Author contributions: N.T.L.: collection and assembly of data, conception and design, and manuscript writing; H.M.M.: collection and assembly of data and manuscript writing; C.D.B., P.N., and G.E.R.: conception and design, financial support, and manuscript writing; J.F.: conception and design, financial support, data analysis and interpretation, and manuscript writing; G.B.N.: conception and design, financial support, data analysis and interpretation, manuscript writing, and final approval of manuscript.
Correspondence: Gerard Nash, Ph.D., School of Clinical and Experimental Medicine, College of Medical and Dental Sciences, University of Birmingham, Birmingham B15 2TT, UK. Telephone: +44-121-414-3670; Fax: +44-121-414-6919; e-mail: firstname.lastname@example.org
Mesenchymal stem cells (MSC) have immunomodulatory properties, but their effects on endothelial cells (EC) and recruitment of leukocytes are unknown. We cocultured human bone marrow-derived MSC with EC and found that MSC could downregulate adhesion of flowing neutrophils or lymphocytes and their subsequent transendothelial migration. This applied for EC treated with tumor necrosis factor-α (TNF), interleukin-1β (IL-1), or TNF and interferon-γ combined. Supernatant from cocultures also inhibited endothelial responses. This supernatant had much higher levels of IL-6 than supernatant from cultures of the individual cells, which also lacked inhibitory functions. Addition of neutralizing antibody against IL-6 removed the bioactivity of the supernatant and also the immunomodulatory effects of coculture. Studies using siRNA showed that IL-6 came mainly from the MSC in coculture, and reduction in production in MSC alone was sufficient to impair the protective effects of coculture. Interestingly, siRNA knockdown of IL-6-receptor expression in MSC as well as EC inhibited anti-inflammatory effects. This was explained when we detected soluble IL-6R receptor in supernatants and showed that receptor removal reduced the potency of supernatant. Neutralization of transforming growth factor-β indicated that activation of this factor in coculture contributed to IL-6 production. Thus, crosstalk between MSC and EC caused upregulation of production of IL-6 by MSC which in turn downregulated the response of EC to inflammatory cytokines, an effect potentiated by MSC release of soluble IL-6R. These studies establish a novel mechanism by which MSC might have protective effects against inflammatory pathology and cardiovascular disease. Stem Cells2013;31:2690–2702
Mesenchymal stem cells (MSC) are multipotent stromal cells found in small numbers in bone marrow and other tissues, which can differentiate into cells of the mesodermal lineages. In addition to their potential to differentiate and repair tissue, MSC may also modulate responses by paracrine actions if administered to sites of disease [1-3]. For instance, in the cardiovascular context, bone marrow-derived MSC have been demonstrated to improve outcomes following acute myocardial infarction and have protective effects against renal and cerebral ischemia/reperfusion in animal models [4-9]. The mechanisms underlying such benefits are not certain but may include effects of MSC on differentiation and repair of the tissue structure, on blood vessel growth, and on immune responses [5, 7, 9-11]. Immunomodulation by MSC was initially exemplified by their inhibition of graft versus host disease in transplantation, but actions against the innate immune response have also been detected [2, 3]. Thus, they can “dampen” the inflammatory responses, delaying apoptosis of activated neutrophils while reducing production of reactive oxygen species  and suppressing responses of natural killer cells . MSC also inhibit the differentiation of monocytes into dendritic cells while increasing release of anti-inflammatory cytokines (interleukin, IL-10) and reducing production of proinflammatory cytokines such as tumor necrosis factor-α (TNF) [14, 15]. MSC may also modulate development of adaptive immune responses relevant to chronic inflammatory disorders by modifying the proliferation, survival, and effector functions of T cells [16-18]. These studies have incubated MSC with leukocytes but have not addressed the possibility that the MSC might interact with endothelial cells (EC) to modulate the recruitment of the leukocytes in the first place.
We have previously found that various stromal cells can modulate endothelial responses [19, 20]. This led us to investigate the possibility that exogenous MSC might ameliorate vascular inflammatory responses, for instance, following ischemia/reperfusion, if delivered via local vasculature, by downregulating leukocyte recruitment. MSC would need first to adhere and possibly integrate in the endothelial monolayer or migrate through it. It has been suggested that MSC use a multistep process to cross endothelium similar to that used by leukocytes, and that MSC “home” to inflamed or damaged tissues more than healthy tissue [21, 22]. Studies in mice have implicated P-selectin in supporting rolling of injected MSC observed in venules of the ear . In vitro, MSC adhered to cytokine-treated EC at very low wall shear stress (0.01 Pa) in one study  and bound in very small numbers after extended perfusion at more physiological venular shear stress (0.1 Pa) in another . In these studies, adhesion was dependent on the leukocyte capture molecules P-selectin or vascular cell adhesion molecule-1 (VCAM-1), and binding to VCAM-1 has also been shown recently to support transendothelial migration of MSC . However, it is worth considering that MSC are much larger than leukocytes, so that MSC emigration in the microcirculation may be assisted by mechanical trapping in the smallest vessels without requirement for specific capture from flow .
