Figure 1a outlines the sequence of events of the air pouch experiments. Air pouches were raised on the backs of 6- to 8-week-old female BALB/c mice (Taconic, Tarrytown, NY, USA) by subcutaneous injection of 3 ml filtered air 17. Pouches were re-inflated on day 3 with an additional 2 ml filtered air. MSU crystals were prepared in accordance with the method proposed by McCarty and Faires 8 and were determined to be free from endotoxin using the Gelclot LAL reagent (Charles River Labs, Wilmington, MA, USA). Aliquots from the same batch were used for all experiments. A suspension of 2 mg MSU crystals in 1 ml sterile endotoxin-free phosphate-buffered saline (PBS) was injected into the pouch on day 6. To verify the time points of peak and natural resolution of inflammation in the pouch, a 50-hour time course experiment was performed during which pouch exudate leukocyte counts were determined at several time points after injection of MSU crystals. In agreement with our previous findings 9, the leukocyte count rose 56-fold from 0 to 9 hours and then subsided, returning close to baseline by 50 hours (Figure 1b). Negative control pouches (n = 5) were injected with 1 ml PBS and harvested at 9 hours. Their mean leukocyte count was similar to that of the pouches at t = 0 hours (0.32 ± 0.18 × 10⁶ cells/pouch at 9 hours versus 0.18 ± 0.09 at 0 hours; p = 0.31, one-tailed t-test). All animal experiments followed internationally recognized guidelines and were approved by the Institutional Animal Care and Research Committee of the Philadelphia VA Medical Center.

Figure 2 shows key steps in the membrane dissection. After killing the animals by carbon dioxide asphyxiation, the apex of the pouch membrane was exposed with a small skin incision (panel b). The membrane was then punctured with a scalpel (panel c). Typically, little free exudate accumulates within the pouch. Leukocytes were therefore lavaged out of the pouch lumen with 2 ml PBS (panel d) and the leukocyte count in the resulting lavage fluid determined with a hemocytometer 10. The pouch membrane was then separated meticulously from adjacent subcutaneous and paraspinal tissues by blunt dissection (panels e to j). Finally, the membrane was grasped with forceps, elevated and cut at the base (panels k and l). Great care was taken to avoid the paraspinal and nuchal tissues, to which the base of the membrane is typically attached. Using a rotatory homogenizer (Omni International, Warrenton, VA, USA), the isolated membranes were homogenized in Trizol medium (Invitrogen, Carlsbad, CA, USA), flash frozen in liquid nitrogen and stored at -70°C until further use.

RNA was extracted over RNeasy spin columns (Qiagen, Valencia, CA, USA) and tested for integrity and quantity on an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA, USA). For unknown reasons, significant RNA degradation ensued when MSU-stimulated membranes, but not control membranes, were frozen and thawed before homogenization (data not shown). All membranes were therefore homogenized in Trizol medium immediately after dissection. Typically, 70 to 90 μg RNA was obtained per pouch.

For Affymetrix microarray analysis, RNA was processed and fluorescently labeled according to standard Affymetrix protocols 11. To control for differences in dissection, RNA from three control or three MSU-stimulated pouch membranes was pooled, processed, labeled and then hybridized to Affymetrix Mo430_2 oligonucleotide microarrays (Affymetrix Inc., Santa Clara, CA). Microarray signals were scanned and analyzed using Affymetrix GCOS software and then imported into the software program GeneSpring version 7.2 (Silicon Genetics, Redwood City, CA, USA) for visualization and filtering. RNA aliquots were also reverse transcribed into cDNA according to standard protocols and analyzed further by qPCR using the TaqMan system and ABI Prism 7000 or 7900 HT sequence detectors (Applied Biosystems, Foster City, CA, USA). Commercially available primer-probe sets (Applied Biosystems) were used.

IL-6 protein concentration in the pouch exudate was determined by enzyme-linked immunosorbent assay (eBioscience, San Diego, CA, USA), after removal of cells and debris by centrifugation.

Mouse peritoneal macrophages were harvested by peritoneal lavage 4 days after intraperitoneal injection of 2 ml aged 3% Brewer's thyoglycollate (Invitrogen Corporation, Carlsbad, CA, USA). Macrophages were allowed to rest in 3 ml tissue culture wells for 1 hour at 37°C (1 × 10⁶ cells/well). After removal of nonadherent cells, cells were grown in medium with or without MSU crystals for the time periods indicated in the figure legends. Previously frozen cells were used for the dose response experiment. Macrophage cultures were more than 95% pure, as verified by flow cytometry for CD11b, major histocompatibility complex class II, and F4/80.

