Fresh TABs with histopathological features of GCA (TAB+) were obtained from twelve newly diagnosed GCA patients. Patients were prospectively enrolled at the Vasculitis Research Unit, Department of Systemic Autoimmune Diseases, Hospital Clinic, University of Barcelona (Spain) and the Department of Internal Medicine and Clinical Immunology, Dijon University Hospital, Dijon (France) after obtaining their written informed consent in accordance with the Declaration of Helsinki (Collection DC-2016-2675). Usual clinical, biological and therapeutic data were recorded at diagnosis. The patients’ main clinical and biological characteristics are summarized in Supplementary Table 1 .

In addition, 9 fresh TABs without features of GCA (negative TABs) were obtained. Among these 9 patients, none had GCA, 4 had polymyalgia rheumatica (PMR) without associated GCA and the other 5 had an inflammatory syndrome unrelated to GCA, PMR or any other vasculitis.

HuMoSCs were generated ex vivo from circulating monocytes, as already described (15). Briefly, peripheral blood mononuclear cells (PBMCs) were obtained by Ficoll density gradient centrifugation. Then, monocytes were purified by Percoll density gradient centrifugation and cultured in X-Vivo medium (Lonza^®) supplemented with GM-CSF (10 ng/mL) and IL-6 (10 ng/mL) (Miltenyi Biotec^®). Cultured cells were harvested after seven days by gentle scraping. At this stage, the cells could be frozen in liquid nitrogen, perhaps transported and thawed within three months. To obtain HuMoSCs, a CD33 positive selection was finally performed according to the Miltenyi kit and protocol for human CD33 isolation (AutoMACS Pro, Miltenyi Biotec^®). The purity of CD33⁺ isolation was checked by flow cytometry (anti-CD33 FITC, clone HIM 3-4, eBioscience) and was always >80%. Otherwise, cells were not kept for subsequent experiments. With this method, even if generated in the presence of IL-6 and GM-CSF, which usually have a proinflammatory effect, these cells have strong immunosuppressive properties in vitro and in an in vivo mouse model of Graft-versus-Host-Disease (15).

For preparation of HuMoSC supernatant, HuMoSCs were cultured (10⁶ cells/mL) in complete RPMI for 48 hours before collection.

Sections of TAB specimens affected by GCA were embedded in MATRIGEL ^® (Dominique Dutscher SAS, Belgium) and cultured as previously described (3). Briefly, surrounding tissue was carefully removed to keep the temporal artery before regular sections of ~1 mm in thickness were cut. These sections were embedded in 25 µL of MATRIGEL^® in a 24-well plate (1 section/well and 2 sections/condition) and then covered by 750 µL of RPMI 1640 medium supplemented with 10% fetal bovine serum, 2 mM L-glutamine, amphotericin B at 2.5 μg/ml and gentamycin at 200 μg/ml. On day 1, HuMoSCs (250.10³ cells/mL) or HuMoSC supernatant (25%) were added. After five days of culture, arterial sections were collected and homogenized in TRIzol reagent using a Minilys homogenizer (Bertin instruments^®) before extraction of total mRNA.

Total RNA was extracted from cultured artery and cDNA obtained by reverse transcription using a random hexamer priming kit (Thermo Fisher Scientific, Applied Bioscience) in a final volume of 100 µL. Then, specific pre-developed TaqMan probes from Thermo Fisher Scientific (TaqMan Gene Expression Assays) were used for PCR amplification (listed in Supplementary Table 2 ). Fluorescence was detected with the CFX96TM Real-Time PCR Detection System and the results analyzed with the CFX ManagerTM software (Bio-Rad Laboratories). Gene expression was normalized to the expression of the endogenous control GUSB using the comparative ΔCt method. mRNA concentration was expressed in relative units with respect to GUSB expression (relative expression).

