Increased lymphangiogenesis in joints of mice with inflammatory arthritis

TNF-Tg mice (Tg 3647 line) in a CBA × C57BL/6 background were originally obtained from George Kollias (The Biomedical Sciences Research Center, Greece). We backcrossed them with C57BL6 mice for seven generations [20]. All TNF-Tg mice used in this study were 4 to 8 months old with severe joint synovitis and bone and cartilage destruction. WT littermates were used as controls. TNF-Tg mice were identified by tail polymerase chain reaction (PCR) genotyping and paw deformation. KRN-TCR-Tg mice were obtained from Drs. Diane Mathis and Christophe Benoist (Harvard University, Cambridge, MA, USA) [23]. K/B × N mice were generated by breeding male KRN-TCR-Tg mice with female non-obese diabetic mice (The Jackson Laboratory, Bar Harbor, ME, USA). Serum was collected from 6- to 12-week-old K/B × N arthritic mice, pooled, and stored at -80°C. To generate mice with serum-induced arthritis (SIA), 4- to 5-week-old BALB/c mice were injected with K/B × N serum intraperitoneally (10 μL/gram body weight) on day 1 and day 3. Paw swelling and redness usually occurred the day after the first injection, peaked at 7 to 14 days, and declined thereafter. The Institutional Animal Care and Use Committee of the University of Rochester (Rochester, NY, USA) approved all studies.

Recombinant murine TNF-α was purchased from R&D Systems, Inc. (Minneapolis, MN, USA). Allophycocyanin-, fluorescein isothiocyanate (FITC)-, and phycoerythrin-anti-mouse CD11b (M1/70) were purchased from eBiosciences (San Diego, CA, USA); FITC-anti-mouse Gr-1 (RB6-8C5) from BD Pharmingen (San Diego, CA, USA); rabbit polyclonal antibody to VEGF-C (H-190) from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA, USA); rabbit polyclonal antibody to LYVE-1 from Abcam (Cambridge, MA, USA); mouse monoclonal antibody to CD31 (Mec13.3) from Biocare Medical LLC (Concord, CA, USA); Alexa Fluor 488 goat anti-hamster immunoglobulin G (IgG) from Molecular Probes Inc. (now part of Invitrogen Corporation, Carlsbad, CA, USA); and Alexa Fluor 546 F(ab')₂ fragment of goat anti-rabbit IgG (H+L) and To-Pro-3 iodid (642/661) were purchased from Invitrogen Corporation.

OCPs were generated from two sources:

CD11b⁺/Gr-1^-/lo OCPs were purified by flow sorting from peripheral blood and bone marrow of TNF-Tg mice and age-matched WT mice by a FACSVantage SE Turbo sorter (Becton, Dickinson and Company, Franklin Lakes, NJ, USA), as we described previously [20, 28]. To obtain enough RNA from samples, bone marrow OCPs were pooled from 7 TNF-Tg mice or 11 WT mice and peripheral blood OCPs were from pooled from 7 TNF-Tg mice or 23 WT mice. Two completely independent experiments were performed. Total RNA was prepared using TRIzol (Invitrogen Corporation), processed, and hybridized to MG-U74Av2 gene chips according to Affymetrix protocols (Affymetrix, Santa Clara, CA, USA). Chips were scanned and analyzed using the GeneTraffic (version 3.2) microarray data analysis software (Lobion Informatics, La Jolla, CA, USA). In each group, WT samples were set as the baseline. Data were presented as the fold increase of samples from TNF-Tg mice over WT samples (baseline sample).

Total RNA was extracted using TRIzol reagent, and cDNA was synthesized by an RNA PCR Core Kit (Applied Biosystems, Foster City, CA, USA). Quantitative PCR amplification was performed with gene-specific primers using an iCycler iQ Multiple-Color Real-Time PCR Detection System (Bio-Rad Laboratories, Inc., Hercules, CA, USA), as we described previously [25, 28]. The primer sequences are listed in Table 1. Expression levels were normalized relative to β-actin in the same sample. For each sample, we first obtained the ΔC_T (difference in threshold cycle values), which is equal to C_T (target gene) – C_T (β-actin). We then used control (for example, phosphate-buffered saline [PBS]-treated) values as baseline to obtain the ΔΔC_T, which is equal to the ΔC_T of the sample minus the ΔC_T of the baseline. Finally, we set the baseline as '1', and thus the fold change of each sample is equal to 2^-ΔΔCT.

