Serum Amyloid A (apoSAA) Expression Is Up-Regulated in Rheumatoid Arthritis and Induces Transcription of Matrix Metalloproteinases

The serum amyloid A (SAA)⁴ protein gene family consists of four members, which are divided into two classes, acute-phase and constitutive SAAs (1). Although acute-phase SAA (A-SAA) encoded by the two genes SAA1 and SAA2 is induced by inflammatory stimuli up to 1000-fold (2), the constitutive SAA4 is only inducible to a small degree (3). There is evidence suggesting that the human SAA3 is a pseudogene (4). Although the major site of A-SAA expression during inflammation is the liver, extrahepatic expression of members of the SAA gene family has been described (5, 6). Multiple functions have been assigned to SAA proteins, such as involvement in cholesterol metabolism (7), interference with platelet function (8), and suppression of the immune response (9). More recently, chemoattractant properties of A-SAA on monocytes, neutrophils, and T-cells have been demonstrated (10, 11). In rabbit synovial fibroblasts, SAA3 induces collagenase (matrix metalloproteinase (MMP)-1) (12). It is an accepted hypothesis that MMPs play a critical role in the pathology of rheumatoid arthritis (RA) (13). Recently expression of A-SAA mRNA has also been detected in human RA synovial tissue and cells (14, 15).

Analyzing the gene expression profile in diseased relative to normal tissue will contribute to a better understanding of the molecular mechanism underlying RA. In this study, we describe the results of gene expression profiling in the rabbit model of Ag-induced arthritis (AIA), which in many aspects, such as histopathology and phenotype of tissue degradation, resembles the human disease. Total RNA from control and diseased synovium was isolated and used to generate cDNAs. Using hybridization of complex amplified cDNA probes to high-density filter arrays and cDNA representational difference analysis (RDA), we were able to identify genes that were differentially expressed in the diseased synovium. The fact that the majority of the differentially expressed genes that have a significant match to a known gene in GenBank had previously been implicated in RA or inflammatory responses in general supports the validity of our experimental approach using the rabbit AIA model. Differentially expressed genes included metalloproteinases such as collagenase-1, collagenase-3, stromelysin, cysteine proteases such as cathepsin D, cell surface molecules like CD44, as well as proteinases that are involved in the activation of proinflammatory cytokines such as IL-1β converting enzyme (F.F., P.W., C.T., R.V., and G.B., unpublished results). However, as the major difference product in rabbit AIA we identified SAA3. Thus we concentrated our further studies on a potential role of SAA in RA.

In this study, we describe a functional analysis of SAA3 and provide evidence that rabbit SAA3 and as well as the human A-SAA induce transcription of MMPs in chondrocytes in vitro suggesting that they probably also play an important role in vivo for the pathogenesis of arthritis.

The data presented in this paper were obtained from experiments performed three times unless stated otherwise.

Animal experiments were performed in accordance with the animal experimentation guidelines and laws of the Swiss Federal and Cantonal Authorities as described in the Basel-Stadt Experimental License No. 1438. Female dwarf Russian rabbits were sensitized intradermally to methylated BSA (mBSA) homogenized 1:1 with CFA on days −28 and −14 (0.5 ml containing 4 mg/ml mBSA). On day 0, rabbits were anesthetized for the intraarticular injections. The right knee received 0.5 ml of 2 mg/ml mBSA in 5% glucose, while the left knee received 0.5 ml of 5% glucose alone. Rabbits (n = 3) were killed on day 7. The synovium of the Ag-challenged and vehicle-injected joint was prepared and total RNA was isolated. Experiments were repeated twice.

Total RNA from control and disease synovium (day 7) was isolated (RNeasy; Qiagen, Hilden, Germany) and reverse transcribed on paramagnetic oligo(dT)₂₅ Dynabeads (Dynal, Oslo, Norway). Due to low amounts of total RNA, the resulting cDNA was amplified by PCR as previously described (16).

The solid-phase library derived from control and diseased synovium was used to generate complex amplified cDNA probes. A cDNA-library from diseased synovium (day 7) was arrayed on high-density filters using a Q-bot picking and spotting robot device (Genetix, New Milton, U.K.) as previously described (17). On each filter, 9216 clones were arrayed in duplicate. Complex amplified cDNA probes were labeled with P-33 using the rediprime kit (Amersham Pharmacia Biotech, Freiburg, Germany) according to the manufacturer’s instructions. Identical high-density filter arrays were hybridized with complex, amplified cDNA probes (2 × 10⁶ cpm/ml) derived from control and disease tissue, respectively. cDNA clones representing differentially expressed genes were sequenced and compared with sequences in GenBank using the Basic Local Alignment Search Tool (BLAST) (Genetics Computer Group, Madison, WI) program.

