Reduced incidence and severity of experimental autoimmune arthritis in mice expressing catalytically inactive A disintegrin and metalloproteinase 8 (ADAM8)


M. D. Zack, Pfizer Global Research and Development, 700 Chesterfield Parkway, Chesterfield, MO 63017, USA.


A disintegrin and metalloproteinase 8 (ADAM8), a catalytically active member of the ADAMs family of enzymes, is expressed primarily on immune cells and thus probably involved in inflammatory responses. ADAM8 is also produced by chondrocytes, and recombinant ADAM8 can induce cartilage catabolism. We therefore decided to test the role of ADAM8 in autoimmune inflammatory arthritis using transgenic mice expressing catalytically inactive ADAM8. Transgenic DBA/1J mice expressing an inactivating point mutation in the ADAM8 gene to change Glu330 to Gln330 (ADAM8EQ) were generated to evaluate the proteolytic function of ADAM8 in an lipopolysaccharide-synchronized collagen-induced arthritis (LPS-CIA) model of autoimmune arthritis. The systemic inflammatory reaction to LPS was also evaluated in these mice. Expression profiling of paw joints from wild-type mice revealed that ADAM8 mRNA levels increased at the onset of clinical arthritis and correlated well with cellular macrophage markers. When subjected to LPS-CIA, ADAM8EQ mice demonstrated decreased incidence and severity of clinical arthritis compared to wild-type mice. Histological examination of paw joints from ADAM8EQ mice confirmed marked attenuation of synovial inflammation, cartilage degradation and bone resorption when compared to wild-type mice. However, transgenic mice and wild-type mice responded similarly to LPS-induced systemic inflammation with regard to mortality, organ weights, neutrophil sequestration and serum cytokine/chemokine production. We conclude that ADAM8 proteolytic activity plays a key role in the development of experimental arthritis and may thus be an attractive target for the treatment of arthritic disorders while minimizing risk of immunocompromise.


Rheumatoid arthritis (RA) is a polyarthritic autoimmune disease that manifests as joint inflammation with thickening of the synovial lining, pannus formation and erosion of cartilage and bone, leading eventually to loss of function. Hand and foot joints are affected most commonly, but any diarthrodial joint can be involved [1,2]. Because of the autoimmune nature of the disease, anti-inflammatory drugs are prescribed commonly for management of swelling and pain. Anti-tumour necrosis factor (TNF)-α therapeutics have provided a tremendous breakthrough in the therapeutic arsenal for RA and other autoimmune diseases [3]. However, these therapies can have serious side effects, either mechanism-based or due to issues that are inherent in injectable protein therapeutics [4,5]. Thus, a need for additional disease modifying anti-rheumatic drugs (DMARDs), in particular for oral DMARDs, still exists.

ADAMs (A disintegrin and metalloproteinase) belong to the metzincin family of enzymes, which are dependent on zinc for catalytic function. Of 20 encoded human genes, approximately half of these proteins are predicted to contain the histidine triad motif (HEXXHXXXXXH) known to co-ordinate zinc that is necessary for proteolytic activity [6,7]. Unlike most matrix metalloproteinases (MMP) and A disintegrin and metalloproteinase with thrombospondin motif (ADAMTS) enzymes, all ADAM genes encode a transmembrane domain, which permits the proteolytic shedding of membrane-bound substrates such as cytokines, adhesion molecules and growth factors. The ADAMs are therefore referred to commonly as sheddases [8], but because the catalytic domains can be removed proteolytically from the cell membrane and still retain catalytic activity, and remnant domains such as disintegrin (DIS), cysteine-rich (CR), and epidermal growth factor (EGF)-like domains also possess independent cell adhesion functionality, ADAMs are now thought to be much more than sheddases [9].

ADAM8 is a proteinase that appears to be restricted to immune cells of myeloid lineage [10]. Recently, it was shown that neutrophils (PMN) isolated from the synovial fluid (SF) from patients with active RA express higher levels of ADAM8 than neutrophils isolated from peripheral blood and the concentration of soluble ADAM8 in the SF correlated directly with the degree of joint inflammation [11]. The presence of ADAM8 increased the ectodomain shedding of membrane-bound L-selectin, leading the authors to propose a critical role for ADAM8-mediated L-selectin shedding in the rheumatic process. Macrophages (MΦ) also express ADAM8 and are thought to play a central role in the pathogenesis of RA owing to their ability to produce various proinflammatory cytokines, chemokines and proteinases [10,12]. Mandelin and colleagues observed increased ADAM8 expression in the synovial-like membranes of loosened hip prostheses which was an inflammatory response characterized by the presence of MΦ-like cells as well as giant cells, but not neutrophils [13]. Recently, Ainola et al. have shown that ADAM8 mRNA and protein are abundant at the bone/cartilage junction in the rheumatoid pannus and that ADAM8 siRNA treatment can attenuate the maturation of osteoclasts in vitro[14]. The potential role for ADAM8 in disease processes that are dependent on PMN and MΦ function may become amplified by the fact that, unlike most other metalloproteinases, ADAM8 is not sensitive to tissue inhibitors of metalloproteinases (TIMP) [15].