To address the above questions, we investigated attachment of MSC to EC using early passage, primary human cells, and the effects this had on subsequent adhesion and migration of flowing neutrophils and lymphocytes. We found poor attachment of MSC to EC from flow. However, we show for the first time that when brought together there was a powerful modulation of EC responses to inflammatory cytokines and downregulation of recruitment of both types of leukocyte. This required crosstalk with signal(s) from EC inducing greatly increased production of IL-6 by MSC that along with soluble IL-6-receptor (sIL-6R) acted on EC to inhibit their ability to recruit leukocytes. Thus there was context-specific development of stem cell function that should dampen responses such as ischemia/reperfusion and could provide therapeutic benefit.
Materials and Methods
Isolation and Culture of EC, MSC, Neutrophils, and Lymphocytes
Human umbilical cords were obtained from the Human Biomaterials Resource Centre (University of Birmingham) after informed consent. Human umbilical vein EC (HUVEC) were isolated from umbilical cords as previously described  and cultured in M199 supplemented with 20% fetal calf serum, 10 ng/mL epidermal growth factor, 35 μg/mL gentamicin, 1 μg/mL hydrocortisone (all from Sigma-Aldrich, Poole, U.K., http://www.sigmaaldrich.com), and 2.5 μg/mL amphotericin B (Life Technologies, Carlsbad, CA, http://www.lifetech.com). Human bone marrow derived MSC were purchased from Lonza and cultured in the manufacturer's recommended medium (MSCGM Mesenchymal Stem Cell Growth Medium BulletKit, Lonza Ltd., Basel, Switzerland, http://www.lonza.com/). The cells were expanded twice with three-way splits before use in the current studies. Based on the manufacturer's information that the cells are supplied in the second passage, this indicates that cells had undergone 11–12 doublings at the time of use. Since senescent cells may secrete cytokines including IL-6 (see below) , we stained for pH-dependent β-galactosidase activity (Cell Signaling Technology, Danver, MA, http://www.cellsignal.com). At the passage used here, we found 4% blue-stained cells, and this increased to 13% after two further passages and 37% after another two passages. Cells could nevertheless be expanded through these further four passages without evident reduction in growth rate, and the cells used here could be differentiated along adipogenic, chondrocyte, and osteogenic lineages using appropriate medium kits (Lonza).
Blood was collected from healthy volunteers into EDTA tubes (Sarstedt Ltd., Leicester, U.K., http://www.lonza.com/) after informed consent. Neutrophils or peripheral blood lymphocytes (PBL) were isolated from blood overlaid on density gradients of histopaque 1077 above histopaque 1119 (Sigma) as described . Leukocytes from upper or lower bands (mononuclear cells or granulocytes, respectively) were washed twice and suspended in phosphate-buffered saline (PBS) containing 1 mM Ca2+, 0.5 mM Mg2+, and 0.15% bovine serum albumin (Sigma) (PBS/BSA) at 106cells per milliliter. For PBL, after the first wash, mononuclear cells were panned on culture plastic to remove monocytes.
Attachment of MSC to EC
Primary HUVEC were dissociated using trypsin/EDTA (Sigma-Aldrich) and seeded in prefabricated channel slides (μ-Slide VI; ibidi GmbH, Martinsried, Germany, www.ibidi.de). Seeding density was chosen to yield confluent monolayers within 24 hours. The confluent monolayers were stimulated for 4 hours with TNF (100 or 1,000 U/mL) or IL-1β (5 × 10−11 to 5 × 10−9g/mL) (both from Sigma) and attached to a perfusion system mounted on the stage of a phase-contrast/fluorescence video microscope enclosed in a Perspex chamber at 37°C, as described [30, 31]. MSC were dissociated with trypsin/EDTA, labeled with Cell-Tracker Green (5 μM for 1 hour; Life Technologies), counted, and adjusted to 0.5 × 106 cells per milliliter and perfused over EC at a chosen wall shear stress for the desired period. Nonadherent cells were washed out with cell-free medium, and the adherent MSC were counted in six fields of known dimensions using fluorescence microscopy and converted to cells per milliliter square. Alternatively, MSC were perfused into the chamber (wall shear stress 0.1 Pa), flow was stopped for 20 minutes, and the number of settled MSC was counted in six fields. Flow was then resumed to washout nonadherent cells. MSC were counted again, and data were converted to a percentage of those that had settled.
Culture of EC with MSC
Two coculture formats were used. HUVEC were seeded in channel slides as above, MSC were injected at chosen concentration and allowed to settle for 60 minutes. Nonadherent cells were then rinsed off with MSC medium and cultured for further 20 hours in that medium. When desired, TNF (100 U/mL; ∼3 nM) or IL-1β (5 × 10−9 g/mL; ∼2.5 nM) was added alone or with interferon (IFN)-γ (10 ng/mL; ∼6 nM; Peprotech, London, U.K., http://www.peprotech.com) and culture continued for a further 4 or 24 hours. The channels were attached to a flow system for assay of leukocyte adhesion, or HUVEC and MSC were retrieved for analysis of gene expression (see below).
Alternatively, MSC and EC were cocultured on opposite sides of filters (six-well format) with 0.4 μm pores, as described . In this case, EC were seeded on the inner surface first and cultured for 24 hours to form a confluent layer. Then MSC (5 × 105 cells) were added to the other surface and cocultured with EC for 20 hours. When desired, TNF (100 U/mL) was then added alone or with 10 ng/mL of IFN-γ for a further 4 or 24 hours. The filters were cut out and attached to a flow system for assay of leukocyte adhesion (see below). Alternatively, cells were retrieved separately from either side of the filter using trypsin for analysis of gene expression (see below).
