Targeting of tolerogenic dendritic cells to heat-shock proteins in inflammatory arthritis

Human blood samples were obtained from healthy controls (HC) and treatment-naïve patients with recent onset arthritis (PsA and RA). Samples were collected with informed consent and following a favourable ethical opinion from local ethics committees. Peripheral blood mononuclear cells (PBMC; from 40 ml EDTA blood per donor) were isolated as previously described [17]. Monocytes were positively selected from PBMC using anti-CD14 microbeads (Miltenyi Biotec) according to manufacturer’s protocol with one minor change: 10 µl instead of 20 µl anti-CD14 beads per 1 × 10⁷ cells was used for cell isolation. CD14-depleted PBMC (hereafter referred to as ‘PBMC’) were collected from the column flow-through and stored for 1 week at − 80 °C in FCS (Gibco) with 10% DMSO (Sigma) and were used for the measurement of HSP-specific T cell responses and the DC/PBMC co-culture experiments (see below).

Immediately after isolation, monocytes were cultured in 24 wells plates (Corning) at 0.5 × 10⁶ cells/ml (total 1 ml/well) for 7 days in CellGenix DC medium (CellGenix) containing penicillin (100 U/ml), streptomycin (100 μg/ml), GM-CSF (50 ng/ml; Immunotools) and IL-4 (50 ng/ml; Immunotools). During this period cells were kept at 37 °C with 5% CO₂. On day 3, half of the medium was substituted by fresh (warm) medium containing GM-CSF (100 ng/ml) and IL-4 (100 ng/ml). For the generation of tolDC, dexamethasone (1 μM; Sigma) was added on days 3 and 6 and 1,25-dihydroxyvitamin D3 (Calcitriol; 0.1 nM; Tocris) and monophosphoryllipid A (MPLA) (1.0 μg/ml; Invivogen) were added only on day 6. Immature DC (imDC) were cultured in the presence of GM-CSF (50 ng/ml) and IL-4 (50 ng/ml). On day 7, 24 h after the last treatment, DC were harvested and washed extensively before functional assays were performed. DC were then resuspended at 4 × 10⁵ cells/ml in X-VIVO-15. DC phenotype was checked using flow cytometry and was consistent with tolDC exhibiting a semi-mature phenotype, expressing low levels of CD83, intermediate levels of CD86 and high levels of HLA-DR and TLR2 (data not shown).

PBMC were thawed, washed and labelled with 0.2 μM carboxyfluorescein succinimidyl ester (CFSE; eBioscience) or 0.2 μM cell proliferation dye eFluor-450 (CTV; eBioscience) in PBS for 10 min at 37 °C. CFSE/CTV was quenched with 10% human serum (HS; Sigma) in HBSS (Lonza). Cells were resuspended at 2 × 10⁶ cells/ml in X-VIVO-15 medium (Lonza) supplemented with 4% HS (final concentration 2%) and plated at 2 × 10⁵ cells per well (96 wells; round bottom; Corning). For each peptide eight wells were prepared. Peptides were added at 10 µg/ml. Cells were cultured for 9 days at 37 °C with 5% CO₂. At the end of the culture, supernatants were collected for cytokine determination. Depletion of CD14 from PBMC did not hamper detection of HSP-specific T cell responses (data not shown).

CFSE or CTV-labelled PBMC were resuspended at 4 × 10⁶ cells/ml in X-VIVO-15 medium (Lonza) supplemented with 8% HS (final concentration 2%) and plated at 2 × 10⁵ cells per well (96 wells; round bottom; Corning). tolDC or, as a control, imDC were added in a 1:10 ratio (i.e. 2 × 10⁴/well), in the absence or presence of PPD (1 µg/ml) CA (1 µg/ml) or a cocktail of the HSP-peptides (HSP40 DnaJP; HSP60 p1 and p3; HSP70 B2; all at 4 µg/ml) Cells were cultured for 6 (IL-10 secretion) or 9 days (all other measurements) at 37 °C with 5% CO₂. TGF-βRI (ALK5) inhibitor (SB-505124; 1 µM; Sigma) was added where indicated. At the end of the culture, supernatants were collected for cytokine determination.