Recently, we published a detailed analysis of the unique autoactivation of ADAM8 [16]. Two major findings from this study were that (i) the ADAM8 pro-domain is preprocessed before autoproteolytic removal occurs and that (ii) the C-terminal domains (DIS, CR and EGF) remain intact and are among the most stable part of the enzyme after the catalytic domain has been removed. It has been shown that C-terminal ADAM8 domains retain adhesive properties and promote cell attachment in the absence of the catalytic and pro-domain [17]. Thus, we hypothesize that ADAM8 has duplicity in function as a catalytically active proteinase and a cell adhesion molecule.

The objective of the current study was to investigate the role of ADAM8 catalytic activity in inflammation in vivo. To properly address this goal mice were generated with a single amino acid mutation, Glu330 to Gln330, known to abolish ADAM8 catalytic activity without disturbing the adhesive role of the DIS, CR and EGF-like domains [17]. Wild-type (wt) and ADAM8EQ mice were compared in an aggressive model of autoimmune arthritis, lipopolysaccharide–synchronized collagen-induced arthritis (LPS–CIA) in terms of incidence and severity. In order to establish whether ADAM8 is essential in acute systemic inflammation, transgenic mice were challenged with intraperitoneal LPS and analysed for differences in organ weight, inflammation as measured by histology and serum cytokine levels.

Material and methods

Laboratory reagents

All common laboratory chemicals were purchased from Sigma (St Louis, MO, USA). All culture media and supplements were purchased from Invitrogen (Carlsbad, CA, USA). TNF-α, interleukin (IL)-3 and macrophage colony-stimulating factor (M-CSF) were purchased from R&D Systems (Minneapolis, MN, USA). Multiplex mouse cytokine/chemokine assay kits were purchased from Millipore (St Charles, MO, USA) and performed using the provided multiplex assay protocols.

Generation of Adam8E330Q point mutant mouse

The ADAM8E330Q targeting vector was generated by polymerase chain reaction (PCR) using DBA/1lacJ mouse genomic DNA as template (Fig. 1). The 5′ homology fragment was created in two fragments, a 2·3 kb fragment including exons 3–10 and a 468 base pairs (bp) fragment containing exons 11–12. Site-directed mutagenesis was used to introduce the E330Q mutation into exon 11. A unique BamHI restriction site was added to each of the two 5′ fragments for ligation together. The amplified 3′ homology fragment was 6·3 kb and included exons 13–22. After sequence verification, each homology fragment was cloned into a backbone plasmid containing a loxP-flanked Pgk-neomycin cassette. Linearized ADAM8E330Q targeting vector was electroporated into DBA/1lacJ embryonic stem (ES) cells and neomycin-resistant colonies were grown and selected using protocols described previously [18]. ADAM8E330Q/+ targeted ES cells were identified by Southern blot analysis. ADAM8E330Q/+ ES cells were electroporated with a cytomegalovirus–P1 bacteriophage (CMV–Cre) plasmid to induce in vitro Cre-mediated removal of the neomycin cassette. Subsequent ADAM8E330Q/+ (-neo) targeted ES cells were microinjected into blastocyst-stage embryos isolated from C57BL/6J females (Jackson Laboratory; Bar Harbor, ME, USA). Male chimeras were identified and back-crossed to DBA/1lacJ females (Jackson Laboratory) to derive germline ADAM8E330Q/+ offspring. Heterozygous males and females were mated to generate homozygous ADAM8E330Q/E330Q mice. The Pfizer Institutional Animal Care and Use Committee reviewed and approved the animal use in these studies. The animal care and use programme is fully accredited by the Association for Assessment and Accreditation of Laboratory Animal Care, International.

Figure 1.

Schematic representation of generation of A disintegrin and metalloproteinase 8 (ADAM8) point mutant mouse. The ADAM8E330Q point mutation was created via homologous recombination in DBA/1lacJ embryonic stem (ES) cells. The loxP flanked Pgk-neomycin cassette, used for ES cell selection, was excised from the targeted ADAM8 allele by P1 bacteriophage (Cre)-mediated recombination using a transiently transfected cytomegalovirus (CMV)–Cre) plasmid.

Bone marrow macrophage (BMM) cultures

Bone marrow-derived MΦ were cultured as described previously [19]. Briefly, femurs from wt and transgenic mice were flushed with 5 ml cold Dulbecco's modified Eagle's medium (DMEM) using a 23-gauge needle. Bone marrow plugs were disrupted via repeat pipetting, centrifuged and resuspended in cold complete media (DMEM, 10% fetal bovine serum, 1% penicillin streptomycin glutamine and 0·5 µM β-mercaptoethanol). Non-adherent mononuclear precursor cells were generated by culturing in the presence of 1 nM IL-3 and 0·44 nM M-CSF. Cells were then differentiated into adherent phagocytes by 3-day culture in absence of IL-3 in complete media. Macrophage cultures were maintained in complete media with 0·80 nM M-CSF. For ADAM8 activity analysis, nearly confluent BMM cultures in 12-well plates (Corning, NY, USA) were stimulated with 100 ng/ml TNF-α in 500 µl fresh complete media for a period of 3 days at 37°C.