Tissues were obtained 9 hours after injecting MSU crystals into the pouch, fixed in formalin for 24 to 48 hours, and embedded in paraffin blocks. Blocks were cut into 5 μm thin sections on a rotatory microtome. Immunoperoxidase staining for IL-6 was performed using a semi-automated immunostaining system (DAKO, Carpinteria, CA, USA) and commercially available polyclonal goat anti-mouse IL-6 IgG (Santa Cruz Biotechnology Inc., Santa Cruz, CA, USA) at 1:100 dilution. Nonspecific goat IgG was used as negative control.

Crucial steps in the dissection procedure are shown in Figure 2 and are also described in the Materials and methods section (see above). Dissected pouch membranes had a gelatinous but also somewhat fibrous consistency and usually weighed 70 to 110 mg. Membranes from MSU-stimulated pouches tended to be firmer and to rupture somewhat less easily during the dissection, probably because of a mild increase in thickness from inflammation 12. Figure 3a illustrates the plane of dissection between the membrane and the overlying subcutaneous tissue. According to our observations, the air pouch membrane originates from longitudinal soft tissue ridges that overlie the paraspinal musculature and from a cape-like, thicker membrane in the nuchal area. Because pieces of these tissues might contaminate the membrane during the dissection and confound a gene expression analysis, we evaluated them histologically (Figure 3b). The nuchal structure was identified as adipose tissue (Figure 3b, left image) and thus originates from the nuchal fat pad. The paraspinal ridge tissue (shown macroscopically in Fig. 2, panel k) turned out to be rich in blood vessels, striated muscle and fascia (Figure 3b, centre) and thus probably was contiguous with the paraspinal muscles. To test the histologic purity of the dissected membranes, haematoxylin and eosin stains were prepared from several membranes. Adipocytes, skeletal muscle and fascia were not observed, confirming that the membranes had been dissected relatively free from surrounding tissue (Figure 3b, right).

As expected, levels of mRNAs encoding resistin (a marker for adipocytes) and dystrophin (skeletal muscle) were by far highest in RNA extracted from the soft tissues overlying the membranes (Figure 3c). In contrast, mRNAs of several mRNAs identified by the microarray analysis of dissected membranes as inducible by MSU crystals (see below) were highest in membrane RNA. Consistent with the observations that IL-6 is also expressed in striated muscle (for instance, Figure 4b) and PUMA-g (protein upregulated on macrophages activated with interferon-γ) in adipocytes 13, the relative soft tissue fractions of these two mRNAs were larger than those of TREM (triggering receptor expressed on myeloid cells)-1 and TREM-2, both of which are predominantly expressed on inflammatory cells.

Relative quantitative PCR (qPCR) analysis for resistin and dystrophin revealed that, despite the histologic absence of fat and skeletal muscle from the dissected membranes, individual noninflamed (control) membranes varied in the amounts of mRNAs encoding these markers. This suggested the presence of small remnants of fat and muscle on the membranes that persisted despite the histologically clean dissection. To correct for this heterogeneity and other potential differences resulting from the dissection, RNA aliquots from three control or three MSU-stimulated membranes (obtained 9 hours after injection of MSU crystals in PBS or PBS alone) were pooled. An aliquot from each pool was then processed according to standard Affymetrix protocols, and the resulting labelled cRNAs hybridized to separate Affymetrix Mo430_2 oligonucleotide microarrays. Of the 45,101 probe sets contained on the microarrays, 21,009 and 19,941 were detected (annotated 'p' in the raw Affymetrix data) in the pooled RNAs from control and MSU membranes, respectively. A total of 5,988 were differentially regulated (Affymetrix flag) in response to MSU crystals. These were filtered on the Affymetrix change P values below 2 × 10^-5 for upregulated and 1 for the downregulated probe sets.

Of these, the 12 most highly over-expressed (linear ratio MSU/control >17) targets are listed in Table 1. Eight (75%) were genes with known proinflammatory functions: histidine decarboxylase (Hdc, the enzyme that catalyzes the conversion of histidine to histamine); the surface receptors PUMA-g (also known as GPR109 or HM74) 14 and TREM-1 15, which are induced on activated macrophages and neutrophils; immunoresponsive gene (Irg)1, a protein that is rapidly induced during monocyte activation with endotoxin 16; prokineticin (PROK)-2, which is a small polypeptide that is involved in macrophage activation, hyperesthesia and other processes 14; IL-6; colony-stimulating factor (CSF)3 (also known as granulocyte colony-stimulating factor); and the adhesion molecule P-selectin. Three genes (25%) encoded less well characterized factors but had potential functions in inflammation. One gene was uncharacterized. Of note, all of the over-expressed genes with proven proinflammatory functions related to innate immunity. Of the 12 most downregulated genes, three (25%) were related to connective tissue constituents, one (granzyme D) to innate immunity, and the others to various processes. The raw microarray data have been deposited in the Gene Express Omnibus microarray repository and are publically available under accession number GSE11498.