RDA was performed essentially as described (18) omitting the third subtractive hybridization round. The second difference product was subcloned into the BamHI site of pBluescript II SK⁺ (Stratagene, La Jolla, CA). Plasmid DNA was prepared and analyzed by sequencing. Resulting sequences were compared with GenBank database using the BLAST program.

Nondecalcified rabbit knee joints (day 7) were fixed in 4% paraformaldehyde. Following an ascending ethanol series, the joints were embedded in methylmethacrylate (Leica HistoDur, Nussloch, Germany) and sections were cut as previously described (19). For in situ hybridization, the plastic was removed by immersion in 2-methoxyethyl acetate. Subsequent steps were performed as described (20) with minor modifications. The antisense and sense riboprobes were labeled with P-33 UTP and SDS was added to a final concentration of 0.5% to the hybridization buffer to facilitate the penetration of the labeled probe into the tissue. After hybridization, slides were dipped in NTB-3 emulsion, dried at room temperature and exposed for 10 days. Exposed slides were developed with D-19 developer and analyzed by microscopy. Experiments were repeated using sections from three different animals.

If not mentioned otherwise, 10 μg total RNA were denatured, separated on a 1% agarose gel, and transferred onto Hybond N⁺ membrane (Amersham Pharmacia Biotech). The filters were hybridized with random primed radioactively labeled cDNA specific for rabbit MMP-13 (21), rabbit SAA3 (22), human MMP-1 (23), and human SAA1 (24) as described (25).

A cDNA fragment representing the mature SAA3 protein with an artificial start codon followed by six amino-terminal histidine residues was amplified by PCR using the following primer combination: 5′-CAGGTCAACTCATATGCATCACCATCACCATCACCGTGAATGGTTAACATTCC-3′ and 5′-CAGGTCAACTCATATGCGTGAATGGTTAACATTCC-3′. The PCR product was digested with NdeI/XhoI and cloned into the bacterial expression vector pET-17b. The correct sequence of the expression construct was confirmed by DNA sequence analysis. Expression of the recombinant protein was induced in Escherichia coli (BL-21 pLys S) by the addition of isopropyl β-d-thiogalactoside (0.4 mM final concentration). rSAA3 protein was purified using a Ni-NTA resin. The eluted protein was further purified by reverse-phase HPLC-chromatography to remove trace amounts of bacterial endotoxin. The protein was lyophilized and resuspended in PBS at a concentration of 0.8 mg/ml. The endotoxin content of the protein was determined by the Limulus amebocyte lysate test as described previously (26) and was below 0.15 endotoxin units/ml. rSAA3 protein was analyzed on an 18% SDS-polyacrylamide gel. The gels were stained with Coomassie Blue R 250. Recombinant human A-SAA protein was purchased from PeproTech (London, U.K.).

Primary chondrocytes were prepared as previously described (26). Cells were cultured in DMEM supplemented with 5% FCS at 37°C in a 5% CO₂ atmosphere until they were 80% confluent. Expression of MMP in chondrocytes was induced either by 100 ng/ml SAA3 or 100 ng/ml human SAA. IL-1β (30 ng/ml) was used as a positive control for MMP induction. After induction for the indicated time, total RNA was prepared using the RNeasy kit (Qiagen) according to the manufacturer’s instructions. TNF-α and IL-1β were purchased from PeproTech.

Immunohistochemistry on paraffin-embedded tissue sections (nondiseased and RA synovium, and nondiseased and osteoarthritic (OA) cartilage) was performed by LifeSpan Biosciences (Seattle, WA) as described previously (20). The monoclonal anti A-SAA Ab (clone REU-82.2; Research Diagnostics, Flanders, NJ) was used as the primary Ab at dilutions of 1:10.

We identified SAA3 as the major difference product in our study analyzing the gene expression profile in AIA. Thus we decided to study its potential role in the disease process of induced arthritis as a model of the human disease. A prerequisite for the further functional characterization was to confirm the differential expression of SAA3 in the diseased joint by an independent method. Using RT-PCR (data not shown) and Northern blot analysis (Fig. 1), a weak signal for SAA3 expression could be detected in the synovium isolated from the control joint. In comparison, SAA3 mRNA expression was highly up-regulated in the diseased synovium.