PDGF-AA, PDGF-BB, fibronectin, COL1α1 and VEGF concentrations were measured in the supernatant of ex vivo TAB cultures by Luminex^® (R&D Systems, Bio-Techne) following manufacturer instructions. Data were acquired and analyzed on a Bio-Plex^® 200 system (Bio-Rad, France).

VSMCs were isolated from negative TABs. Briefly, healthy arteries were embedded in MATRIGEL^® to ensure prolonged survival and then covered by DMEM supplemented with 10% fetal bovine serum, 2 mM L-glutamine, amphotericin B at 2.5 μg/ml and gentamycin at 200 μg/ml. Cells were treated with trypsin-EDTA (Gibco) between each passage. VSMCs were used after 4–7 doubling passages.

VSMCs were cultured in 6-well plates with or without human PDGF-AB (R&D Systems, 20 ng/mL) to stimulate VSMC migration [6], and HuMoSC supernatant at 10% or 25%. After 72 hours, the medium was removed and replaced by fresh medium with the corresponding additives and molecules. VSMCs reached confluence after six days of culture. One scratch per well was done using a sterile 10 µL pipette tip. Three pictures per well at different levels were automatically taken using phase-contrast microscopy (Axio Observer 7, Zeiss) every 6 hours for 48 hours. The videos were analyzed by the CellImaP Platform (University of Burgundy, Dijon, France).

VSMCs were plated in flat-bottom 96-well plates at 4,000 cells/well in triplicate wells and incubated at 37°C in 5% CO₂ for 1 to 7 days. VSMCs were cultured with fresh DMEM (10% FBS), with or without human PDGF-AB (R&D Systems, 20 ng/mL) to stimulate VSMC proliferation [6], HuMoSC supernatant at 10% or 25% and Dexamethasone at 0.5 μg/ml. After 2, 4 and 7 days of culture, cells were fixed and stained with 0.2% crystal violet (Sigma-Aldrich) in 20% methanol for 20 minutes. Wells were washed, air-dried and solubilized in 1% sodium dodecyl sulfate (SDS). Optical density was measured with a plate reader (Multiskan Ascent, Thermo Scientific) at 620 nm wavelength to assess the proliferation of VSMCs.

Cell lysates were obtained using modified radioimmunoprecipitation assay (RIPA) buffer supplemented with protease inhibitors (Complete, Boehringer Mannheim, Mannheim, Germany), NaF (50 mM), PMSF (1 mM), aprotinin, leupeptin, pepstatin (1 ng/ml) and orthovanadate (2 mM) (Sigma-Aldrich). The following primary antibodies were used for immunodetection: rabbit anti-total AKT, anti-phospho-AKT (Thr 308) and anti-phospho AKT (Ser 473) and used at 1:1000 dilution (cell signaling), rabbit anti-total Erk1/2 (dilution 1:2000) and rabbit anti-phospho-Erk1/2 (Thr202/Tyr204) (dilution 1:1000) (cell signaling), mouse anti β-actin used at 1:5000 dilution (Sigma-Aldrich) and rabbit anti-vinculin used at 1:30000 dilution (Abcam). Goat anti-rabbit IgG HRP-linked (cell signaling) and goat anti-mouse IgG HRP-conjugate (Bio-Rad) were used as secondary antibodies at 1:2000 dilution.

Cultured artery sections were fixed in cold paraformaldehyde (4%), followed by exposure to increasing concentrations of saccharose (15–30%). Sections were then embedded in optimal cutting temperature OCT (Sakura, Flemingweg, The Netherlands) and stored at -80°C. Then, 10 μm cryostat sections were fixed, blocked and incubated with primary antibodies against human phospho-S6 ribosomal protein (p-rpS6, Ser 235/236, Cell Signaling, at 1:200 dilution) and α-smooth muscle actin (polyclonal, Abcam, at 1:500 dilution), and then donkey secondary antibodies were conjugated with fluorochromes A488 (green) and A647 (red) (Abcam, at 1:300 dilution). We used DAPI to stain nuclei (blue). Visualization was performed with a Leica TCS SP8 fluorescence microscope available at the DimaCell Platform (INRAE, University of Burgundy, Dijon, France).

