Human inflammatory synovial fibroblasts induce enhanced myeloid cell recruitment and angiogenesis through a hypoxia-inducible transcription factor 1α/vascular endothelial growth factor–mediated pathway in immunodeficient mice
Hyperplasia and phenotypic changes in fibroblasts are often observed in chronic inflammatory lesions, and yet the autonomous pathogenic contribution of these changes is uncertain. The purpose of this study was to analyze the intrinsic ability of fibroblasts from chronically inflamed synovial tissue to drive cell recruitment and angiogenesis.
Fibroblasts from patients with rheumatoid arthritis (RA) or osteoarthritis (OA), as well as fibroblasts from healthy synovial tissue and healthy skin, were cultured and subcutaneously engrafted into immunodeficient mice. Cell infiltration and angiogenesis were analyzed in the grafts by immunohistochemical studies. The role of vascular endothelial growth factor (VEGF), CXCL12, and hypoxia-inducible transcription factor 1α (HIF-1α) in these processes was investigated using specific antagonists or small interfering RNA (siRNA)–mediated down-regulation of HIF-1α in fibroblasts.
Inflammatory (OA and RA) synovial fibroblasts, compared with healthy dermal or synovial tissue fibroblasts, induced a significant enhancement in myeloid cell infiltration and angiogenesis in immunodeficient mice. These activities were associated with increased constitutive and hypoxia-induced expression of VEGF, but not CXCL12, in inflammatory fibroblasts compared with healthy fibroblasts. VEGF and CXCL12 antagonists significantly reduced myeloid cell infiltration and angiogenesis. Furthermore, targeting of HIF-1α expression by siRNA or of HIF-1α transcriptional activity by the small molecule chetomin in RA fibroblasts significantly reduced both responses.
These results demonstrate that chronic synovial inflammation is associated with stable fibroblast changes that, under hypoxic conditions, are sufficient to induce inflammatory cell recruitment and angiogenesis, both of which are processes relevant to the perpetuation of chronic inflammation.
Fibroblasts are ubiquitous mesenchymal cells with vital functions during development and adulthood. They synthesize the extracellular matrix components of connective tissues needed for homeostasis and reparative responses. During development, interactions between mesenchymal and other cell lineages are necessary for the formation of many organs, and fibroblasts are sufficient to provide the positional cues required for the induction and development of the different tissues (1, 2). In the adult, multiple evidence points to specialized fibroblasts as a major force in the regulation of cell homing, migration, and differentiation of highly dynamic cell populations, such as cells of the immune system or the bone marrow (3, 4). With regard to the pathologic functions of fibroblasts, besides their contribution to tissue damage or repair, fibroblasts seem to play critical roles in orchestrating the homing, growth, or function of other cell types, such as inflammatory or cancer cells (5–8).
The contribution of a pathologic fibroblast phenotype to the development of tumors is widely recognized (7–9). Tumor stroma has the capacity to induce the recruitment of bone marrow–derived myeloid cells that, in concert with cancer fibroblasts, induce a strong angiogenic response fostering tumor growth (8, 10). Hypoxia plays a relevant role in this process through the activation of hypoxia-inducible transcription factor (HIF), consistent with the role of HIF as a tumor-progression factor (11–13). Two important HIF-responsive fibroblast factors are vascular endothelial growth factor (VEGF) and the chemokine CXCL12, both of which can synergize in the recruitment and retention of myeloid cells, a critical cell element in the angiogenic response of tumor stroma (11, 14).
