Axial spondyloarthropathy (SpA) is a family of polygenic inflammatory diseases for which the pathophysiology is complex, with much remaining unknown. Ankylosing spondylitis is the most common form of SpA. Study of gene expression using microarrays offers a novel approach to determining pathogenesis of diseases. Analysis of peripheral blood in patients with systemic lupus erythematosus using this technique has led to the discovery that many lupus patients have an upregulation of genes induced by type I interferons 1.

The present study utilizes a methodology that incorporates an experimental design consisting of primary and validation datasets of subjects, a comprehensive microarray platform, and robust statistical techniques to investigate the presence of a SpA gene expression signature and the presence of novel biomarkers of disease.

Gene expression microarray analysis was performed on whole blood collected from two independent sets of SpA and control subjects. Our analysis of Set 1 identified 556 probe sets that were upregulated and 962 probe sets that were downregulated in subjects with SpA compared with healthy control subjects. Because some transcript levels were evaluated by multiple probe sets on the microarray chip, the chosen probe sets corresponded to 369 upregulated gene transcripts and 721 downregulated gene transcripts. In Set 2, 704 probe sets (550 gene transcripts) were upregulated; 14 probe sets (7 gene transcripts) were downregulated in patients with SpA relative to the control subjects. Heat maps illustrate differences between the groups [see Additional data file 1]. There were 124 probe sets (92 known and 10 unidentified gene transcripts) that were classified in both sets as upregulated in SpA subjects; 10 probe sets (3 known and 2 unidentified gene transcripts) were downregulated in both sets [see Additional data file 2].

We conducted a literature search using National Center for Biotechnology Information databases, including PubMed 16, on all significantly over- or underexpressed gene transcripts to determine their biological functions. Within the group of transcripts identified in both sets, there were 20 gene transcripts involved in immunity or inflammation that might constitute part of the immune signature in SpA. Table 4 presents these transcripts with functional annotations. In particular, we found upregulation of IL-1 receptors and the downregulation of a potential regulator of the IL-1 pathway, NLRP2. Other upregulated transcripts of interest included 'secreted protein acidic and rich in cysteine' (SPARC) and secretory leukocyte peptidase inhibitor (SLPI). Four gene transcripts that have a role in bone remodeling, including kremen 1, were differentially expressed (Table 5). These might form part of a bone remodeling signature for SpA.

Because of a disproportionate number of females in the control group, we conducted a post hoc analysis of variance on the effect of gender. Four models based on different effects of gender and disease state on gene expression were considered. Akaike's information criterion was used to select the model that best fits the data for each probe set. For 14 of the 24 genes included in this secondary analysis, Akaike's information criterion selected the model that assigned the principle expression differences to the disease state after correcting for a gender effect (model 3 in the methods section). The model selected for the remaining 10 genes (marked with an asterisk in Tables 4 and 5) also included an interaction effect of disease state and gender (model 4). For these genes, male subjects with SpA had higher fold-changes than both control subjects and female subjects with SpA, and we cannot exclude a possible effect of gender on the level of transcript expression. However, as an example, even if the downregulation of NLRP2 is a result of the male predominance in the disease group, it would nonetheless represent a novel insight into the male predisposition to SpA.

There are few published studies of gene expression in SpA. Our study reveals a number of genes that are differentially expressed in peripheral blood of patients with SpA and that can be related to the current understanding of its pathogenesis. Our study differs from prior studies in a variety of methodological ways including the number of transcripts studied (more than 47,000 per subject), the exclusion of patients on disease-modifying medications, the use of whole blood, which avoids the potential artifact induced by isolating leukocytes or leukocyte subsets, and pathway analysis in silico. Use of a primary dataset and an independent validation dataset provides additional robustness. Utilizing a false discovery rate calculation limits the possibility of false positives due to chance alone.

Almost all of the transcripts identified as having increased or decreased expression [see Additional data file 2] deserve comment with regard to the pathogenesis of SpA, but space precludes such a thorough discussion. We have selected a small number of transcripts for additional comment. The detection of a set of gene transcripts that may have a role in the immune response and are differentially expressed in both datasets suggests the presence of an 'immune signature' in SpA. Prior work has strongly implicated the IL-1 family in the pathogenesis of SpA. Increased IL-1β mRNA has been found in peripheral blood profiling in individuals with spondyloarthropathy 17. Genetic studies have found that polymorphisms in the IL-1 gene family are associated with ankylosing spondylitis 18 and psoriatic arthritis 19. The finding that both IL-1 receptor (IL-1R) 1 and IL-1R2 are increased at a transcript level suggests a possible correlation with a genetic association between ERAP1 (ARTS1) polymorphisms and ankylosing spondylitis 20; ERAP1 is a proteinase believed to lessen immune responses by cleaving receptors for cytokines including IL-1. Triggering receptor expressed on myeloid cells (TREM)-1 has also previously been implicated in the pathogenesis of ankylosing spondylitis 21. The detection of transcripts that have independently been implicated in SpA adds to the credibility of gene expression microarray analysis as a technique to identify causal factors in this disease.

SLPI has not previously been implicated in the pathogenesis of SpA. SLPI, however, downregulates the synthesis of TNFα 22 and, as such, may well play an important role in the pathogenesis of this disease that often responds markedly to TNF inhibition. SPARC, which is also known as osteonectin, has been implicated in the pathogenesis of scleroderma 23, but not SpA. SPARC could logically be listed as a contributor to bone remodeling (see below), but it also negatively regulates dendritic cell migration and T cell activation 24.

