Notch ligands are biomarkers of anti-TNF response in RA patients

Abstract

Notch and its ligands play a critical role in rheumatoid arthritis (RA) pathogenesis. Hence studies were conducted to delineate the functional significance of the Notch pathway in RA synovial tissue (ST) cells and the influence of RA therapies on their expression. Morphological studies reveal that JAG1, DLL4, and Notch1 are highly enriched in RA ST lining and sublining CD68⁺CD14⁺ MΦs. JAG1 and DLL4 transcription is jointly upregulated in RA MΦs reprogrammed by TLR4/5 ligation and TNF, whereas Syntenin-1 exposure expands JAG1, DLL4, and Notch1 expression levels in these cells. Single-cell RNA-seq data exhibit that JAG1 and Notch3 are overexpressed on all fibroblast-like synoviocyte (FLS) subpopulations, in parallel, JAG2, DLL1, and Notch1 expression levels are modest on RA FLS and are predominately potentiated by TLR4 ligation. Intriguingly, JAG1, DLL1/4, and Notch1/3 are presented on RA endothelial cells, and their expression is mutually reconfigured by TLR4/5 ligation in the endothelium. Synovial JAG1/JAG2/DLL1 or Notch1/3 transcriptomes were unchanged in patients who received disease-modifying anti-rheumatic drugs (DMARDs) or IL-6R Ab therapy regardless of disease activity score. Uniquely, RA MΦ and endothelial cells rewired by IL-6 displayed DLL4 transcriptional upregulation, and IL-6R antibody treatment disrupted RA ST DLL4 transcription in good responders compared to non- or moderate-responders. Nevertheless, the JAG1/JAG2/DLL1/DLL4 transcriptome was diminished in anti-TNF good responders with myeloid-pathotype and was unaltered in the fibroid-pathotype except for DLL4. Taken together, our findings suggest that RA myeloid Notch ligands can serve as markers for anti-TNF responsiveness and trans-activate Notch receptors expressed on RA FLS and/or endothelial cells.

INTRODUCTION

Rheumatoid arthritis (RA) is an inflammatory autoimmune disease, in which elevated migration of myeloid and lymphoid cells and their crosstalk with fibroblast-like synoviocytes (FLS) advances neovascularization and joint damage [1, 2]. Notch ligands (JAG1/JAG2/DLL1/DLL4) and receptors (Notch1/3) are implicated in diverse facets of RA and preclinical pathology. Distinct from soluble ligands, Notch ligands, and their receptors are transmembrane proteins with large extracellular domains [3]. Notch signaling is activated by ligand binding from neighboring cells, conversely, Notch ligands binding to the receptor within the same cell impairs Notch-mediated function [4]. In the canonical pathway, upon ligation, Notch receptors are cleaved twice, initially by ADAM17 and later by γ-secretase complex, and as a result Notch intracellular domain (NICD) is released into the nucleus where it controls transcriptional responses.

The use of γ-secretase inhibitor or Notch anti-sense transgenic mice disrupts Notch-potentiated arthritogenicity in collagen-induced arthritis (CIA) by mitigating joint inflammation via TNF, CCL2, and IL-12 reduction and NF-κB deactivation [5]. Others have shown Notch1 and Notch3 intracellular domains (N1ICD and N3ICD) are overexpressed in RA synovial tissue (ST) FLS in response to hypoxia [6]. Consequently, HIF1α knockdown disrupted the expression of Notch1 and Notch3 as well as N1ICD and N3ICD in RA FLS exposed to hypoxia [6]. Distinctly, while Notch1 was involved in RA FLS migration during hypoxia, Notch3 was responsible for promoting RA FLS survival [6].

More recent studies have indicated that Notch1, Notch3, and their ligands JAG1, JAG2, and DLL4 facilitate FLS and endothelial cell cross-signaling [7]. Intriguingly, JAG1 and DLL4 can reprogram RA FLS to THY1^high phenotype, thus Notch landscape is enriched with THY1^high compared to THY1^low FLS subtypes [7]. The investigators revealed that the proximity of endothelial cells to RA FLS played a pivotal role in RA FLS THY1^high remodeling, fostered by Notch ligands and receptors. In line with these observations, KxB/N-induced arthritis was attenuated in Notch3−/− or arthritic mice who received Notch3 neutralizing antibodies compared to the control group [7].

Conversely, in CIA, soluble JAG1 intramuscular administration delayed arthritis onset by suppressing CD8⁺ T cell differentiation without altering CD4⁺ T cell development [8]. Other groups have exhibited that γ-secretase inhibitor expands Treg remodeling in CIA [9]. Although all Notch receptors were expressed in RA peripheral blood, the frequency of Notch3 in activated T cells was exclusively linked with disease activity score [10]. Extending these observations, stimulation with anti-CD3/anti-CD28 antibodies escalated the expression of Notch1 and its downstream target (HES) in RA T-cells [10]. Moreover, in individuals with RA susceptibility locus rs874040, stimulation with JAG-Fc or DLL-Fc fusion protein reconfigured the naïve T cells to Th17 cells [11]. These findings suggested that Notch ligation is involved in RA effector T-cell pathology.