IL-6 is known to be raised in gout 17, but it had not been studied in the MSU-stimulated murine air pouch. qPCR validation of its expression in the 50-hour time course experiment summarized in Figure 1 was therefore chosen to establish proof of principle of the microrray approach. Confirming the upregulation at 9 hours that had been detected with the microarrays (Table 1), IL-6 mRNA was 12.9-fold higher at this time point than at t = 0 hours (Figure 4a). However, peak induction occurred much earlier (1 hour) and was much higher (108-fold). In agreement with the observation that the pouch leukocyte count had returned close to its original level at the end of the 50-hour time course (Figure 1b), IL-6 mRNA also returned to near baseline by 50 hours. The mRNAs encoding IL-1β, tumour necrosis factor (TNF)-α, and the immediate early transcription factor early growth response (Egr)-1 18 followed similar kinetics (Figure 4a). These findings confirmed induction of IL-6 mRNA in the membrane and identified the 1-hour time point as the peak of a rapid, early surge in proinflammatory cytokine and immediate early gene transcription.

In agreement with the induction of IL-6 mRNA in the membrane, analysis by enzyme-linked immunosorbent assay revealed that IL-6 protein levels at 9 hours were 8.7-fold higher in exudates from MSU-stimulated pouches than in exudates from control pouches (Figure 4b). The spatial expression of IL-6 in the inflamed air pouch was then determined immunohistochemically, using sections containing the pouch membrane adhering to the overlying skin. Consistent with previous reports of IL-6 expression in striated muscle 19, specific staining was seen in muscle fibres of the lamina muscularis of the subcutaneous tissue (Figure 4b, centre; marked 'm'), whereas the signal in hair follicles was not specific (marked 'f', arrows). In the inflamed membrane, specific IL-6 staining was seen in a multitude of cells, including mononuclear and polymorphonuclear cells and fibroblasts (Figure 4b, right).

qPCR demonstrated dramatic induction of these mRNAs throughout the air pouch time course (Figure 4c). Confirming the microarray results, all of these mRNAs were elevated at 9 hours. Interestingly, the extent and kinetics of their induction differed; whereas maximal induction of Irg1 (177-fold), PROK-2 (136-fold), TREM-1 (39-fold) and Hdc (120-fold) occurred at 4 hours, PUMA-g mRNA was upregulated 120-fold and peaked at 9 hours.

To validate the results of the microarray analysis further, we determined kinetics of two mRNAs that were downregulated by MSU crystals, according to the array analysis (Figure 4d). TREM-2 is a homologue of TREM-1 that is downregulated during macrophage activation 20, and the microarray analysis had revealed an 81% decrease of its mRNA 9 hours after MSU crystal injection. Indeed, the time course demonstrated a 79% decrease in TREM-2 mRNA, with a nadir between 9 and 18 hours and subsequent recovery to near baseline level at 50 hours. In contrast, mRNA encoding granzyme D mRNA, which was 96% downregulated at 9 hours according to the microarrays, decreased steadily and reached 6% of its original level at 50 hours. As negative controls, we measured expression of two mRNAs that were not among the differentially regulated genes on the arrays (Figure 4d). First, leukaemia/lymphoma-related factor (LRF; also known as FBI-1 and OCZF), a transcription factor important in cellular transformation 21, osteoclastogenesis 22 and regulation of the HIV-1 promoter 2324. Levels of its mRNA demonstrated a statistically insignificant twofold increase at 4 hours. Second, mRNA encoding Nab2, a co-regulator of Egr-1 25, exhibited only minor fluctuations. To exclude their origin from remnant adipose or muscle tissue, qPCR results for all of the above mRNAs at 9 hours were also correlated with resistin and dystrophin mRNA expression, but no significant associations were found.

