This was a prospective study of consecutive patients with a final diagnosis of RA based on the 2010 ACR/EULAR criteria (20) as well as age- and sex-matched healthy individuals (HC). Detailed medical history including disease duration, prior, and current treatments was obtained from each patient. Each patient also underwent full clinical assessment with determination of disease activity according to the simplified disease activity index (SDAI) (21) and the disease activity score 28 (22). Whenever available, synovial fluid samples were obtained from patients undergoing routine joint aspiration.

Peripheral venous blood or synovial fluid was drawn from each individual, and PBMCs were isolated by Histopaque density gradient centrifugation. The total cell number was determined by a Beckmann Coulter. Cells were cultured at 1 × 10⁶ cells/ml in RPMI 1640 containing 10% fetal calf serum, 2 mM l-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin in the presence of 20 U/ml human recombinant IL-2 (SIGMA, Vienna, Austria) and initial stimulation with 10 μg/ml plate-bound anti-CD3 Ab.

Surface and intracellular staining of freshly isolated PBMCs was performed using appropriate combinations of antibodies for detection of CD3, CD4, CD28, CD25, CD127, FoxP3, CTLA-4, IL-10, TNF-α, Th1-type cytokines IL-2 and IFN-γ, Th2-type cytokine IL-4 and Th17-type cytokine IL-17 (all Becton Dickinson, San Diego, CA, USA), and γH2AX (Cell Signaling, Danvers, MA, USA). Surface staining for 20 min was followed by permeabilization for 30 min and intracellular staining for 30 min according to a routine protocol. Golgi transport was inhibited by brefeldin A (10 ng/ml) or monensin (10 ng/ml) 4 h prior to cytokine staining. Appropriate isotype controls were used. Stained cells were analyzed on a FACS Canto II (Becton Dickinson). Data are analyzed with DIVA software and FlowJo.

For functional assays, CD4⁺ T-cells were isolated by positive selection of PBMCs labeled with magnetic-bead conjugated antihuman CD4 mAbs using MACS MultiSort Kit and autoMACSPro according to manufacturer’s instructions (Miltenyi). Purified CD4⁺ T-cells were then separated into the CD28⁺CD25⁺CD127^dim (conventional regulatory), CD28⁺CD25⁻ (conventional non-regulatory), and CD28⁻CD25⁺CD127^dim (senescent Treg-like) fractions by another sorting step using FACS technology (FACS Aria). In order to obtain sufficient numbers of RA Tregs, pre-enrichment of PBMCs was necessary and resulted in upregulation of CD25. Therefore, isolation of Treg subsets was based on CD25 positivity as well as absence of CD127. For CD28 downregulation experiments, Tregs were isolated using CD4⁺CD25⁺CD127^dim/− Reg. T Cell isolation Kit II (Miltenyi) and autoMACSPro. For validation, flow cytometry was then performed to determine purity (>90%) of selected cells.

RNA from enriched subsets was extracted using RNeasy Protect Mini Kit (Qiagen, Germantown, MD, USA) according to the manufacturer instructions. RNA was used for reverse transcription with First Strand cDNA Synthesis Kit for RT-PCR AMV (Roche) according to the manufacturer regulations. cDNA was diluted 1:5 for PCR using AmpliTaq Gold™ DNA Polymerase (Applied Biosystems), 1× PCR Gold Buffer (Applied Biosystems), 2.5mM MgCl₂ (Applied Biosystems), 0.4 mM dNTP Polymerization Mix (GE Healthcare), 0.5 μM TCR C β 5′FAM labeled primer (Ingenetix, Vienna, Austria), and 0.5 μM unlabeled TCR V β primers (Ingenetix) according to Hingorani et al. (23). Primer sequences are listed in Table S2 in Supplementary Material. Cycle conditions were a denaturation step at 94°C for 6 min, 35 cycles of 94°C for 1 min, 59°C for 1 min, and 72°C for 1 min, a final annealing step at 72°C for 7 min. After amplification, the PCR product was supplemented with GeneSCan™-500 TAMRA™ Size Standard (Applied Biosystems) and HI-DI Formamide (Applied Biosystems). Electrophoresis was performed with ABI Prism 310 Genetic Analyzer (Applied Biosystems), and 310 Data Collection Software. Analysis was done by GeneScan^® Software (Applied Biosystems). Calculations included peak count (Complexity score) and single peak area in percent of whole peak area.

