A Distinct Inflammatory Gene Expression Profile in Patients with Psoriatic Arthritis

ORIGINAL ARTICLE A distinct inflammatory gene expression profile in patients with psoriatic arthritis

AK Stoeckman1, EC Baechler1, WA Ortmann1, TW Behrens1, CJ Michet2 and EJ Peterson1 1Department of Medicine, University of Minnesota Medical School, Center for Immunology, University of Minnesota, Minneapolis, MN, USA and 2Division of Rheumatology, Mayo Clinic, Rochester, MN, USA

Psoriatic arthritis (PsA) is a systemic inflammatory condition featuring polyarthritis associated with psoriasis. Apart from clinical indicators, few biomarkers exist to aid in the diagnosis and management of PsA. We hypothesized that whole blood gene expression profiling would provide new diagnostic markers and/or insights into pathogenesis of the disease. We compared whole blood gene expression profiles in PsA patients and in age-matched controls. We identified 310 differentially expressed genes, the majority of which are upregulated in PsA patients. The PsA expression profile does not significantly overlap with profiles derived from patients with rheumatoid arthritis or systemic lupus erythematosus. Logistic regression identified two lymphocyte-specific genes (zinc-finger protein 395 and phosphoinositide-3-kinase 2B) that discriminate PsA patients from normal controls. In addition, a highly coregulated cluster of overexpressed genes implicated in protein kinase A regulation strongly correlates with erythrocyte sedimentation rate. Other clusters of coregulated, yet suppressed genes in PsA patient blood include molecules involved in T-cell signaling. Finally, differentially expressed genes in PsA fall into diverse functional categories, but many downregulated genes belong to a CD40 signaling pathway. Together, the data suggest that gene expression profiles of PsA patient blood contain candidate novel disease markers and clues to pathogenesis. Genes and Immunity (2006) 7, 583–591. doi:10.1038/sj.gene.6364334; published online 14 September 2006

Keywords: microarray; gene expression; psoriatic arthritis; lymphocytes; cytokine signaling

Introduction trials of TNF-a inhibitors in PsA have shown effective- ness in controlling skin manifestations and in alleviating Psoriatic arthritis (PsA) is a painful, often disabling musculoskeletal symptoms,6 a significant number inflammatory arthritis associated with psoriasis.1 Psor- of PsA patients fail to respond to TNF inhibition. iasis affects approximately 2% of the Caucasian popula- Given the expense and potential toxicity of anti-TNF tion, and arthritis occurs in about 11% of both male and agents, clinical predictors of response to treatment female patients with psoriasis.2 Despite considerable are urgently needed. heterogeneity in the presentation of arthropathy and the Recently, studies utilizing microarray technology to extent of skin disease, genetic studies support the notion survey gene expression profiles have generated new that PsA is a distinct disease entity with a strong hypotheses concerning the pathogenesis of complex heritable component.3 While a number of genetic loci autoimmune disorders.7 Analyses of differential gene have been associated with psoriasis and PsA,4 the expression in peripheral blood cells of patients with number of genes linked to both psoriasis and PsA is disorders such as rheumatoid arthritis (RA) and systemic limited.5 Few serologic or proteomic biomarkers exist to lupus erythematosus (SLE) have identified several gene aid in diagnosis or monitoring of disease activity. ‘signatures’, groups of genes that are differentially Histopathologic changes in psoriatic plaques include expressed in patients as compared to controls.8–10 In activation and expansion of keratinocytes, and accumu- some cases, differentially expressed genes correlate with lation of T cells, B cells, macrophages and neutrophils.1 particular disease manifestations, with disease activity Activation of both CD4 and CD8T cells has been scores, or with responsiveness to treatment regimens. implicated in the pathogenesis of the skin and joint These observations serve as ‘proof of principle’ that for phenotype. One of the major inflammatory molecules systemic inflammatory diseases in general, peripheral implicated in PsA pathogenesis is the cytokine tumor blood gene expression profiling can be applied to disease necrosis factor-alpha (TNF-a).1 Although recent clinical diagnosis and monitoring, and to the identification of novel therapeutic targets. The goal of this study was to examine the utility of Correspondence: Dr EJ Peterson, Department of Medicine, Uni- gene expression profiling as a hypothesis generating and versity of Minnesota, 6-122 Hasselmo Hall, 312 Church Street, diagnostic tool in PsA. The results suggest that PsA Minneapolis, MN 55455, USA. E-mail: [email protected] patients exhibit a peripheral blood gene expression Received 28 February 2006; revised 5 July 2006; accepted 10 July profile distinct from both healthy controls and from 2006; published online 14 September 2006 that observed in other systemic autoimmune states Gene-expression profiling in psoriatic arthritis AK Stoeckman et al 584 (e.g. the ‘interferon signature’ observed in SLE8,10,11 or the sets for any one gene, our focus was narrowed by ‘monocyte signature’ of RA9). Additionally, logistic removing genes that had an absolute difference between regression analysis identified two lymphocyte-specific patient and control mean intensity values of less than genes that most accurately discriminate patients from 500, and by excluding genes encoding hypothetical controls in this discovery sample. Highly coregulated proteins and non-descript mRNAs. Of approximately genes that correlate most closely with inflammation 12 070 genes expressed in whole blood, we identified include two members of the RAS oncogene family a group of 310 genes that were differentially expressed (RAB13 and RAB32) which are implicated in the in PsA patients (Figure 1). As we expected, all of the inhibition of protein kinase A (PKA) activity. A number patients clustered together by unsupervised hierarchical of differentially expressed genes observed in PsA clustering, while the controls clustered independently. blood are commensurate with previous observations Among the 310 differentially expressed genes, 236 were of disease dependency upon T-cell function, but also upregulated in patients vs controls, while 74 were suggest possible roles for genes involved in apoptosis, downregulated. These results suggest that peripheral cell adhesion, chemokine/cytokine signaling, G-protein blood gene expression profiles can distinguish PsA from coupled signaling and B-cell-mediated immunity. normal controls. Further, both the costimulatory molecule CD40 and its ligand are among a number of adaptive immune cell genes downregulated in PsA patients compared to controls.