The induction of TREM-1 in macrophages by MSU crystals has been reported 26. To test whether macrophages were also potential sources of Irg1, PROK-2, Hdc and PUMA-g, kinetics of these mRNAs were determined during 18 hours after the addition of MSU crystals to cultured mouse peritoneal macrophages (Figure 5). As in the tissue, MSU crystals induced sharp, early peaks in TNF-α, IL-6, IL-1β, and Egr-1 mRNAs (Figure 5a). MSU crystals also upregulated Irg1, PROK-2, Hdc, and PUMA-gmRNAs (Figure 5b). However, with the exception of PUMA-g, their levels peaked somewhat later than in the membranes. As illustrated in Figure 5c, these mRNAs were, to various extents, induced more highly in the membranes than in the macrophages. As in the membranes, MSU crystals caused a decrease in TREM-2 mRNA in the macrophages and no significant changes in Nab2 and LRF mRNAs (data not shown).

The induction of IL-1β, IL-6, TREM-1, PUMA-g, Hdc, PROK-2 and Irg1 by MSU crystals was also quantified in a dose response experiment (Figure 6). IL-1β and IL-6 mRNAs were induced maximally at a crystal concentration of 375 μg/ml medium, as were PUMA-g, Hdc and PROK-2 mRNAs. Induction of TREM-1 mRNA peaked at a slightly lower, and that of Irg1 mRNA at the lowest crystal concentration.

We validated the induction of six of the most highly over-expressed genes by qPCR. How might these function in the pathogenesis of crystal-induced inflammation? PUMA-g is a G-protein-coupled transmembrane receptor that was initially identified in a microarray screen of macrophages activated with TNF-α and interferon-γ 29. In addition to activated macrophages, it is also expressed on neutrophilic granulocytes and adipocytes. On the latter, it is a major receptor for the cholesterol-lowering agent nicotinic acid 13. Binding of this ligand to PUMA-g is also responsible for the cutaneous flushing response that is frequently observed in patients taking nicotinic acid 30, and it has been proposed that cutaneously expressed PUMA-g leads to the release of prostaglandins D₂ and E₂, which in turn cause local vasodilatation 30. These results suggest that in MSU crystal inflammation PUMA-g transmits signals in macrophages, and perhaps also granulocytes, which potentiate release of prostaglandins and thus enhance the inflammatory process. Further studies with cultured cells or PUMA-g^-/- mice will now be important in elucidating the function of PUMA-g in crystal-induced inflammation.

TREM-1, too, is induced on the surface of activated macrophages and neutrophils, and has been shown to play important roles in the systemic manifestations of sepsis 26. Interestingly, Murakami and coworkers 28 showed that TREM-1 mRNA in the pouch exudate peaked before the maximum accumulation of leukocytes, thus indicating activation of TREM-1 transcription early on. It has been proposed that one function of TREM-1 is to amplify Toll-like receptor (TLR) signaling 26. Considering that TREM-1 mRNA reached maximum levels after the initial cytokine mRNA surge (Figure 4a) and that recognition of MSU crystals via TLR signaling is believed to be among the earliest events in crystal inflammation 31, one may postulate that the role of TREM-1 in crystal inflammation also relates to potentiation of signals transmitted by TLRs.

Irg1 was isolated originally from a cDNA library made from endotoxin-stimulated macrophages 16, and was later found to be expressed in macrophages infected with mycobacteria 3233. Its induction by MSU crystals in peritoneal macrophages (Figures 5 and 6) suggests that it may play general roles in macrophage activation.

Prokineticins are evolutionary highly conserved small secreted polypeptides that play roles in various tissues, including the central nervous system, the gastrointestinal tract, the haematopoietic system and the vasculature 143435. PROK-2 was recently shown to augment macrophage chemotaxis and proinflammatory cytokine synthesis 36. Interestingly, it also causes hyperesthesia, including by local injection into rodent paws, and mice lacking the PROK receptor PKR1 have diminished pain responses 37. It is therefore intriguing to speculate that PROK-2 contributes to the symptoms of MSU crystal inflammation by enhancing both inflammation and pain in affected joints.

The enzyme Hdc converts histidine to histamine. Histamine is elevated in the joint fluid of patients with gout 38 and in the MSU crystal stimulated rat air pouch, where it is believed to be the result of mast cell degranulation 12. Our results, however, suggest that histamine may also arise from increased Hdc levels during MSU crystal inflammation. Consistent with our observation that MSU crystals induced Hdc mRNA synthesis in macrophage culture, with reports of Hdc mRNA expression in neutrophils 39, and with the well documented presence of mast cells in the air pouch membrane 12, all of these three cell types are probably highly inducible sources of Hdc in the pouch membrane. Thus, Hdc and histamine may contribute significantly to the evolution of MSU crystal inflammation in this model, but also in gouty synovitis in humans.