ADAM8 activity assay

Conditioned media from stimulated BMM cultures were concentrated twofold by centrifugation (Millipore, St Charles, MO, USA). Total protein content was analysed by Bradford assay (Biorad, Hercules, CA, USA). Conditioned media were then normalized to total protein prior to activity assay. ADAM8 activity assay was performed as described previously [16]. Briefly, peptide substrate (DABCYL-HGDQMAQKSK-FAM-NH2) was purchased from Biozyme Inc. (Apex, NC, USA) and a 3-mM stock was made by solubilizing in dimethylsulphoxide (DMSO). Conditioned media were tested in 384-well plate format (Corning, #3705) in a final reaction volume of 30 µl. Final assay concentrations were 50 mM Tris pH 7·4, 150 mM NaCl, 5 mM CaCl2 and 1 µM substrate. Plates were analysed for 30 min at 37°C using a SpectraMax plate reader (Molecular Devices, Sunnydale, CA, USA) with 495 nM excitation and 519 nM emission.

LPS-boost model of collagen-induced arthritis

Arthritis was induced in 8–12-week-old male DBA/1J mice (Harlan, Indianapolis, IN, USA) by injection of 100 µg bovine type II collagen (CII) (Western Institute for Biomedical Research, Salt Lake City, UT, USA) in complete Freund's adjuvant (Sigma) at the base of the tail, as described previously [20]. On day 28, mice were injected intraperitoneally with 10 µg LPS (Sigma) in 100 µl sterile saline. Following LPS injection, mice were given moist food and monitored daily for their overall health. Mice were monitored for arthritis on days 34, 42 and 46 after immunization. Severity of clinical arthritis was assessed using a score of 1–3 for each paw (resulting in a maximum score of 12 per mouse) as described previously [20]. Briefly, score 1 = any redness or swelling of digits of the paw; score 2 = gross swelling of the whole paw or deformity; and score 3 = joint ankylosis. Severity was expressed as the average severity score for each group. Incidence was expressed as the percentage of mice with a disease score ≥ 1. On day 46, mice were killed by asphyxiation and blood was collected by cardiac puncture and allowed to clot, after which serum was collected. Hind paws were collected in 10% formalin for histological analysis.

Enzyme-linked immunosorbent assay (ELISA) for serum anti-type II collagen antibodies

Costar (#3590) enzyme immunoassay/radioimmunoassay (EIA/RIA) plates were prepared by adding 100 µl of a 0·01 mg/ml solution of bovine type II collagen diluted in 0·15 M phosphate buffer pH 7·4 and incubating overnight at 4°C. Plates were then blocked with phosphate buffered saline (PBS, pH 7·4) with 1% bovine serum albumin for 2 h. After three plate washes with PBS/0·05% tween, 100 µl mouse serum [diluted 1 : 1500 in PBS/0·1% bovine serum albumin (BSA)] were added to each well. Plates were incubated for 2 h at room temperature and washed four times (PBS/0·05% Tween) before addition of detection antibody. Horseradish peroxidase-labelled anti-mouse antibody (Promega #W4021, Madison, WI, USA) was added at 1:7500 dilution and incubated in the dark for 1 h at room temperature. After six plate washes, plates were developed by the addition of tetramethylbenzidine (TMB) substrate and read on a Spectramax plate reader at 450 nM absorbance.

Histological staining and scoring

Organs were embedded in paraffin and cut into 5 µm sections and then stained with haemotoxylin and eosin (H&E) according to standard procedure for analysis. For immunohistochemical (IHC) analysis of liver and spleen, slides were deparaffinized and rehydrated through graded ethanols and pretreated with Borg Solution (Biocare Medical, Concord, CA, USA) for 3 min. Endogenous peroxidase activity was blocked with 3% H2O2 for 15 min. Slides were then washed with 50 mM Tris, 150 mM NaCl, 0·05% Tween, pH 7·4 (TBST). Slides were then incubated with 200 µl anti-CD3 (Accurate Chemical & Scientific Corp., Westbury, NJ, USA), anti-F4/80 (Serotec, Raleigh, NC, USA) or control immunoglobulin (Dako #X0903; Dako, Glostrup, Denmark) for 30 min at recommended dilutions. Secondary horseradish peroxidase-labelled antibodies were incubated for 30 min. Slides were washed with TBST before the addition of DAB+ (Dako) substrate and subsequent washing prior to H&E counterstain. Hind paws from CIA studies were sectioned sagitally through the centre of the joint, stained, and then scored for number of joints affected, inflammation, cartilage damage and bone damage on a scale of 0–4 as described previously [21,22]. A total combined joint score was generated for each animal by summing the number of joints affected, the inflammation score and the cartilage damage score for both the right and left hind paws. Bone damage was not included as a component of the total joint score, as in most cases this value was identical to the cartilage score. All scoring was performed by a trained observer who was blinded for clinical severity scores.