The impact of Notch ligands and receptors has been examined in connection with RA FLS, endothelial cells, and T cells. However, the influence of RA therapies on Notch ligands/receptors or their functional significance in RA macrophages (MΦs) is undefined. Morphological and expression studies show for the first time that JAG1, DLL4, and Notch1 are highly enriched in RA ST lining and sublining CD68⁺CD14⁺ MΦs. Interestingly, transcription of JAG1 and DLL4 is mutually upregulated by TLR4/5 ligands and TNF, whereas Syntenin-1 can expand JAG1, DLL4, and Notch1 transcription in RA monocyte-differentiated MΦs. Unexpectedly, treatment with disease-modifying anti-rheumatic drugs (DMARDs) or IL-6R antibody (Ab) is ineffective on RA ST JAG1/JAG2/DLL1 or Notch1/3 transcriptome by RNA-seq analysis. Uniquely, DLL4 transcription levels are responsive to IL-6 stimulation in RA MΦs and endothelial cells, and IL-6R Ab therapy represses RA ST DLL4 transcriptome in good responders. Nonetheless, JAG1/JAG2/DLL1/DLL4 transcription was dysregulated in anti-TNF good responders with myeloid-pathotype and was unchanged in the fibroid-pathotype except for DLL4. Aside from JAG1 which is overexpressed on all FLS subpopulations (F1 to F4, by single-cell RNAseq), JAG2 and DLL1/DLL4 are modestly presented on RA FLS but their expression could be enhanced by TLR4 ligation and TNF exposure in these cells. Taken together our study reveals that expression of Notch ligands on RA MΦs can function as biomarkers for anti-TNF response and potentially trans-activate Notch receptors expressed on RA FLS and/or endothelial cells.

METHODS AND MATERIALS

Cells.

Peripheral blood samples from rheumatoid arthritis (RA) patients were collected according to the protocol approved by the University of Illinois at Chicago Institutional Ethics Review Board. RA patients were diagnosed according to the 1987 revised criteria of ACR [12]. All patients gave written informed consent before blood was drawn. Peripheral blood mononuclear cells (PBMCs) were isolated by density gradient centrifugation using Ficoll-Paque PREMIUM (GE Healthcare) and subsequently used for further analysis. FLS from fresh RA ST were isolated by mincing and digestion in a solution of dispase, collagenase, and DNase [13]. Cells were used between passages 3 and 9. Human umbilical vein endothelial cells (HUVECs) were purchased from Lonza and used between passages 3 and 9 [14].

Cell stimulation.

RA macrophages (MΦs), RA FLS, or HUVECs were serum starved overnight and then were either untreated (PBS) or treated with LPS/IFNγ (100 ng/mL, Sigma, R&D Systems), TNFα (100 ng/mL, R&D Systems), IL-1β (100 ng/mL, R&D Systems), IL-6 (100 ng/mL, R&D Systems), flagellin (100 ng/mL, Invivogen), Syntenin-1 (1000 ng/mL, NKMAX Co.), or GM-CSF (100 ng/mL, Biolegend) for 6h. Cells were subsequently harvested in TRIzol reagent (Life Technologies) for mRNA isolation and quantification.

Immunohistochemistry.

RA ST formalin-fixed, paraffin-embedded samples were sectioned and stained for colocalization of Jag1 (1:50), Jag2 (1:50), DLL1 (1:50), DLL4 (1:50), Notch1 (1:50), and Notch3 (1:50) on CD68⁺ or CD14⁺ RA macrophages (1:100), VWF⁺ endothelial cells (1:1000) and Vimentin⁺ RA FLS (1:1000). Fluorescence secondary anti-rabbit (1:200) and anti-mouse (1:200) antibodies were utilized to visualize staining.

Normal (NL), osteoarthritis (OA), and RA ST formalin-fixed, paraffin-embedded samples were sectioned and stained for H&E using Jag1 (1:50), Jag2 (1:50), DLL1 (1:50), DLL4 (1:50), Notch1 (1:50), and Notch3 (1:50) antibodies, were scored on a scale of 0–5 in a blinded manner (0 = normal appearance, 1 = minimal changes, 2 = mixed appearance, 3 = moderate changes, 4 = marked changes, and 5 = severe changes), and distinguished within the synovial lining, sublining, and vasculature.

Quantitative RT-PCR.

According to the manufacturer’s instructions, RNA was isolated using a TRIzol reagent (Life Technologies). RNA was reverse transcribed to cDNA, and qRT-PCR analysis was performed using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems) and TaqMan Gene Expression Master Mix (Applied Biosystems). Predesigned IDT primers were used (Table 1). Data are presented as fold change (2^−ΔΔCt) normalized to the housekeeping gene (β-actin) and compared to the untreated control. Data were acquired with the QuantStudio5 (Applied Biosystems) qRT-PCR machine.

Bulk and scRNAseq transcriptome analysis.

The RNA-seq dataset GSE198520 was accessed using the web interface https://www.ncbi.nlm.nih.gov/geo/geo2r/?acc=GSE198520/ deposited by Wang et al. [15] to evaluate the expression of JAG1, JAG2, DLL1, DLL4, Notch1, and Notch 3 in RA synovium biopsied from 46 RA patients 12 weeks after treatment with anti-TNF. Patients in this cohort fulfilled the 2010 ACR/EULAR RA Classification Criteria and were enrolled at the Centre for Experimental Medicine and Rheumatology, Barts and The London School of Medicine, Queen Mary University of London, UK. RA patients exhibited clinically defined synovitis and fit the criteria for UK NICE guidelines (failure of at least 2 csDMARDs and DAS28 >= 5.1) to start anti-TNF therapy. Following enrollment, patients underwent minimally invasive US-guided synovial biopsy of the most inflamed joint (ultrasound synovial thickening score >= 2). The patient data were grouped and displayed based on whether the patients were considered non-responders, moderate responders, or good responders to anti-TNF therapy. Response to therapy was evaluated using ACR/EULAR DAS28 response criteria defined as good response (ΔDAS [DAS28 at baseline – DAS28 at 12 weeks after treatment] >1.2 with DAS28 at 12 weeks ≤ 3.2), moderate response (DAS28 change >1.2 with DAS28 at 12 weeks >3.2, or DAS28 change 0.6–1.2 with DAS28 at 12 weeks ≤5.1), or nonresponse (DAS28 ≤0.6, or DAS28 change 0.6–1.2 with DAS28 at 12 weeks > 5.1) [16]. Data were further separated based on whether the patient had myeloid or fibroid pathotype and whether they were non-responders, moderate responders, or good responders to anti-TNF treatment.