TREM-2 is a cell surface receptor similar to TREM-1. It promotes differentiation of macrophages into osteoclasts 40 but inhibits macrophage activation and is downregulated during this process 20. The decrease in TREM-2 mRNA in both membranes and cultured macrophages suggests that macrophage activation by MSU crystals leads to an increase in TREM-1 but to a decrease in TREM-2 mRNA levels. Granzyme D has been detected in murine cytotoxic T cells 41 and endometrial natural killer (NK) cells 42. The roles of these cell types in urate crystal inflammation have not been studied. However, it is known that uric acid can activate CD8⁺ T cells under some circumstances 43, and in preliminary studies we have detected mRNA encoding the NK cell lectin-like receptor A2 in the air pouch membrane (Mayer CT, Schumacher HR, Pessler F, unpublished data). Thus, the persistent downregulation of granzyme D throughout the time course suggests that MSU crystals may modulate cytotoxic T cell or NK cell activity in the membrane. Alternatively, one must consider that dilution of the membrane RNA pool by mRNAs contained in immigrating inflammatory cells or by otherwise highly upregulated mRNAs might also cause – at least in part – artifactual decreases in mRNAs expressed by resident membrane cells.

The results of the present study also shed new light on the kinetics of transcriptional regulation in the air pouch membrane. We observed an early, steep rise in proinflammatory cytokine mRNA synthesis that essentially paralleled the synthesis of mRNA encoding the immediate early transcription factor Egr-1. Similarly rapid kinetics were observed in cultured macrophages (Figure 5). Indeed, inclusion of the 30-minute time point revealed that increasing levels of mRNAs encoding Egr-1 and all three cytokines could be detected this early. It is currently believed that TLR-mediated activation of the Nalp3 containing inflammasome, leading to proteolytic cleavage of pre-existing pro-IL-1β to IL-1β, is among the first steps in MSU crystal inflammation 3144. Our findings suggest that activation of the proinflammatory cytokine genes encoding IL-1β, IL-6 and TNF-α, as well as Egr-1 and probably other immediate early transcription factors, is among the next, early events in reprogramming the transcriptional machinery in MSU crystal inflammation, both in an intact tissue and in cultured cells. TREM-1, Irg1, PROK-2, Hdc and PUMA-g mRNA levels all peaked after this dramatic burst in early cytokine transcription. These factors may therefore play roles in perpetuating or amplifying this early inflammation. We had performed the microarray analysis 9 hours after MSU crystal injection, the peak of leukocyte accumulation in the pouch lumen. However, results of the time course analysis demonstrate that major transcriptional reprogramming in the air pouch membrane occurs several hours earlier. Future RNA-based studies into events leading to the peak phase of inflammation in the membrane should therefore focus on relatively early time points up to 4 hours.

Gene expression results obtained from the dissected membranes do not allow one to determine whether a given overexpressed gene was induced in the resident membrane cells or imported by infiltrating cells. It is therefore important to validate differential expression of a given gene in a purified cell population, as was done here with cultured macrophages. Even though we validated the induction of all six targets in cultured macrophages, their fold induction was lower in these cells than in the membranes (Figure 5). This suggests that events in the membranes due to cell flux, expression in cells other than macrophages (for example, neutrophils), or more efficient gene regulation due to cell-cell interactions contribute significantly to regulation of these genes in the intact membranes. Another caveat is that the microarray analysis was not replicated, which precluded an extended statistical analysis. However, the unusually high degrees of over-expression and under-expression apparently compensated for this lack of replication; they allowed us to choose a small number of genes for qPCR validation strictly by the high magnitudes of their expression changes.

It is also important to remember that inflammation extends beyond the membrane into the loose, fibroblast-rich tissue between the membrane and the striated muscle layer (Figure 3a and Schiltz and coworkers 12). The air pouch membrane therefore does not contain the entire inflammation that develops in the air pouch, and inflammatory cells may remain in the air pouch wall if the membrane is removed with the blunt dissection method described here. Furthermore, MSU crystals cause swelling of the air pouch membrane 45. This is most likely due to the influx of inflammatory cells and associated interstitial edema and must be considered when evaluating the membrane histologically.

Our findings suggest that a more extensive, replicated microarray analysis involving higher numbers of air pouch membranes may be a powerful tool to identify genes that are differentially active in various aspects of inflammation and innate immunity. In particular, it should lend itself well to the identification of genes that act at different time points in the evolution or resolution of inflammation, genes that are regulated in response to treatment with anti-inflammatory agents, or genes that reflect changes in the resident connective tissue cells in response to inflammation.