An important contribution of the hyperplasia of fibroblasts to chronic inflammation and tissue destruction has also been proposed, particularly in patients with rheumatoid arthritis (RA), in whom there is sufficient evidence to indicate an association between abnormal fibroblast phenotype and chronic inflammation (15). The crosstalk between fibroblasts and infiltrating leukocytes seems necessary for arthritis development, such that specific targeting of synovial fibroblasts is sufficient to reduce the inflammatory process (16). Furthermore, tumor necrosis factor receptor I expression by stromal cells is necessary and sufficient to drive arthritis in experimental models (17). Macrophage infiltration and angiogenesis are critical processes in the pathogenesis of chronic arthritis, and their indirect down-regulation by different therapies has been demonstrated as an early and reliable marker of clinical response (18–22). Inflammatory stroma in chronic RA mirrors several features of cancer stroma, including a severely reduced oxygen concentration, enhanced macrophage infiltration and angiogenesis, and a pseudotumoral growth and invasion of neighboring tissues (23–26). Therefore, we have hypothesized that chronic inflammatory and tumor environments could similarly favor myeloid cell recruitment and angiogenesis through phenotypic changes in stromal fibroblasts (27).
On the basis of this hypothesis, we used engraftment of fibroblasts obtained from patients with chronic arthritis into immunodeficient mice. Our results demonstrated that inflammatory fibroblasts display an enhanced ability to induce the recruitment of myeloid cells and angiogenesis, and these changes were correlated with increased VEGF expression. The process could be blocked at different levels, including via the inhibition of the HIF-1α/VEGF axis and the chemokine CXCL12.
MATERIALS AND METHODS
Human cells and tissues.
Synovial tissue samples were obtained from 11 patients with RA and 9 patients with osteoarthritis (OA) at the time of knee prosthetic replacement surgery, and from 7 adult donors without a history of joint disease, from whom macroscopically healthy joints were obtained at necropsy or at elective knee arthroscopic surgery for meniscal tears. Healthy skin was obtained from 4 individuals during cosmetic surgery procedures. The study was approved by the Ethics Committee of the Hospital 12 de Octubre, and the tissue samples were obtained after the subjects had provided their informed consent. Fibroblast cultures were established by explant growth in 10% fetal calf serum/Dulbecco's modified Eagle's medium and used between passages 3 and 9. Hypoxic cultures (exposed to an atmosphere of 0.5% O2) were obtained in a cell incubator under controlled anaerobic (CO2/N2) conditions.
Fibroblast implants in immunodeficient mice.
Fibroblast implants were prepared by suspension of 0.5 × 106 fibroblasts in 500 μl of Matrigel (BD Biosciences, San Jose, CA). Matrigel was injected subcutaneously into the back skin of Athymic Nude-Foxn1nu mice (Harlan-Ibérica, Barcelona, Spain). After 7 days, the skin containing the Matrigel plugs was excised and snap-frozen. Frozen sections were fixed in 4% paraformaldehyde and examined by hematoxylin and eosin staining or used for immunolabeling studies.
Matrigel sections were analyzed by immunofluorescence or immunoperoxidase labeling with an anti-CD31 (platelet endothelial cell adhesion molecule) antibody (Santa Cruz Biotechnology, Santa Cruz, CA), anti–α-smooth muscle actin (anti–α-SMA) antibody (clone 1A4; Sigma-Aldrich Química, Madrid, Spain), or phycoerythrin-labeled anti-CD11b antibody (clone M1/70; BD PharMingen, San Jose, CA). Sections were counterstained with 4′,6-diamidino-2-phenylindole or hematoxylin. Secondary antibodies, either biotinylated (Vector Laboratories, Burlingame, CA) or labeled with Alexa 488 (Invitrogen, Eugene, OR), were used. Quantitative data were obtained by counting the number of CD31-positive blood vessels or the total number of positively staining cells per area in digitalized images covering the whole Matrigel area, using ImageJ software (http://rsb.info.nih.gov/ij).
Mice were treated by intraperitoneal (IP) administration of the specific CXCL12/CXCR4 receptor antagonist bicyclam AMD3100 (Sigma-Aldrich Química) or the anti-human VEGF monoclonal antibody (mAb) bevacizumab (Roche Farma S.A., Madrid, Spain). The HIF-1α transcriptional antagonist chetomin (Alexis Biochemicals, San Diego, CA) was incorporated into the Matrigel matrix prior to the suspension of fibroblasts.