The web interface https://peac.hpc.qmul.ac.uk/ developed by Lewis et al. [17] was used to evaluate the expression of JAG1, JAG2, DLL1, DLL4, Notch1, and Notch3 in synovial tissues from early RA in the PEAC study. The Pathobiology of Early Arthritis Cohort consisted of 90 rheumatoid arthritis patients fulfilling 2010 ACR/EULAR RA Classification Criteria. All patients had clinically defined synovitis with symptoms that were present for less than 12 months. Patients receiving corticosteroids, sDMARDs, or biologic therapies were excluded. Detailed methods for the generation of data used in this web interface have been published previously [17, 18]. Gene transcript expression levels are expressed as variance stabilizing transformation transformed read counts using the Bioconductor package DESeq2. The raw RNA-Seq data have been deposited at ArrayExpress accession E-MTAB-6141. The patient data were grouped based on response to therapy using the ACR/EULAR DAS28 response criteria as described above. The web interface https://r4ra.hpc.qmul.ac.uk/ developed by Rivellese et al. [19] was used to evaluate the expression of JAG1, JAG2, DLL1, DLL4, Notch1, and Notch3 in synovial tissue from RA patients that were treated with Tocilizumab. A cohort of 164 patients aged 18 years or over who fulfilled the 2010 American College of Rheumatology/European Alliance of Associations for Rheumatology (EULAR) classification for RA and were eligible for treatment with rituximab therapy according to UK NICE guidelines (patients who failed or were intolerant to csDMARD therapy and at least one biologic therapy) were included in the trial. Initially, a synovial biopsy was taken of a clinically active joint at the beginning of the trial. Patients were then randomized to tocilizumab treatment administered at 8 mg/kg tocilizumab at 4-week intervals. Detailed methods for the generation of data used in this web interface have been published previously [19]. The patient data were grouped based on response to therapy using the ACR/EULAR DAS28 C reactive protein (CRP) response criteria as described above.

The single-cell RNA sequencing data from Zhang et al. [20] was accessed from the Broad Institute Single Cell portal at the following URL: https://singlecell.broadinstitute.org/single_cell/study/SCP279/amp-phase-1. This cohort consisted of RA patients who were older than 18 years and had at least one inflamed joint who were recruited from 10 sites in the network. These patients were undergoing elective surgical procedures when synovial biopsy specimens were obtained.

The single-cell RNA sequencing data from Wei et al. [7] was accessed from the Broad Institute Single Cell portal at the following URL: https://singlecell.broadinstitute.org/single_cell/study/SCP469/synovial-fibroblast-positional-identity-controlled-by-inductive-notch-signaling-underlies-pathologic-damage-in-inflammatory-arthritis. A cohort of RA patients that fulfilled the ACR 2010 Rheumatoid Arthritis classification criteria were included. Synovial tissue samples were acquired when the patients underwent either joint replacement or synovectomy procedures.

Statistical Analysis.

For comparison among multiple groups, one-way ANOVA followed by Tukey’s multiple comparisons test was employed using GraphPad Prism9 software. For RNA-seq data, the p-value was adjusted for multiple comparisons by FDR. For comparisons between the two groups, a Mann-Whitney test for unpaired data was utilized. When comparing RNA-Seq data against continuous or ordinal variables, the Spearman rank correlation test was used, and Spearman rho and p-values were shown. p < 0.05 was considered statistically significant.

RESULTS

JAG1 is expressed in RA synovial tissue MΦs and its expression is responsive to anti-TNF therapy

Differential expression of JAG1 was determined in RA compared to osteoarthritis (OA) and normal (NL) STs. Morphological studies revealed that JAG1 was enriched in RA compared to OA and NL ST lining and sublining cells (Fig. 1B). Nonetheless, JAG1 was similarly expressed on RA and OA endothelial cells which were higher than those on NL counterparts (Fig. 1B). RA synovial JAG1 transcriptome was unaffected by DMARD or IL-6R Ab (Tocilizumab) therapy (Fig. 1C–D). On the contrary, the JAG1 transcriptome was negated in anti-TNF good responders (ΔDAS28>1.2) compared to those with non- (ΔDAS28≤0.6) or moderate-response (ΔDAS28≤1.2&>0.6) (Fig. 1E). Notably, the reduced JAG1 transcriptome in the anti-TNF good responders was observed in myeloid but not fibroid pathotypes (Fig. 1F–G). Authenticating the RNA-seq data, immunofluorescence (IF) studies uncovered that JAG1 was presented on RA ST CD68⁺CD14⁺ MΦs (Fig. 1H, Suppl. Fig. 2A). TLR4/5 ligation, TNFα, Syntenin-1 (Syn), and GM-CSF stimulation (but not IL-1β) were found to be responsible for reprogramming naïve cells into RA MΦs that displayed JAG1 transcriptional upregulation (Fig. 1A,I, Suppl. Fig. 2F). Moreover, JAG1 was highly expressed in all 4 subpopulations (F1 to F4) of RA lining and sublining FLS and VWF⁺ RA endothelial cells as determined by IF and single-cell RNA-seq (Fig. 1J–K). Unlike RA FLS, endothelial cells reconfigured by TLR4, TNFα, or IL-1β activation exhibited amplified JAG1 transcription (Fig. 1A,L). Overall, our findings demonstrate that while JAG1 is expressed on RA MΦs, FLS, and endothelium, its expression in myeloid cells is influenced by TLRs and specific monokines and can be indicative of anti-TNF responsiveness.