Following the confirmation that SAA3 is up-regulated in the diseased joint, we analyzed the spatial expression pattern of SAA3 by in situ hybridization of whole rabbit knee joints. SAA3 mRNA was detected in cells infiltrating into the inflamed joint (Fig. 2, A and B). SAA3 expression could also be detected in areas where pannus formation starts (Fig. 2, A and B). Furthermore, strong expression of SAA3 was found in the meniscus and in articular chondrocytes (Fig. 2, C–E), which has not been reported before. No signal was obtained with the sense riboprobe (data not shown). In contrast to the Northern blot analysis, no SAA3 expression could be detected by in situ hybridization in the control joint (data not shown).

To confirm our results from the in situ hybridization experiment, we isolated primary rabbit articular chondrocytes (RACs) from a healthy animal. RACs were treated for 24 h with the indicated concentration of IL-1β, IL-1β plus TNF-α, or PMA. RNA was isolated and analyzed by Northern blot analysis, using a probe specific for rabbit SAA3 (Fig. 3, upper panel). In resting primary chondrocytes, SAA3 expression was below the detection limit. Upon stimulation for 24 h with either IL-1β alone or in combination with TNF-α, SAA3 expression was strongly induced. PMA did not induce expression of SAA3 in primary chondrocytes.

Previously, it was shown that SAA3 induces interstitial collagenase (MMP-1) in rabbit synovial fibroblasts in an autocrine or paracrine manner (12). Using in situ hybridization, we extended this observation and found SAA3 transcription in vivo also in rabbit chondrocytes. To directly test whether SAA3 protein can induce MMP expression in chondrocytes, we expressed recombinant rabbit SAA3 in E. coli and stimulated chondrocytes with the purified protein. The purified SAA3 protein had the expected molecular mass of 12 kDa and was over 98% homogeneous as determined by Coomassie blue staining of an 18% SDS-PAGE gel (data not shown). Purified SAA3 protein at a concentration of 100 ng/ml with an endotoxin concentration of less than 1.5 fg/ml was used for stimulation of chondrocytes. This concentration of endotoxin is significantly below MMP-inducing endotoxin concentrations in chondrocytes as previously shown (26).

RACs were treated with the indicated final concentration of rSAA3 protein for the indicated time and RNA was isolated and analyzed by Northern blot (Fig. 4,A). Using a rabbit-specific probe, MMP-13 mRNA expression was strongly increased after 4 h decreased at 8 h and peaked again at 24 h. As a positive control for MMP-13 expression, RACs were treated with IL-1β. The intensity of the MMP-13 transcript in cells treated with 100 ng/ml rSAA3 or 30 ng/ml IL-1β was comparable. We also examined human articular chondrocytes (HACs) for MMP induction using rSAA3 as an inducing agent. Similar to rabbit chondrocytes, transcription of MMP-1 was induced by 100 ng/ml rSAA3 in human primary chondrocytes 4 and 8 h after stimulation (Fig. 4 B), however, less pronounced when compared with 30 ng/ml IL-1β. This may be explained by species differences using rabbit SAA3 for the induction of human chondrocytes. We have not been successful in detecting human MMP-13 transcripts in HACs after stimulation with rSAA3. This could reflect the lower abundance of MMP-13 transcript in comparison to the MMP-1 RNA in human chondrocytes.

Because the human SAA3 is defined as a pseudogene (4) we examined human nondiseased synovium and synovium derived from RA patients as well as normal cartilage and OA cartilage for expression of other members of the SAA gene family. Using a mAb specific for acute A-SAA (detecting both SAA1 and SAA2) high levels of A-SAA protein were detected in the inflamed synovium of RA patients (Fig. 5,B). In addition, A-SAA protein was also detected in OA cartilage although at lower expression levels than in the RA synovium (Fig. 5,D). In nondiseased tissue, A-SAA expression could not be detected (Fig. 5, A and C).

Regulation of A-SAA in chondrocytes was examined directly by treating HACs with 30 ng/ml IL-1β for the indicated time points and analyzing A-SAA expression by Northern blot. SAA1 and SAA2 are highly homologous on the nucleotide level as well as on the amino acid level (Fig. 6). Therefore, the possibility exists that the probe used in the Northern blot analysis will not distinguish between SAA1 and SAA2 transcripts. A-SAA transcripts were detected after 8 h of treatment and were increasing over a 48-h period (Fig. 6). Using RT-PCR we could amplify a single band from HACs stimulated with IL-1β for 24 h (data not shown). Subsequently, the PCR product was subcloned and sequenced. The sequence analysis revealed that the PCR product was 100% identical with the previously published sequence of SAA1 (24). Under the PCR conditions we used, we have not been able to amplify a cDNA for SAA2 (27, 28).