For comparison, parallel cultures of HUVEC or MSC alone were carried out at the same cell densities in channel slides or on filters, as required. In some experiments, a neutralizing antibody against IL-6 (clone 6708, 5 μg/mL), against IL-6 receptor (clone 17506; 5 μg/mL), or against transforming growth factor-beta (TGF-β; clone 9016, 10 μg/mL) (all from R&D Systems, Abingdon, U.K., http://www.rndsystems.com) was added during coculture and cytokine stimulation. In others, HUVEC were cultured alone for 20 hours with 10 ng/mL human recombinant IL-6 (Immunotools, Friesoythe, Germany, www.immunotools.de) before addition of TNF for a further 4 hours.
Conditioned media (CMed) were collected, typically before and after cytokine treatment of cells. CMed were used to pretreat EC before addition of cytokines and leukocyte adhesion assays, or their content was analyzed (see below). In some experiments, neutralizing antibody against IL-6 or IL-6R (CD126) (both 5 μg/mL; R&D) was added to CMed, or CMed were depleted of sIL-6R by affinity column (see below).
Leukocyte Recruitment to EC from Flow
Adhesion assays were performed using μ-Slides or filters incorporated into a custom-made flow chamber  using the same flow system as for MSC adhesion [30, 31]. A 4-minute bolus of leukocytes was perfused over the cell surface at a wall shear stress of 0.05 Pa followed by cell-free wash buffer. Video recordings were made of a series of microscope fields along the centerline of the flow channel after 6 minutes of washout. Videomicroscopic recordings were analyzed off-line using a computerized image analysis system (ImagePro; Media Cybernetics, Inc. Rockville, MD, USA, http://www.mediacy.com/). Leukocytes adherent to the cell surface were counted and classified as rolling adherent (spherical cells moving over the surface much slower than free-flowing cells), stationary adherent (typically with distorted shape and actually migrating slowly on the surface), or phase-dark transmigrated cells . The total number adherent was converted to number/mm2 per 106 perfused.
Analysis of Content of CMed and Depletion of sIL-6R
Concentrations of cytokines in CMed were measured using MILLIPLEX MAP multiplex immunoassay (Millipore, Billerica, MA, http://www.millipore.com) according to manufacturer's instructions. An initial screen was made for 13 cytokines potentially linked to inflammation (including IL-6, IFN γ-induced protein 10 IP-10/CXCL10 and IL-8/CXCL8); subsequent analyses used a single assay for IL-6. Concentrations were measured on a LX100 machine (Luminex Corp., USA, http://www.luminexcorp.com/) and calibrated against titrations of recombinant standard using STarStation software (ACS, USA, http://www.appliedcytometry.com/). Separately, we measured TGF-β using a commercial ELISA (R&D, Quantikine ELISA), which relies on acid activation of samples, giving a value for total content rather than measuring endogenous activated agent. Values for supernatants from monocultures or cocultures are given after subtraction of values for the basal medium alone.
To remove sIL6-R, CMed were incubated with TrueBlot R anti-mouse immunoglobulin IP beads (Bioscience, http://www.ebioscience.com/) conjugated with human IL-6R antibody as recommended by the manufacturer. Supernatant was collected after centrifugation and stored −80°C until required. sIL-6R was eluted from the IP beads by centrifugation after Laemmli buffer treatment. CMed (with or without depletion) or eluates were run on Pre-cast NuPage 10% Bis-Tris gel (Invitrogen, Carlsbad, CA, http://www.invitrogen.com). The proteins were transferred from gel onto Immun-blot membranes (Bio-Rad, www.bio-rad.com) which were probed with mouse antibody against IL-6R (R&D). Membranes were then incubated with horse radish peroxidase-conjugated secondary goat-anti-mouse IgG (Amersham Biosciences, U.K., http://www.amersham.com), and proteins were visualized by Pierce ECL Western blotting substrate (Thermo Scientific, http://www.thermoscientific.com/).
Modification of Gene Expression by Small-Interfering RNA
Expression of IL-6 or IL-6R by HUVEC or MSC was reduced using small-interfering RNAs (siRNA) as described . siRNA against IL-6 or IL-6R, or universal negative control, nonspecific siRNA were purchased from Thermo Scientific. Transfection of oligonucleotides was achieved using Lipofectamine RNAiMAX reagent (Invitrogen) and 35 nM siRNA. HUVEC within μ-Slides or MSC cultured in plates were transfected for 4 hours and cultured for further 20 hours. MSC were retrieved from the plates and seeded on EC in μ-slides for 24 hours and then treated with cytokines as above.
Analysis of Gene Expression by Quantitative Polymerase Chain Reaction
In studies of effects of siRNA, mRNA was isolated from EC or MSC retrieved from μ-slides using the RNeasy Mini Kit (Qiagen, Crawley, U.K., http://www1.qiagen.com). For studies of effects of cytokines, mRNA was isolated from EC which had been cultured on filters with or without MSC on the opposite side, using the same kit. mRNA was reverse transcribed and analyzed by quantitative polymerase chain reaction (qPCR) using Quanti-Tect probe RT-PCR kit according to manufacturer's instructions (Qiagen) as described . Primers were bought as Taqman Gene Expression Assays (Applied Biosystems, Warrington, U.K., http://www.appliedbiosystems.com) for IL-6, IL-6R, suppressor of cytokine signaling 3 (SOCS3), E-selectin, inter-cellular adhesion molecule-1 (ICAM-1), and VCAM-1. Samples were amplified using the 7500HT Real-Time PCR machine and analyzed using the software package SDS 2.2 (Applied Biosystems). Data were initially expressed as relative expression units compared to mRNA for 18S in the same samples.