Membrane labeling was performed on HuMoSCs. Antibodies were incubated for 30 minutes at 4°C in PBS with 0.5% FBS. Antibodies are listed in Supplementary Table 3 . Data were acquired on a BD Bioscience Canto II cytometer and analyzed with FlowJo^® v10.

HuMoSC chemotaxis was assessed using Boyden chambers with 10 μm pore polycarbonate filters. The lower chamber was filled with undiluted supernatants of control TAB or GCA-TAB cultivated for five days in MATRIGEL. A total of 30.10³ HuMoSC/well in 50 µL of 10% FBS RPMI medium were loaded in the upper chamber. In selected chambers, an antagonist of CCR2 (Calbiochem, used at 100 nM) or CCR5 (maraviroc, R&D Systems, used at 1 µM) was added. After 4.5-hour incubation, the polycarbonate membrane was stained with hematoxylin and eosin and washed before counting number of cells/field.

Data are expressed as numbers (%) for categorical variables and median (IQR) or mean (SEM) for continuous variables. Mann–Whitney or paired Wilcoxon tests were performed to compare continuous variables, as appropriate. Statistical significance was set at P<0.05 (two-tailed). GraphPad Prism^® was used for statistical analyses.

Figure 1 depicts analysis of the effect of HuMoSCs or their supernatant in a model of ex vivo cultures of temporal arteries ( Figure 1A ). The mRNA expression of genes of the main markers of vascular inflammation in cultured arteries is reported in Figures 1B–G : proinflammatory cytokines (IL1B and IL6), markers of proinflammatory monocytes/macrophages and Th1 cells (CCL2 and CCR2), markers of Th1 cells (CXCR3) and markers of activated antigen-presenting cells (HLADRA). In the presence of HuMoSCs, we observed a significant decrease in the level of expression of CXCR3, CCL2 and CCR2 genes and a tendency for a decrease in HLADRA (P=0.054). HuMoSC supernatant had a relatively similar effect with a tendency to be less powerful than HuMoSCs.

Proteins that compose the extracellular matrix were also assessed in the model of ex vivo cultured arteries. The effect of HuMoSCs or their supernatant was similar even if the latter was less powerful: dramatic decrease in the mRNA and protein levels of the alpha-1 chain of collagen-1 ( Figures 2A, C ), significant decrease in the mRNA level of COL3A1 ( Figure 2B ) and a tendency to a decrease in the mRNA and protein levels of fibronectin ( Figures 2D, E ).

Angiogenic factors were assessed by measuring the mRNA level of VEGFA and the concentration of VEGF in the supernatant of ex vivo cultures of arteries. By contrast with HuMoSC supernatant, HuMoSC cells had no impact on the mRNA level of VEGFA ( Figure 2F ). However, both HuMoSC cells and supernatant induce a dramatic decrease in the secretion of VEGF ( Figure 2G ), suggesting an anti-angiogenic effect of HuMoSCs.

Taken together, these results suggest an ability of HuMoSCs to inhibit vascular remodeling, prompting us to investigate this therapeutic potential in a single-cell model using VSMC cultures as these cells are the main actor of vascular remodeling in GCA (5, 9, 16). To retain a single model and avoid any allogeneic stimulations, further experiments were performed in the presence of HuMoSC supernatant rather than HuMoSCs.

PDGF plays an important role in the activation of vascular remodeling related to VSMC proliferation and migration in GCA (6). We therefore used this factor to stimulate VSMC migration and proliferation in in vitro experiments. We clearly demonstrated that HuMoSC supernatant inhibits in a dose-dependent manner the migration of VSMCs previously activated with PDGF-AB. Even after 48 hours of culture, a treatment with HuMoSC supernatant kept the scratch open ( Figure 3A ).