Figure 1. Synovial JAG1 expression was suppressed in anti-TNF good responders.

A) Graphical summary of Notch ligand/receptor expression induced by TLR-signaling or cytokines in RA MΦs, FLS, or endothelial cells. B) Synovial tissue sections from NL, OA, and RA were stained for JAG1 (original magnification x10) and the staining was scored on a 0–5 scale (n = 4). C) Relative expression of JAG1 was determined by RNA-seq [17] in ST biopsies from RA non-responsive (ΔDAS28 ≤ 0.6, n = 23), moderate (ΔDAS28 ≤ 1.2 & > 0.6, n = 29) and good responders (ΔDAS28 > 1.2, n = 29) to DMARD therapies. D) RA STs from patients (n = 81) that were either non-responders, moderate responders, or good responders to anti-IL6R Ab (Tocilizumab) therapy were assessed by RNA-seq [19] for JAG1 expression. E-G) RA STs from patients (n = 46) that were non-responders, moderate responders, or good responders to anti-TNF therapy were assessed by RNA-seq [15] for differences in JAG1 expression between groups. RA STs from patients with F) myeloid pathotype (n=21) or G) fibroid pathotype (n=8) were further delineated based on response to therapy and JAG1 expression. H) RA STs were fluorescently stained for JAG1 expression on CD68⁺ RA MΦs, Vimentin⁺ RA FLS, and VWF⁺ endothelial cells (n=3, original magnification x20). I) RA MΦs were stimulated with 100 ng/ml inflammatory factors or 1000 ng/ml Syntenin-1 (SYN1) (6h) and transcription of JAG1 was assessed by qRT-PCR (n=5–13). J) scRNA-seq analysis was performed on RA ST (n = 18) [20]. Clustering revealed 4 fibroblast subpopulations [three types of THY1⁺ sublining fibroblasts (F1-CD34⁺, F2-HLA-DRA^hi, and F3-DKK3⁺) and CD55⁺ lining fibroblasts (F4)] and 4 MΦ subpopulations (M1-IL1B⁺, M2-NUPR1⁺, M3-C1QA⁺, and M4-IFN-activated). Distinct patterns of expression are displayed amongst cell types and subpopulations for JAG1. K) Normalized expression levels of JAG1 are displayed for RA lining and sublining FLS and endothelial cells based on scRNA-seq data from Wei et al. [7]. L) Endothelial cells were stimulated with 100 ng/ml inflammatory factors (6h) and transcription of JAG1 was assessed by qRT-PCR (n=6). Data are presented as mean ± SEM; differences were determined by the Mann-Whitney test, one-way ANOVA with either Tukey post-test, or the p-value was adjusted for multiple comparisons by FDR. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001

Synovial JAG2 and DLL1 transcription levels are markers for anti-TNF good response in patients with myeloid pathotype

Histological studies delineate that JAG2 and DLL1 are differentially overexpressed on RA compared to OA and NL STs (Fig. 2A, 2B, 2D, 2E). We show that JAG2⁺ cells reside in RA ST sublining and endothelial cells, whereas DLL1⁺ cells are predominately presented on RA ST lining compared to OA and NL counterparts (Fig. 2B, 2E). Immunofluorescence and/or RNA-seq studies further indicate that JAG2 is largely expressed on sublining CD68⁺CD14⁺ MΦs and VWF⁺ endothelial cells over FLS (Fig. 2C, 2G, Suppl. Fig. 1D–E). However, DLL1 is similarly expressed on RA ST CD68⁺CD14⁺ MΦs, Vimentin⁺ FLS, and VWF⁺ endothelial cells (Fig. 2F, 2H, Suppl. Fig. 1I–J). Consistent with JAG1, synovial JAG2 and DLL1 were unchanged in response to DMARDs and Tocilizumab therapy irrespective of DAS28 (Suppl. Fig. 1A–B, 1F–G). In patients with anti-TNF good response (ΔDAS28>1.2), synovial JAG2 and DLL1 transcription levels were downregulated in myeloid, unlike fibroid pathotype, despite both being elevated by TNFα and TLR4 stimulation in RA FLS (Fig. 1A, 2I–L, 2N, 2P). In RA MΦs, TNF and/or TLR4/Syntenin-1 exposure differentially escalated JAG2 or DLL1 transcription (Fig. 2M, 2O, 1A). Nonetheless, DLL1 transcription was mutually augmented by TLR4 ligation in RA MΦs, FLS, and endothelial cells (Fig. 2O–Q, 1A). Taken together, MΦs in RA ST are vital for JAG2 or DLL1-mediated pathology and their expression can be dysregulated in anti-TNF-responsive patients.

Figure 2. JAG2 and DLL1 expression on synovial tissue acts as a biomarker for anti-TNF good responders.