Cells were cytospun on glass slides and fixed by cold methanol at -20°C for 10 minutes. After washing with PBS, cells were incubated in 0.1% Triton and blocked with 1% bovine serum albumin/PBS for 30 minutes at room temperature. The fixed cells were stained with a mixture of FITC-anti-CD11b and anti-VEGF-C antibody followed by Alexa Fluor 546 F(ab')₂ fragment of goat anti-rabbit IgG (H+L) and by To-Pro-3 iodid.

Joint sections from TNF-Tg and WT mice were fixed in 4.5% phosphate-buffered formalin, decalcified in 14% ethylenediaminetetraacetic acid, and embedded in paraffin wax. Deparaffinized sections were quenched with 3% hydrogen peroxide and treated for antigen retrieval for 30 minutes. Adjacent serial sections were then stained with anti-LYVE-1 or anti-CD31 antibodies. Lymphatic vessels were quantified by a point-counting method, as we described previously [29]. For each mouse, two sections were cut at 250 μm apart and the area and size of LYVE-1⁺ lymphatic vessels were measured within the synovial tissue. The size and area of lymphatic vessels were expressed per square millimeter of synovium.

Raw 264.7 osteoclast/macrophage precursor cells were serum-starved overnight and then treated with 10 ng/mL of TNF for 30 and 60 minutes. Nuclear extract preparation and electrophoretic mobility shift assay (EMSA) were performed as described previously [30]. The following double-stranded oligonucleotides were used in this study according to the published mouse VEGF-C promoter sequence [18] (only the top strands are shown): VEGF-C nuclear factor-kappa B (NF-κB)-like sequence (WT), 5'-GCCCAGGGGGGTCCCCGGGAGG-3'; mutated VEGF-C NF-κB-like sequence (mutation underlined), 5'-GCCCAGGGGATTCTCCGGGAGG-3'; and SP-1-binding sequence (control; Invitrogen Corporation), 5'-CGAGCCGGCCCCGCCCATC-3'. For competition assays, binding reactions were pre-incubated with unlabeled oligonucleotides for 15 minutes at room temperature.

A detailed methodology of this new technique in mice will be published in a separate manuscript [31]. Briefly, mice were positioned with the right leg in a custom-designed murine knee coil and scanned in a Siemens 3 Tesla clinical magnet (Siemens AG, Munich, Germany). A high-resolution fat-suppressed T1-weighted sequence (Sagittal T1-weighted FLASH [fast low-angle shot], repetition time = 45 ms, echo time = 9.03 ms, 192 × 192 pixels, 20 mm × 20 mm field of view, 32 slices of 0.16-mm slice thickness, flip angle = 25°, 1 signal average, time: 8:28) was then initiated. An intravenous injection of Gd-DTPA contrast agent (Omniscan; Amersham Health, now part of GE Healthcare, Little Chalfont, Buckinghamshire, UK) was given, and a post-contrast high-resolution scan (same parameters) was collected. Analysis was performed with Amira 3.1 (TGS; Mercury Computer Systems, Inc., Chelmsford, MA, USA). An image registration and subtraction algorithm on the pre- and post-contrast images in Amira generated an image of the voxels of contrast enhancement. From this image, a three-dimensional (3D) region of interest of muscle tissue was used as a measure of delivered contrast agent, and a threshold of enhancing synovial tissue was generated from this value. Lymph nodes were traced manually and thresholded to define the margin between the node and surrounding fat. Tissue volumes (3D) were reconstructed using a surface generation module in Amira.

Data are presented as mean ± standard deviation of three culture dishes, and all experiments were performed at least twice. Statistical analyses were performed with Statview statistical software (SAS Institute Inc., Cary, NC, USA). Differences among more than two groups were compared using one-way analysis of variance followed by the Mann-Whitney U test. P values of less than 0.05 were considered statistically significant. Each experiment was repeated at least twice with similar results.