Microarray analysis

Mouse paw homogenates were prepared by snap-freezing in liquid nitrogen, powdered in an SPEX6750 freezer mill (Metuchen, NJ, USA) and resuspended in TriZol reagent (1–2 ml per 100 mg powder) by inversion for 1 h. Fluorescent labelling of amplified RNA (aRNA) was performed using a modified protocol with the Kreatech ULS aRNA Fluorescent Cy3/Cy5 Labeling Kit (Kreatech, Amsterdam, the Netherlands). Two µl labelling buffer and 4 µl dye were added to 5 µg aRNA in a total volume of 20 µl. Once combined, samples were incubated for 15 min at 85°C. After incubation, samples were returned to room temperature and spun down to collect condensate. All samples were labelled with each dye to allow for generation of fluoro-reverse duplicates. Labelled samples were purified using Zymo Research RNA Clean and Concentrate spin columns according to the manufacturer's protocol (Orange, CA, USA). Samples were eluted in 100 µl RNase/DNase free water. Labelled samples were quantitated on the Nanodrop ND-1000 spectrophotometer (Thermo Scientific, Wilmington, DE, USA). Seven hundred and fifty nanograms of each labelled aRNA from a comparison were added to a total volume of 230 µl water and 9 µl fragmentation buffer from Agilent Technologies (Santa Clara, CA, USA). Hybridization mixtures were then incubated for 30 min at 60°C. Hybridization mixtures were cooled to room temperature and spun down to collect condensate. Two hundred and forty µl 2X Hi-RPM hybridization buffer (Agilent Technologies) were added to each mixture. These probes were then hybridized to Agilent Technologies 4 × 44 K mouse whole genome microarrays for 18 h at 65°C in the Agilent Sure-Hyb hybridization chambers. Chambers were disassembled in a 0·6× sodium chloride/sodium citrate buffer (20× SSC; 3 M sodium chloride, 0·3 M sodium citrate, pH 7·0) with 0·005% Triton X-102 wash bath at room temperature and washed for 10 min in wash bath 1. Microarrays were transferred to a 0·01× SSC with 0·005% Triton X-102 wash for 5 min at 40°C. Microarrays were then dried with filtered compressed nitrogen and scanned on the Agilent Technologies DNA Microarray Scanner at 5 µm resolution. Data were extracted from the scanned image using Agilent Technologies Feature Extraction Software version 9·5. Extracted data were processed through in-house analysis tools to remove background, balance signal, calculate Cy3/Cy5 ratios (fractional balanced differential expressions) and incorporate fluoro-reverse relationships. Several molecular markers of cellular differentiation (CD), employed commonly to establish cellular phenotype, were chosen for characterization [23]. T cells were identified by the presence of CD3, CD4 or CD8 mRNA. Monocytes were identified by the presence of CD33 and CD14. MΦ were identified by F4/80, sialoadhesin, macrophage scavenger receptor-1 (Msr1) and integrin αM subunit (Itgam). Natural killer cell were identified by the presence of CD56 mRNA. B cells were identified by the presence of CD19, CD37 and CD79b transcripts. Mast cells were identified by chymase 1 and beta 2 tryptase [24]. PMNs were characterized by the presence of myeloperoxidase, elastase-2 and lactotransferrin transcripts [25].

Protocol for LPS challenge

Male mice were injected intraperitoneally with LPS at 100 µg/25 g or 300 µg/25 g in sterile saline. Blood was collected after 24 h by cardiac puncture prior to animal takedown by asphyxiation. Serum was collected via centrifugation and prepared for chemokine analysis. Organs collected for histological analysis were immediately placed in 10% neutral buffered formalin.

Statistical analysis

All statistical analysis was performed using LabStats Microsoft Excel add-in. This statistical package was generated by Pfizer Biostatistics and Reporting (Sandwich, Kent, UK) and Tessells Support Services plc (Abingdon, Oxon, UK).


Mutation of Glu330 abolishes ADAM8 activity

Mutation of a critical glutamic acid residue (Glu330) has been shown to inactivate the proteolytic activity of recombinant mouse ADAM8 in vitro without affecting adhesive properties of the DIS/CR/EGF domains [17]. We generated this same mutation in vivo using DBA/1 mice. Presence of the mutation was confirmed by PCR of tail biopsies (not shown). ADAM8EQ transgenic mice appeared phenotypically normal, with no differences in organ mass, breeding patterns or obvious behaviour. In order to confirm that the Glu330 mutation effectively abolished ADAM8 peptidase activity in vivo, we used a previously described ADAM8 substrate to investigate peptidolytic activity produced by TNF-α-stimulated BMM from wt and transgenic animals [26]. It has been demonstrated that this peptide is degraded most efficiently by ADAM8 compared to multiple members of the MMP and ADAM families of metalloenzymes [26]. In cells derived from wt mice, conditioned media from TNF-α stimulated BMM produced ADAM8 activity, as analysed in an ADAM8 peptide activity assay. As expected, BMM-conditioned media from ADAM8EQ mice were not able to degrade the ADAM8 substrate (data not shown).

