Ruxolitinib inhibits IFNγ-stimulated Sjögren’s salivary gland MSC HLA-DR expression and chemokine-dependent T cell migration

Abstract

Objectives

Sjögren’s disease (SjD) is a systemic autoimmune disease characterized by focal lymphocytic infiltrate of salivary glands (SGs) and high SG IFNγ, both of which are associated with elevated lymphoma risk. IFNγ is also biologically relevant to mesenchymal stromal cells (MSCs), a SG resident cell with unique niche regenerative and immunoregulatory capacities. In contrast to the role of IFNγ in SjD, IFNγ promotes an anti-inflammatory MSC phenotype in other diseases. The objective of this study was to define the immunobiology of IFNγ-exposed SG-MSCs with and without the JAK1 & 2 inhibitor, ruxolitinib.

Methods

SG-MSCs were isolated from SjD and controls human subjects. SG-MSCs were treated with 10 ng/ml IFNγ +/– 1000 nM ruxolitinib. Experimental methods included flow cytometry, RNA-sequencing, chemokine array, ELISA and transwell chemotaxis experiments.

Results

We found that IFNγ promoted expression of SG-MSC immunomodulatory markers, including HLA-DR, and this expression was inhibited by ruxolitinib. We confirmed the differential expression of CXCL9, CXCL10, CXCL11, CCL2 and CCL7, initially identified with RNA sequencing. SG-MSCs promoted CD4⁺ T cell chemotaxis when pre-stimulated with IFNγ. Ruxolitinib blocks chemotaxis through inhibition of SG-MSC production of CXCL9, CXCL10 and CXCL11.

Conclusions

These findings establish that ruxolitinib inhibits IFNγ-induced expression of SG-MSC immunomodulatory markers and chemokines. Ruxolitinib also reverses IFNγ-induced CD4⁺ T cell chemotaxis, through inhibition of CXCL9, -10 and -11. Because IFNγ is higher in SjD than control SGs, we have identified SG-MSCs as a plausible pathogenic cell type in SjD. We provide proof of concept supporting further study of ruxolitinib to treat SjD.

Rheumatology key messages.

• IFNγ promotes salivary gland mesenchymal stromal cell (MSC) HLA-DR expression and chemokine production, resulting in CD4+ T cell chemotaxis.

• Ruxolitinib, a selective JAK1&2 inhibitor, inhibits IFNγ-induced HLA-DR expression, chemokine production and CD4+ T cell chemotaxis.

• SG-MSCs are a plausible novel contributor to Sjögren’s disease pathophysiology through their response to pathologic IFNγ.

Introduction

Salivary and lacrimal gland inflammation are hallmark features of Sjögren’s disease (SjD) and central to SjD pathogenesis because it promotes autoantibody production and B-cell lymphoproliferation [1]. This inflammation is characterized by T cell predominant lymphocytic infiltrates and upregulated IFN-induced gene expression, driven by high local IFNα and IFNγ levels (i.e. IFN signature) [2, 3]. Glandular type II IFN signature (IFNγ) predominates the type I interferon signature and is associated with greater salivary gland (SG) focus scores [4] and lymphoma risk [5].

Residing within SG stroma are mesenchymal stromal cells (MSCs), a cell population responsible for tissue repair and stem cell niche functionality in other organs such as bone marrow, bowel and adipose [6]. We previously reported that SjD and control SG-MSCs have similar immunobiology characteristics and provided a fundamental description of SG-MSC’s response to IFNγ [7]. The objective of this study was to fully characterize the physiologic response of SG-MSCs to IFNγ and the downstream JAK/STAT pathway in vitro using pharmacological modulation with the clinically approved JAK1&2 inhibitor ruxolitinib.

Methods

Patient recruitment and MSC isolation

We recruited subjects during visits for labial SG biopsy (IRB# 2018–1815). SjD subjects met 2016 ACR/EULAR criteria [8]. Control subjects were referred for labial salivary gland biopsy for sicca but had normal serology (ANA ≤ 1:160 and negative anti-SSA antibody), focus score <1 and did not have an autoimmune disease. SjD subjects (n = 17) averaged 51 years old and were mostly Caucasian females (95%). Control subjects (n = 12) averaged 55 years old and were also majority Caucasian (100%) females (92%). SjD and control subjects had similar age (P = 0.4), sex (P = 0.8) and race (P = 0.5). Among SjD subjects, the majority were anti-SSA antibody positive (95%), anti-nuclear antibody positive (71%) and had rheumatoid factor (60%). Fewer had low C3 (6%) or C4 (12%). The average focus score was 1.7 among SjD subjects. Detailed demographics are reported in Supplementary Tables S1 and S2 (available at Rheumatology online). MSCs were isolated from labial SGs as previously described and as in the Supplementary Methods, available at Rheumatology online [7, 9]. All experiments were performed on MSCs before passage eight. Five SjD subjects were recruited for peripheral blood mononuclear cell (PBMC) transwell studies. The average age of these subjects was 59, all were female, and most were Caucasian (80%). All SjD subjects recruited for the blood draw were anti-SSA antibody positive and had normal complement, rheumatoid factor and focus score.