The reduced expression of Nod-Like receptor family, pyrin domain containing 2 (NLRP2 or NALP2) is a novel observation and is especially intriguing. NLRP2 is a component of some inflammasomes 25 and is a member of the NLR family of proteins many of which function as danger-associated molecular pattern receptors of the innate immune system. Polymorphisms in other NLR and related genes have been implicated in diseases that share clinical features with SpA, including Behçet's disease, Crohn's disease, and psoriatic arthritis. Polymorphisms or mutations in genes encoding components or regulators of inflammasomes are associated with several autoinflammatory diseases. NLRP2 functions as an intracellular pattern recognition receptor whose downstream function includes activation of caspase 1 and inhibition of nuclear factor kappa B, both of which lead to regulation of IL-1β (Figure 1) 2627. The downregulation of NLRP2 may therefore lead to upregulation of IL-1β, which in turn may regulate IL-1R expression 27. There is no a priori reason to believe that the expression of a gene such as NLRP2 is affected by gender. If NLRP2 is indeed under expressed in males, this downregulation may be an important clue to the male predominance in this disease.

Ossification is the hallmark of SpA, but there is also ongoing bone resorption with up to 56% of patients becoming osteopenic and a significant proportion becoming osteoporotic 28. The wnt-catenin pathway and its primary regulator Dickkopf 1 (DKK-1) regulate the balance between osteoblast and osteoclast function 29. The upregulation of kremen1 in our data suggests negative regulation of the wnt-catenin pathway via its interaction with DKK-1. The net effect of this and other factors may be bone resorption 30. Endogenous bone morphogenetic protein 6 (BMP6) has been described in a mouse model of enthesis ossification and shown to promote osteoblast differentiation. Inhibition of BMP6 prevents the onset and progress of an SpA-type model of arthritis 31.

This study has some limitations. Firstly, although the diagnosis of SpA was made using a validated method, it was not possible to grade disease activity because most patients were attending an eye clinic. Patients did not routinely have X-rays or MRI scans of the pelvis. However, nearly 100% of the patients had inflammatory lower back pain confirmed by a rheumatologist. Secondly, the control group consisted of volunteers with females outnumbering males. There was a predominance of males in the SpA group as is expected in this condition. Gender differences were apparent for a number of differentially expressed genes located on sex chromosomes. These gender-linked genes could be readily identified on the basis of their chromosomal location and they are not known to contribute to inflammation [see Additional data file 2]. A post hoc analysis was conducted on the transcripts selected as having higher or lower expression levels in SpA subjects to identify those that were also influenced by gender.

Statistical tests on the effect of gender and/or disease on gene expression revealed that disease, rather than the disproportionate number of males in the group with SpA, accounted for the differences in gene expression. However, gender does play a role in SpA, because the vast majority of patients with SpA are male. For some transcripts the overexpression or underexpression of a particular transcript in SpA is more apparent in males. The directional consistency of differences revealed by the initial SAM analysis and the secondary analysis add further support to our findings.

Despite the limitations mentioned above, this study has clearly identified a number of novel and intriguing potential contributors to SpA. Gene expression microarray may elucidate pathogenesis, facilitate diagnostic specificity, correlate with pharmacologic responsiveness, and predict prognosis. We based this study in an ophthalmology clinic to test the hypothesis that patients with SpA and active uveitis would express genes in peripheral blood to distinguish those with uveitis from those without uveitis. Our initial evaluation of this hypothesis indicates that a larger database is necessary to determine if such differences exist. This goal will require large databases with careful accrual of clinical data. We believe that the present study represents an important step toward understanding the molecular mechanisms of SpA.

BMP: bone morphogenetic protein; CEL: cell fluorescence intensity; DKK-1: Dickkopf-1; GCOS: GeneChip Operating System; GC-RMA: GC Robust Multiarray Analysis; IL-1: interleukin-1; IL-1R: interleukin-1 receptor; NLRP2 (NALP2): Nod-Like Receptor family, pyrin domain containing 2; OHSU: Oregon Health & Science University; PM: perfect match; SAM: Significance Analysis of Microarrays; SLPI: secretory leukocyte peptidase inhibitor; SpA: axial spondyloarthropathy; SPARC: secreted protein acidic and rich in cysteine (also known as osteonectin); TNF: tumor necrosis factor; TREM-1: triggering receptor expressed on myeloid cells 1.

CH has an equity interest (less than $5,000) in Affymetrix Inc. None of the other authors has any competing interests.

CA recruited subjects and obtained informed consent, drew blood, conducted clinical data entry, and reviewed the manuscript. CH conducted experimental design, supervised microarray assays, conducted data interpretation, and contributed to the manuscript. DC conducted statistical analysis, and contributed to the manuscript. JL recruited subjects and obtained informed consent, drew blood, conducted clinical data entry, and reviewed the manuscript. JR conducted experimental design, examined patients, conducted data interpretation, edited the manuscript, and supervised the entire project. JS conducted experimental design, provided oversight for human subject research, conducted data interpretation, and edited the manuscript. SP conducted experimental design and database design, oversaw RNA extraction, conducted data interpretation, and contributed to manuscript editing. SS examined patients, analyzed data, and drafted the manuscript. TD extracted RNA from blood samples, and reviewed the manuscript. TM contributed to the manuscript.