Effect of ADAM8 catalytic inactivation on the development of CIA

Clinical arthritis (severity score ≥ 1) developed in 14 of 15 wt mice (93%) by day 34 (Fig. 2a), with an average clinical severity score of 5·5 (Fig. 2b). The disease did not progress any further up to day 46. In contrast, transgenic mice demonstrated a significantly lower incidence: by day 34, only seven of 15 ADAM8EQ mice were arthritic (47%), with an average clinical severity score of 2·5. By day 46, 53% (eight of 15) transgenic mice were arthritic with an average severity score of 2·1. Statistical comparison of ADAM8EQ mice to wt mice confirmed that these differences were significant (P < 0·05) at all time-points, both in terms of incidence and severity. Additional statistical analysis of severity scores limited to those mice demonstrating some level of disease revealed a trend (P = 0·075, one-sided t-test) in which ADAM8EQ mice had less severe disease than wt mice (Fig. 2c).

Figure 2.

A disintegrin and metalloproteinase 8 (ADAM8EQ) have reduced incidence and severity in lipopolysaccharide (LPS)-synchronized collagen-induced arthritis when compared to wild-type (wt) controls. (a) Clinical incidence scores are reported for days 34, 42 and 46 after immunization in wt mice (n = 14/15; black bars) and ADAM8EQ transgenic mice (n = 7/15; striped bars). Significance (*) was determined by one-way analysis of variance (anova, P < 0·05). (b) Clinical severity is reported at days 34, 42 and 46 in wt mice (black bars) and ADAM8EQ transgenic mice (striped bars). Significance (*) was determined by anova, P < 0·005. (c) Severity scores for arthritic mice only in wt (n = 14, solid bar) and transgenic mice (n = 7, striped bars). Significance was evaluated by one-way t-test, P = 0·075.

Production of anti-CII antibodies

All immunized mice, both wt and ADAM8EQ, generated similar amounts of circulating anti-collagen II antibody as analysed by ELISA on day 46 (data not shown).

Histological scoring of arthritis in hind paws

Sections from all groups were scored based upon number of joints affected, inflammation, bone resorption and cartilage degradation. Each hind paw was scored for inflammation and cartilage degradation on a scale of 0–4, with 4 being the most severe. The right and left hind paw scores were then combined to generate an inflammation and cartilage score for each animal. Inflammation and cartilage scores were then combined with number of joints affected to create a total joint score. A summary of histological scoring is presented in Table 1. All parameters scored were significantly less severe (P < 0·05) in ADAM8EQ mice when compared to their wt counterparts. A representative section from each group is shown in Fig. 3, illustrating clearly that the wt mice develop severe synovial inflammation and cartilage erosion (Fig. 3c), while a significant number of ADAM8EQ transgenic mice are protected (Fig. 3d). Additional histological analysis of joint scores from individual affected paws (Fig. 3e) revealed that arthritic paws from ADAM8EQ mice were significantly less severe than those of arthritic wt mice (P < 0·05, one-sided t-test, Fig. 4).

Table 1.  Histological evaluation of joint inflammation in LPS-synchronized CIA.
mean (s.e.m.)
mean (s.e.m.)
  1. Hind paws were prepared and scored as described in Materials and methods. Total joint score is the sum of the following three parameters: number of joints affected, inflammation and cartilage degradation. Data are presented as the mean scores for the given parameter and the associated P-value generated by one-way t-test analysis between groups. ADAM8EQ, A disintegrin and metalloproteinase 8; s.e.m., standard error of the mean.

Total joint score13·3 (2·8)4·9 (2·3)0·029
Joints affected4·8 (1·3)1·3 (0·7)0·034
Bone resorption4·9 (0·9)1·9 (0·9)0·026
Inflammation3·7 (0·7)1·6 (0·7)0·043
Cartilage degradation4·9 (0·9)2·0 (0·9)0·033
Figure 3.

Haemotoxylin and eosin staining of mouse hind paw joints in lipopolysaccharide-synchronized collagen-induced arthritis (LPS–CIA). A sagittal section through the middle proximal interphalangeal joint of representative mice at 40× magnification. (a) Wild-type (wt) control and (b) A disintegrin and metalloproteinase 8 (ADAM8EQ) control joints. (c) Representative wt arthritic; (d) representative ADAM8EQ mouse protected from CIA and (e) representative ADAM8EQ not protected in CIA. Inlays of (c), (d) and (e) are stained with toluidine blue to highlight the proteoglycan content in articular cartilage.