Ethics approval: This study was approved by the University of Wisconsin Health Sciences IRB (IRB# 2018–1815 & 2015–0156) and complies with the Declaration of Helsinki.

Flow cytometry

After stimulation with recombinant IFNγ (10 ng/ml; Peprotech, Rocky Hill, NJ, USA) or ruxolitinib (1000 nM; INCB-18424, LC Laboratories, Woburn, MA, USA) for 48 h, SG-MSCs were harvested for flow cytometry, washed and subjected to Ghost Red 780 viability staining per the manufacturer’s instructions (Tonbo Biosciences, San Diego, CA, USA). For MSCs, we used antibodies against HLA-DR, CD274 PD-L1 and CD54 ICAM. For PBMCs, we used antibodies against CD1c, CD3, CD4, CD8, CD11c, CD14, CD16, CD19, CD56, CD123 and HLA-DR. For CD4⁺ T cells, we used antibodies against CD3, CD4, CD8, CD14, CD45 and CXCR3. We used fluorescence minus one for controls. Flow results were analysed with FCS Express 7 (De Novo software, Pasadena, CA, USA). Further details are provided in the Supplementary Methods, available at Rheumatology online.

Western blot

Procedures for protein isolation and western blot are in the Supplementary Methods, available at Rheumatology online. We used primary antibodies against indoleamine 2,3 dioxygenase (IDO) (rabbit; MilliporeSigma, Burlington, MA, USA), STAT1 (rabbit; Cell Signalling Technology [CST], Danvers, MA, USA), pSTAT1 (rabbit; CST), pSTAT2 (rabbit; CST) and GAPDH (rabbit; CST). We used a secondary antibody to HRP-conjugated goat anti-rabbit (Bethyl Laboratories, Montgomery, TX, USA).

RNA-sequencing

We performed transcriptomic analysis using biologic replicates of control and SjD as previously described and in the Supplementary Methods (n = 4, each; available at Rheumatology online) [7]. Analysis was carried out on a total of 16 samples: (i) control IFNγ treated (n = 4); (ii) control IFNγ and ruxolitinib treated (n = 4); (iii) SjD IFNγ treated (n = 4); and (iv) SjD IFNγ and ruxolitinib treated (n = 4). We used DAVID 6.8 online tool to determine gene ontology (GO) of differentially expressed genes (website: https://david-d.ncifcrf.gov/) [10, 11].

Chemokine array

SG-MSCs were serum starved for 24 h, then cultured with (i) vehicle; (ii) ruxolitinib 1000 nM; (iii) IFNγ (10 ng/ml); or (iv) both ruxolitinib and IFNγ for 48 h. After 48 h, we collected conditioned media for analysis. Conditioned media from one SG-MSC cell line in all four treatment conditions was concentrated 10× before application of the media to a human chemokine array kit membrane per the manufacturer’s recommendations (R&D Systems, Minneapolis, MN, USA).

ELISA

We used conditioned media for ELISAs for CCL2, CXCL8, CXCL9, CXCL10, CXCL11, CXCL12, CXCL16, midkine, TIG2 (Peprotech) and CCL7 (Sigma, St Louis, MO, USA) per the manufacturer’s recommendations (n = 4–6 SjD and n = 3–5 controls).

