Fibroblasts, once believed to be relatively inert cellular components of the tissue, are now believed to play an important role in tissue homeostasis, in particular, in regulating leukocyte recruitment in inflammation 4, 10
. Here we have demonstrated for the first time that fibroblasts can stimulate recruitment of lymphocytes, or limit this response in EC, depending on the tissue from which the fibroblasts were obtained. When fibroblasts from the synovium of patients with RA or OA were cultured with EC, adhesion of flowing lymphocytes was induced even in the absence of added cytokines. In contrast, dermal fibroblasts from the same donors, and fibroblasts from non-inflamed synovium, had little effect on adhesion to otherwise unstimulated EC, but actually reduced the number of lymphocytes binding in the presence of cytokines TNF-α+IFN-γ. In the cytokine-stimulated model, the fibroblasts from inflamed synovium were less consistent in their effects, with those from RA patients tending to be stimulatory, but those from OA patients being inhibitory. The responses induced by synovial or dermal fibroblasts from RA patients appeared to be largely caused by soluble mediators. These may have been made more effective by close juxtaposition of cells, but direct contact between cells was not essential. Interestingly, IL-6 had opposing effects depending on the milieu: it was essential for the stimulatory effects of synovial fibroblasts in the absence of added cytokines, but in the presence of TNF-α+IFN-γ, IL-6 along with TGF-β mediated the inhibitory effects of dermal fibroblasts.
Adhesion of lymphocytes to EC cultured with RA synovial fibroblasts was essentially abolished by blockage of α4
-integrin and greatly reduced by inhibition of CXCR4, the receptor for CXCL12 (SDF-1). This is consistent with previous studies showing that interaction between α4
-integrin and VCAM-1 can support capture of flowing lymphocytes by EC and that activation of this integrin is required to stabilise adhesion 20, 21
. Although CXCL12 has been shown to fulfil such an integrin-activating and adhesion stabilising role when added exogenously 22
, this is the first report of activity of endogenous, cell-generated CXCL12. We previously found that CXCL12 was constitutively expressed and secreted by rheumatoid synovial fibroblasts 16
. Since CXCL12 mRNA was barely detectable in EC here, and there are no reports of endogenous CXCL12 activity in HUVEC, it is highly likely that CXCL12 was transferred from the synovial fibroblasts and presented by the EC during co-culture. We found only slight up-regulation of VCAM-1 in co-cultures, while CXCR4 ligation was required for most of the adhesion seen in synovial co-cultures. This suggests that the receptors used for initial capture from flow were only slightly increased and that activation of lymphocyte integrins by CXCL12 from synovial fibroblasts was the critical factor for the elevation of stable adhesion. An analogous situation occurred in our previous studies, where synovial fibroblasts induced selectin-dependent adhesion of neutrophils to EC 9
. There, selectin expression was again barely detectable and stable adhesion required activation of CXCR2 on the neutrophils.
When EC were stimulated with TNF-α+IFN-γ, lymphocyte adhesion was greatly reduced by combined blockade of domains 1 and 4 of VCAM-1. Blockade of α4
-integrin on lymphocytes caused a similar degree of inhibition of adhesion. Thus for co-culture, it is likely that the interaction between α4
-integrin and VCAM-1 was the main basis of lymphocyte adhesion. Fibronectin has been shown previously to contribute to adhesion of lymphoid cells lines to HUVEC, but not PBL 23
. We attempted to quantify the level of fibronectin on the surface of our cultures by immunofluorescence labelling and flow cytometry, but did not detect a signal above that obtained with non-specific control antibody (data not shown). Thus, given the efficacy of the combined VCAM-1 blockade, no other ligand for PBL integrin need be postulated here. The finding that blockade of domain 4 of VCAM-1 was more effective than domain 1 was unexpected based on some previous reports. Domain 1 has been suggested to act as the main capture receptor from flow 24, 25
, but domain 4 may be important in adhesion strengthening for activated lymphocytes 26
. The data presented here mainly represent firmly adherent lymphocytes that have remained attached for minutes, and so our data suggest that domain 4 may be important for maintaining adhesion. E-selectin has also been shown to make a contribution to adhesion of PBL to TNF-α-treated HUVEC 27
and there was evidence of a minor contribution to adhesion here. Combination of antibodies against E-selectin and domain 1 of VCAM-1 had an effect greater than either antibody alone, suggesting both assisted in capture.
In the cytokine model, activation through CXCR3 was important to stabilise adhesion, and EC showed strong mRNA expression of three potential ligands for CXCR3 (CXCL9, 10 and 11). These results are consistent with previous reports demonstrating that CXCR3 ligands promote lymphocyte adhesion to TNF-α+IFN-γ-stimulated HUVEC 13
. Cultured synovial fibroblasts have been shown to secrete CXCL9, CXCL10 and CXCL11 following stimulation with IFN-γ alone 28
or in combination with TNF-α in vitro 29
and this might contribute to the augmentation of adhesion in synovial co-cultures. However, down-regulation of adhesion by co-culture with dermal fibroblasts was not associated with reduction in the expression of these chemokines. Although surface expression of VCAM-1 or E-selectin showed a tendency to be reduced, this was evident in the co-cultures with synovial fibroblasts as well. Thus while inhibition of the response to cytokines in the dermal co-cultures might arise from changes in secretion or presentation of chemokines, no firm conclusion can be made.