Because the protein sequence of rabbit SAA3 and human A-SAA are also highly homologous, we hypothesized that stimulation of HACs with human rA-SAA might have a similar effect on MMP transcription as stimulation of HACs with rabbit SAA3. HACs were stimulated with human rA-SAA protein, total RNA was prepared and analyzed for MMP-1 (Fig. 7,A) and MMP-13 (Fig. 7,B) expression by Northern blot analysis. Normalized to the amount of RNA loaded on the gel, as shown by staining of 28S and 18S rRNA (Fig. 7,C), no difference in MMP-1 expression between controls and 4-h stimulated cells was observed but clearly, MMP-1 mRNA levels are increased at 8 h before they resume to baseline levels at 24 h (Fig. 7,A). Interestingly, the signal obtained for MMP-1 mRNA was severalfold higher at the 24-h time point when A-SAA protein was used at a concentration of 50 μg/ml (Fig. 7,A). The latter concentration is similar to the concentration of SAA1 found in the synovial fluid of RA patients (1). It has also been reported that this concentration triggers chemotaxis in PMNs or monocytes (29). This concentration of A-SAA protein also yielded detectable MMP-13 mRNA induction (Fig. 7,B). We compared the MMP-inducing effect of A-SAA to the level of MMP induction by IL-1β (Fig. 7 C). The induction of MMP-1 transcription, normalized to the amount of RNA loaded on the gel in IL-1β stimulated HACs is around 2-fold more than the amount of transcripts induced by A-SAA. A-SAA concentrations found in synovial fluid of RA patients are significantly higher (1) than the concentration used for stimulating HACs.

To identify new drug targets for a complex human disease such as RA, a better understanding of the molecular mechanisms underlying the onset of the disease and disease progression is required. Comparison of the gene expression profile in normal and diseased tissue offers a powerful approach to study the molecular mechanisms underlying RA. However, access to human material, in particular tissue biopsy samples from healthy donors, is limited. For these reasons, we decided to analyze gene expression patterns in a relevant animal model (rabbit AIA). We established a gene expression profile of AIA in rabbits using amplified complex cDNA probes derived from control and disease synovium for hybridization to an arrayed cDNA library representing the pattern of gene expression in the diseased situation (F.F., R.V., and G.B., unpublished results). In addition we used cDNA RDA, a method that was previously successfully used and was found to predominantly lack false positive hits (18). Of note, for the majority of the differentially expressed known genes such as MMP-1, MMP-3, MMP-13, CD44, cathepsin D, IL-1-converting enzyme, etc. (F.F., P.W., C.T., R.V., and G.B., unpublished results) a putative role in the context of RA or an inflammatory response has been suggested previously. This demonstrates the feasibility of our experimental approach to start with very low amounts of total RNA to identify genes that are differentially expressed in AIA.

As the major difference product in AIA we identified SAA3, which was confirmed by Northern blot analysis demonstrating that SAA3 transcription is augmented in the diseased rabbit synovium.

Using in situ hybridization on whole rabbit knee joints, we identified cell type(s) expressing SAA3 in vivo. In addition to infiltrating cells and the area where pannus formation starts, SAA3 mRNA expression was most abundant in the tip of the meniscus and in articular chondrocytes. Extrahepatic expression of SAA has been described previously (5, 6, 12, 14, 15, 30, 31, 32); however, expression of A-SAA in chondrocytes has not been described. This novel A-SAA mRNA expression pattern was confirmed by Northern blot analysis of isolated rabbit chondrocytes. In resting primary chondrocytes, SAA3 expression could not be detected. Upon stimulation with proinflammatory cytokines like IL-1β and TNF-α, SAA3 transcription is induced. In contrast to previous results that reported PMA-inducible SAA3 expression in synovial fibroblasts (31, 33), SAA3 could not be induced by the same stimulus in primary rabbit chondrocytes. Previously, SAA3 protein was shown to induce MMP-1 expression in an autocrine manner in isolated rabbit fibroblasts (12, 31). In extension of these studies, we showed that recombinant rabbit SAA3 protein induced transcription of MMP-13 in primary rabbit chondrocytes. We also provide evidence for induction of MMP-1 transcription by recombinant rabbit SAA3 in primary human chondrocytes (Fig. 4 B).