For microarray analysis, RNA was extracted from MSC cultured alone on filters, EC cultured alone on filters, and from each separately from cells cocultured on opposite sides of filters, labeled with Cy3 dye as per protocol in the Low Input Quick Amp Labeling Kit (Agilent Technologies 5190-2305, Palo Alto, CA, http://www.agilent.com). A specific activity of greater than 6.0 was confirmed by measurement of 260 nm and 550 nm wavelengths with a NanoDrop ND-1000 Spectrophotometer. Labeled RNA (600 ng) was hybridized for 16 hours to Agilent SurePrint G3 Human 8 × 60K microarray slides. After hybridization, slides were washed and scanned with a High Resolution C Scanner (Agilent Technologies). Feature extraction was performed using Agilent Feature Extraction Software. Statistical analysis of differential gene expression (coculture vs. monoculture for a given cell) was carried out using linear models for microarray data (LIMMA) . Probes that pass the statistical analysis (LIMMA p < .05) were filtered for fold change, to provide those that showed >1.5 difference between conditions.
Evaluation of Protein Expression by Flow Cytometry
For studies of effects of cytokines, EC which had been cultured on filters with or without MSC on the opposite side were detached with nonenzymic dissociation buffer and washed with PBA/BSA. Cells were incubated with phycoerythrin-conjugated mouse monoclonal antibody (mAb) against E-selectin, apocyanin-conjugated mAb against ICAM-1, or fluorescein isothiocyanate-conjugated mAB against VCAM-1 (clones 68-5H11, HA58 or 51-10C9, respectively, all from BD Bioscience, Oxford, U.K., http://www.bdbiosciences.com) or isotype- and label-matched mouse IgG1 control (DakoCytomation Ltd., Ely, U.K., http://www.dakocytomation.com) diluted 1: 50 in PBS containing 2% normal goat serum for 1 hour on ice. Cells were then washed and fixed with 2% formaldehyde. The median fluorescence intensities (MFI) were measured using a FACScan flow cytometer, and background values for non-specific antibodies were subtracted (Cyan, Becton Dickinson Ltd, Oxford, U.K., http://www.bd.com).
Effects of multiple treatments were tested using ANOVA, followed by post hoc comparisons to control by Dunnett test or between treatments by Bonferroni test, as appropriate. Samples exposed to individual treatment or inhibitors were compared to untreated controls by paired t test. All tests were performed using the computer program Minitab (Minitab Inc., State College, PA, http://www.minitab.com).
Effects of MSC on Leukocyte Recruitment by EC
We tested several configurations in which MSC might modulate the proinflammatory responses of EC. First, EC and MSC were cultured on opposite sides of 0.4 μm pore filters (with or without added cytokines). The filters were incorporated in a flow chamber, and neutrophils or PBL were perfused over them. Levels of adhesion to unstimulated EC were very low, and coculture with MSC tended to reduce them further (Fig. 1A, 1B). Levels of adhesion to stimulated EC were much higher, and importantly, coculture significantly reduced the levels of adhesion of neutrophils or PBL to EC that had been treated with cytokines (TNF or TNF+IFN, respectively; Fig. 1A, 1B).
Although this demonstrated potency of MSC, the model did not obviously mimic the therapeutic use of MSC injected in the blood, unless MSC attach to EC and rapidly move into the subendothelial space. We thus perfused MSC over EC that had been cultured in flow channels to see if they would adhere and then be able to modify leukocyte recruitment. Perfusion of MSC at wall shear stress of 0.05 Pa (the low end of venular shear) resulted in negligible adhesion to stimulated or unstimulated EC. If we reduced flow to 0.01 Pa and then washed out at 0.05 Pa, adhesion remained barely detectable for unstimulated EC, but we did observe dose-dependent adhesion to EC treated with cytokines (Fig. 1C). On the other hand, if we perfused MSC, stopped flow and allowed them settle on untreated EC, and then washed out, there was consistent adhesion, with 30% ± 8% of settled cells attaching (mean ±SEM, n = 8). Adhesion in this static system did not increase if the EC were pretreated with TNF (data not shown).