Previous studies demonstrated an increase in circulating OCPs in patients [32] and animals [20] with arthritis and that OCP frequency is reduced in response to anti-TNF therapy, suggesting that OCPs may play important roles in the pathogenesis of arthritis [33]. To screen for novel genes that are differentially expressed by OCPs between arthritic and normal mice, we performed microarrays on RNA isolated from CD11b⁺/Gr-1^-/lo OCPs from peripheral blood mononuclear cells and bone marrow pooled from TNF-Tg and WT mice. The purity of CD11b⁺/Gr-1^-/lo cells was confirmed by FACS. In our initial bioinformatic screen, we focused on genes encoding angiogenic factors because they are critical for development of inflammation and bone erosion in arthritic joints. Among more than 50 known angiogenic factors (including matrix metalloproteinases, adhesion molecules, enzymes, and growth factors), we found that expression levels of PDGF-B, PDGF receptor β, and VEGF-C were significantly increased in circulating OCPs of TNF-Tg mice (approximately six-fold in TNF-Tg over WT OCPs) (Figure 1). Expression levels of these genes were also increased in bone marrow OCPs but to a lesser extent. Since TNF is known to induce PDGF signaling [34], we decided to explore the possibility that VEGF-C has a novel role in inflammatory arthritis similar to its role in metastatic cancer [35, 36].

To determine whether TNF directly upregulates expression of VEGF-C by OCPs, we cultured WT spleen cells with M-CSF for 3 days to generate OCPs, as we described previously [25]. The rationale for using spleen cells rather than bone marrow cells is that spleen-derived OCPs are closer to circulating OCPs than bone marrow OCPs in terms of their osteoclast-forming potency [37] and the increased VEGF-C expression level in circulating OCPs is more than three times higher than in bone marrow OCPs (Figure 1). TNF treatment increased VEGF-C mRNA levels by three- to four-fold in a time- and dose-dependent manner, beginning at between 4 and 8 hours, suggesting transcriptional regulation (left panel in Figure 2a). It also increased VEGF-C protein expression at 24 hours (right panels in Figure 2a). TNF did not significantly affect expression of other VEGF members (Figure 2b). Dose-response experiments demonstrated that a low dose of TNF (0.5 ng/mL) is sufficient to stimulate VEGF-C protein expression, supporting an in vivo role for TNF to induce VEGF-C expression (Figure 2c).

We also examined whether IL-1, another cytokine whose levels are increased significantly in joints of TNF-Tg mice, affects VEGF-C levels in OCPs. Similar to TNF, IL-1 treatment (by 24 hours) also increased VEGF-C mRNA by two- to three-fold, but IL-1 + TNF did not have an additive effect (Figure 2d). This suggests that TNF and IL-1 may use a similar signaling pathway to regulate VEGF-C expression.

NF-κB is a transcription factor that mediates induction of genes by TNF in many cell types, including OCPs [38]. NF-κB also mediates heregulin-beta-1-induced VEGF-C expression in human breast cancer cells [39]. We found that there is a putative NF-κB-binding element, GGGGTCCC, at the -108/-99 region of the mouse VEGF-C promoter, which is at a location similar to that in the human promoter [18]. Thus, to determine whether TNF stimulates binding of NF-κB proteins to this element in OCPs, we treated Raw 264.7 cells with TNF for 30 and 60 minutes and performed Western blot analysis and EMSA on nuclear extracts from the cells. TNF treatment for 30 minutes increased nuclear translocation of NF-κB p65 and p50 proteins (Figure 3a) and specific binding of NF-κB to the VEGF-C DNA probe (Figure 3b). To further confirm the involvement of NF-κB in TNF-induced VEGF-C expression, we treated WT OCPs with an NF-κB inhibitor (Calbiochem, now part of EMD Biosciences, Inc., San Diego, CA, USA) and found that it inhibited TNF-induced VEGF-C expression in a dose-dependent manner (Figure 3c). These data support TNF induction of VEGF-C expression by activation of NF-κB.