Figure 4.

Total joint scores for arthritic mouse paws. Average total joint score is shown in mice with clinical arthritis only is shown for wild-type mice (n = 16, solid bar) and transgenic mice (n = 7, striped bars). Error bars represent the standard error of the mean. Significance (*) was evaluated by one-way t-test, P = 0·015.

Characterization of ADAM8 expression in the course of LPS–CIA

The LPS-synchronized CIA model of autoimmune arthritis reaches maximal disease incidence and severity by day 32, 4 days after intraperitoneal injection of LPS. At day 40, the inflammatory response begins to resolve and paw swelling subsides. Thus, we chose to characterize this critical inflammatory period in terms of relative gene expression by analysing mRNA abundance in the hind paw from days 27 to 42. Whole paw homogenates were used to extract mRNA from wt DBA/1 mice in LPS-synchronized CIA at eight time-points (days 27, 28, 29, 32, 34, 36, 40 and 42). Animals (five mice/time-point) were killed from day 27 (1 day prior to LPS injection) to day 42 (14 days after LPS injection) and mRNA extracts were profiled using a whole mouse genome array chip. We focused on cell-associated transcripts described in the literature to understand more clearly the nature of the infiltrating cells seen in the paw of animals in the LPS–CIA model. It is likely that few, if any, transcripts are expressed exclusively in a single cell type under all conditions. Thus, we focused on groups of genes to serve as a molecular ‘fingerprint’ to investigate the relative abundance of immune cells. A panel of genes, including F4/80, sialoadhesin, macrophage scavenger receptor-1 (Msr1) and integrin αM subunit (Itgam), all rose markedly at day 29, immediately after the LPS boost, while sialoadhesin peaked between days 29 and 36 (Fig. 5a), a time-frame at which disease severity reaches a maximum. This consensus of genes indicates that the MΦ is a key player in the inflammatory process. ADAM8 expression was also increased during this time-frame and peaked at day 32, 4 days after LPS administration, and then returned to baseline level by day 42 (Fig. 5a). This correlation in expression suggests that macrophages could be a source of ADAM8 in this model. Gomez-Gaviro et al. have suggested that neutrophils may be a source of ADAM8 in RA [11], and therefore we studied the expression profile of a panel of neutrophil markers during the course of this animal model. It was found that markers for neutrophils (Mpo: myeloperoxidase; Ela2: elastase-2; and Ltf: lactotransferrin) are not up-regulated between days 27 (pre-LPS) and 42, suggesting that in the LPS–CIA model, neutrophils are probably not a significant source of ADAM8 production (Fig. 5b). Interestingly, PMN markers appear to be decreasing between days 27 and 28, prior to injection of LPS when there is no clinical evidence of disease.

Figure 5.

Relative expression of cellular differentiation markers in whole mouse paw homogenates in lipopolysaccharide-synchronized collagen-induced arthritis (LPS-CIA). (a) Relative cell transcripts associated with the macrophage phenotype, F4/80, sialoadhesin, macrophage scavenger receptor-1 (Msr1) and integrin αM subunit (Itgam) are plotted along with A disintegrin and metalloproteinase 8 (ADAM8) transcript levels; (b) relative cell transcripts associated with the neutrophil phenotype, myeloperoxidase (Mpo), elastase-2 (Ela2) and lactotransferrin (Ltf) are plotted along with ADAM8 transcript levels.

ADAM8EQ mice have a normal response to LPS challenge

In order to assess the role of ADAM8 proteinase activity during a systemic acute inflammatory response, ADAM8EQ mice and their wt counterparts were challenged with a high dose (300 µg/25 g) or low dose (100 µg/25 g) of LPS, administered intraperitoneally. No lethality was observed after LPS administration within any group. Comparison of organ weights after LPS challenge revealed modest increases in relative lung and spleen weight in wt mice as well as their transgenic counterparts (Table 2). Histological examination of spleen, lung and liver tissue 24 h after LPS challenge revealed no differences in neutrophil sequestration between wt or mutant mice after either dose of LPS (data not shown), while IHC revealed no differences in MΦ (F4/80) or T cell staining (CD3) in liver or spleen (data not shown). Analysis of serum cytokines/chemokines showed a significant increase in regulated upon activation, normal T cell expressed and secreted (RANTES), IL-10, monocyte chemoattractant protein-1 (MCP-1) and granulocyte-colony stimulating factor (G-CSF) 24 h after injection of LPS (Fig. 6). Again, there were no significant differences between transgenic and wt groups. These cytokines were selected from a panel of 22 cytokines measured in this study, which can be found in its entirety in Table S1.

Table 2.  Comparison of relative organ weights after LPS-induced systemic inflammation.
GroupnLPS (µg/25 g)Body weight (g)Liver (%BW)Lung (%BW)Spleen (%BW)
  1. Mice were given LPS, 100 or 300 mg per 25 g intraperitoneally as described in Materials and methods. After 24 h, organs were harvested and immediately weighed. Weights are presented as the mean relative organ weight plus or minus the standard error of the mean.