A) Synovial tissue sections from NL, OA, and RA were stained for JAG2 (original magnification x10), and B) the staining was scored on a 0–5 scale (n = 4). C) RA STs were fluorescently stained for JAG2 expression on CD68⁺ RA MΦs and VWF⁺ endothelial cells (n=3, original magnification x20). D) Synovial tissue sections from NL, OA, and RA were stained for DLL1 (original magnification x10), and E) the staining was scored on a 0–5 scale (n = 4). F) RA STs were fluorescently stained for DLL1 expression on CD68⁺ RA MΦs and VWF⁺ endothelial cells (n=3, original magnification x20). G-H) Normalized expression levels of G) JAG2 and H) DLL1 are displayed for RA lining and sublining FLS and endothelial cells based on scRNA-seq data from Wei et al. [7]. I-L) RA STs from patients (n = 46) that were non-responders, moderate responders, or good responders to anti-TNF therapy were assessed by RNA-seq [15] for differences in I) JAG2 and K) DLL1 expression between groups. RA STs from patients with myeloid pathotype (n=21) were further assessed based on response to therapy and I) JAG2 and L) DLL1 expression. M) RA MΦs and N) RA FLS were stimulated with 100 ng/ml inflammatory factors (6h) and transcription of JAG2 was assessed by qRT-PCR (n=5–10). O) RA MΦs, P) RA FLS, or Q) endothelial cells were stimulated with 100 ng/ml inflammatory factors or 1000 ng/mL Syntenin-1 (SYN1) (6h), and transcription of DLL1 was measured by qRT-PCR (n=4–6). Data are presented as mean ± SEM; differences were determined by the Mann-Whitney test, one-way ANOVA with either Tukey post-test, or the p-value was adjusted for multiple comparisons by FDR. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001

Expression of DLL4 is modulated by IL-6R Ab and anti-TNF therapies

Uniquely, the expression of DLL4 is enriched in RA compared to OA and NL lining, sublining, and endothelial cells (Fig. 3A). Immunofluorescence and/or RNA-seq studies further substantiate DLL4 overexpression on CD68⁺ MΦs, Vimentin⁺ FLS, and VWF⁺ endothelial cells in RA STs (Fig. 3B–C). Transcription of DLL4 was jointly advanced in RA MΦs and endothelial cells reprogrammed by TLR4 and IL-6 but not IL-1β, while TLR5 ligation, TNFα or Syntenin-1 stimulation also upregulated myeloid DLL4 mRNA levels (Fig. 3D–E, 1A, Suppl. Fig. 2I). In contrast to DMARDs, IL-6R Ab and anti-TNF therapies in good responders restrained DLL4 transcription (Fig. 3F–H). Distinct from other Notch ligands, DLL4 transcriptome was downregulated in the ST of myeloid and fibroid pathotypes in anti-TNF good responders (Fig. 3I–J). Altogether, our data indicate that ST DLL4 transcription is exceptionally responsive to anti-IL-6R and anti-TNF therapies in patients with myeloid and/or fibroid RA pathotypes.

Figure 3. DLL4 expression is uniquely mitigated in anti-IL6R Ab and anti-TNF good responders.

A) Synovial tissue sections from NL, OA, and RA were stained for DLL4 (original magnification x10) and the staining was scored on a 0–5 scale (n = 4). B) RA STs were fluorescently stained for DLL4 expression on CD68⁺ RA MΦs, Vimentin⁺ RA FLS, and VWF⁺ endothelial cells (n=3, original magnification x20). C) Normalized expression levels of DLL4 are displayed for RA lining and sublining FLS and endothelial cells based on scRNA-seq data from Wei et al. [7]. D) RA MΦs and E) endothelial cells were stimulated with 100 ng/ml inflammatory factors or 1000 ng/ml Syntenin-1 (SYN1) (6h) and transcription of DLL4 was assessed by qRT-PCR (n=5–8). F) Relative expression of DLL4 was determined by RNA-seq [17] in ST biopsies from RA non-responsive (ΔDAS28 ≤ 0.6, n = 23), moderate (ΔDAS28 ≤ 1.2 & > 0.6, n = 29) and good responders (ΔDAS28 > 1.2, n = 29) to DMARD therapies. G) RA STs from patients (n = 81) that were either non-responders, moderate responders, or good responders to anti-IL6R Ab (Tocilizumab) therapy were assessed by RNA-seq [19] for DLL4 expression. H-J) RA STs from patients (n = 46) that were non-responders, moderate responders, or good responders to anti-TNF therapy were assessed by RNA-seq [15] for differences in H) DLL4 expression between groups. RA STs from patients with I) myeloid pathotype (n=21) or J) fibroid pathotype (n=8) were further delineated based on response to therapy and DLL4 expression. Data are presented as mean ± SEM; differences were determined by the Mann-Whitney test, one-way ANOVA with either Tukey post-test, or the p-value was adjusted for multiple comparisons by FDR. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001

DMARDs and RA biotherapies are ineffective on synovial Notch1 and Notch3 expression

Similar to DLL4, histological studies display that Notch1 is highly presented in RA compared to OA and NL lining, sublining, and endothelial cells (Fig. 4A–B). Nonetheless, Notch3 is greatly expressed on RA relative to OA and NL lining and endothelial cells (Fig. 4C–D). Single-cell RNA-seq data corroborate that Notch1 is expressed on all RA MΦ subtypes (M1-M4), while both Notch1 and Notch3 are presented on all FLS subpopulations (Fig. 4E–H). Based on single-cell RNA-seq data, Notch3 expression on RA ST endothelial cells was lower relative to Notch1 and other Notch ligands. Morphological studies revealed overexpression of Notch1/3 on CD68⁺CD14⁺ MΦs, Vimentin⁺ FLS, and VWF⁺ endothelial cells in RA STs (Fig. 4I–J, Suppl. Fig. 2D–E). Whereas both Notch1/3 transcription was mutually upregulated by flagellin in endothelial cells, their induction in RA MΦs and FLS was distinctively reconfigured (Fig. 4K–P, 1A). In contrast to Notch ligands, the expression of Notch1 and Notch3 was unaffected in response to DMARDs, IL-6R Ab, and anti-TNF in RA STs (Fig. 4Q–T, Suppl. Fig. 1K–P). Overall, our data suggest dysregulation of myeloid JAG1/2 and DLL1/4 expression by anti-TNF therapy in good responders can interfere with the transactivation of Notch1/3.