Transwell assay

SG-MSCs were plated on the bottom of a 5 or 8 μm 12-well transwell system. After growing to confluence, the SG-MSCs were serum starved in α-MEM containing 1% FBS for 24 h. We added media containing vehicle, IFNγ (10 ng/ml), or both IFNγ (10 ng/ml) and ruxolitinib (1000 nM) to SG-MSCs. After 24 h, the MSCs were washed. Peripheral blood was obtained from healthy donors (n = 4) and SjD subjects (n = 5) and PBMCs were isolated using a Ficoll gradient. CD4⁺ negative selection was performed using EasySep^TM Cell separation (Stemcell Technologies, Vancouver, BC, Canada). The PBMCs or CD4⁺ T cells were counted and re-suspended and IL2 (100 IU/ml) was added for cultures of >12 h. Media also contained GolgiPlug Protein Transport Inhibitor (BD Biosciences, San Jose, CA, USA) or a monoclonal antibody to CCL2, CCL7, CXCL9, CXCL10 or CXCL11 (ThermoFisher, Waltham, MA, USA) per the manufacturer’s recommendations. The media containing PBMCs or CD4⁺ T cells was added to the top of the transwell system. After either 5 or 12 h, cells remaining in the top well of the transwell system were aspirated, washed and stained for flow cytometry.

Scratch assay

Control (n = 3) and SjD (n = 3) SG-MSCs were grown to confluence and treated with vehicle, ruxolitinib (1000 nM), IFNγ (10 ng/ml) or both IFNγ and ruxolitinib as described in the Supplementary Methods available at Rheumatology online. After scratch placement, we performed automated thresholding with TScratch software [12].

Real-time quantitative PCR

RNA and cDNA were generated as previously described and as in the Supplementary Methods and Supplementary Table S3, available at Rheumatology online [7]. We normalized to GAPDH and graphed cycle times as fold-change expression of SjD over control using 2^−ΔΔCT.

Statistical analysis

Comparison between two continuous variables was performed using Mann–Whitney U test for non-parametric distributions, Welch’s t test for normal distributions with unequal variance, and Student’s t test for normal distributions with equal variances. Comparison between multiple continuous variables was performed using ANOVA with Fisher’s LSD. We used Prism software (Graphpad, San Diego, CA, USA).

Results

Ruxolitinib inhibits expression of IFNγ-induced immunomodulatory markers

MSCs expand regulatory T cells and suppress cytotoxic T cells. IFNγ potentiates these effects through expression of proteins including IDO, PD-L1 and ICAM-1 [13–17]. In contrast, IFNγ can have pro-inflammatory effects on MSCs, prompting bone marrow MSCs to express major histocompatibility (MHC) class I and II, thereby acting as antigen presenting cells [18]. Accordingly, we set out to determine the effects of IFNγ on SG-MSC immunomodulatory proteins and if this effect is reversed by the JAK1&2 inhibitor ruxolitinib.

We found IFNγ (10 ng/ml) treatment increased HLA-DR and ICAM expression in SG-MSCs (Supplementary Fig. S1, available at Rheumatology online). To evaluate the dose-range of ruxolitinib required to reverse characteristic IFNγ-induced MSC immunobiologic markers, we cultured SG-MSCs with 10 ng/ml of IFNγ and ruxolitinib from 0.1 nM to 1000 nM. Statistically significant suppression of ICAM was detectable with 500 nM ruxolitinib. Near absolute suppression of IFNγ-induced upregulation of HLA-DR, PD-L1, ICAM and IDO occurred with 1000 nM ruxolitinib (Fig. 1A–E).

Fig. 1.

Ruxolitinib inhibits expression of IFNγ-induced immunomodulatory markers

Cells from three SG-MSC cell lines were serum starved for 24 h then treated with 10 ng/mL of IFNγ ± ruxolitinib (0.1–1000 nM) for 48 h. After 48 h, cells were washed and stained with antibodies to HLA-DR, PD-L1 and ICAM-1 for flow cytometry. (A) Representative gating strategy and expression of HLA-DR, PD-L1 and ICAM (CD54) after stimulation with IFNγ ± ruxolitinib; (B–D) median fluorescence intensity (MFI) of MSC immunomodulatory markers HLA-DR, PD-L1, ICAM-1; (E) Representative image and analysis of IDO western blot. SG-MSCs were serum starved × 24 h, then treated with vehicle (n = 8), 10 ng/mL IFNγ (n = 8), ruxolitinib 1000 nM (n = 3), or both IFNγ and ruxolitinib (n = 8) for 48 h. Protein was isolated and antibodies to IDO and GAPDH were used. IDO expression of each condition relative to the GAPDH loading control was estimated by densitometry analysis. IDO expression was significantly enhanced in IFNγ treated conditions; values are means (s.e.m.). *=P < 0.05, **=P < 0.01, ***=P < 0.001, ns=non-significant.