Results Whole blood gene expression profiles of PsA patients are distinct from profiles of age- and sex-matched controls We analyzed gene expression profiles in whole blood obtained from 16 PsA patients and 15 normal controls. Patient demographic and clinical data are summarized in Table 1. The following criteria were applied to generate a list of genes that were differentially expressed between PsA patients and normal controls: (a) P-value o0.00001 by an unpaired Student’s t-test for the difference in mean expression level between groups; (b) greater than a twofold change in the mean expression level between the two groups. Initially, we obtained a list of 578 differentially regulated transcripts. After filtering multiple probe

Table 1 Demographic and clinical data for PsA patient subjects

Gender Male n ¼ 7 Female n ¼ 9 Age 44 years712 (range ¼ 15–65 years) Disease duration 13 years712 (range ¼ 1–43 years) MD global disease activity score 4.772.7 (range 0–10) Patients with active psoriasis n ¼ 15 Swollen joints (number) 4.474.8 (range 0–15)

Total white blood cell count ( Â 103/ml) 7.472.2 (range ¼ 4.2–12.8) Lymphocyte count 2.070.8 Monocyte count 0.7270.22 Peripheral mononuclear cell 4.571.8 Figure 1 Gene expression profiles of peripheral blood from 15 count (PMN) controls and 16 PsA patients. Hierarchical clustering of 310 Erythrocyte sedimentation rate 29.3733.5 (range ¼ 2–107) differentially expressed genes is shown. Each row indicates a single (mm/h) gene, and each column represents one subject. Before log2- transformation and clustering, individual data points were ex- Medications at study entry pressed as the ratio of intensity value to the mean of control Any DMARD n ¼ 10 intensity values. Expression value intensities are depicted according Methotrexate n ¼ 7 to the color-indicator key and range from À3 to 3 on a log2 scale. Prednisone n ¼ 1 Red indicates genes that are upregulated and green indicates genes Infliximab or Etanercept n ¼ 5 that are downregulated relative to mean expression levels. Vertical Azathioprine n ¼ 1 bars represent groups of genes that are highly coregulated and Sulfasalazine n ¼ 2 either correlate with erythrocyte sedimentation rate (A) or are implicated in adaptive immune cell function (B). Patient columns Values represent means7s.d. are indicated in red above the cluster.

