as in the general CD4 population around half of the CXCR3+ cells also expressed CXCR5 and to a lesser extent CCR6 (data not shown).
Citrulline-specific T cells are present in lymph node biopsies of arthralgia and early RA patients
Next, we applied the multi-tetramer assay to lymph node biopsies from HLA-DRB1*04:01-positive individuals. Included in this part of the study were arthralgia patients with elevated RF and ACPA levels as well as early RA patients. Samples from one patient with undifferentiated arthritis and one healthy control were also included in this analysis (see Additional file 1: Table S1.2)
Due to the limited cell numbers that can be retrieved from a core needle biopsy we decided against the general ex vivo protocol but opted instead for analysing cells after in vitro expansion using PHA and IL-2. We could detect citrulline-reactive T cells in all of the tested individuals albeit with varying number of specificities (Figure 3). All five RA patients had vimentin- and α-enolase-specific T cells while CILP/fibrinogen-specific T cells were only found in four patients. Three of the five arthralgia patients displayed reactivity against all citrullinated peptides. The remaining two had no CILP/fibrinogen-specific T cells, but both displayed vimentin-reactive and one also α-enolase-specific T cells. Also for the healthy control we could demonstrate the presence of vimentin- and α-enolase-specific T cells, however no CILP/fibrinogen-reactive ones. The patient diagnosed with undifferentiated arthritis on the other hand had all T cell reactivities we tested for. It should be noted that since these cells required cell culture expansion, we could not calculate the original frequencies nor analyse the unmanipulated phenotypes.
Citrulline-specific T cells decline in early RA patients achieving drug-induced remission
Finally, we applied the multi-tetramer assay to peripheral blood samples from ten patients enrolled in the LURA study at the Karolinska University Hospital. This cohort comprises 134 newly diagnosed patients with duration of patient-reported symptoms ≤1 year and naïve to treatment with disease-modifying anti-rheumatic drugs (DMARDs) and oral glucocorticoids (GC). The selection of the patients for this substudy was based on HLA-DRB1*04:01-positivity and the availability of blood samples from both baseline and six months follow-up.
T cells specific for both influenza and citrullinated peptides were found to be part of the peripheral T cell repertoire at both baseline and following six months of standard medication with methotrexate in this early RA cohort (Figure 4a). Citrulline-specific T cells were detected in all but one patient, who interestingly turned out to be negative for both RF as well as ACPAs. The specificity that was detected least often was towards vimentin, with 3/10 and 4/10 positive patients for baseline and follow-up, respectively. Both α-enolase- and CILP/fibrinogen-specific cells were found in 50-60% of the patients at both time points (Figure 4b). The general frequencies determined for influenza-specific as well as autoreactive T cells were in line with previous results and ranged from 10 to 230 and 1 to 7 tetramer-positive cells per million CD4+ cells, respectively. Notably, three of the four patients for whom the frequency of citrulline-specific T cells declined from baseline to follow-up, were also patients that achieved a DAS28-value <2.6 which corresponds to minimal disease activity and clinical remission (Figure 4a, coloured in green and table 1). Another patient could, due to a missing DAS28-value at follow-up, not be added to the group of responders, although we could demonstrate a decrease in citrulline-reactive T cells also for this patient (Figure 4, coloured in grey and table 1).
Subsequent analysis of CD45RA and CCR7 expression revealed a significant decrease of central memory type T cells after six months amongst CILP/fibrinogen-specific cells. Along the same line we detected a general trend towards lower proportions of memory type T cells for both α-enolase- and CILP/fibrinogen-specific cells and concurrently higher proportions of naïve as well as terminal effector memory T cell phenotypes (Figure 4c and Additional file 1: Figures S2c, right panel, and S3a). Conversely, vimentin-specific T cells displayed a trend for increased central memory and decreased naïve T cells at six months follow-up (Figure 4c and Additional file 1: Figure S3a).
To investigate the activation status of these cells in more detail we analysed expression of CD25 and CD38, markers of recent T cell activation and detected no significant changes upon comparing baseline and follow-up for neither of the two markers. As expected for patients not subject to ongoing infections or recent vaccination, the level of expression on the influenza-specific control as well as the general CD4+ population was very low for both markers. However, trends are pointing towards lower proportion of CD25+ cells among both α-enolase- and CILP/fibrinogen-specific T cells after six months of treatment. For CD38 we found the opposite feature with an increased expression on both α-enolase- and vimentin-specific memory cells after six months (see Additional file 1: Figure S3b).
Similarly, we examined the surface expression of the chemokine receptors CCR6, CXCR3 and CXCR5 and found a general decrease in the frequency of CXCR5+ cells for all autoreactive as well as the influenza-specific T cells at follow-up. In general, CXCR5, compared to CXCR3 and CCR6, is the chemokine receptor that was most prominently expressed on tetramer-positive cells with mean levels of up to 40% for citrulline-specific and 54% for influenza-specific populations (see Additional file 1: Figure S3b) while for the general CD4+ population it was only around 10%. In the case of CXCR3 the average expression level did not exceed 22% and was lower at follow-up for CILP/fibrinogen- and vimentin-specific cells. Among α-enolase-reactive cells on the other hand we could not detect changes in the proportion of CXCR3+ cells. The frequency of CCR6-expressing cells was generally highest among vimentin-specific cells (mean between 17 and 33%) but decreased from baseline to follow-up while it increased for both α-enolase- and CILP/fibrinogen-specific cells (see Additional file 1: Figure S3b).
Among all citrulline-specific cells we observed a number of cells co-expressing two or three of the chemokine receptors. Here it was most common to find CXCR5 in combination with either CXCR3 or CCR6 (data now shown). In contrast to this we found only a few influenza-specific cells co-expressing these chemokine receptors with a majority of CXCR3+ cells carrying CCR6 and/or CXCR5, while among CXCR5+ cells the proportion of cells carrying a second chemokine receptor was rather small (data not shown).