For cell surface staining: cells were washed in flow cytometry buffer (PBS (Lonza) supplemented with 3% fetal calf serum (FCS; Gibco), 1 mM EDTA (Fisher Scientific) and 0.01% sodium azide (Sigma)) before incubating them for 30 min on ice in flow cytometry buffer containing antibodies and 4 µg/ml human immunoglobulin (Ig)G (Grifols). Cells were washed and resuspended in flow cytometry buffer before analysis. For intracellular cytokine staining (ICS): cells were first stimulated with PMA (50 ng/ml) and ionomycin (1 μg/ml) for 5 h in the presence of brefeldin A (1 μg/ml; all from Sigma Aldrich) at 37 °C, 5% CO₂. Cells were then surface stained as described before, washed in flow cytometry buffer and fixed/permeabilised using cytofix/cytoperm buffer (BD Biosciences) for 30 min on ice. Cells were washed twice in 1× perm wash buffer (BD Biosciences) and stained for 30 min in 1× perm wash buffer containing antibodies and 8% mouse serum. Cells were washed once in 1× perm wash buffer and once in flow cytometry buffer before resuspending them in flow cytometry buffer. Data were collected on an LSRfortessa X20 (BD Biosciences) and analysed using FlowJo (Tree Star Inc). Additional file 1: Table S1 depicts antibodies and live/dead dyes used for analysis.

Cells from the PPD-DC/PBMC co-cultures were harvested and a total of 100,000 CFSE⁻CD4⁺DAPI⁻ cells per sample were sorted into RLT buffer (Qiagen) supplemented with 1% β-mercaptoethanol (Sigma) using a FACSAria-fusion sorter (BD Biosciences). Lysates were stored for up to 4 months at − 80 °C before isolation of RNA. RNA was isolated using an RNeasy Micro Kit (Qiagen; including DNase step) according to the manufacturer’s protocol. A total of 100 ng of RNA was used for gene expression profiling by the nCounter technology platform (NanoString; Human Immunology Panel), performed according to the manufacturer’s protocol. Data was analysed using nSolver™ Analysis Software version 4.0 and R.

Cells from PPD-DC/PBMC co-cultures were harvested on day 6 and a minimum of 25,000 CFSE⁻CD4⁺ (unfractionated proliferated CD4⁺ T-cells from control-PPD cultures) or CFSE⁻CD4⁺LAG3⁺CD49b⁺ (proliferated Tr1 cells from tolDC-PPD cultures) sorted into X-VIVO-15 with 20% HS, using a FACSAria-fusion sorter. Sorted cells were rested for 2 days in X-VIVO-15 supplemented with 2% HS and 10 IU/ml of IL-2 (Proleukin; 25,000 cells/well; 96-well round-bottom). Cells were subsequently washed and restimulated with 10 µg/ml platebound anti-CD3 (OKT3; Biolegend) and 1 µg/ml soluble anti-CD28 (CD28.2; Biolegend) in X-VIVO-15 supplemented with 2% HS (total 100 µl; 96-well flat-bottom). Supernatants were collected after 72 h for cytokine determination.

Cytokine production was determined in supernatants by Meso Scale Discovery (MSD; U-Plex (IL-10, IFNγ, IL-4, IL-17A, GM-CSF) or by sandwich ELISA from BD (IL-10).

The following statistical analyses were performed using Prism 5: repeated measures analysis of variance (ANOVA) with Bonferroni correction for comparisons between multiple groups, paired Student t-test for comparisons between two groups.

Our initial studies investigated whether immunodominant pan-DR-binding peptides from bacterial HSP40 (dnaJP1), mycobacterial (myc)-HSP60 (p1 and p3) and myc-HSP70 (B29) [14, 15, 18] could be recognised by peripheral blood CD4⁺ T-cells of healthy donors and IA patients (Table 1). We included both RA and PsA patients, because patients from both these IA groups participated in the phase I tolDC safety trial [3].

As shown in Fig. 1b, 78% of healthy donors, 73% of RA patients and 90% of PsA patients responded to at least one of the HSP peptides tested (Fig. 1b; see ‘All HSP’ which depicts the best HSP response for each donor). In the majority of cases, donors responded to one HSP only, but in around 10% of cases there was a detectable response to all four HSP peptides (Additional file 2: Fig S1).

To study the inflammatory nature of HSP-specific T-cell responses, we measured the secretion of the pro-inflammatory cytokines IFNγ, GM-CSF, IL-17A and IL-4 and the anti-inflammatory cytokine IL-10 in the supernatants of the HSP cultures. PBMC from both healthy donors and RA/PsA patients produced significant levels of pro-inflammatory cytokines in response to HSP (Fig. 1c).