Genes and Immunity Gene-expression profiling in psoriatic arthritis AK Stoeckman et al 585 The PsA gene expression profile shows little overlap with gene TNFa and IL-1b are apical molecules in cytokine expression profiles of other inflammatory autoimmune diseases cascades that play roles in a number of autoimmune or with cytokine-driven gene profiles inflammatory conditions, including PsA.4 We hypo- PsA shares a number of clinical features with other thesized that differentially expressed genes in peripheral systemic inflammatory syndromes and arthropathies blood of PsA patients would be regulated by these such as RA and SLE. To learn whether these conditions cytokines. In order to identify gene expression patterns exhibit overlapping gene expression profiles, the PsA characteristic of cytokine stimuli, profiles were deter- gene list was compared to differentially expressed gene mined for normal PBMCs incubated with either medium lists generated from recent expression profiling studies alone or with saturating concentrations of IL-1b or TNF in RA9 and SLE.8 Direct comparison of raw gene chip (EC Baechler and TW Behrens, unpublished results). data between these groups was not possible due to Comparison of differentially expressed genes induced by different methods of RNA preparation and utilization of IL-1b or TNF revealed that these stimuli regulate highly earlier generations of Affymetrix chips for the RA and overlapping sets of genes, consistent with work indicat- SLE investigations. In order to compare the gene lists; we ing significant cross-talk between these proinflammatory determined a common denominator to which all disease cytokines (data not shown).12 When the cytokine- lists were compared. This denominator included 5877 stimulated gene lists were compared to the PsA genes that were present on all chips used for analysis, and patient gene list, no greater overlap was observed with were expressed in whole blood. Using w2 analysis, either cytokine than what was expected by chance significance of overlap was determined between the alone. The paucity of overlapping genes between PsA filtered disease lists. Despite clinical observations that and cytokine-stimulated PBMCs may suggest that the both PsA and RA are systemic inflammatory disorders reported heightened activity of TNFa and IL-1b in PsA manifesting similar pannus-producing arthropathies, only patients13 is not clearly reflected in whole blood gene three genes were found to be coordinately and differen- expression profiles. tially expressed in both disorders (RAB13, RAB32, Fc fragment of IgG binding protein). This overlap is not Individual genes have high discriminatory potential for disease greater than what would be expected by chance alone. diagnosis One of the overlapping genes, RAB32, belongs to the Currently, specific and sensitive blood-derived diagnos- monocyte ‘signature’ recently described for RA.9 When tic tests for PsA are not available. We employed logistic the PsA list was compared to a list of genes reported to be regression analysis to explore the potential utility of differentially regulated in SLE, no significant overlap was individual differentially expressed genes in the discri- observed. The single overlapping gene did not belong to mination of PsA patients from normal controls. Figure 2 either the interferon8,10 or the granulocyte ‘signatures’ shows individual expression levels for the four genes recently associated with clinical SLE10 (not shown). While identified by this analysis to have the highest discrimi- technical differences between the current study and the natory potential: zinc-finger protein 395 (ZNF395), dead RA or SLE profiling experiments limit interpretation, box polypeptide 28, pecanex-like 3, and phosphoinosi- these comparisons do not suggest similarity between tide-3-kinase, class 2, beta polypeptide (PI3KC2B). Dead blood cell gene expression patterns observed in these box polypeptide 28 and pecanex-like 3 are expressed autoinflammatory syndromes and PsA. ubiquitously, while expression of ZNF395 and PI3KC2B

Zinc Finger Protein 395 Phosphoinositide-3-kinase Class 2 beta 2 1.5 1 1 0.5 0 0 -0.5 -1 -1 -2 -1.5 -2 -3 Log2(expression) Log2(expression) -2.5 -4 -3 Controls PsA Controls PsA

DEAD-box Polypeptide 28 Pecanex-like 3 2 1.5 1 1 0.5 0 0 -0.5 -1 -1 -2 -1.5 -2 -3 -2.5

Log2(expression) -3 Log2(expression) -4 -3.5 -5 -4 Controls PsA Controls PsA Figure 2 Logistic regression analysis of differentially expressed genes. The four genes with the highest discrimination value are shown. The graphs indicate the level of gene expression as log2 of intensity (expression) values. The vertical line separates the control individuals from the PsA patients.

Genes and Immunity Gene-expression profiling in psoriatic arthritis AK Stoeckman et al 586

Figure 3 Detail view of highly coregulated gene clusters. (a) Upregulated gene cluster designated by vertical bar A in Figure 1. Patient columns are indicated in red above the cluster. Patient expression levels of genes highlighted in blue significantly (Po0.05) correlate with ESR. (b) Downregulated gene cluster designated by vertical bar B in Figure 1. Patient columns are indicated by the horizontal bar below the cluster. Genes emboldened and underlined are implicated in adaptive immune cell function.

is observed specifically in lymphocytes (Novartis atlas, individual genes and other clinical parameters, none http://symatlas.gnf.org/SymAtlas/). ZNF395 has been of these genes fit into highly coregulated groups. implicated in regulation of neuronal cell genes, and These results suggest that RAB family members PI3KC2B has been implicated in epidermal differentia- implicated in PKA signaling are coregulated and tion, but no functions for these genes have been correlate with a measure of systemic inflammation in described in hematopoietic cells.14,15 These results sug- PsA whole blood. gest that detection of whole-blood expression levels of a The list shown in Figure 3b contains individual down- small group of genes may be useful for confirming a PsA regulated genes that are tightly coregulated (Figure 1, clinical diagnosis. vertical bar B). Many of these genes are involved in T- or B-cell development or function, including CD74, Coregulated gene clusters are enriched for genes that are Fyn, RUNX3 and the IL-2 receptor ‘common’ gamma implicated in adaptive immune cell function and for genes chain. Thus, a number of genes implicated in lymphocyte with expression levels that correlate with erythrocyte functions are coregulated and suppressed in PsA sedimentation rate blood. Suppression of these genes is not likely indicative To determine whether groups of coregulated, differen- of significant reduction in PsA lymphocyte numbers, tially expressed genes in PsA correlate with clinical para- since none of the patients display lymphopenia (Table 1; meters, correlation coefficients (r-values) were calculated normal range for lymphocyte count is 0.9–2.9 Â 103/ml), for clinical criteria vs gene expression values. We and since other lymphocyte-specific genes are upregu- observed enrichment of upregulated genes exhibiting lated in patients vs controls (Figure 4). positive correlation with erythrocyte sedimentation rate (ESR) within a highly coregulated group of genes Differentially expressed genes in PsA can be classified into (Figure 1, vertical bar A). Specific genes in this group are several functional categories listed in Figure 3a; genes exhibiting significant correla- To highlight potential functional pathways represented tion coefficients (Po0.05, corresponding to r40.6) with among the differentially expressed genes in PsA patients, patient ESR values are highlighted in blue. To determine we utilized public, annotated databases to group the significance of the correlations between gene expres- them into broad functional categories. Organization of sion and clinical features, P-values were generated by genes based on participation in cellular processes random permutation of the dataset. Among the coregu- according to the Panther Classification System (Applied lated genes are several RAB family members (small Biosystems) led to the observation that multiple differ- molecular weight G-proteins), two of which – RAB13 and entially expressed genes in PsA are implicated in RAB32 – have been implicated in protein kinase A (PKA) functions including apoptosis, cell adhesion, cytokine/ cellular localization and inhibition of PKA signaling.16,17 chemokine signaling, G-protein signaling and adaptive Although significant correlations were found between immunity. Figure 4 illustrates subsets of genes that were