Transduction of fibroblasts with lentiviral–green fluorescent protein (GFP) and lentiviral–small interfering RNA (siRNA).
To track the fibroblasts implanted into the excised Matrigel plug sections, fibroblasts were first transduced with GFP-expressing lentiviral particles, obtained by cotransfection of 293T cells with pRRL.eGFP transfer vector, pMDLgag/ pol-RRE, pRSV-Rev packaging vectors, and pMD2-VSVg envelope vector (28). Supernatants were harvested 48 hours after transfection, filtrated through a 0.45-μm filter, and directly used for fibroblast transduction. GFP expression was directly examined by fluorescence microscopy of fibroblast cultures before incorporation of the fibroblasts into the Matrigel implants.
HIF-1α–targeting siRNA, as previously described (29), and a control siRNA containing the same scrambled sequence were cloned on a pRNAT lentiviral transfer vector (GenScript, Piscataway, NJ). The efficiency of HIF-1α targeting was checked by Western blotting with specific HIF-1α antibodies (BD PharMingen) and by quantitative reverse transcription–polymerase chain reaction (RT-PCR) for VEGF messenger RNA (mRNA) expression in fibroblasts exposed to an atmosphere of 0.5% O2 or treated with 300 μM CoCl2.
Real-time quantitative RT-PCR.
Total RNA was extracted, and 1 μg was used for first-strand complementary DNA synthesis. Quantitative PCR analysis was performed on a Roche LightCycler instrument using SYBR Green PCR Master Mix (Roche Diagnostics, Mannheim, Germany) according to the manufacturer's recommendations. The following sequences of primers were used: for CXCL12, sense 5′-TCTGAGAGCTCGCTTGAGTG-3′ and antisense 5′-GTGGATCGCATCTATGCATG-3′; for VEGF, sense 5′-GGTGAAGTTCATGGATGTCT-3′ and antisense 5′-GCTGTAGGAAGCTCATCTCT-3′; and for β-actin, sense 5′-CTACCTCATGAAGATCCTCAC-3′ and antisense 5′-GTCCACGTCACACTTCATGATG-3′. For relative quantification, we compared the amount of target mRNA normalized to that of the endogenous reference, using the 2 formula, where Ct is the mean of the threshold cycle at which the amplification of the PCR product is initially detected.
Data were analyzed using Prism software (GraphPad Software, San Diego, CA). Either Student's t-test or the Mann-Whitney test was used, as appropriate, to determine statistically significant differences in the quantitative PCR or histologic data. In all analyses, P values less than 0.05 were considered significant.
Induction of enhanced myeloid cell recruitment and angiogenesis by inflammatory fibroblasts in immunodeficient mice.
To examine the intrinsic ability of human inflammatory fibroblasts to induce cell recruitment and homing in vivo, we implanted Matrigel plugs containing synovial fibroblasts derived from human inflammatory (OA and RA) synovial tissue by subcutaneous injection into immunodeficient mice. After 7 days, the implants containing inflammatory fibroblasts showed a dense cellular infiltrate composed of vessel-like structures, some of which displayed a central lumen (Figure 1A). In contrast, control Matrigel implants, which lacked fibroblasts, remained acellular, and no vascular structures could be identified (Figure 1A). These structures in the inflammatory fibroblast–derived implants were identified as mature blood vessels composed of CD31-positive endothelial cells and perivascular anti–α-SMA–positive pericytes, and were surrounded by a strong perivascular mononuclear cell infiltrate showing positivity for the mouse CD11b myeloid cell marker (Figures 1B and C).
Engrafting of implanted fibroblasts was tracked by transduction of fibroblasts with lentiviral–GFP prior to engraftment. GFP expression was detectable in >90% of the transduced fibroblasts in tissue culture. GFP-positive fibroblasts were identified in the Matrigel implants, wherein they represented a low proportion of the cells. GFP-positive fibroblasts had a scattered distribution without a blood vessel disposition (Figure 1D) and did not show colocalization with CD31 immunolabeling, thus discarding the possibility of differentiation of implanted cells into endothelial cells (results not shown).