To test anti-inflammatory effects of lumen-delivered MSC, we allowed them to attach and then cocultured them for 20 hours before addition of cytokines and performance of adhesion assays. Under these conditions, confocal microscopy indicated that the adherent MSC incorporated into the endothelial monolayer but did not appear to move underneath it (Supporting Information Fig. 1), although others have observed that MSC are able to migrate under TNF-treated EC within hours . MSC were added at a range of concentrations, to obtain final ratios of adherent MSC to EC ranging from 1:1,000 to 1:25 based on the results of the “static” adhesion experiments noted above. Phase-contrast and fluorescence images for such a typical series of cultures are shown in Supporting Information Figure 2; on average MSC seeded at 1:25 EC occupied 8% of the surface area based on digital analysis of fluorescent images using ImagePro. In a series of such cultures, MSC caused a dose-dependent reduction in neutrophil adhesion to EC that had been stimulated with TNF or with IL-1 (Fig. 2A, 2C). Coculture also caused a reduction in the percentage of adherent neutrophils that went on to migrate through the endothelial monolayer (Fig. 2B, 2D). Similar trends were seen with PBL adhesion and migration on EC that had been treated with TNF + IFN for 24 hours (Fig. 2E, 2F). We also examined the effect on neutrophil adhesion and migration if MSC were added to EC after they had already been treated with TNF for 3 hours. When the culture was prolonged for only 1 hour further (4-hour TNF), MSC did not modify adhesion or migration significantly (ratio with:without MSC was 0.92 ± 0.17 for adhesion, and 0.85 ± 0.06 for transmigration; n = 4, NS). This showed, among other things, that the presence of MSC themselves was not the reason for loss of adhesion and migration. On the other hand, if culture with TNF was extended to 24 hours, the level of adhesion and migration in the controls was less than at 4 hours (ratio of 24 hour:4 hour was 0.61 ± 0.06 for adhesion, and 0.79 ± 0.09 for transmigration; mean ± SEM, n = 5, p < .05 and NS, respectively vs. 4 hours), and this effect was stronger when MSC were present (ratio of 24 hours:4 hours was 0.34 ± 0.06 for adhesion and 0.56 ± 0.11 for transmigration; n = 5, p < .01 and p < .05 vs. 4 hours; p < .01 and NS vs. values for 24 hours without MSC). Thus MSC appeared to accelerate recovery from a prior stimulus.
Role and Source of IL-6 in Suppression of Recruitment
Since MSC in direct contact with EC, or separated from them by a filter both reduced leukocyte recruitment, we tested effects of supernatants from cocultures on EC monocultures. Pre-exposure to supernatant from cocultures (but critically, not from cultures of EC or MSC alone) inhibited neutrophil adhesion and migration when EC were treated with TNF (Fig. 3A, 3B). We thus screened supernatant with a multiplex bead assay, and the only marked difference between monocultures and coculture was in the concentration of IL-6; EC and MSC alone released similar detectable levels of cytokine, but the concentration from cocultures was about 10-fold higher (Fig. 3C). We thus retested activity of the supernatant after addition of a neutralizing antibody against IL-6, which in the Luminex assay reduced levels of measureable IL-6 to the basal levels in monoculture (Fig. 3C). In this case, the suppressive effects of the supernatant were no longer significant (Fig. 3A, 3B). Next, IL-6-neutralizing antibody was added to cocultures throughout. In its presence, the inhibitory effects of coculture were again reduced; this was the case for neutrophils recruited by EC treated with TNF (Fig. 4A, 4B) or PBL recruited by EC treated with TNF+IFN (Fig. 4C, 4D). We also added a blocking antibody against IL-6R to cocultures and found again that the suppressive effects were reduced (Fig. 4E, 4F). Finally, we added a similar level of recombinant IL-6 (10 ng/mL) to EC monoculture before TNF; there was a reduction in adhesion (28% ± 13%; mean ± SEM, n = 5, NS) but more significantly migration (39% ± 8%; mean ± SEM, n = 5, p < .01) of neutrophils (as we have published previously; ). Thus IL-6 was identified as a major mediator of the immunosuppressive effects when MSC were cocultured with EC.
We next investigated the source of IL-6 by using siRNA to reduce expression in each cell type; siRNA targeting IL-6 mRNA reduced mRNA levels by >60% for the EC or MSC individually, compared to nonspecific siRNA control (mean from four or three experiments, respectively). In cocultures, the high level of released IL-6 was reduced if the MSC had been treated with siRNA targeting IL-6 but not if EC had been treated; treating both cells was no better than MSC alone, and treating both cells with negative control siRNA did not reduce IL-6 release (Fig. 5A). Consistent with this finding, treating MSC with siRNA targeting IL-6 negated the effect of coculture on neutrophil adhesion and transmigration, but treating EC did not impair immunosuppression; combined treatment of MSC and EC gave similar results to MSC alone (Fig. 5B, 5C).
Role of sIL6-R in Suppression of Recruitment
The above results suggested that MSC were acted upon by EC to produce IL-6 that then acted on EC to modify their responses. We used siRNA against IL-6R to investigate whether IL-6 indeed operated through IL-6R expressed dominantly on EC. siRNA targeting IL-6R reduced mRNA levels by >85% for the EC or MSC individually, compared to nonspecific siRNA control (mean from four or three experiments, respectively). Treatment of EC or MSC, or both EC and MSC, tended to inhibit the effect of coculture on neutrophil adhesion and migration (Fig. 6A, 6B). Interestingly, the effect of treating MSC with siRNA appeared greater than treating EC, although this trend was not statistically significant.
For the studies of the effects of siRNA on neutrophil recruitment (Figs. 5B, 5C, 6A, 6B), the paired controls shown were treated with nonspecific siRNA. These were considered to be the more appropriate control and we did not test EC without siRNA in those assays. Unpaired comparison of recruitment between EC treated with nonspecific siRNA (Figs. 5, 6) and untreated EC in the earlier studies (e.g., Fig. 3) gave values for adhesion of 476 ± 19 and 512 ± 46/mm2 per 106 perfused, and for transmigration of 57 ± 4% and 35 ± 4% (mean ± SEM, n = 7 and 12, respectively). While these results suggest there may have been some nonspecific effect of siRNA treatment per se on transmigration, or variation between HUVEC or neutrophils over the course of the work, they do not negate the direct comparisons between specific and nonspecific siRNA.