Genes and Immunity Gene-expression profiling in psoriatic arthritis AK Stoeckman et al 587 Apoptosis G Protein Signaling

caspase 10 RAB37 HIV-1 Tat int2 RABGAP1 MCF2L CASP2/RIPK1 adap RAB32 NACHT / PYD1 HRAS-like supp 3 MCL-1 A kinase anchor 6 B-cell CLL ARHGEF12 Reg. of Fas-apop reg of G pro sign 6 TRAIL-rece RAB7 GNAO1 FAST kinase P2RY4 BCL-2 tr.factor RAB13 NACHT / PYD2 RAB40A -5.5 -3.5 -1.5 0.5 2.5 4.5 v-crk CT10 Average Fold Change ras hom gene fam A RALGD Cell Adhesion -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0 PTK2 protyr kinase2 Average Fold Change sema domain

protocadherin g-C3 T&B Cell-Mediated Immunity Preg-spec b-1gly4 MHC II Doa cadherin-like 26 Ig lambda variable

catenin delta 1 CD84 antigen Ig heavy gamma 1 parvin, alpha CD74 antigen selectin P ligand CD3E antigen

actinin, alpha 1 CD3Z antigen runt-rel tr. factor 3 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 Average Fold Change Fyn MAPKKKK 1 Cytokine/Chemokine Signaling Fc frag of IgG bp interferon alpha 17 CD40 ligand

Cbl-interactor Sts-1 CD40 PI3-kinase gamma Leptin receptor -6.0 -4.0 -2.0 0.0 2.0 4.0 inhibin beta C Average Fold Change erythropoietin recep

IL1 recep acc prot

TGF beta 1

IL2-recept gamma

TNFr-assoc 1 -4.5 -3.5 -2.5 -1.5 -0.5 0.5 1.5 2.5 3.5 4.5 Average Fold Change

Figure 4 Differentially expressed genes in PsA can be classified into diverse functional categories. Functional classification of selected differentially expressed genes based on the Panther Classification System. Bars indicate the average fold change in gene expression level for PsA patients vs controls. classified into the above-mentioned pathways and cell peripheral blood numbers has recently been linked indicates the mean fold change in gene expression to methotrexate treatment in SLE patients,18 we found between PsA patients and healthy controls. no significant differences in mean CD40 or CD40LG Further pathway analysis was conducted using the expression levels when comparing patients treated with Ingenuity Pathway Analysis Network Generation Algo- methotrexate or with other DMARDs and untreated rithm (www.ingenuity.com). This system groups a patients (data not shown). Therefore, it is unlikely selection of genes into networks based on Ingenuity’s that DMARD treatment contributes to the observed Pathways Knowledge Base (IPKB). The IPKB constructs downregulation of CD40 expression in PsA patients. In putative functional networks based on primary summary, the pathway analysis demonstrates that literature evidence for protein–protein interactions. Of differentially expressed genes in PsA fall into a number the 310 differentially expressed genes, 146 genes could of functional categories, and that many genes encoding be assigned to networks based on IPKB classification. members of a network of proteins implicated in CD40 A network centered on CD40 signaling contained a signaling are downregulated in PsA patient blood. substantial number (20) of differentially expressed PsA genes. Within this network, CD40, its ligand (CD40LG), and seven other genes were downregulated in PsA Discussion patients (Table 2). Thus, over 10% of the total number of downregulated genes in PsA are present in the CD40- Compared to healthy controls, PsA patients display centered network. Using real-time RT-PCR, we verified whole blood gene-expression profiles characterized by a the microarray-based observations of downregulation of predominance of upregulated genes. This profile does a number of genes in this pathway, including CD40, not significantly overlap with profiles derived from other CD40LG and CD74 (data not shown). Although CD40 is inflammatory autoimmune diseases, although a few PsA expressed predominantly on B cells and a decrease in B genes are regulated in common with genes differentially