The extent of cell infiltration and that of angiogenesis were variable in the Matrigel implants containing the different fibroblast lines but were more reproducible in different implants of the same line in the different mice. The maximal responses were induced by RA fibroblasts, but OA fibroblasts also induced a significantly increased response compared with fibroblasts obtained from noninflammatory tissues (healthy skin or synovial tissue), as shown in Figures 2A and B.
Increased constitutive and hypoxia-induced expression of VEGF, but not CXCL12, in inflammatory synovial fibroblasts.
Previous studies have identified VEGF and CXCL12 as 2 fibroblast-derived factors critical to myeloid cell recruitment and angiogenesis in other settings (11, 14). To analyze whether differences in the expression of CXCL12 or VEGF account for the observed differences between healthy and inflammatory fibroblasts, we quantified VEGF and CXCL12 mRNA expression in the different groups of fibroblasts. Significantly higher constitutive VEGF mRNA expression was observed in inflammatory RA and OA fibroblasts compared with healthy fibroblasts, whereas similar levels of constitutive CXCL12 mRNA expression were observed in all groups (Figure 3A). No significant differences in VEGF or CXCL12 mRNA levels were observed between OA and RA synovial fibroblasts (Figure 3A). We therefore used RA synovial fibroblasts in further studies.
Since avascular Matrigel implants represent a critically hypoxic environment similar to that in inflammatory tissue (22–24), we analyzed VEGF expression in response to hypoxia in the different groups of fibroblasts. In fibroblasts incubated under an atmosphere of 0.5% O2, transient cytoplasmic and nuclear accumulation of HIF-1α was observed (Figure 3B), similar to that reported in other cell types (30). HIF-1α accumulation was closely followed by a strong induction of VEGF mRNA expression (Figure 3C). Although the magnitude of the response to hypoxia was similar in inflammatory (RA and OA) synovial fibroblasts and healthy synovial fibroblasts, the absolute expression of VEGF was also significantly higher under hypoxic conditions in both RA and OA synovial fibroblasts compared with healthy synovial fibroblasts (Figure 3A). A slightly lower level of VEGF mRNA was detected in hypoxic OA fibroblasts compared with hypoxic RA fibroblasts, but the difference was not statistically significant (Figure 3A). Expression of CXCL12 mRNA showed a detectable, but weaker, induction by hypoxia (Figure 3C), but the absolute expression levels of CXCL12 were similar in all groups of synovial fibroblasts (Figure 3A).
Enhanced angiogenesis has been shown to correlate with increased CXCL12 expression and myofibroblast differentiation in some tumors (8). Although a low proportion of myofibroblasts (anti–α-SMA–positive cells) could be demonstrated by immunofluorescence labeling in some fibroblast cultures, we failed to detect a correlation between the distribution of anti–α-SMA–positive cells and CXCL12 expression (results not shown), nor did we detect a correlation between CXCL12 expression and the angiogenic capacity of the different fibroblast groups, as shown in Figures 2 and 3.
Inhibition of inflammatory fibroblast–induced cell infiltration and angiogenesis by VEGF or CXCL12 antagonists.
To confirm the roles of VEGF and CXCL12 derived from inflammatory synovial fibroblasts in myeloid cell recruitment and angiogenesis, we analyzed the effect of their specific antagonists in vivo. The specific anti-human VEGF mAb bevacizumab was used to neutralize the expression of fibroblast-derived VEGF, since, at the selected dose, it has been shown to neutralize expression of human, but not mouse, VEGF in human tumor xenografts in mice (31). A strong inhibitory effect was achieved by a single IP injection of bevacizumab at a dose of 5 mg/kg at the time of Matrigel injection. The inhibitory effect included a significant reduction in cell infiltration as well as a decrease in the number of new vessels in the Matrigel implants (Figure 4).