An effect of siRNA acting through MSC might occur if another role for MSC was to provide sIL-6R to potentiate the effect of IL-6 on the EC. We thus collected supernatant and used beads conjugated with antibody against IL-6R to extract the receptor. The supernatant which had been cleared of sIL6R was less potent in suppressing neutrophil adhesion or migration (Fig. 6C, 6D). When eluates from the beads were subject to Western blot, there were clear bands showing IL6-R presence, with each cell type contributing (Fig. 6E). Thus IL6-R released by MSC helped potentiate the effect of the IL-6.
Role of TGF-β in Crosstalk Between EC and MSC
The increase in release of IL-6 by MSC in coculture and functional effects on EC indicate that coculture modifies both cell types; while IL-6 may be the major functional modifier of EC, some other agent(s) must act on the MSC. TGF-β has been identified in previous studies as a modulator of leukocyte recruitment generated in cocultures . We thus added a neutralizing antibody against TGF-β to cocultures. In its presence, TNF-induced neutrophil adhesion remained suppressed by coculture (albeit slightly less than in its absence), but suppression of transmigration was inhibited and no longer significant (Fig. 7A, 7B). We also found that when antibodies against TGF-β and IL-6 were both added to cocultures, adhesion and transmigration returned to levels close to those for EC cultured alone (Fig. 7A, 7B). The concentrations of TGF-β were measured in supernatants from monocultures and cocultures by ELISA. This showed that the level in cocultures was similar to the sum of the levels in the monocultures (concentrations for MSC, EC, and MSC+EC were 83 ± 24, 358 ± 72 and 450 ± 83 pg/mL, respectively; mean ±SEM from three to eight experiments). Thus the effects in coculture were presumably dependent on increase in activation of the latent form rather than increased secretion. We also measured IL-6 release in these experiments. In addition to upregulation of IL-6 release in the unstimulated 20-hour cocultures, 4-hour TNF caused a further increase of IL-6 secretion by monocultures and cocultures, with cocultures again releasing more than the sum of monocultures (Fig. 7C). Interestingly, neutralizing TGF-β tended to reduce IL-6 levels, both for unstimulated cocultures, and even more so for the TNF-treated cocultures. Since this suggests that TGF-β may play a role in regulating IL-6 production, we added recombinant TGF-β to MSC for 24 hours; indeed supernatant concentration of IL-6 increased from 720 ± 30 to 1,300 ± 177 pg/mL (mean ± SEM from four experiments; p < .05 by paired t test).
Evidence of Crosstalk Between EC and MSC from Analysis of Gene Expression
To investigate cellular crosstalk further, we carried out gene array analysis, comparing EC in monoculture versus coculture, and MSC in monoculture versus coculture in four independent experiments. We found >900 endothelial genes and >400 MSC genes modified in coculture compared to monoculture (fold change >1.5; p < .05). One noticeable responder was SOCS3, an IL-6-inducible gene with relevance to inflammation , which was upregulated in the EC by coculture. We verified this by qPCR and found 3.4-fold increase in coculture that was lost if MSC were treated with siRNA against IL-6 (mean from two experiments).
Role of Adhesion Molecules and Chemokines in Modification of Leukocyte Recruitment
The effects of coculture on leukocyte recruitment suggest that cytokine-induced upregulation of adhesion molecules or chemokines might have been inhibited. We analyzed secretion of IL-8 or IP-10, potent activators of neutrophils or T cells, respectively. These were upregulated by treatment with TNF or TNF+IFN, respectively (data not shown), but coculture did not modify the levels found in the supernatant; IL-8 = 17.2 ± 1.7 versus 17.5 ± 1.7 ng/mL for TNF-treated monoculture versus coculture (mean ± SEM, n = 4); IP-10 = 3.6 ± 1.6 versus 3.1 ± 1.0 ng/mL for TNF+IFN-treated monoculture versus coculture (mean ± SEM, n = 3). We also measured EC mRNA for adhesion molecules E-selectin, ICAM-1, and VCAM-1 and their surface expression on EC. Treatment with cytokines increased mRNA expression of E-selectin, ICAM-1, and VCAM-1, but coculture with MSC did not inhibit this upregulation (e.g., E-selection expression for TNF-treated coculture relative to monoculture was 1.2 ± 0.5, mean ± SEM, n = 5). At protein level, after treatment with TNF for 4 hours, upregulation of ICAM-1 but not E-selectin (i.e., receptors relevant for neutrophil recruitment) was inhibited; MFI for ICAM-1 for TNF-treated coculture relative to monoculture was 0.52 ± 0.15 (mean ± SEM, n = 5, p < .01 by paired t test), for E-selectin the ratio was 1.6 ± 0.4 (n = 5, not significant). After treatment with TNF+IFN for 24 hours, upregulation of ICAM-1 and VCAM-1 (i.e., receptors relevant for lymphocyte recruitment) was inhibited, although only the latter reached statistical significance; MFI for ICAM-1 for TNF-treated coculture relative to monoculture was 0.63 ± 0.11 (mean ± SEM, n = 4, p = .10 by paired t test), for VCAM-1 the ratio was 0.28 ± 0.15 (n = 3, p < .05 by paired t test).