Genes and Immunity Gene-expression profiling in psoriatic arthritis AK Stoeckman et al 588 Table 2 PsA differentially expressed genes belonging to a CD40 blood samples to PBMC isolation procedures before gene signaling network expression profiling. Reagents and methods used in the current study Gene name Fold changea Entrez represent several departures from previous studies. First, (patients/controls) number the HG-U133Plus 2.0 chip contains a greatly expanded collection of transcripts compared to the HGU95A chip MHC II DO alpha 3.4 3111 used in the Koczan study or the Atlas Human Array Leptin receptor 2.6 3953 Amyloid beta precursor protein 2.5 351 filter 7740-1 used by Gu et al. Second, in the current HRAS-like suppressor 3 2.5 11145 study, PAXgene collection tubes containing RNA stabi- Kinesin 5A 2.5 3798 lizing reagents were used for preparation of total RNA Erythropoietin receptor 2.2 2057 from whole blood of patients. Baechler et al.21 have MAPK8 interacting protein 1 2.2 9479 shown that the regulation of hundreds of peripheral Collagen XXV alpha 1 2 84570 blood genes is sensitive to ex vivo handling, transport and Adrenomedullin receptor 1.9 11318 Transcription factor 20 (AR1) 1.9 6942 PBMC purification procedures. Our use of PAXgene V-crk like 1.9 1399 tubes avoided potential artifacts introduced by the in CD74 antigen À2.1 972 vitro manipulation of blood cells. Use of later-generation DNAJ homolog B6 À2.1 10049 hybridization platforms and whole blood RNA sources, Heat shock 70 kDa protein 8 À2.1 3312 however, limits the ability to compare profiles with those Glucose regulated protein, 58 kDa À2.3 2923 from earlier studies. As recently published profiling MAP4K1 À2.5 11184 Upstream transcription factor 2 À2.5 7392 studies in RA and SLE used purified PBMCs as an CD40 ligand À2.9 959 RNA source and generated hybridization data on earlier CD40 À3.6 958 generations of chips,8,9 it is a challenge to compare SWI/SNF-related chromatin reg A4 À4.2 6597 results with those of the current PsA study. Never- theless, these studies employed filtering procedures to Fold change is expressed as average patient expression value generate differentially expressed gene lists similar to vs average control value. those utilized in the current study, and it is notable aAll genes met the criteria of Po0.00001 by unpaired Student’s that list comparison does not suggest significant t-test. overlap among differentially expressed genes in these conditions and in PsA. As both TNF and IL-1 are strongly upregulated in PsA expressed in RA and SLE patient PBMCs or in cytokine- synovial fluid,22 and TNF inhibitors have recently shown treated PBMCs. In keeping with existing hypotheses dramatic efficacy in a subset of patients with PsA,6 the concerning the pathogenesis of PsA, the differentially observed lack of overlap between expression profiles expressed genes encode molecules implicated in T-cell from PsA blood and from TNF or IL-1 treated PBMCs is signaling. A highly coregulated subset is enriched for surprising. It is possible that in PsA patients, altered genes that positively correlate with ESR, and thus transcription engendered by these cytokines may be reflect inflammation. A small panel of lymphocyte- occurring primarily in microenvironments of affected specific genes with potential to discriminate PsA patients tissues (e.g. patient joints and skin) rather than in from controls was identified. Finally, differentially circulating blood cells. Certainly, elevated immuno- expressed genes in PsA blood fall into diverse functional reactivity and biological activity of TNFa in the lesions networks, but a high proportion of downregulated genes of psoriasis patients have been described.23 However, as can be classified as part of a CD40-centered pathway. with the RA and SLE differentially expressed gene Previous studies of peripheral blood in PsA or lists, technical differences between the generation of the psoriasis patients utilizing gene expression profiling PsA and of the cytokine-treated PBMC gene lists suggest are limited. Koczan et al.19 recently reported on micro- that comparative analysis of expression profiles should array experiments conducted on isolated PBMCs from 11 be interpreted cautiously. psoriasis patients. Eighteen differentially expressed In addition to a comparison analysis between the PsA genes (all upregulated) in the ‘disease’ stage compared differentially expressed gene list and gene lists from to the ‘cured’ (post-treatment) stage were reported. None other diseases or cytokine treatment, we performed of the PsA differentially expressed genes in the current logistic regression analysis in order to identify genes study were among the genes reported by Koczan, with the best discriminatory potential between PsA possibly due to ex vivo PBMC purification procedures patients and controls. All four genes (ZNF395, pecanex- used by the latter or to the lack of comparison against like 3, PI3KC2B, and dead box polypeptide 28) with the normal controls. In a smaller microarray-based study by strongest ability to discriminate patients from controls Gu et al.,20 14 genes were found to be upregulated in the were downregulated in the PsA patients. Pecanex-like 3 peripheral blood cells of a majority of six PsA patients and dead box polypeptide 28 exhibit ubiquitous expres- compared to controls. Although a number of the over- sion, while both ZNF395 and PI3KC2B show highest expressed genes found in PsA patients fit into relevant expression levels in peripheral blood CD19 þ B cells inflammation pathways (e.g. proinflammatory cytokines (Novartis atlas, http://symatlas.gnf.org/SymAtlas/). and receptors, signaling molecules), only one gene Determination of expression levels for small panels of reported in that study (IL-2 receptor gamma) was also genes like the four listed above may have application in found to be differentially expressed in the current confirming diagnosis of or in monitoring disease activity study. Such differences from the current study may be in PsA. However, validation of these genes as tools for explained by the usage of a restricted 588-gene chip, clinical application will require longitudinal studies in inclusion of only six patients with PsA, and subjection of larger, well-characterized PsA patient populations. Such