Ruxolitinib reverts the characteristic transcriptomic profile of IFNγ-treated SG-MSCs

After defining upregulation of SG-MSC immunomodulatory markers, we sought a global transcriptomic profile of SjD (n = 4) and control (n = 4) SG-MSCs after IFNγ and ruxolitinib treatment. We cultured SG-MSCs with IFNγ (10 ng/ml) or both IFNγ (10 ng/ml) and ruxolitinib (1000 nM) for 48 h and performed RNA sequencing on these samples. CCL7, CXCL9, -10 and -11 were in the top 50 differentially transcribed genes in both SjD and control SG-MSCs treated with IFNγ compared with those treated with IFNγ and ruxolitinib simultaneously (Fig. 2A and B). SjD vs control SG-MSCs treated with IFNγ had 100 significantly differentially regulated genes, whereas those treated with both had only 11 differentially regulated genes (Fig. 2C and D).

Fig. 2.

Ruxolitinib reverts the characteristic transcriptomic profile of IFNγ-treated SG-MSCs

SG-MSCs were serum starved for 24 h in αMEM and 1% FBS then treated with IFNγ (10 ng/mL) ± ruxolitinib (1000 nM). Four treatment groups were included in the experiment: (i) IFNγ-treated SjD SG-MSCs (n = 4); (ii) IFNγ-treated control SG-MSCs (n = 4); (iii) both IFNγ and ruxolitinib-treated SjD SG-MSC (n = 4); and (iv) both IFNγ and ruxolitinib-treated control SG-MSC (n = 4). After 48 h, the cells were trypsinized and frozen for RNA isolation and RNA-sequencing. Heat maps displayed differentially regulated genes filtered with an adjusted P-value <0.05 and log2 fold change ≥2 for upregulated genes and ≤–2 for down-regulated genes. Gene ontology was performed on genes with an adjusted P-value < 0.05. (A) Differentially expressed genes for SjD SG-MSCs treated with IFNγ vs both IFNγ and ruxolitinib; (B) differentially expressed genes for control SG-MSCs treated with IFNγ vs both IFNγ and ruxolitinib; (C) differentially expressed genes for SjD vs control SG-MSCs treated with IFNγ; (D) differentially expressed genes for SjD vs control SG-MSCs treated with both IFNγ and ruxolitinib; (E) gene ontology analysis shows top three enrichment clusters and the associated p-values within the clusters for SjD SG-MSCs treated with IFNγ vs both IFNγ and ruxolitinib; (F) gene ontology analysis of the top three enrichment clusters and associated P-values within the clusters for control SG-MSCs treated with IFNγ vs both IFNγ and ruxolitinib; (G) gene ontology analysis of the top three enrichment clusters and associated P-values within the clusters for control vs SjD IFNγ treated SG-MSCs. ES, enrichment score. *P < 0.05, **P < 0.01, ***P < 0.001.

GO analysis of SG-MSCs treated with IFNγ compared with those treated with both IFNγ and ruxolitinib showed the top three enriched clusters of differentially expressed genes were composed of (i) interferon signalling; (ii) MHC processing and presentation; and (iii) response to interferon (Fig. 2E and F). Interestingly, GO analysis comparing SjD to control SG-MSCs treated with IFNγ revealed extracellular region, extracellular matrix (ECM) organization, cell adhesion and ion binding, among others, as the top enriched differentially expressed gene clusters (Fig. 2G). These findings indicate that SjD and control SG-MSCs might respond differentially to IFNγ but are similar in the absence of IFNγ stimulation. In all cases, ruxolitinib blocked the MSC’s transcriptomic response cascade to IFNγ stimulation.

Ruxolitinib inhibits IFNγ-induced SG-MSC production of chemokines CCL2, CCL7, CXCL9, CXCL10 and CXCL11

Analysis of the most differentially regulated genes between SG-MSCs treated with IFNγ and those treated with both IFNγ and ruxolitinib identified MSC-secreted chemokines as a biologically relevant group. We used a chemokine array to qualitatively identify the major chemokines produced with IFNγ stimulation of SG-MSCs. We identified upregulated chemokines of interest including CCL2, CCL7, CXCL9, CXCL9, CXCL10, CXCL11, CXCL12, CXCL16, TIG2 and midkine (Fig. 3A–E).

Fig. 3.