MSCs have been shown to have a number of immunomodulatory effects, for instance, modifying the responses of various leukocyte subsets on contact [11-17]. However, perhaps surprisingly, their ability to modulate the key step of leukocyte recruitment from the circulation has not been tested until now. This gap in our knowledge may have arisen because of lack of adequate models, either in vitro or in vivo. Adhesive interactions between MSC and endothelium have been evaluated [23, 24] but not the subsequent effects of MSC on leukocyte recruitment in inflamed vessels. Here, we present an in vitro coculture model, which showed the clear ability of MSC to modify endothelial responses underlying leukocyte adhesion and migration. In setting up the model, we assessed first the potential means of delivery of MSC. MSC adhered poorly from flow, and as reported by others , only at very low wall shear stress. However, they bound efficiently when allowed longer contact with EC and then incorporated in the monolayer. Considering that these human MSC averaged 20 μm in diameter (data not shown), adhesion of the MSC from flow may not be necessary because they are likely to initially become lodged in terminal arterioles and capillaries . Moreover, infused MSC may generate an instant blood-mediated inflammatory reaction with procoagulant activity that may promote deposition in microvessels . While we found that MSC could act on EC from the luminal or abluminal sides of the endothelial monolayer, we considered an approach where cells were delivered from the lumen and incorporated in the endothelial layer more relevant to therapeutic intervention. We thus allowed MSC to bind to EC, then cultured them together, and used flow-based assays of leukocyte recruitment to test the ability of MSC to downregulate the endothelial response to inflammatory cytokines.
Using this model, we found that the numbers of leukocytes attaching, and the ability of the cells that did attach to transmigrate were downregulated after coculture of EC with MSC. The inhibitory effect was not specific to one cytokine or type of leukocyte. Soluble mediator(s) played a large role, and supernatant from cocultures (but not monocultures) could largely reproduce the downregulation of recruitment. Indeed, the inhibitory effect was dependent on a great increase in IL-6 release, mainly from the MSC. IL-6 alone had an inhibitory effect, and its neutralization greatly reduced the effects of supernatants or coculture. Evidently, the EC acted on the MSC to release IL-6, which then acted back on the EC. Not only this but also MSC provided sIL-6R which potentiated their effect. Removal of sIL-6R from supernatant reduced its bioactivity and siRNA against IL-6R was effective when it was applied to the MSC as well as EC, even though IL-6 itself acted on the EC to modify their responses. Thus while IL-6R on EC was presumably used, MSC provided enough sIL-6R to be effective even when receptor expression on EC was reduced; the combination of both sources of receptor allowed the greatest immunomodulatory effect.
The implication of these results is that EC “talk” to MSC as well as vice versa. Supernatant from MSC alone did not modify leukocyte recruitment, but culture with EC induced MSC to greatly increase IL-6 production and generate an active supernatant, and coculture modified hundreds of genes in MSC as well as in EC. Interestingly, the anti-inflammatory effect of coculture was less if TGF-β was neutralized, and levels of IL-6 released were also reduced, making TGF-β a likely mediator of changes in MSC induced by coculture. The total level of TGF-β in coculture supernatants was not greater than the sum of that produced by MSC and EC. However, TGF-β in its latent complex can be constitutively released by many cells including EC, and the key to its bioactivity is transformation to the active form by proteolytic modification or binding to surface molecules such as αv-integrins and thrombospondin [39, 40]. In the current context, coculture may promote its activation in a manner similar to that when EC are cultured with pericytes . Moreover, we found that TGF-β could induce increased secretion of IL-6 by MSC, in line with previous studies of IL-6 secretion by fibroblasts . Thus activation of TGF-β in coculture, leading to increased IL-6 release was probably one pathway contributing to immunomodulation. However, since neutralization of TGF-β and IL-6 had a greater effect than the TGF-β alone, not all activity can be attributed to action of TGF-β through IL-6. The other promoter of anti-inflammatory effects delivered by MSC in coculture was sIL-6R, but immunoblotting of supernatants did not suggest that its release was dependent on coculture, and it is likely that endogenous production was adequate for the response.
The end result of crosstalk between MSC and EC was to strongly influence the endothelial response to cytokines such as TNF, IL-1, and IFN-γ. Crosstalk continued under cytokine excitation. Both cells released more IL-6, but again, amplification of release occurred in cytokine-treated cocultures, and this was reduced by neutralization of TGF-β. This is consistent with the ability of EC to upregulate release and activation of TGF-β upon cytokine stimulation. Thus while changes in gene expression and IL-6 release by MSC in coculture were not dependent on addition of cytokines, and supernatant reduced effect of cytokines on EC without MSC being present, cytokines may also affect EC-MSC crosstalk. Results are not consistent with MSC simply taking up cytokines, or responding to cytokines by releasing agent(s) that then modulated endothelial response. However, protective effects of coculture may continue and indeed be increased under inflammatory stimulus. Our current hypothesis regarding crosstalk arising from the above discussion is summarized in Supporting Information Figure 3. This includes the likelihood that additional, as yet undiscovered pathways contribute to crosstalk, for instance, release of proteolytic enzymes which activate TGF-β, as we observed in cocultures of EC with secretory smooth muscle cells .