Genes and Immunity Gene-expression profiling in psoriatic arthritis AK Stoeckman et al 589 studies should also be adequately powered to address consent was obtained from patients who provided blood important clinical issues like prediction of response to samples. The study included 16 Caucasian patients with treatment with biological response modifiers. PsA. At the time of sample collection, 94% had active T cells have long been recognized as key players in the psoriasis and 75% had active joint disease. Clinical pathogenesis of PsA.24 We identified a number of genes data, including swollen joint counts, disease activity involved in T-cell immunity which were included in a score and laboratory results (erythrocyte sedimentation highly coregulated gene cluster: CD3e, IL-2 receptor rate, C reactive protein levels and peripheral blood cell (‘common’) gamma chain, RUNX3 and Fyn. CD3e is a counts), were obtained at the time of sample acquisition. transmembrane adaptor protein that couples T-cell A single observer (CJM) performed all clinical assess- receptor ligand recognition with activation of protein ments and chart review. Healthy (self-reported) controls tyrosine kinases.25 Fyn is a membrane-associated protein were matched for age (within 3 years of patient ages) and tyrosine kinase required for optimal Th2 cytokine race. The controls included three males and 12 females. production, suppression of Th1-mediated autoimmune disease, and maintenance of T-cell anergy.26–28 Loss of RUNX3 transcription factor activity in mice is associated RNA isolation and labeling with autoimmune disease29–31 as well as the skewing of Venous blood (10 ml) was collected in PAXgene sample CD4/CD8 T-cell development.32 The observed down- tubes (PreAnalytiX). Total RNA was obtained using regulation of these T-cell-expressed genes in PsA blood the PAXgene Blood RNA Kit per manufacturer’s agrees with both the negative regulatory functions instructions. RNA was quantified using the Agilent assigned to many of these genes and with the model of RNA 6000 Nano Assay Protocol. Typical yield of total PsA as a T-cell-dependent inflammatory disease. RNA was 40 mg per 10 ml of blood. Subsequent cDNA/ A subset of highly coregulated genes is enriched for cRNA synthesis was performed with 1 mg of total RNA upregulated genes whose expression levels significantly using the MessageAmp aRNA Kit (Ambion). The cRNA correlate with ESR, a non-specific measure of the was labeled with Biotin-11-CTP (Perkin Elmer) and systemic inflammatory response. Two of these genes Biotin-16-UTP (Roche) before hybridization. are RAB family members shown to be involved in 16,17 negative regulation of PKA activity. As signaling Microarray analysis and quality control through PKA downregulates T-cell responses to anti- m 33 Labeled cRNA (15 g) was fragmented and hybridized gen, the inhibition of PKA by RAB13 and RAB32 could to Affymetrix HG-U133 Plus 2.0 chips in the Bio- lead to enhanced T-cell activity. This is consistent with medical Image Processing Laboratory at the University the hyperactivity of T cells observed in inflammatory of Minnesota. The Affymetrix HG-U133 Plus 2.0 chip autoimmune diseases such as PsA. contains greater than 54 000 probe sets representing B cells may also play a role in the pathogenesis of 24 approximately 38 500 genes. Genedata Expressionist PsA. Ingenuity pathway analysis of the differentially Refiner was used to assess the quality of chip hybridiza- expressed PsA genes revealed the suppression of tion. Each chip was scaled to an average intensity of 1500 numerous genes within a network centered on CD40, a to adjust for global differences in hybridization. Factors TNF receptor superfamily member required for B-cell 34 used to determine successful chip hybridization in- function. Although increased signaling through CD40 cluded levels of background and noise, percent of probes has previously been shown to be important in the 0 0 35–37 considered positive, and 3 –5 ratios of control genes development of psoriasis, RA and SLE we observed (actin and GAPDH). the downregulation of CD40, CD40 ligand and a number of CD40 signaling intermediates in whole blood of PsA patients. Decreased numbers of B cells in the circu- Data acquisition and cluster analysis lating blood of PsA patients does not likely account for Genedata Expressionist (5.0.6) was used to obtain gene decreased CD40 expression, since other B-cell specific expression values (intensities) for each gene. The genes (e.g. Ig heavy chain) are upregulated in PsA following criteria were applied to generate a list of patients. Additionally, evidence in the literature indicates genes that were differentially expressed between PsA no decrease in the absolute number of B cells in the patients and normal controls: (A) P-value of o0.00001 by peripheral blood of PsA patients vs healthy controls.38,39 an unpaired Student’s t-test; (B) greater than a twofold Selective recruitment of B cells or other CD40-expressing change in the mean gene expression between the two leukocytes (e.g. CD11c þ dendritic cells) out of the blood groups. The list was further filtered by removing genes and into sites of pathology is an alternative explanation, that had an absolute difference between patient and given observations of enhanced CD40 expression in control mean intensity values that was less than 500, and psoriatic lesions.40,41 by excluding hypothetical proteins and non-descript mRNAs. The expression values of the differentially