In humans, the SAA3 gene does not appear to be functional (4). We speculated that another member of the SAA gene family might compensate for the possible loss of a functional SAA3 gene in human. We provide evidence using RT-PCR and DNA sequence analysis that human chondrocytes stimulated with IL-1β express A-SAA mRNA (data not shown). Northern blot analysis showed that A-SAA mRNA was induced by the proinflammatory cytokine IL-1β in human primary chondrocytes. We further provide evidence by immunohistochemistry that A-SAA is highly expressed in RA synovium, confirming previous results showing expression of A-SAA mRNA in RA synovial tissue and cells (14, 15). In extension of these results we demonstrate A-SAA expression in OA cartilage. It is important to point out that rabbit SAA3 and human A-SAA are highly homologous at the amino acid level (Fig. 6). Moreover, in extension of previous results (12, 34) our study provides evidence that recombinant human A-SAA protein is able to induce transcription of MMP-1 and MMP-13 in human chondrocytes. Similarly, it has been demonstrated that A-SAA protein is able to induce expression of MMP-2 and MMP-3 in isolated human synovial fibroblasts (34). Although we show in this study up-regulation of MMPs only at the RNA level in our experimental in vitro system (rabbit and human primary chondrocytes), we assume that the level of MMP mRNA correlates with the amount of protein. This assumption is consistent with the generally accepted concept that MMPs are mainly regulated on the transcriptional level (35) as well as at the posttranslational level, including the activation of the latent enzyme (36) and inhibition of the activated enzyme through complex formation with tissue inhibitors of metalloproteinases or TIMPS (37). Our results clearly demonstrate that human A-SAA and rabbit SAA3 are capable of inducing transcription of MMPs in chondrocytes ex vivo.

At a concentration of 50 μg/ml, A-SAA is able to induce chemotaxis of human macrophages, which is abrogated by the addition of pertussis toxin, suggesting the interaction of SAA with a G protein-coupled receptor (29). Recently formyl peptide receptor-like 1 has been identified as the receptor mediating the chemotactic activity of A-SAA for human phagocytic cells (38). It has also been demonstrated that recombinant human A-SAA protein induces IL-1β synthesis in THP-1 cells at 30 μg/ml (39). In contrast, the A-SAA concentration required to induce MMP transcription ex vivo in chondrocytes is much lower (100 ng/ml). In this study we did not analyze the actual levels of SAA protein produced in our tissue culture model. However, there is data demonstrating that in the acute inflammation phase, A-SAA serum levels can reach concentrations of up to 2 mg/ml (40). Furthermore, the concentration of A-SAA protein in the synovial fluid of RA patients can reach up to 50 μg/ml (1). The concentrations found in vivo in the serum and in the synovial fluid of RA patients are similar to the concentrations that are able to induce chemotaxis and cytokine synthesis, and above concentrations of A-SAA sufficient to induce MMP transcription ex vivo. In addition to known proinflammatory cytokines such as IL-1β and TNF-α our data suggest that locally produced A-SAA may contribute to enhanced MMP synthesis in chondrocytes and synovial fibroblasts and could be of physiological relevance for disease progression. Thus, interfering with A-SAA induced expression of disease relevant MMPs such as MMP-1 and MMP-13 might open up new avenues for inhibiting degradation of cartilage in arthritis. It was recently reported that in the rabbit synovial fibroblast cell line HIG-82, IL-1 and IL-6 synergistically induce rabbit SAA2 expression (41). The primary structure of the human rA-SAA protein we used in our studies corresponds to SAA1α except the addition of a methionine at the NH₂ terminus, the substitution of aspartic acid for asparagine at position 60 and histidine for arginine at position 71 (Fig. 6). Although these substitutions have previously reported to occur in natural variants of SAA proteins (11, 42) this protein has also been classified as a hybrid form between SAA1 and SAA2. Thus, future studies will now focus on evaluating the relative importance of the different SAA family members compared with each other and to other proinflammatory cytokines for onset and progression of RA.

We thank Sebastian Meier-Ewert and Hans Lehrach for helping us with generating the high-density filter arrays of the cDNA library; Paul Ramage and Rene Hemmig for the purification of the rSAA3 protein; Mauro Zurini for performing the endotoxin test; Ernst Böhnlein and Uwe Junker for helpful discussion and reading of the manuscript; Andrew MacKenzie and Jörg Eder for helpful discussions; and Brian Richardson for his support and interest in this work.