DNA from enriched subsets was extracted using QIAamp DNA Blood Mini Kit (Qiagen, Germantown, MD, USA) according to the manufacturer instructions. Telomere lengths were measured using DNA from PBMCs as well as T-cell subsets by quantitative real-time PCR analysis and LightCycler FastStart DNA Master SYBR Green I (Roche, Vienna, Austria) as previously described (24).

CD4⁺ T-cells were isolated always from the same healthy donor (JF) since RA effector cells were reported to be resistant to Treg-mediated suppression (5). Cells were incubated with CellTrace Violet (Life Technologies) and cultured in the presence or absence of CD28⁺CD25⁺CD127^dim or CD28⁻CD25⁺CD127^dim RA T-cells at a 1:1 ratio. Autologous CD28⁺CD25⁺CD127^dim served as a control. Cells were stimulated by adding 10 μg/ml anti-CD3 Ab and incubated at 37°C for 3 days and analyzed by flow cytometry. To elucidate the impact of TNF-α and IFN-γ on the suppressive potential of different subsets, we added neutralizing antibodies (R&D systems) at 10 μg/ml. A normal goat IgG control was also added at the start of culture accounting for any non-specific changes.

Freshly isolated PBMCs were stimulated with plate-bound anti-CD3 (10 μg/ml) or PHA (1 μg/ml) and cultured for 18 h. Afterward cells were harvested and stained for Annexin V and viability dye (ebioscience, San Diego, CA, USA) as well as surface markers according to the manufacturer’s protocol. Annexin V⁺ cells represent apoptotic cells, viability dye⁺ cells represent necrotic cells, and double positive cells represent late apoptotic cells.

All statistical analyses were performed using the SPSS program, version 23 (Chicago, IL, USA). In case of a normal distribution (tested with the Kolmogorov–Smirnov test) of the continuous variables, the mean and SD are shown, and we used the two-sided Student’s t-test (comparison of two groups) or ANOVA (comparison of three or more groups) for comparisons. In case of a non-parametric distribution, the median and range are shown, and we used the Mann–Whitney U and the Kruskal–Wallis tests to assess differences between groups. Correlation between variables was evaluated by the Spearman’s rank correlation coefficient. Paired data were compared with the Wilcoxon test.

This study was approved by the Institutional Review Board of the Medical University Graz, and written informed consent was obtained from each individual prior to inclusion in the study.

We know that in RA, (1) a proportion of T-cells lack the co-stimulatory molecule CD28, (2) the loss of CD28 reflects early T-cell senescence and is partially caused by pro-inflammatory stimuli, and (3) Tregs undergo a similar development to non-Tregs from a naïve-like to a memory-like status. We therefore investigated whether expression of CD28 is reduced on FoxP3⁺ T-cells (which is the most specific marker for Tregs) from RA patients.

In RA patients but not in controls, we observed a FoxP3⁺ T-cell subset lacking the expression of CD28. The prevalence of circulating CD4⁺CD28⁻FoxP3⁺ T-cells was higher in RA patients compared to healthy individuals [0.7% of total CD4⁺ (range 0–19.2) vs. 0.2% (0–17); p = 0.029; Figures 1A,B], whereas the frequency of CD4⁺CD28⁺FoxP3⁺CD25⁺ Tregs was equal in both groups [5% of total CD4⁺ (3.2–7.3) vs. 4.9% (3.2–9.5); p = 0.988; Figure 1C]. The intensity of FoxP3 expression was similar in CD28⁻ and CD28⁺ subsets (MFI: 601.8 ± 200.4 vs. 634.8 ± 196.5, p > 0.05, Figures 2A,B). In synovial fluid samples from RA patients, we noted that 3.8% (1.6–18.8) of CD4⁺FoxP3⁺ T-cells were negative for CD28 (n = 8, data not shown).