expressed genes were log2 transformed and then used Patients and methods to perform unsupervised hierarchical clustering with the use of the program CLUSTER. Clustering results Study participants were viewed with the TREEVIEW software.42 To deter- This study was approved by the Institutional Review mine the number of genes expressed in whole blood, Boards at the Mayo Clinic and the University of intensity values for the 54 675 probe sets were examined, Minnesota. All patients were recruited through the and genes with average values below 600 for either Rheumatology clinic at the Mayo Clinic (Rochester, patients or controls were removed. Multiple probe sets MN, USA), and all patients met the American College for a single gene, hypothetical proteins, and non-descript of Rheumatology criteria for diagnosis. Informed written mRNAs were then eliminated.

Genes and Immunity Gene-expression profiling in psoriatic arthritis AK Stoeckman et al 590 Comparison analysis 13 Ritchlin C, Haas-Smith S, Hicks D, Cappuccio J, Osterland C, In order to appropriately compare differentially ex- Looney R. Patterns of cytokine production in psoriatic pressed gene lists between diseases, a common gene list synovium. J Rheumatol 1998; 25 (8): 1544–1552. was determined. Of the 12 070 HG-U133 Plus 2.0 genes 14 Harada K, Truong AB, Cai T, Khavari PA. The class II expressed in whole blood, the corresponding probe phosphoinositide 3-kinase C2beta is not essential for sets not also represented on the previous generations of epidermal differentiation. Mol Cell Biol 2005; 25 (24): chips (those used in RA, SLE and cytokine studies) were 11122–11130. eliminated. In addition, we removed genes affected by 15 Tanaka K, Shouguchi-Miyata J, Miyamoto N, Ikeda JE. Novel 21 nuclear shuttle proteins, HDBP1 and HDBP2, bind to neuronal ex vivo handling and PBMC purification procedures. cell-specific cis-regulatory element in the promoter for the All differentially expressed disease gene lists were human Huntington’s disease gene. J Biol Chem 2004; 279 (8): filtered according to this common list. Significance for 7275–7286. gene overlap was determined by chi square analysis. 16 Alto NM, Soderling J, Scott JD. Rab32 is an A-kinase anchoring protein and participates in mitochondrial dy- Logistic regression analysis namics. J Cell Biol 2002; 158 (4): 659–668. Logistic regression analysis to determine genes with high 17 Kohler K, Louvard D, Zahraoui A. Rab13 regulates PKA discriminative potential for affected status was carried signaling during tight junction assembly. J Cell Biol 2004; 165 out using the glm function in R’s statistical package and (2): 175–180. specifying a binomial model with a logit link. 18 Bohm I. Decrease of B-cells and autoantibodies after low-dose methotrexate. Biomed Pharmacother 2003; 57 (7): 278–281. 19 Koczan D, Guthke R, Thiesen HJ, Ibrahim SM, Kundt G, Krentz H et al. Gene expression profiling of peripheral Acknowledgements blood mononuclear leukocytes from psoriasis patients identi- fies new immune regulatory molecules. Eur J Dermatol 2005; 15 The authors thank Dr Kathy Moser and Jason Bauer for (4): 251–257. helpful discussion, Catherine Slattery for technical 20 Gu J, Marker-Hermann E, Baeten D, Tsai WC, Gladman D, assistance with sample preparation, and the patients Xiong M et al. A 588-gene microarray analysis of the who participated in the study. Funding was provided peripheral blood mononuclear cells of spondyloarthropathy patients. Rheumatology 2002; 41: 759–766. by the Minnesota Partnership for Biotechnology and 21 Baechler EC, Batliwalla FM, Karypis G, Gaffney PM, Moser K, Medical Genomics, and the National Institutes of Health. Ortmann WA et al. Expression levels for many genes in human peripheral blood cells are highly sensitive to ex vivo incuba- tion. Genes Immun 2004; 5: 347–353. References 22 Partsch G, Steiner G, Leeb BF, Dunky A, Broll H, Smolen JS. Highly increased levels of tumor necrosis factor-alpha and 1 Klippel JH. Primer on the Rheumatic Diseases, 11th edn. Arthritis other proinflammatory cytokines in psoriatic arthritis synovial Foundation, 1997. fluid. J Rheumatol 1997; 24 (3): 518–523. 2 Gelfand JM, Gladman DD, Mease PJ, Smith N, Margolis DJ, 23 Ettehadi P, Greaves MW, Wallach D, Aderka D, Camp RD. Nijsten T et al. Epidemiology of psoriatic arthritis in the Elevated tumor necrosis factor-alpha biological activity in population of the United States. J Am Acad Dermatol 2005; 53 psoriatic skin lesions. Clin Exp Immunol 1994; 96: 146–151. (4): 573–577. 