Similarly, in mice treated with the CXCR4 receptor antagonist bicyclam AMD3100 by daily IP injection of 300 μg, a significant reduction in both cell infiltration and angiogenesis was observed (Figure 4). Therefore, either the VEGF or CXCL12/CXCR4 antagonist was sufficient to reduce myeloid cell infiltration and angiogenesis, supporting the relevance and complementary roles of both fibroblast factors in this process.
Inhibition of inflammatory fibroblast–induced cell infiltration and angiogenesis by the HIF transcriptional antagonist and by HIF-targeting siRNA.
Since both fibroblast-derived factors, VEGF and CXCL12, seemed to contribute to cell infiltration and angiogenesis and both are inducible by hypoxia (12, 13, 25), we investigated the effect of either targeting the HIF-dependent transcriptional response with the use of the small molecule chetomin or blocking the expression of HIF-1α by transduction of fibroblasts with specific lentiviral-siRNA (29, 32).
Treatment of RA fibroblast cultures with the HIF transcriptional antagonist chetomin induced significant cytotoxicity at concentrations of ≥50 nM (results not shown), although these higher concentrations have been previously reported to be noncytotoxic in other cell types (32). At the noncytotoxic concentration of 10 nM, chetomin was still able to suppress the induction of VEGF mRNA in response to hypoxia (Figure 5A). In Matrigel implants containing RA fibroblasts and 10 nM chetomin, both significantly reduced cell infiltration and significantly reduced angiogenesis were observed (Figure 5B).
Similarly, specific targeting of HIF-1α by lentiviral siRNA transduction of RA fibroblasts was able to reduce both HIF-1α accumulation and induction of VEGF mRNA expression (Figure 5C). Matrigel implants containing HIF-1α siRNA–transduced RA fibroblasts showed significant reductions in both cell infiltration and angiogenesis compared with that in control siRNA–transduced fibroblasts (Figure 5D).
The potential contribution of synovial fibroblasts to chronic inflammation has been long recognized (15–17). Under the influence of exogenous microbial products, cytokines, or other proinflammatory stimuli, they can release a variety of mediators, such as cytokines, chemokines, or matrix metalloproteinases, that contribute to inflammation and tissue damage. Previous observations suggest that RA synovial fibroblasts acquire an abnormal and heritable phenotype, which includes an enhanced ability to invade and destroy cartilage and a perturbed expression of cytokines or chemokines (33–35). Our results show that in the absence of additional stimuli, and under hypoxic conditions ex vivo, fibroblasts in the setting of inflammatory arthritis induce the recruitment of myeloid cells and the development of blood vessels, a capacity previously observed in cancer stroma, with wide implications in the perpetuation of chronic inflammation (11, 36, 37).
Multiple potentially chemotactic factors have been found to be overexpressed by inflammatory fibroblasts, and this could contribute to leukocyte recruitment (35, 38). Increased constitutive VEGF expression partly explains the observed inflammatory fibroblast phenotype. The strong effect of the VEGF antagonist not only in angiogenesis but also in myeloid cell recruitment supports the local relevance of this factor as the prime mover of these processes. Local and systemic levels of VEGF have been found to be increased in RA, and plasma VEGF levels are rapidly down-regulated by effective therapies (18–20). However, increased VEGF expression has not been described previously in cultured RA fibroblasts, possibly due to the use of OA fibroblasts, and not healthy fibroblasts, as controls (25). Interestingly, both OA and RA fibroblasts were characterized by an enhanced proangiogenic capacity, pointing to the possibility of a common response occurring in chronic inflammatory stroma, rather than a disease-specific process. OA is considered a milder inflammatory disease, but significant inflammation and vascular remodeling also seem to contribute to joint destruction (39, 40).