Figure 4. Synovial Notch1 and Notch3 receptor expression was unaffected by anti-TNF therapy.

A-D) Synovial tissue sections from NL, OA, and RA were stained for A) Notch1 and C) Notch3 (original magnification x10), and the B) Notch1 and D) Notch3 staining was scored on a 0–5 scale (n = 4–6). E) scRNA-seq analysis was performed on RA ST (n = 18) [20]. Clustering revealed 4 fibroblast subpopulations [three types of THY1⁺ sublining fibroblasts (F1-CD34⁺, F2-HLA-DRA^hi, and F3-DKK3⁺) and CD55⁺ lining fibroblasts (F4)] and 4 MΦ subpopulations (M1-IL1B⁺, M2-NUPR1⁺, M3-C1QA⁺, and M4-IFN-activated). Distinct patterns of expression are displayed amongst cell types and subpopulations for Notch1. F) Normalized expression levels of Notch1 are displayed for RA lining and sublining FLS and endothelial cells based on scRNA-seq data from Wei et al. [7]. G) scRNA-seq analysis was performed on RA ST (n = 18) [20]. Distinct patterns of expression are displayed amongst cell types and fibroblast and MΦ subpopulations for Notch3. H) Normalized expression levels of Notch3 are displayed for RA lining and sublining FLS and endothelial cells based on scRNA-seq data from Wei et al. [7]. I-J) RA STs were fluorescently stained for I) Notch1 and J) Notch3 expression on CD68⁺ RA MΦs, Vimentin⁺ RA FLS, and VWF⁺ endothelial cells (n=3, original magnification x20). K) RA MΦs, L) RA FLS, and M) endothelial cells were stimulated with 100 ng/mL inflammatory factors or 1000 ng/ml Syntenin-1 (SYN1) (6h), and transcription of Notch1 was assessed by qRT-PCR (n=5–6). N) RA MΦs, O) RA FLS, and P) endothelial cells were stimulated with 100 ng/ml inflammatory factors (6h) and transcription of Notch3 was assessed by qRT-PCR (n=6). Q-T) RA STs from patients (n = 46) that were non-responders, moderate responders, or good responders to anti-TNF therapy were assessed by RNA-seq [15] for differences in Q) Notch1 and S) Notch3 expression between groups. RA STs from patients with myeloid pathotype (n=21) were further assessed based on response to therapy and R) Notch1 and T) Notch3 expression. Data are presented as mean ± SEM; differences were determined by the Mann-Whitney test, one-way ANOVA with either Tukey post-test, or the p-value was adjusted for multiple comparisons by FDR. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001

DISCUSSION

Our study uncovers that Notch ligands are enriched on RA synovial CD68⁺CD14⁺ MΦs and their expression is mainly upregulated by TLRs and inflammatory cytokines during RA monocyte reprogramming into MΦs. Intriguingly, JAG1 and Notch3 are overexpressed on all FLS subpopulations, concurrently, JAG2, DLL1, DLL4, and Notch1 expression levels are rather modest on RA FLS and can be primarily amplified by TLR4 ligation. Conversely, all Notch ligands and receptors were presented in RA endothelial cells and their expression could be further expanded via TLR4/5 activation. RNA-seq studies characterized that neither DMARD nor anti-IL-6R Ab (antibody, Tocilizumab) affects synovial Notch ligands (except for DLL4), while anti-TNF good responders intercepted ligand transcription in the myeloid but not fibroid pathotypes (aside from DLL4). Overall, our findings indicate that myeloid expression of Notch ligands can act as anti-TNF response biomarkers and are upregulated in response to inflammatory stress signals in RA MΦs.

In RA and preclinical models, Notch ligands (JAG1/JAG2/DLL1/DLL4) and receptors (Notch1/3) are implicated in FLS and/or endothelial cell-mediated angiogenesis and inflammatory phenotype in part through VEGF [21], and this manifestation is further advanced by HIF1α-driven hypoxia [6]. Notch signaling is also involved in the early disease phase by expanding Th1/Th17 cell differentiation [22] and in the later stages accelerates osteoclast formation via RANKL and NFATc1 [23]. Nevertheless, the significance of RA MΦs in Notch signaling and their responsiveness to DMARD or biotherapy are undefined.

Uniquely, JAG1/JAG2/DLL1/DLL4 and Notch1/3 are highly expressed on CD68⁺CD14⁺RA MΦs by immunofluorescence staining. Distinct from the histological studies, where Notch family members are widely presented, single-cell RNA-seq of RA ST reveals that Notch1 is singularly expressed by all MΦ-subpopulations (SC-M1-M4) [20]. Notably, as IL-1β does not impact the frequency of Notch ligands in RA MΦs, but the combination of LPS/IFNγ strongly advances JAG1, DLL1, and DLL4 expression in myeloid cells, our data suggest that these family members belong to the M4 rather than M1 MΦ subpopulation [20].