Moreover, HuMoSC supernatant also induces dose-dependent inhibition of the proliferation of VSMCs previously activated with PDGF-AB ( Figure 3B ). Interestingly, the inhibitory effect of HuMoSC supernatant was much stronger than that of dexamethasone.

Taken together, these results suggest that HuMoSC supernatant have an anti-vascular remodeling effect through its ability to interfere with the PDGF pathway.

Given these data, we evaluated the activity of HuMoSCs in the PDGF pathway. In the model of ex vivo cultures of temporal arteries, we observed a mild decrease in the level of PDGFA expression in the presence of HuMoSC supernatant ( Figure 4A ), which was not confirmed at a protein level ( Figure 4C ). Additionally, HuMoSCs decreased more significantly the mRNA expression of PDGFB ( Figure 4B ), but paradoxically PDGF-BB concentration increased in the presence of HuMoSCs or their supernatant ( Figure 4D ). Taken together, these results do not support a major effect of HuMoSCs or their supernatant on PDGF-AB production.

PDGF are expressed as dimers (AA, AB and BB) that bind to their receptors, which are also composed of two chains: PDGF-Rαα and PDGF-Rαβ bind all types of PDGF dimers whereas PDGFR-ββ only links PDGF-BB. Then, it activates several intracellular pathways inducing the phosphorylation of Akt and Erk and ultimately the activation of the mTOR pathway, resulting in cellular proliferation (17).

The study of ex vivo cultures of TAB revealed that treatment with HuMoSCs, and to a lesser extent HuMoSC supernatant, resulted in a significant decrease in the level of expression of both PDGF receptor chains ( Figures 4E, F ).

Further experiments revealed that HuMoSCs or their supernatant had no immediate effect on the phosphorylation of Erk1, Erk2 and Akt (Thr 308 and Ser 473) ( Figure 4G ).

Taken together, these results suggest that HuMoSCs, or their supernatant, may act via a decrease in PDGF receptor expression more than through a rapid decrease in activation of the PDGF signaling pathway.

mTOR acts by activating the p70 ribosomal S6 kinases: S6K1 and S6K2. S6K1 phosphorylates the S6 protein on the 40S ribosomal subunit and leads to cell proliferation. Therefore, we evaluated mTOR activity by measuring the mean fluorescence intensity of phospho-S6 ribosomal protein in cultured arteries. We observed that PDGF induced activation of the mTOR pathway in the media and the neointima ( Figure 5A ), and that was strongly decreased in the presence of HuMoSCs or HuMoSC supernatant ( Figure 5B ). Interestingly, this effect seemed stronger in the neointima than in the media ( Figure 5C ). By contrast, mTOR was poorly activated in the adventitia ( Supplementary Figure 1 ).

We then explored how HuMoSCs might be recruited to GCA lesions to support the cells’ therapeutic potential. Flow cytometry analysis of HuMoSCs showed that they predominantly express CCR5 and to a lesser extent CCR2, very few CXCR3 and no CCR7 ( Figure 6A ). In addition, mRNA levels of chemokines that link CCR5 (i.e. CCL3, CCL4 and CCL5) were strongly expressed in GCA-TAB but not in healthy TAB ( Figure 6B ). Then, we demonstrated that this interaction between CCR5 and its ligands explains the rapid recruitment of HuMoSCs within the arterial wall visualized when HuMoSCs are added in TAB cultured ex vivo in MATRIGEL ( Figures 6C, D ).We therefore performed Boyden chamber migration assays in the presence of CCR2 and/or CCR5 antagonists, which showed that HuMoSCs migrate more toward the lower chamber when it contains supernatant from GCA-TAB, whereas they hardly migrate toward supernatant from healthy TAB (P<0.001). Migration decreased significantly in the presence of a CCR2 antagonist (P=0.005) and even more so in cases of CCR5 blockade (P<0.001), suggesting that HuMoSCs are able to be recruited to the arterial wall through an interaction between CCR5 and its ligands, and to a lesser extent between CCR2 and its ligands, which are all produced within arteries affected by GCA ( Figure 6E ).