Alongside FoxP3, CD25 and a low expression of CD127 are characteristic for Tregs (26, 27). CD4⁺ CD28⁻FoxP3⁺ T-cells were positive for CD25 and expressed low levels of CD127; both markers, however, were reduced relative to CD4⁺CD28⁺FoxP3⁺ Tregs: CD25, MFI: 328 (199–2,068) vs. 1,998 (1,456–2,946); p < 0.001; Figure 2C and CD127, MFI: 454 (111–905) vs. 1,206 (906–1,961); p < 0.001; Figure 2D. CTLA-4, PD-1, and CD15s are other typical markers of Tregs, all of them were found on CD4⁺CD28⁻FoxP3⁺: CTLA-4 expression was similar in CD4⁺CD28⁻FoxP3⁺ T-cells and CD4⁺CD28⁺FoxP3⁺ Tregs [MFI: 123.5 (103–350) vs. 122 (104–648); p = 0.499; Figure 2E], whereas PD-1 was higher [MFI: 449.5 (167–811) vs. 258.5 (221–480); p = 0.017; Figure 2F] and CD15s lower [MFI: 864.5 (647–914) vs. 1,102.5 (690–1,251); p = 0.028; Figure 2G] in the former compared to the latter cell population. CCR6, another molecule normally expressed by Tregs was almost absent on CD4⁺CD28⁻FoxP3⁺T-cells [0.2% (0–7.5) of CD4⁺CD28⁻FoxP3⁺ T-cells were positive for CCR6⁺ vs. 39.8% (28.9–55.2) of CD28⁺ Tregs; p = 0.012, Figure 2H] (28).

Collectively, these analyses suggest that CD4⁺CD28⁻FoxP3⁺ T-cells have a Treg-like phenotype.

To investigate whether the novel CD28⁻ Treg-like subset revealed signs of cellular senescence in addition to the loss of CD28, we undertook the following experiments: first, we tested the accumulation of γH2AX foci, which represent repair-proof double-strand breaks in DNA (29). Second, we performed TCR spectratyping because TCR diversity is known to diminish along with T-cell aging (30). Third, we tested the accumulation of senescence-associated β-galactosidase (SABG) another known biomarker of cellular senescence (31). Fourth, we measured the telomere length given that a shrinkage of telomeres is considered to indicate replicative senescence (32).

As depicted in Figure 3A, CD28⁻ Treg-like cells showed higher γH2AX mean fluorescence intensity than CD28⁺ Tregs [MFI: 6,422 (2,952–258,589) vs. 4,875 (2,875–7,743), p = 0.046]. Similarly, TCR diversity was reduced in CD28⁻ Treg-like cells [84 (36–104)] compared to their CD28⁺ counterparts [115 (109–125); p = 0.037; Figure 3B]. SABG expression, moreover, was likewise enhanced in CD28⁻ Treg-like cells [MFI: 1,117 (671–1,263) vs. 499 (474–591); p = 0.043; Figure 3C].

Interestingly, telomere length was similar in CD28⁻ Treg-like cells [6.11 kbp (5.46–6.19)] and CD28⁺ Tregs [5.89 (5.6–6.17), p = 0.373, Figure 3D]. From previous studies we know, however, that the telomere length of Tregs is reduced already (compared to non-Tregs), and a further shrinkage of telomeres is thus unlikely to occur (13).

Clinical characteristics of RA patients and controls are depicted in Table S1 in Supplementary Material. Two (3.3%), 8 (13.1), and 32 (52.5) out of the 61 RA patients with available SDAI values had high, moderate, or low disease activity, respectively; 19 (31.1) patients were in clinical remission (33). Five (6%) RA patients had early disease (≤2 years’ duration).

Frequencies of CD28⁻ Treg-like cells (but not those of CD28⁺ Tregs) correlated with age in RA patients and controls (corr_coeff = 0.416, p < 0.001 and corr_coeff = 0.557, p < 0.001, respectively; Figure S1A in Supplementary Material). Neither the prevalence of CD28⁻ Treg-like cells nor those of CD28⁺ Tregs was linked with disease duration, acute phase reactants, clinical variables, or treatment (Figures S1B,C in Supplementary Material and data not shown).