Soluble mediators, in particular IL-6, played important roles in the modulatory effects of fibroblasts. Rheumatoid synovial fibroblasts are known to secrete a range of pro-inflammatory cytokines, including IL-6 30
. Here, RA synovial co-cultures consistently secreted higher concentrations of IL-6 than dermal co-cultures in the absence of cytokines. Neutralisation of IL-6 significantly inhibited lymphocyte adhesion to EC cultured with synovial fibroblasts, as was the case in earlier studies of neutrophils adhesion to similar co-cultures 9
. However, when we added exogenous IL-6 to EC alone, this did not induce an increase in lymphocyte adhesion. This is consistent with work by Modur et al
, who found that addition of soluble IL-6Rα in addition to IL-6 was required to induce adhesion of neutrophils to EC. Previously, we did not detect sIL-6R in co-culture supernatants 9
. In fact, although the co-receptor Gp130 is widely expressed among different cells, IL-6Rα is believed to be restricted in expression to mononuclear leukocytes, fibroblasts and hepatocytes 31, 32
. Nevertheless, we examined IL-6Rα mRNA expression by qPCR and found detectable levels, albeit much lower than those for Gp130. Neither was modified by co-culture with synovial or dermal fibroblasts. This led us to evaluate the surface expression of IL-6Rα protein on EC by flow cytometry, where very low levels were detected in two cultures but labelling was not above background in two others. Again, this was the same for mono-cultures or co-cultures. The foregoing indicates that it is unlikely that the IL-6 acting alone directly induced increased adhesion to EC in the synovial co-cultures. IL-6 may have induced the synovial fibroblasts to increase the production of CXCL12 and/or other agents promoting lymphocyte recruitment. However, another possibility is that EC in the modified phenotype induced by synovial co-culture were more responsive to IL-6 than simple mono-cultures even though there was no evident change in the expression of receptors (see below).
In contrast, IL-6 took on an inhibitory role in conjunction with TGF-β, in the reduced response to cytokines induced by co-culture of EC with dermal fibroblasts. Interestingly, we found that addition of exogenous IL-6 greatly reduced the response of endothelial mono-cultures to TNF-α+IFN-γ. Such an inhibitory effect of IL-6 has not been reported in EC previously to our knowledge and suggests that the low levels of IL-6Rα, which we detected, could modify cellular responses at least. TGF-β has been reported to suppress the secretion of chemokines, including CXCL10, from HUVEC stimulated with TNF-α+IFN-γ 33
. In addition, pre-incubating human dermal microvascular EC with TGF-β inhibited their response to TNF-α or IL-1, and abrogated mononuclear leukocyte adhesion 34
. In vivo
, intra-tracheal injection of IL-6 with TGF-β inhibited neutrophil extravasation into LPS-induced lung inflammation 35
, indicating that the combination can act as an inhibitor of acute inflammation. Here, IL-6 may have modulated responses of EC to cytokines and may also have induced dermal fibroblasts to generate inhibitory accessory signals, including TGF-β, which reduced the ability of EC to bind lymphocytes (e.g
. by altering chemokine trafficking and/or presentation as noted above). Synovial fibroblasts released at least as much IL-6 as dermal fibroblasts, but the endothelial response to TNF-α+IFN-γ was not decreased in their presence. Thus it appears either that the co-cultured EC were in an altered phenotype that had a different response to IL-6 (as suggested above) or that secondary inhibitory agents were not released by the synovial cells. In any case, the results indicate that a single agent such as IL-6 can have quite different effects on an inflammatory response depending on the local environment and presence of other growth factors or cytokines.
Because the effects of IL-6 were disparate in the different types of co-culture, and changes in leukocyte recruitment were not simply attributable to changes in the expression of adhesion molecules or chemokines by the EC, we wondered whether the EC took on distinctive phenotypes when cultured with synovial or dermal fibroblasts. To investigate this we carried out gene microarray analysis for endothelial mono-cultures and co-cultures. The analysis indicated that although the differently cultured EC were distinguishable by their gene expression profiles, the synovial fibroblasts modulated a much greater number of genes than the dermal fibroblasts, and they induced a genomic state that was much more distinct from the EC cultured alone. This suggests that the fibroblasts from inflamed tissue not only provided CXCL12 and IL-6 but induced a phenotype in the EC, which was responsive to IL-6 and adhesive to lymphocytes. On the other hand, dermal fibroblasts left EC in a state that was not intrinsically responsive to IL-6. Instead, we hypothesise that the dermal fibroblasts provided IL-6, which moderated the response to cytokine treatment (an effect reproduced in mono-cultures), rather than establishing a new, non-responsive phenotype in advance.
In conclusion, these studies reveal that stromal cells can modulate the inflammatory response mediated by EC and that this modulation may itself depend on the state of the stromal cells in question (e.g
. their own exposure to a chronically-inflamed environment). Not only fibroblasts but hepatocytes and SMC can also modulate the ability of overlying EC to recruit leukocytes 7–9
. The effects may be directly stimulatory, or modulatory of responses to other agonists. It appears that fibroblasts from non-inflamed stroma can regulate the cytokine-sensitivity of vascular endothelium, while fibroblasts associated with chronic inflammation bypass this and develop a directly inflammatory phenotype. In the context of RA, the effects of synovial fibroblasts in co-culture are consistent with reports that CXCL12 is expressed highly in the synovium of the rheumatoid joint and may promote T-cell recruitment 36–38
. IL-6 has been implicated as a modulator of RA in humans 39
and in animal models 40
, while in rats, inhibition of α4
-integrin reduced the accumulation of T cells into arthritic joints 41
. Thus the model described here appears relevant to the pathology of RA in particular. More generally, co-cultures of appropriate primary cells, with or without added cytokines, offer a means to study physiological processes regulating inflammation in human tissue. The results in such systems may be complex as illustrated by the effects of IL-6 here, which depended on the local milieu and other released factors. It is thus difficult to predict a priori
which soluble mediators should be added, e.g
. to model inflammation using endothelial mono-cultures, and further studies are needed to fully define the regulatory or stimulatory repertoires of stromal cells.