Wild-type5024·60 (1·07)4·732 (0·343)0·589 (0·041)0·244 (0·0115)
ADAM8EQ5025·04 (1·36)4·953 (0·390)0·545 (0·073)0·270 (0·037)
Wild-type510022·44 (3·25)5·090 (0·195)0·601 (0·056)0·409 (0·013)
ADAM8EQ510022·62 (0·64)5·399 (0·328)0·647 (0·058)0·384 (0·024)
Wild-type1030021·10 (1·08)5·310 (0·245)0·653 (0·025)0·3911 (0·029)
ADAM8EQ1030022·11 (2·18)5·233 (0·129)0·624 (0·025)0·396 (0·027)
Figure 6.

Serum chemokine analysis after systemic lipopolysaccharide (LPS) challenge. Selected chemokines known to potentiate MΦ were analysed by multiplex immunoassay according to the manufacturer's directions 24 h after intraperitoneal LPS injection, 100 or 300 µg/25 g body weight. Chemokine levels (pg/ml) in wild-type mice (black bars) and A disintegrin and metalloproteinase 8 (ADAM8EQ) mice (striped bars) are shown, with the y-axis minimum set to the assay detection limit.


ADAM8 is a proteinase whose catalytic function has not yet been elucidated in vivo despite several linkages to pathologies such as cancer, asthma and rheumatoid arthritis [11,12,27]. We have found recently that ADAM8 is produced by osteoarthritic chondrocytes, can degrade fibronectin to produce a fibronectin neoepitope seen in osteoarthritic cartilage and recombinant ADAM8 can induce cartilage catabolism ex vivo (manuscript submitted). Thus, ADAM8 may be implicated in osteoarthritis. Beyond chondrocytes, the expression of ADAM8 is restricted largely to cells of the immune system, thus making ADAM8 an intriguing target for the treatment of inflammatory disorders. ADAM8 null mice have been generated with no obvious abnormalities, but the potential physiological roles of C-terminal domains have made data interpretation difficult [28].

In the current work, we describe a transgenic mouse in which a critical catalytic site glutamic acid (Glu330) is mutated to a glutamine, a mutation that has been demonstrated to abolish the catalytic activity of ADAM8 without affecting adhesive functions [17]. We have used this genetically modified mouse to investigate the catalytic function of ADAM8 in autoimmune arthritis and acute inflammation.

Genetic manipulation of the ADAM8 gene in DBA/1 mice, rendering the protein catalytically inactive, resulted in a significantly attenuated response in the LPS-synchronized CIA model. After 46 days, ADAM8EQ mice displayed 53% disease incidence compared to 93% of wt mice, while the disease in the transgenic mice that developed arthritis was significantly milder, as observed histologically. This demonstrates that ADAM8 catalytic activity contributes to the inflammatory process within the joint space. However, because this mutation is intended to mimic complete abolition of ADAM8 activity, one would predict an ‘all or none’ response, and it is therefore surprising that half the immunized transgenic mice developed arthritis. One potential explanation is that a developmental redundancy, which is always a concern using genetically modified mice, would somehow circumvent ADAM8-regulated inflammatory pathways. A second and more likely explanation for the lack of complete protection in CIA is that ADAM8 is simply one component in the complex orchestration of an inflammatory response.

In an effort to address the nature of the ADAM8 role in LPS-synchronized CIA, we utilized tissue expression profiling to analyse whole paw homogenates in order to track relative changes in gene expression after LPS boost. Similar to traditional CIA, this model is characterized by synovial inflammation, cartilage degradation and bone resorption [29,30]. However, intraperitoneal injection of LPS 28 days after CII immunization produces a more robust and synchronized inflammatory response with an incidence greater than 90% in wt DBA/1 mice [31,32]. We found that by day 32, 4 days after the LPS boost, there is a rise in MΦ-associated mRNA levels (F4/80, Msr1, sialoadhesion and Itgam) in the inflamed joint that is concomitant with an increase in ADAM8 mRNA levels. These levels agree well with previous in-house and published clinical disease measurements showing maximal clinical disease at day 32, 4 days after LPS injection [32]. Because of the robust and consistent nature of this aggressive autoimmune model, it is likely that the MΦ is the cellular source of ADAM8 in arthritic paws. Furthermore, histological examination revealed an absence of inflammatory cells in immunized transgenic mice, indicating that ADAM8 catalytic activity is agonistic to the inflammatory process by facilitating the recruitment of inflammatory cells to the joint space. Tissue expression profiling experiments suggest that the cellular infiltrate is likely to be made up of MΦ or MΦ-like cells. Thus, ADAM8 catalytic activity may regulate MΦ function in the joint space, although the precise mechanism remains unclear. Recently, we have shown that ADAM8 degrades fibronectin, an important component of the extracellular matrix (ECM) that permits cell spreading and attachment (Arthritis & Rheumatism accepted manuscript). Hence, it is plausible that ADAM8 is a part of the MΦ proteolytic machinery necessary to navigate through the ECM to the site of inflammation, as we have seen no effects on proinflammatory cytokine production due to ADAM8 inactivation. Lastly, ADAM8 has been demonstrated to promote osteoclast formation, which is involved actively in bone resorption which may occur via interactions with α9β1 integrins [14,33,34]. However, Choi et al. have shown this to be mediated through the non-catalytic disintegrin domain, which should not be altered in the ADAM8EQ mouse [33]. Taking all this into consideration, future experiments will focus on identifying the mechanism by which ADAM8 facilitates the influx of inflammatory cells into the joint space.