Concurrently, the Notch1 ligand, JAG1, promotes myeloid cell infiltration in invasive breast cancer, and perhaps in RA, JAG1/DLL4 expressed on lining and sublining MΦs and FLS can attract Notch1 expressing monocytes to the joint [24]. Others have shown that Notch1 or Notch3-mediated NICD release and translocation remodels naive cells to inflammatory MΦs that display advanced IRF8 transcription [25] accompanied by higher NF-κB and MAPK signaling [26–28]. The MΦs reprogrammed by N1ICD also exhibit hypermetabolic activity coupled with misregulated oxidative phosphorylation leading to elevated levels of HIF1α and ROS [29]. Extending these findings, DLL4 negates differentiation of pro-repair M2 MΦs rewired by IL-4 via IRF5 and STAT1 transcriptional upregulation [30].

Recent studies have shown that DMARDs such as methotrexate (MTX) impair Notch1 signaling by reducing the Nedd4 interaction with Numb in RAW cells [31]. Conversely, MTX and Notch siRNA combination therapy was more effective in attenuating arthritic joint inflammation compared to MTX alone; suggesting that they may have diverse mechanisms of action [32]. Similarly, RNA-seq data revealed that the expression of Notch ligands and receptors was comparable in DMARD nonresponsive patients compared to those with moderate and good responses. Unlike DMARD therapy, which is ineffective on Notch family members, ST JAG1/JAG2/DLL1/DLL4 transcripts were diminished in RA patients with good response to anti-TNF therapy (ΔDAS28>1.2) compared to non- (ΔDAS28≤0.6) or moderate responders (ΔDAS28≤1.2&>0.6). We found that TNF exposure increases the transcription of JAG1, JAG2, and DLL4 on RA MΦs, while JAG2 and DLL1 transcriptome were upregulated on RA FLS. Anti-TNF good responders with myeloid-pathotype had reduced JAG1/JAG2/DLL1/DLL4 transcripts, implying that MΦs are primarily responsible for the responsiveness, given that except for DLL4, Notch ligand expression was unaltered in the fibroid-pathotype. Consistent with our data, JAG2 expression was increased in RA compared to normal FLS in response to TNF stimulation [33].

Remarkably, Notch1 transcripts were suppressed in the myeloid but not fibroid pathotype with a good anti-TNF response and unaffected in RA ST of the same group. Moreover, despite Notch3 elevated transcription in TNF-reprogrammed RA MΦs, Notch3 transcripts were unaffected in the RA ST of anti-TNF good responders. In agreement, others have delineated that Notch1 signaling is potentiated by TLR4 and TLR7 ligation and γ-secretase inhibitor deactivates this manifestation via TRAF and NF-κB pathways [34, 35].

Anti-IL-6R Ab exclusively downregulated DLL4 expression in good compared to non-responders in contrast to anti-TNF therapy that was efficacious for all Notch ligands. Distinct from DLL4 transcription which was potentiated by IL-6 in RA MΦs and endothelial cells and restricted by Tocilizumab in RA ST, Notch3 and Notch1 transcription levels were unchanged in Tocilizumab responders despite being enriched by IL-6 in RA FLS or endothelial cells. On the other hand, while HIF1α advances Notch3 transcription in breast cancer cells, the combination of γ-secretase inhibitor and Tocilizumab can more effectively obstruct tumor growth compared to γ-secretase inhibitor alone [36]. Moreover, IL-6/STAT3 activation has been shown to expand JAG1/Notch3 signaling leading to hypoxia resistance and elevated breast cancer cell invasion [37, 38].

Moreover, TLR4/TLR5 ligands, TNF, and Syntenin-1 amplify DLL4 and JAG1 expression suggesting both TNF alone and TNF produced by TLR signaling can modulate Notch ligand expression, while GM-CSF or IL-6 uniquely potentiates JAG1 or DLL4 transcriptome respectively by remodeling RA monocytes into inflammatory MΦs. Nonetheless, TNF-reconfigured RA MΦs exclusively accentuate JAG2 transcription. Additionally, TLR4 ligand exposure upregulates DLL1, and Syntenin-1 stimulation increases DLL1 and Notch1 expression in RA MΦs. Others have documented that administration of anti-DLL1 Ab interferes with myeloid DLL1 ligation to arthritic joint fibroblasts to suppress IL-6, GM-CSF, and MMP3 expression [39].

TLR4 or TLR5 ligands and IL-6 were responsible for the induction of Notch1 or Notch3 on RA FLS or endothelial cells. TLR4 ligation, TNF, and IL-1β enhanced JAG1 expression, while TLR4 activation exclusively elevated DLL1 expression in endothelial cells. When the frequency of Notch family members was evaluated, JAG2, DLL4, and Notch1 were more broadly expressed on RA compared to OA or NL endothelial cells. Immunofluorescence studies further characterized that endothelial cells predominantly express Notch1, yet Notch3 is mainly expressed on RA FLS lining and sublining. Substantiating these observations, single-cell RNA-seq and IF staining of RA VWF⁺endothelial cells from multiple RA cohorts display that all Notch family members are presented on these cells, while JAG1 and Notch3 are also widely expressed on all FLS-subpopulation clusters (SC-F1-F4).

Notch ligands and receptors can interact within the same cell (cis) or between different cell types (trans)[40]. Earlier studies delineate that while cis-interaction disrupts Notch signaling, trans-ligation fosters an inflammatory landscape. Others show that Notch1/3 are represented on THY1^high-FLS cluster [7] and can potentially ligate to JAG1, DLL1, or DLL4 expressed on endothelial cells, or as we postulate RA MΦs. Intriguingly, TLR5 ligation augments endothelial Notch1/3 expression, concurrently, boosting myeloid JAG1/DLL4 transcription perhaps increasing the likelihood of trans-activation. In line with these observations, good responders to anti-TNF with myeloid-pathotype predominately restrained JAG1/JAG2/DLL1/DLL4 transcription. Strikingly, TNF-transgenic (TNF-Tg)-induced arthritis was attenuated by Notch1 inhibitor by shifting inflammatory cells to pro-repair MΦs [41], indicating a cross-regulation between Notch and TNF pathways.