IL-6 was identified as the major agent acting on EC to reduce leukocyte recruitment. Although IL-6 is commonly referred to as an “inflammatory cytokine” this is not strictly correct. Its level is raised in many conditions involving tissue injury or infection as part of the acute phase response that follows the initial insult and is a protective process [44, 45]. On the other hand, its continuous elevation may play a role in the exacerbation of chronic inflammatory disease, such as rheumatoid arthritis . It has also been proposed, for example, that IL-6 plays a role in transforming the early phase of neutrophil recruitment to the later phase of mononuclear recruitment . We previously reported that IL-6 could have proinflammatory or anti-inflammatory effects in other coculture models, depending on the context and cell types present . In addition, we showed that induction of IL-6 production in EC alone, by infection with Kaposi's sarcoma herpes virus, could downregulate neutrophil migration . Here, IL-6 was provided in combination with sIL-6R by MSC. In solution, sIL-6R can form a complex with IL-6 which can bind the ubiquitous transmembrane coreceptor gp130 (CD130) to generate signals in cells lacking or having low expression of the transmembrane form of IL-6R. With the combination, we observed marked downregulation of adhesion and transmigration of adherent lymphocytes as well as neutrophils, the most potent effect we have observed to date. One notable gene modification was upregulation of SOCS3 in the EC. SOCS3 is a known IL-6-regulated gene which itself downregulates cellular responses to cytokines , and which we previously found to play a role in modification of neutrophil recruitment by viral infection . Here we found that coculture inhibited the upregulation of ICAM-1 and VCAM-1 on the endothelial surface. Our own work and that of others has shown that these receptors contribute to adhesion and migration of neutrophils and lymphocytes under the stimulatory regimes used here [19, 49]. Thus, their loss explains, in part at least, changes in recruitment observed. E-selectin, which is a major capture receptor for neutrophils, was not reduced by coculture, and at mRNA level, all of the receptors remained elevated by cytokines. In addition, release of promigratory chemokines IL-8 and IP-10 was not suppressed by coculture. Thus, there was not a blanket suppression of cytokine signaling but more subtle modification of translation or presentation of mediators of adhesion and migration. In vivo, such effects on EC may be augmented by ability of infused MSC to shift the balance from proinflammatory to anti-inflammatory cytokine release by T cells . A more direct effect on attachment is also possible, as we previously found evidence that close juxtaposition of fibroblasts reduced lymphocyte adhesion to EC through enzymatic modification of the endothelial surface . Analysis of genes modulated by coculture and SOCS3 and screening of supernatants for mediators specific to cocultures should be useful to identify further mechanisms of modulation of leukocyte recruitment or agents active in crosstalk.
The inhibitory effects of MSC were clearly dependent on the numbers added to EC but did not require addition of large numbers of stem cells. Addition of the highest number after TNF treatment of EC, shortly before leukocyte perfusion, did not inhibit recruitment, indicating that effects could not be attributed to physical obstruction by the MSC. Interestingly, we saw neutrophils interact adhesively with the MSC that were visible, but their effects remained inhibitory of adhesion overall. We also noted that area coverage by the MSC at the highest concentration was only about 8% of the monolayer (e.g., Supporting Information Figure 2). Moreover, MSC exerted inhibitory effects when cultured on the opposite side of a filter. Thus their effects were through crosstalk with the EC rather than disturbance of the endothelial integrity.
The results described here have implications for the therapeutic use of MSC. If delivered from the blood to the vessel wall, MSC could inhibit development of an inflammatory leukocyte infiltrate. Our studies focused on delivery in advance of a single cytokine treatment which does not obviously represent the case of responding to an acute insult such as ischemia/reperfusion. Not surprisingly, addition of MSC 3 hours after TNF treatment could not reduce adhesion 1 hour later; at 3 hours the adhesive response for neutrophils is nearly fully developed . However, later addition did accelerate resolution of the response to TNF, suggesting a protective effect could arise from treatment after injury. Moreover, chronic inflammatory pathologies may involve continuous generation of cytokines, and even acute insult may cause continuing generation of mediators, with the potential to develop a vicious proinflammatory feedback over time . In these cases, delivery of MSC should help resolve or inhibit full development of the damaging response. It is also notable that MSC cultured below but separated from the EC were anti-inflammatory. This suggests that endogenous MSC might naturally fulfil a regulatory role in inflammation, and/or that injection of MSC into tissue (instead of intravenously) could act locally to down regulate leukocyte recruitment.
It is also evident that the responses of MSC are highly context dependent. Contact with EC was required to generate anti-inflammatory mediators. Experimentally, this means that actions of MSC are best studied in realistic, multicellular models as described here. These findings also suggest the possibility that MSC delivered to the “right” milieu to channel their behavior, or indeed lodged in the lung capillaries after i.v. injection , may operate as a “factory” for the production of immunomodulatory agents which could act remotely. Alternatively, direct local application of IL-6 and/or associated mediators identified in coculture models might have benefit (obviating the need for delivery of cells themselves). Indeed, the potential for therapeutic use of such a “secretome” has gained recent attention . In summary, MSC have been suggested to be protective via a variety of cellular effects [1-3]. This study reveals another potent mode of action that might for instance ameliorate the outcome of acute cardiac events where animal studies suggest MSC may be therapeutic [5-7] and clinical trials are on-going .
This study shows that endothelial cells and mesenchymal stem cells engage in cross-talk through soluble mediators, leading to down-regulation of the responses of the endothelial cells to inflammatory cytokines. The resultant inhibition of leukocyte recruitment may represent a mechanism by which mesenchymal stem cells can limit inflammatory responses and vascular damage. The study also suggests a paradigm for context specific immunomodulatory effects of mesenchymal cells where interaction with endothelium causes mutual modification of cell phenotype.
This work was supported by The British Heart Foundation (grant number RG_10–153).
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors indicate no potential conflicts of interest.