To examine the role of ADAM8 catalytic activity in the systemic inflammatory response, we subjected transgenic mice to two doses of LPS challenge and measured organ weights, analysed serum cytokines and evaluated tissues by histology after injection. There was no mortality at either dose of LPS in either wt or transgenic animals. This is noteworthy because other ADAMs proteases are known to shed the p55 TNF receptor (p55TNFR), and ADAM8 has been reported to cleave a peptide based on the p55TNF receptors in vitro[35]. Xanthoulea et al. found that p55TNFR shedding is essential for survival in LPS-induced systemic inflammation [36]. We used an identical dose of LPS in our studies, thus our data suggest that ADAM8 is not a principal sheddase of the p55TNFR and is not indispensable for survival in acute sepsis. Both doses of LPS produced a significant increase in spleen weight as a percentage of body weight in wt and transgenic animals. While this suggested that both doses of LPS were able to induce an inflammatory response, there was no difference between wt and transgenic mice in terms of spleen weight. Histological examination of lung, liver and spleen revealed no overt differences between wt and ADAM8EQ mice. Both wt and ADAM8EQ transgenics demonstrated similar PMN sequestration in these tissues after LPS challenge, suggesting that ADAM8 catalytic activity is not critical for PMN recruitment to the lung, liver or spleen. Lastly, we examined G-CSF, IL-10, MCP-1 and CCL-5 via immunoassay in response to LPS challenge to investigate any abnormalities in the MΦ signalling cascade. While LPS caused a significant elevation in analytes tested after 24 h, there were no differences found between wt and ADAM8EQ mice. Interestingly, the higher dose of LPS produced an inflammatory response that was much lower than animals treated with a lower dose of LPS after 24 h. This facet of the data is hard to reconcile, but may be explained by a more rapid clearance of inflammatory cytokines/chemokines in response to a higher dose of LPS. Due to the similarities between ADAMEQ and wt animals, we conclude that ADAM8 inactivation does not sacrifice immunocompetence in an LPS-driven model of systemic inflammation.

TNF-α and IL-1β inhibition have both demonstrated therapeutic efficacy in this model of arthritis [32]. In addition, multiple groups have proven p38 kinase inhibitors to be therapeutically effective in LPS–CIA and traditional CIA models, likely mediated through modulation of TNF-α production [29,37]. As stated earlier, our transgenic mice only allow ADAM8 catalytic activity to be tested in a prophylactic mode with significant decrease in disease incidence and progression, but not complete protection. Thus, it remains to be established whether or not ADAM8 inhibition using selective pharmacological inhibitors will ameliorate the arthritic response in CIA. However, because ADAM8 inactivation does not alter the chemokine/cytokine response to inflammation and does not demonstrate immunocompromise in response to LPS challenge, ADAM8 inhibition may represent a novel pathway to modulate inflammation, via suppressing the ability of MΦ to extravasate to areas of inflammation, with a desirable safety profile.

In conclusion, we have explored the function of ADAM8 catalytic activity in vivo by generating a point mutation at a critical glutamic acid within the ADAM8 gene (Glu330Gln) known to abolish proteolytic activity. Transgenic mice display no overt abnormalities and have a normal breeding pattern. We demonstrated for the first time, in vivo, that ADAM8 catalytic activity is proinflammatory in an experimental arthritis model, LPS-synchronized CIA. We observed an obvious decrease in incidence and severity of disease both morphometrically and histologically. In addition to histological observations, we have utilized whole tissue expression profiling data to demonstrate that the relative increase in ADAM8 mRNA correlates well with mRNA for cellular MΦ markers and clinical severity. In addition, we subjected these mice to acute endotoxin challenge and observed no abnormalities when compared to wt controls with regard to organomegaly, neutrophil sequestration and chemokine/cytokine production. Thus, we propose that inhibition of ADAM8 proteolytic activity may provide a safe option for treatment of inflammatory disorders such as RA, where MΦ are central to disease progression.


The authors would like to thank Sheri Cox, Chris Bollinger, Fernando Happa and Denise Lay for their contributions to histology preparation and slide delivery.


All authors own stock or stock options in Pfizer.