Ruxolitinib inhibits SG-MSC IFNγ-induced expression of CCL2, CCL7, CXCL9, CXCL10 and CXCL11

SG-MSCs were serum starved × 24 h, then treated with vehicle, ruxolitinib (1000 nM), IFNγ (10 ng/mL) or both IFNγ and ruxolitinib. After 48 h, the conditioned media were collected for all of the samples. For the chemokine array, the conditioned media for one sample in all four conditions was concentrated to 10× and applied to the chemokine array membrane per the manufacturer’s specifications. For ELISA, conditioned media (1×) from four to eight of the samples was used or the procedure per the manufacturer’s recommendations. (A–D) Chemokine array results for conditioned media from vehicle-treated, ruxolitinib-treated, IFNγ-treated and IFNγ and ruxolitinib-treated SG-MSCs; (E) schematic identifying location of each chemokine on the array; (F–O) ELISAs for CXCL9, CXCL10, CXCL11, CCL2, CCL7, CXCL8, CXCL12, CXCL16, TIG2 and midkine; (P) western blot for STAT1, pSTAT1 and pSTAT2 SG-MSCs treated with IFNγ ± ruxolitinib 1000 nM. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

To quantitatively confirm the chemokines identified by the array, we performed ELISAs. IFNγ-stimulated MSCs had significantly increased protein levels of CCL2, CCL7, CXCL9, CXCL10 and CXCL11. Simultaneous treatment with ruxolitinib completely inhibited this effect, implicating the critical involvement of JAK-STAT signalling in this process (Fig. 3F–J). In contrast to the enhanced secretion of these chemokines, we observed that CXCL8 significantly decreased in control IFNγ-treated SG-MSCs, with a similar trend seen in SjD (Fig. 3K). Although not significant when analysed separately, when SjD and control conditioned media samples were combined, CXCL16 production was significantly different between vehicle and IFNγ-treated MSCs (P < 0.01; Fig. 3L). There was no difference in CXCL12, CXCL16, TIG2 and midkine (Fig. 3M–O). SG-MSCs treated with IFNγ ± ruxolitinib showed IFNγ increased pSTAT1, which was inhibited with ruxolitinib (Fig. 3P). Accordingly, pSTAT1 is the likely mechanism by which IFNγ promotes chemokine expression.

IFNγ amplifies SG-MSC-induced CD4⁺ T cell chemotaxis through CXCL9, -10 and -11 interaction with CXCR3

Inflammatory cell infiltrates, a typical feature of SjD SGs, are mostly composed of CD4⁺ T cells in early lesions but also CD8⁺ T cells and B cells [19]. To investigate the effect of SG-MSCs on PBMC migration, we cultured control SG-MSCs (n = 1) in the bottom of a transwell system until the monolayer became confluent and then treated them with vehicle, IFNγ or IFNγ and GolgiPlug Protein Transport Inhibitor (Fig. 4A). After 24 h, SG-MSCs were washed and human PBMCs from healthy donors (n = 4) were placed in the top well of the transwell. After 5 h, cells remaining in the top well were collected for flow cytometry to determine under which conditions migration occurred most. We found that IFNγ stimulation increased the migration of PBMCs from the top well to the bottom well, indicating IFNγ pre-treated SG-MSCs promote PBMC migration (Fig. 4B), and increased migration was reversed with GolgiPlug Protein Transport Inhibitor.

Fig. 4.

IFNγ amplifies SG-MSC-induced CD4⁺ T cell chemotaxis through CXCL9, -10 and -11

(A) Representative experiment schematic; (B) SG-MSCs (n = 1) were cultured in the bottom of a 5 μm pore transwell system until confluent then they were treated with vehicle, IFNγ (10 ng/mL) or IFNγ and GolgiPlug Protein Transport Inhibitor. After 24 h, the SG-MSCs were washed well and PBMCs from four separate healthy controls were placed in the top well of the transwell system. After 5 h, we collected cells from the top well, stained, and used for flow cytometry; (C) SjD (n = 3) and control (n = 3) SG-MSCs were grown to confluence on the bottom of an 8 μm pore transwell system. We treated the SG-MSCs with IFNγ (10 ng/mL) or IFNγ and ruxolitinib (1000 nM) for 24 h, washed them well, and added PBMCs from one healthy donor to the top well of the transwell system. After 5 h, we harvested the cells for flow cytometry; (D–E) SG-MSCs (n = 1) were grown to confluence on the bottom well of a 5 μm transwell system and pre-treated with or without IFNγ (10 ng/mL). After 24 h, we washed the SG-MSCs well and then we added from healthy controls PBMCs (n = 4) to the top well of the system with media containing neutralizing antibodies to CCL2, CCL7, CXCL9, CXCL10, CXCL11, all neutralizing antibodies combined, or GolgiPlug. After 12 h, the remaining PBMCs that had not migrated were harvested from the top well and subjected to flow cytometry; (F–H) SjD MSCs (n = 1) were pre-treated with or without IFNγ as above and, after 24 h, we applied purified CD4⁺ T cells negatively selected from three SjD donors to the top well of the system. We used media with neutralizing antibodies to CCL2, CCL7, CXCL9, CXCL10, CXCL11, all neutralizing antibodies combined. After 12 h, the remaining CD4⁺ T cells that had not migrated were harvested from the top well for flow cytometry. *P < 0.05.