To examine whether OCPs at the site of inflammation express VEGF-C, we isolated cells from joints of TNF-Tg mice and WT littermates using a sequence enzyme digestion method [26, 27]. We obtained, on average, a total of 1.5 × 10⁶ cells from the four paws of a TNF-Tg mouse. FACS analysis revealed that approximately 60% of joint cells from TNF-Tg mice are CD11b⁺ and almost all of them are Gr-1^- (Figure 4a). Thus, in subsequent experiments, we used only CD11b as a single marker for OCPs. To determine whether CD11b⁺ cells express VEGF-C, we performed double immunofluorescence staining using anti-CD11b and VEGF-C antibodies in a cytospin preparation of these cells. There were much more CD11b and VEGF-C double-stained cells in the joints of TNF-Tg mice than in WT mice (Figure 4b). We also found that most of the VEGF-C⁺ cells are CD11b⁺, whereas no CD11b^- cells are VEGF-C^-, suggesting that VEGF-C is produced in arthritic joints by OCPs. To examine whether these joint CD11b⁺ cells specifically produce more VEGF-C than other VEGFs, we cultured them with M-CSF for 3 days, harvested adherent cells, and performed real-time reverse transcription-PCR to assess the expression level of VEGF-A, VEGF-C, and VEGF-D. The rationale for using these M-CSF-dependent adherent cells is that they are composed mainly of OCPs and monocytes, as we described previously [25]. We found that M-CSF-dependent cells that were derived from joints of TNF-Tg mice expressed much higher levels of VEGF-C compared with those of WT mice (Figure 4c) whereas VEGF-A and VEGF-D levels were similar, suggesting that OCPs/monocytes are the major source of VEGF-C in arthritic joints of TNF-Tg mice. To confirm that there is elevated TNF or IL-1 expression in the joints of TNF-Tg mice, we measured the expression levels of TNF and IL-1 mRNA and demonstrated that they were remarkably increased compared with levels in WT joints (Figure 4d), which supports the hypothesis that local increased cytokines stimulate VEGF-C production.

Since VEGF-C is a specific lymphatic growth factor [40] and VEGF-C expression is increased in OCPs, we sought to determine whether there was increased lymphangiogenesis in joint sections from TNF-Tg mice compared with WT mice using immunohistochemistry and an antibody to the lymphatic endothelial cell marker, LYVE-1. We also assessed, as a positive control, blood vessels lined by endothelial cells using a CD31 antibody because these are known to be increased in TNF-Tg versus WT joints [41]. These results clearly demonstrated large LYVE-1⁺ lymphatic vessels compared with smaller CD31⁺ vessels in the pannus of TNF-Tg mice (Figure 5a). Histomorphometric quantitation of the area and number of these lymphatic vessels confirmed a significant increase in TNF-Tg versus WT joints (Figure 5b). To assess the functional significance of this increased lymphangiogenesis, we used a new in vivo MRI technology that we have developed [31] to see whether the draining popliteal lymph nodes from the knee and ankle joints of TNF-Tg mice with synovitis are increased versus WT littermates. Figure 6a demonstrates the remarkable differences in 2D MRI and 3D reconstructed images of the synovium and popliteal lymph nodes of TNF-Tg versus WT mice. Volumetric analyses confirmed that TNF-Tg mice have significantly larger synovial and popliteal lymph node volumes compared with WT mice (Figure 6b) and these were confirmed with histology (Figure 6c).

To demonstrate that increased lymphangiogenesis is not limited to the arthritis in TNF-Tg mice and to determine the association between joint inflammation and lymphatic vessel formation during the course of arthritis induction, we used mice with K/B × N SIA. This SIA model has clearly distinguishable phases of disease progression and uniform severity of joint lesions among animals compared with TNF-Tg mice. The SIA mice were sacrificed at 0, 14, and 35 days after K/B × N serum injection, and joint samples were subjected to histology and LYVE-1 immunostaining. At day 14 after serum injection, mice developed severe joint inflammation with pannus formation and inflammatory cell infiltration, which is accompanied with increased lymphatic vessel area and number (Figure 7). By day 35, the inflammation declined, but the lymphatic vessel formation had increased further (Figure 7) to a degree similar to that seen in TNF-Tg mice (Figure 5). These data suggest that increased lymphangiogenesis is a common phenotype of inflammatory-erosive arthritis.