Although CXCL12 was not overexpressed by inflammatory fibroblasts, the antagonist of its specific CXCR4 receptor was also able to down-regulate cell recruitment and angiogenesis. Therefore, both VEGF and CXCL12 fibroblast-derived factors seem required for the recruitment of myeloid cells and angiogenesis. In a transgenic model of organ-specific VEGF/CXCL12 overexpression, both factors have shown complementary roles in the recruitment and retention of perivascular myeloid cells necessary for the angiogenic response (14). We have previously identified CXCL12 as one of the factors responsible for the angiogenic activity in RA synovial fluid (26). The results of previous studies showing that myofibroblastic differentiation was associated with increased CXCL12 expression explain the enhanced angiogenic response of some cancer fibroblasts (8). Although fibroblasts are the main source of CXCL12 in RA synovium, we did not find a correlation between increased constitutive CXCL12 expression, myofibroblast differentiation, and the proangiogenic capacity of inflammatory fibroblasts, which is consistent with the results of previous studies (25, 26, 41). Therefore, the pathologic differentiation of fibroblasts seems heterogeneous and can lead to a proangiogenic phenotype through different molecular pathways. In different cancer types, stable overexpression of platelet-derived growth factor or fibroblast growth factor type 2 have also been found to contribute to angiogenesis (42, 43).
Deletion of HIF-1α in cells of the myeloid lineage in murine models of inflammation has been shown to reduce cell infiltration as a result of metabolic changes that may limit the myeloid cell migratory capacity (44). VEGF and CXCL12 function as HIF transcriptional targets under hypoxic conditions (13, 45), providing an additional link between hypoxia and inflammatory cell infiltration. Our results show that the response of resident cells to hypoxia is also critical to the recruitment of inflammatory cells through the synthesis of chemotactic and proangiogenic factors such as VEGF. Therefore, targeting of HIF-mediated responses could also revert the contribution of stromal fibroblasts to chronic arthritis perpetuation. In our model, both inhibition of HIF transcriptional activity and specific siRNA targeting of HIF expression had similar effects. A large variety of HIF-targeting compounds have been described, but none of these compounds has demonstrated sufficient specificity (46). In our experimental model, we used the small molecule chetomin, which is the most potent HIF/P300 transcriptional inhibitor, when tested at nanomolar pharmacologic and noncytotoxic concentrations, identified so far (32). Although it lacks nonspecific transcriptional effects, the specificity and additional biologic activities of chetomin have not been thoroughly evaluated.
Current targeted therapies for RA are directed against T cells, B cells, or macrophage cytokines, and all reduce macrophage and lymphoid cell infiltration of the synovial membrane to a similar extent (47–49). However, within a short time after therapy withdrawal, the disease almost invariably relapses. Despite the remission of inflammation and cell infiltration, a hyperplasic stroma with an expanded vascular bed may remain, and this is correlated with further damage to the bone and cartilage tissue (22, 50–52). Therapies specifically targeted to fibroblasts or vascular responses are not available for RA. Targeting stromal cell–derived proangiogenic factors and targeting HIF transcriptional responses represent alternative approaches to reduce the contribution of fibroblasts to the chronic inflammatory response.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Pablos had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Santiago, Galindo, Pablos.
Acquisition of data. Del Rey, Izquierdo, Caja, Usategui, Santiago, Galindo, Pablos.
Analysis and interpretation of data. Del Rey, Izquierdo, Caja, Usategui, Pablos.
We thank Drs. F. J. Blanco and M. J. López-Armada (Hospital Juan Canalejo, Madrid, Spain) and members of the Servicio de Traumatología (Hospital 12 de Octubre) for providing the synovial tissue samples, Dr. J. C. Segovia (CIEMAT, Madrid, Spain) for providing the lentiviral vectors, Dr. P. M. Chumakov (Cleveland Clinic Foundation, Cleveland, OH) for providing the HIF-1α siRNA vector, and Dr. J. C. Ramírez (CNIC, Madrid, Spain) for help with the lentiviral methods.