Next, we investigated the effect of ruxolitinib on IFNγ-induced PBMC migration. We repeated the transwell experiment as above but performed the experiment using SjD (n = 3) and control (n = 3) SG-MSC cell lines to determine whether SG-MSCs from disease or control donors differentially affected PBMC migration. SG-MSCs were pre-treated with IFNγ or IFNγ and ruxolitinib, washed, and PBMCs from one healthy subject were added to the top well of the transwell system. After 5 h, the cells were collected for flow cytometry analysis. IFNγ pre-treated SG-MSCs increased migration of PBMCs but when ruxolitinib was added to the system, there was suppression of migration (Fig. 4C). No difference in PBMC migration between control and SjD SG-MSCs was observed.

Because IFNγ pre-treated SG-MSCs induced PBMC chemotaxis, and GolgiPlug and ruxolitinib inhibits this effect, we postulated that JAK-STAT signalling induced chemokine production may be responsible for observed PBMC migration. Therefore, we sought to determine the mechanisms driving these phenomena. Control SG-MSCs (n = 1) were grown to confluence on the bottom of the transwell system with or without IFNγ. After 24 h, we washed the SG-MSCs and added healthy control PBMCs (n = 4) to the top well of the system with media containing single neutralizing antibodies (anti-CCL2, anti-CCL7, anti-CXCL9, anti-CXCL10 and anti-CXCL11) in different wells, all neutralizing antibodies combined, or GolgiPlug. After 12 h, PBMCs from the top well were harvested and subjected to flow cytometry analysis. CD4⁺ T cells migrated more significantly in response to IFNγ than no treatment and anti-CXCL9, -CXCL10 and -CXCL11 neutralizing antibodies inhibited IFNγ -induced CD4⁺ T cell migration (Fig. 4D). No significant effect on CD8⁺ T cell migration was observed (Fig. 4E). Migration of NK cells was observed in all conditions except with the GolgiPlug (Supplementary Fig. S2A, available at Rheumatology online). We also observed a trend towards increased CD19⁺ B-cell migration with IFNγ-pre-treated SG-MSCs (P = 0.08) that improved most with anti-CXCL9 neutralizing antibody (Supplementary Fig. S2B, available at Rheumatology online).

In order to approximate in vivo disease conditions, we used PBMCs (n = 3) and purified CD4⁺ T cells (n = 3) from SjD subjects in a transwell system with SjD MSCs on the bottom well. We found CD4⁺CXCR3⁺ T cells have increased chemotaxis to IFNγ-treated SG-MSCs. There was a trend towards reduced chemotaxis of CD4⁺CXCR3⁺ T cells with the addition of neutralizing antibodies to CXCL-9, -10 and -11 as well as a significant improvement when all of the neutralizing antibodies were added (Fig. 4F). CD4⁺CXCR3^– T cells showed less robust chemotaxis to IFNγ-treated SG-MSCs and did not respond to the addition of CXCL-9, -10 or -11 neutralizing antibodies (Fig. 4G). When analysed as the proportion CD4⁺CXCR3⁺ T cells/total T cells remaining in the top well, we appreciated significant inhibition of migration in response to CCL7, CXCL-9, -10 and -11 (Fig. 4H). Analysis of SjD PBMCs chemotaxis yielded a trend similar to control PBMCs. Chemotaxis occurred more in CD4⁺ T cells towards IFNγ-treated MSCs and CCL7, CXCL-9, -10 and -11 inhibited migration, though not reaching significance (Supplementary Fig. S2C–G, available at Rheumatology online).

Once we had determined the effects of IFNγ on SG-MSC-induced chemotaxis of PBMCs, we sought to define the effect of IFNγ on MSC movement. To accomplish this, we grew SjD (n = 3) and control (n = 3) SG-MSCs to confluence, scratched the surface and measured the rate of wound healing. We found that IFNγ led to a non-significant trend in slower SG-MSC migration that was reversed when MSCs were cultured with IFNγ and ruxolitinib in both SjD (P = 0.2) and control (P = 0.3) SG-MSCs (Supplementary Fig. S3A–C, available at Rheumatology online). There was no difference in SG-MSC migration when comparing SjD and control SG-MSCs.