Previous studies reported that the downregulation of CD28 on T-cells is driven by pro-inflammatory signals including TNF-α or IL-15, which are also highly expressed in RA patients (34, 35). To test if these agents mediate the generation of CD28⁻ Treg-like cells in vitro, we isolated CD4⁺CD25^highCD127^dim/− Tregs from healthy individuals and cultured them in the presence or absence of IL-15 or TNF-α. We observed a significant decrease of CD28 on Tregs stimulated with TNF-α [3,295 (1,293–16,853)] compared to mock stimulation [5,628.5 (1,782–16,559); p = 0.025, Figure 4A]. A decrease of CD28 was also observed following incubation with IL-15. This difference, however, did not reach statistical significance [4,483 (713–15,309); p = 0.138]. Moreover, the reduced expression of CD28 was robust following elongated stimulation with TNF-α but not IL-15 (Figure 4B). Expression of CD25, CD127, and FoxP3 was not different following either way of stimulation and remained in Treg-specific expression levels (Figure 4C).

Next, we examined the cytokine profile of CD28⁻ Treg-like cells from RA patients. CD28⁺ Tregs are known to produce high levels of IL-10, which is one way how they inhibit effector cells (36). CD28⁻ non-Tregs acquire a senescence-associated secretory phenotype, which is characterized by the predominate release of Th1 cytokines such as TNF-α and IFN-γ (37).

CD28⁻ Treg-like cells yielded a more pronounced production of IL-10 than CD28⁺ Tregs following stimulation with anti-CD3 [4% (1.5–13.1) of CD28⁻ Treg-like cells vs. 0.55% (0.3–1.6) of CD28⁺ Tregs were positive for IL-10; p = 0.005, Table 1]. At the same time, however, cytokines characteristic for Th1 (IL-2, TNF-α, and IFN-γ), Th2 (IL-4), and Th17 lineages (IL-17) were also more frequently produced by CD28⁻ Treg-like cells compared to CD28⁺ Tregs (all p = 0.005). The largest difference was observed concerning the production of IL-4 (14-fold).

Conventional senescent CD4⁺ T-cells produced lower amounts of IL-4 and IFN-γ than CD28⁻ Treg-like cells, however, no difference was found regarding the production of IL-10. For details see Table 1.

The suppression of effector T-cells is a defining feature of regulatory cells. Tregs from mice with a conditional knockout for CD28 are non-suppressive, and in RA patients, Treg function has been reported (at least by some authors) to be defective (5, 19). Therefore, we studied the ability of CD28⁻ Treg-like cells to suppress the proliferation of non-Tregs. As depicted in Figure 5A, the suppressive function of CD28⁻ Treg-like cells from RA patients was compromised compared to that of CD28⁺ Tregs [median reduction of proliferation of stimulated non-Tregs: −2.2% (−8.7.1 to 77.7) vs. 32.7% (−0.4 to 77.9); p = 0.008].

Given that TNF-α and IFN-γ were shown to interfere with Treg function in vitro and that CD28⁻ Treg-like cells produced high levels of these cytokines (38, 39), we tested whether the suppressive capacity of CD28⁻ Treg-like cells was restored by the blockade of TNF-α or IFN-γ. The addition of neutralizing antibodies had no effect on the suppressive function of CD28⁻ Treg-like cells or CD28⁺ Tregs (Figures 5B,C).

Regulatory T-cells from CD28⁻ deficient mice have a pronounced proliferative/survival disadvantage (19). Therefore, we analyzed the proliferative capacity and apoptosis induction of CD28⁻ Treg-like cells. Upon stimulation with anti-CD3, we observed a lower rate of cell division of CD28⁻ Treg-like cells compared to CD28⁺ Tregs [53.1% CFSE^high (range 14.5–93.6) vs. 5.4% (0.9–30.6), respectively; p = 0.008; Figure 5D]. In addition, we performed Annexin V staining assays in order to analyze apoptosis induction this subset. In the CD28⁻ Treg-like subset, increased frequencies of early apoptotic [7% (range 2–25.7) vs. 1% (0.5–2.5); p = 0.005], necrotic [1.9% (0–9.2) vs. 0.2% (0.1–0.6); p = 0.011], and late apoptotic cells [17.1% (7.5–51.7) vs. 2.6% (1.3–8.4); p = 0.005] were observed upon stimulation with anti-CD3 compared to the CD28⁺ Treg population. Overall, more than 30% of CD28⁻ Treg-like cells were apoptotic or necrotic upon activation compared to only 5% of CD28⁺ Tregs (Figure 5E).