The T-cell modulatory effects of human tolDC are usually studied in (tol)DC/T-cell co-culture models. However, because tolDC need to be able to regulate T-cell responses in the context of other immune cells, we assessed whether tolDC could regulate autologous CD4⁺ T-cells in the antigen-specific PBMC culture system described above. To this aim, we initially used the recall antigen PPD. T-cell responses to PPD were assessed in the absence or presence of tolDC. As a control, we used another DC population with known, but unstable, tolerogenic function—immature monocyte-derived DC (imDC). Mature monocyte-derived DC as a control were also considered, but we found that these mature DC induced very high background (i.e. no antigen added) proliferation of CD4, CD8 and NK cells, which was not the case when imDC or tolDC were added to the PBMC cultures (data not shown). We compared the immune-related gene-expression profile of PPD-specific T-cells activated in the absence or presence of tolDC or imDC. The gene expression profiles of the differentially activated PPD-specific T-cell groups clustered well together, indicating clear differences between treatment groups at the mRNA level (Fig. 2a). We identified 83 differentially expressed genes (DEGs) from a total of 579 genes in PPD-specific T-cells activated in the presence of tolDC (tolDC-PPD T-cells) as compared to PPD-specific T-cells activated in the absence of monocyte-derived DC (control PPD T-cells) (Benjamini–Hochberg false discovery rate 10%). Of these 83 genes, 24 DEGs were also found in tolDC-PPD T-cells as compared to PPD-specific T-cells activated in the presence of imDC (imDC-PPD T-cells). Ten additional DEGs could be identified in tolDC-PPD T-cells as compared to imDC-PPD T-cells (Fig. 2b and Additional file 3: Table S2).

Among the tolDC-PPD versus control PPD T-cell identified DEGs, 20 genes could either be identified as type 1 regulatory T-cells (Tr1)-specific genes or genes that have been described as inducers of Tr1-specific genes (Fig. 2c, [19‐30]).

Recently, Gagliani et al. [19], described two surface antigens, LAG3 and CD49b, as being highly and stably expressed on Tr1 cells. Moreover, co-expression of only these two surface proteins allowed for the identification of Tr1 cells. To confirm that tolDC induce a Tr1 phenotype in PPD-specific T-cells, we measured the co-expression of LAG3 and CD49b and several other Tr1-specific proteins identified in the gene expression analysis, by flow cytometry. TolDC induced significant upregulation of the LAG3, TIM3, CD86 and PD-1 proteins as compared to control PPD cultures and the combined expression of LAG3 and CD49b could identify tolDC-induced PPD-specific Tr1 cells (Fig. 2d).

A hallmark of Tr1 cells is the high secretion of IL-10 in the absence of IL-4 [19, 31‐33]. As shown in Fig. 3a, significantly higher levels of IL-10 were produced in the tolDC-PPD cultures as compared to the imDC-PPD or control PPD cultures. We furthermore showed that re-stimulation of sorted, proliferated LAG3⁺CD49b⁺CD4⁺ Tr1 cells produced high levels of IL-10 but no IL-4 (Fig. 3b).

Overall, these data clearly indicate that tolDC induce a Tr1 phenotype in autologous antigen-specific CD4⁺ T-cells.

Since we established that HSP-specific T-cells are pro-inflammatory in both patients and healthy controls, and that the co-expression of LAG3 and CD49b can be used to identify tolDC-activated antigen-specific Tr1 cells, we used these markers to study whether tolDC induced a Tr1 phenotype in HSP-specific T-cells of IA patients. tolDC-activated HSP T-cells had significantly higher levels of LAG3/CD49b, coinciding with decreased levels of GM-CSF and IFNγ and increased IL-10 production as compared to control PPD T-cells (Fig. 4a, b and d). In addition, tolDC also induced a Tr1 phenotype in response to a third (control) antigen, Candida albicans (CA; Fig. 4c). Thus, tolDC are capable of inducing Tr1 responses in autologous antigen-specific CD4⁺ T-cells, irrespective of the antigen.

As our antigen-specific T-cell activation model using PBMC allowed us to check the behaviour of other types of immune cells, we had observed high levels of NK-cell proliferation. Interestingly, bystander NK-cell proliferation was largely inhibited by tolDC in HSP-cultures (Fig. 5a, b) and CA-cultures (Fig. 5c). We have previously shown that tolDC produce high amounts of TGFβ [17]. It is known that TGFβ can hamper NK-cell proliferation and activation [34, 35] and we therefore investigated whether blocking of the TGFβ-receptor could reverse the tolDC-induced suppression of NK-cell activation. Indeed, addition of SB-505124, a small molecule inhibitor of TGF-βRI, restored NK-cell proliferation in tolDC-CA cultures (Fig. 5d).

Altogether, our findings demonstrate that tolDC can convert pro-inflammatory antigen-specific CD4⁺ T-cells in IA (RA and PsA) patients into anti-inflammatory Tr1 cells, which are characterised by expression of LAG3/CD49b and IL-10 production. In addition, tolDC-derived TGFβ blocks bystander NK-cell activation.