The use of knockout mice or Ab against Notch3 alleviated K/BxN arthritis [7], and the Notch intracellular domain inhibitor mitigated CIA [6]. Others elucidated that γ-secretase inhibitor and Notch1 knockdown can ameliorate preclinical arthritis in part by suppressing neutrophil migration and FLS-mediated T effector cytokines [5]. Considering most investigations of the Notch family members were conducted in experimental models, it is unclear which RA cell types are involved in trans-interaction between the Notch ligands and receptors. Our data highlight the significance of myeloid JAG1/DLL4 trans-activating Notch1 on RA endothelial cells or myeloid JAG2/DLL4 trans-activating Notch3 on RA FLS. We show for the first time, the importance of Notch ligands on RA MΦs and underscore that Notch ligands can function as biomarkers for anti-TNF response in RA patients.

Supplementary Material

Supplementary 1

Supplemental Figure 1. JAG2, DLL1, and Notch1/3 expression was unaffected by DMARDs or anti-IL6R Ab. A) Relative expression of JAG2 was determined by RNA-seq [17] in ST biopsies from RA non-responsive (ΔDAS28 ≤ 0.6, n = 23), moderate (ΔDAS28 ≤ 1.2 & > 0.6, n = 29) and good responders (ΔDAS28 > 1.2, n = 29) to DMARD therapies. B) RA STs from patients (n = 81) that were either non-responders, moderate responders, or good responders to anti-IL6R Ab (Tocilizumab) therapy were assessed by RNA-seq [19] for JAG2 expression. C) RA STs from patients with fibroid pathotype (n=8) that were non-responders, moderate responders, or good responders to anti-TNF therapy were assessed by RNA-seq for JAG2 expression [15]. D) scRNA-seq analysis was performed on RA ST (n = 18) [20]. Distinct patterns of expression are displayed amongst cell types and fibroblast and MΦ subpopulations for JAG2. E) RA STs were fluorescently stained for JAG2 expression on Vimentin⁺ RA FLS (n=3, original magnification x20). F) Relative expression of DLL1 was determined by RNA-seq [17] in ST biopsies from RA non-responsive (ΔDAS28 ≤ 0.6, n = 23), moderate (ΔDAS28 ≤ 1.2 & > 0.6, n = 29) and good responders (ΔDAS28 > 1.2, n = 29) to DMARD therapies. G) RA STs from patients (n = 81) that were either non-responders, moderate responders, or good responders to anti-IL6R Ab (Tocilizumab) therapy were assessed by RNA-seq [19] for DLL1 expression. H) RA STs from patients with fibroid pathotype (n=8) that were non-responders, moderate responders, or good responders to anti-TNF therapy were assessed by RNA-seq for DLL1 expression [15]. I) scRNA-seq analysis was performed on RA ST (n = 18) [20]. Distinct patterns of expression are displayed amongst cell types and fibroblast and MΦ subpopulations for DLL1. J) RA STs were fluorescently stained for DLL1 expression on Vimentin⁺ RA FLS (n=3, original magnification x20). K) Relative expression of Notch1 was determined by RNA-seq [17] in ST biopsies from RA non-responsive (ΔDAS28 ≤ 0.6, n = 23), moderate (ΔDAS28 ≤ 1.2 & > 0.6, n = 29) and good responders (ΔDAS28 > 1.2, n = 29) to DMARD therapies. L) RA STs from patients (n = 81) that were either non-responders, moderate responders, or good responders to anti-IL6R Ab (Tocilizumab) therapy were assessed by RNA-seq [19] for Notch1 expression. M) RA STs from patients with fibroid pathotype (n=8) that were non-responders, moderate responders, or good responders to anti-TNF therapy were assessed by RNA-seq for Notch1 expression [15]. N) Relative expression of Notch3 was determined by RNA-seq [17] in ST biopsies from RA non-responsive (ΔDAS28 ≤ 0.6, n = 23), moderate (ΔDAS28 ≤ 1.2 & > 0.6, n = 29) and good responders (ΔDAS28 > 1.2, n = 29) to DMARD therapies. O) RA STs from patients (n = 81) that were either non-responders, moderate responders, or good responders to anti-IL6R Ab (Tocilizumab) therapy were assessed by RNA-seq [19] for Notch3 expression. P) RA STs from patients with fibroid pathotype (n=8) that were non-responders, moderate responders, or good responders to anti-TNF therapy were assessed by RNA-seq for Notch3 expression [15]. Data are presented as mean ± SEM; differences were determined by one-way ANOVA. The p-value was adjusted for multiple comparisons by FDR. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001

Supplementary 2

Supplemental Figure 2. Notch ligands and receptors are expressed on CD14⁺ RA MΦs. A-E) RA STs were fluorescently stained for A) JAG1, B) JAG2, C) DLL1, D) Notch1, and E) Notch3 expression on CD14⁺ RA MΦs (n=3, original magnification x20). F-K) RA MΦs were stimulated with 100 ng/ml IL-1β (6h) and transcription of F) JAG1, G) JAG2, H) DLL1, I) DLL4, J) Notch1, and K) Notch3 was assessed by qRT-PCR (n=6–7). Data are presented as mean ± SEM; differences were determined by the Mann-Whitney test. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001

DATA AVAILABILITY

All data generated or analyzed during this study are included in this paper and its supplementary information files.