CXCL9, CXCL11 and CXCR3 are upregulated in SG tissue

Because IFNγ pretreated SG-MSCs promote CD4⁺ T cell migration in vitro through CXCR3, we tested in vivo conditions by qPCR on whole SG tissue. CXCL9 and CXCL11 were increased in SjD SG tissue compared with control SGs. CXCL10 expression showed a non-significant trend towards greater expression in SjD compared with control SGs (Fig. 5A). CXCR3 was also increased in SjD SG tissue (Fig. 5B). ACKR1 and ACKR3, decoy receptors for CXCL11, were similarly expressed in SjD and control SGs (Fig. 5B) [20].

Fig. 5.

CXCL9, CXCL11 and CXCR3 are upregulated in SG tissue

Real-time quantitative PCR (qPCR) was completed on RNA from seven control and nine SjD SGs for the indicated genes. Data show fold change (FC) in the SjD sample vs the average healthy control delta CT value. (A) FC of CCL2, CCL7, CXCL9, CXCL10 and CXCL11 in SjD SG tissue compared with control. (B) FC of receptors CXCR3, ACKR1 and ACKR3 in Sjd SG tissue compared with control. Each dot represents the average of triplicate repeats. Significance was analysed using ΔCTs for comparison via Student’s unpaired t test. *P < 0.05.

Discussion

IFNγ promotes a pathophysiologic response in SG-resident MSCs that is reversed with the selective JAK1&2 inhibitor, ruxolitinib. IFNγ stimulates SG-MSCs to produce chemokines that increase CD4⁺ T cell chemotaxis through interaction with CXCR3, potentiating a positive feedback loop involving IFNγ-producing CD4⁺ T cells (Fig. 6). Further, IFNγ promotes the expression of HLA-DR, potentially allowing SG-MSCs to act as antigen presenting cells. Overall, IFNγ promotes a pathophysiologic response in SG-resident MSCs that was suppressed by the application of the selective JAK1&2 inhibitor, ruxolitinib. Thus, we describe a biologically plausible physiologic response to pathologic IFNγ in SG-MSCs that is potentially treatable with ruxolitinib.

Fig. 6.

SG-resident mesenchymal stromal cells promote PBMC chemotaxis and IFNγ amplifies chemotaxis further

SG-resident MSCs promote a positive feedback loop driving T cell infiltrate in SjD. (A) Local IFNγ, higher in SjD, drives SG-MSCs to express HLA-DR and CXCL9, 10 and 11. HLA-DR expression by SG-MSCs suggests the possibility they drive local T cell proliferation with an APC-like function; (B) increased CXCL9, -10 and -11 increases CD4⁺ T cell chemotaxis to the gland; (C) increased proliferation and migration of T cells amplifies IFNγ levels, creating a positive feedback loop contributing to SjD pathogenesis. Ruxolitinib interrupts the positive feedback loop, supporting further study of ruxolitinib in SjD.

IFNγ treated SG-MSCs promote CD4⁺ T cell migration through secretion of CXCL9, -10 and -11, and this process is reversed by ruxolitinib. Increased CD4⁺ T cell migration is a salient finding because early SG infiltrate is predominantly composed of CD4⁺ T cells [19]. Previously, other cell types have been implicated in aberrant T cell migration in SjD. For example, Aota et al. showed that CXCL10 production by SjD ductal epithelial cells is reversed with baricitinib, another selective JAK1&2 inhibitor [21]. They also showed the baricitinib reversed IFNγ-induced T cell migration in conditioned media derived from SG epithelial cells treated with IFNγ. Indeed, baricitinib is currently being studied as a therapy in SjD (NCT05016297) but, unlike ruxolitinib, baricitinib is associated with thrombosis [22]. Interestingly, anti-CXCR3 antibody failed to improve lymphocytic infiltrate but improved salivary flow in mouse models, emphasizing the complexity of SjD pathogenesis [23].

We confirmed SG-MSCs produce CXCL9, -10 and -11 upon IFNγ exposure. Multiple studies have reported involvement of CXCL10, -9 and -11 in SjD. We found a trend towards increased SG tissue CXCL10 akin to what others have previously described; however, possibly due to a smaller sample size, this finding did not achieve significance [24]. Interestingly, CXCL10 is increased in preclinical SjD, potentially reflecting CXCL10’s earlier involvement in SjD pathogenesis [25]. Further supporting the role of CXCL10 in sialadenitis, MRL/lpr mice injected with a CXCL10 antagonist had reduced Th1 CXCR3⁺ infiltrate and parenchymal destruction, leading to lower IFNγ production [26]. In addition to CXCL10, CXCL9 is increased in SjD whole salivary glands and within salivary gland epithelial cells in response to IFNγ [27, 28]. In SjD, most infiltrating periductal lymphocytes in SG tissue are CXCR3⁺, indicating roles for CXCL9, -10 and -11 [27]. CXCL11 is located in the ductal epithelium adjacent to lymphoid infiltrates in SjD but is absent in control SGs [29]. Serum CXCL11 has been identified as predictive of SjD when combined with low concentrations of sCD163 [30]. In addition to gland tissue, CXCL9, -10 and -11 are present in tear film and on the eye surface in SjD [31].

We found that stimulated SG-MSCs enhanced the production of CCL2. CCL2 is higher in SjD SG tissue, serum and saliva than in controls [25, 32]. Moreover, SjD subjects with SG germinal centres have higher CCL2 levels than those without [33, 34]. As with CXCL9, -10 and -11, CCL2 is expressed in ductal structures, and its expression is increased in vitro when SG epithelial cells are treated with IFNγ [35]. Interestingly, and relevant to SjD epidemiology notable for female predominance and perimenopausal disease onset, aromatase knockout mice show increased sialadenitis and CCL2 in gland tissue [36]. This potentially indicates that post-menopausal status might drive CCL2 expression and sialadenitis in SjD.

Less is known regarding CCL7 and CXCL8 in SjD pathogenesis; CCL7 is increased in submandibular glands of NZB/W mice treated with poly (I: C) [37]. CXCL8 is increased in SjD plasma and saliva but does not correlate with focus score [34, 38]. We found CXCL8 is reduced upon IFNγ stimulation in controls, but this effect is less pronounced in SjD. It is possible that the diminished CXCL8 reduction in response to IFNγ might also be involved in SjD pathogenesis by impairing phagocytosis.

In addition to chemokine production, MSCs act as homeostatic immunomodulators through pleiotropic cell–cell interactions. In bone marrow, adipose and gingiva, IFNγ induces MSCs to expand regulatory T cells and suppress cytotoxic T cell proliferation, and this process is mediated by factors including IDO and PD-L1 [15, 17, 39, 40]. On the other hand, IFNγ-stimulated MSCs are known to promote immune activation through upregulation of MHC class II and antigen-presenting cell function, promoting an adaptive immune response [18, 41–44]. Our results using SG-MSCs agree with previous observations in bone marrow and adipose MSCs: IFNγ-treated SG-MSCs express increased HLA-DR, IDO, ICAM and PD-L1. Therefore, it is possible that, under inflammatory conditions SG-MSCs act as antigen-presenting cells and exacerbate, as opposed to ameliorate, the local immune environment. We found that the JAK1&2 inhibitor ruxolitinib reverses the expression of these markers, potentially representing a therapeutic modality towards regulating SG-MSC immunomodulatory function.

We previously reported that SjD and control culture adapted SG-MSCs behave similarly, aside from increased expression of ECM markers in SjD (a fibroblast-like phenotype) [7]. Here, we performed RNA-seq to distinguish key differences between IFNγ-treated SjD and control SG-MSCs. IFNγ-treated SjD and control SG-MSCs differentially respond to IFNγ. In contrast, SG-MSCs treated with both ruxolitinib and IFNγ show little differential gene expression between SjD and controls. Top differentially expressed genes included ECM and motility enrichment clusters, analogous to our previously reported data. IFNγ might amplify differences in ECM and motility, a process reversed by ruxolitinib. We found that IFNγ created a trend towards slower migration in wound healing; however, further studies are needed to determine whether this finding imparts impaired SG-MSC immunomodulatory or regenerative capacity. Because of the clinical heterogeneity of SjD, additional studies should be performed to determine whether our findings can be applied to other SjD subtypes.

In conclusion, we identified SG-MSCs as additional players in the complex pathogenesis of SjD. IFNγ-treated SG-MSCs promote chemotaxis of CD4⁺ T cells through CXCL9, -10 and -11 expression and by acting as antigen-presenting cells. Importantly, we identified ruxolitinib, a JAK1&2 inhibitor without known thrombotic side effects, as a drug capable of modifying the pro-inflammatory response of SG-MSCs to IFNγ.

Data availability statement

Data available on request.

Supplementary data

Supplementary data are available at Rheumatology online.