24 Panayi GS. Immunology of psoriasis and psoriatic arthritis. 3 Rahman P, Elder JT. Genetic epidemiology of psoriasis and Baillieres Clin Rheumatol 1994; 8: 419–427. psoriatic arthritis. Ann Rheum Dis 2005; 64 (Suppl 2): ii37–ii39. 25 Call ME, Wucherpfennig KW. The T Cell receptor: critical role 4 Ritchlin CT. Pathogenesis of psoriatic arthritis. Curr Opin of the membrane environment in receptor assembly and Rheumatol 2005; 17: 406–412. function. Annu Rev Immunol 2005; 23: 101–125. 5 Tomfohrde J, Silverman A, Barnes R, Fernandez-Vina MA, 26 Fukushima A, Yamaguchi T, Ishida W, Fukata K, Udaka K, Young M, Lory D et al. Gene for familial psoriasis suscept- Ueno H. Mice lacking the IFN-gamma receptor or fyn develop ibility mapped to the distal end of human chromosome 17q. severe experimental autoimmune uveoretinitis characterized Science 1994; 264 (5162): 1141–1145. by different immune responses. Immunogenetics 2005; 57 (5): 6 Nash PT, Florin TH. Tumour necrosis factor inhibitors. Med J 337–343. Aust 2005; 183 (4): 205–208. 27 Davidson D, Shi X, Zhang S, Wang H, Nemer M, Ono N et al. 7 Moser KL, Gaffney PM, Grandits ME, Emamian ES, Machado Genetic evidence linking SAP, the X-linked lymphoprolifera- DB, Baechler EC et al. The use of microarrays to study tive gene product, to Src-related kinase FynT in T(H)2 autoimmunity. J Investig Dermatol Symp Proc 2004; 9 (1): 18–22. cytokine regulation. Immunity 2004; 21 (5): 707–717. 8 Baechler EC, Batliwalla FM, Karypis G, Gaffney PM, Ortmann 28 Welke J, Zavazava N. P59(fyn) is upregulated in anergic CD8+ WA, Espe KJ et al. Interferon-inducible gene expression T cells. Hum Immunol 2002; 63 (10): 834–843. signature in peripheral blood cells of patients with severe 29 Alarcon-Riquelme ME. Role of RUNX in autoimmune lupus. Proc Natl Acad Sci USA 2003; 100 (5): 2610–2615. diseases linking rheumatoid arthritis, psoriasis and lupus. 9 Batliwalla FM, Baechler EC, Xiao X, Li W, Balasubramanian S, Arthritis Res Ther 2004; 6: 169–173. Khalili H et al. Peripheral blood gene expression profiling in 30 Brenner O, Levanon D, Negreanu V, Golubkov O, Fainaru O, rheumatoid arthritis. Genes Immun 2005; 6 (5): 388–397. Woolf E et al. Loss of Runx3 function in leukocytes is 10 Bennett L, Palucka AK, Arce E, Cantrell V, Borvak J, associated with spontaneously developed colitis and gastric Banchereau J et al. Interferon and granulopoiesis signatures mucosal hyperplasia. Proc Natl Acad Sci USA 2004; 101 (45): in systemic lupus erythematosus blood. J Exp Med 2003; 197 16016–16021. (6): 711–723. 31 Bae SC, Lee YH. Phosphorylation, acetylation and ubiquitina- 11 Crow MK, Kirou KA, Wohlgemuth J. Microarray analysis of tion: The molecular basis of RUNX regulation. Gene 2006; interferon-regulated genes in SLE. Autoimmunity 2003; 36 (8): 366 (1): 58–66. 481–490. 32 Kohu K, Sato T, Ohno S, Hayashi K, Uchino R, Abe N et al. 12 Feldman M, Maini RN. Anti-TNFalpha therapy of rheumatoid Overexpression of the Runx3 transcription factor increases the arthritis: what have we learned? Annu Rev Immunol 2001; 19: proportion of mature thymocytes of the CD8 single-positive 163–196. lineage. J Immunol 2005; 174: 2627–2636.

Genes and Immunity Gene-expression profiling in psoriatic arthritis AK Stoeckman et al 591 33 Torgersen KM, Vang T, Abrahamsen H, Yaqub S, Tasken K. 38 Koreck A, Suranyi A, Szony BJ, Farkas A, Bata-Csorgo Z, Molecular mechanisms for protein kinase A-mediated mod- Kemeny L et al. CD3+CD56+ NK T cells are significantly ulation of immune function. Cell Signal 2002; 14 (1): 1–9. decreased in the peripheral blood of patients with psoriasis. 34 Toubi E, Shoenfeld Y. The role of CD40-CD154 interactions in Clin Exp Immunol 2002; 127: 176–182. autoimmunity and the benefit of disrupting this pathway. 39 Spadaro A, Scrivo R, Moretti T, Bernardini G, Riccieri V, Autoimmunity 2004; 37: 457–464. Taccari E et al. Natural killer cells and gamma/delta 35 Desai-Mehta A, Lu L, Ramsey-Goldman R, Datta SK. T cells in synovial fluid and in peripheral blood of Hyperexpression of CD40 ligand by B and T cells in human patients with psoriatic arthritis. Clin Exp Rheum 2004; 22: lupus and its role in pathogenic autoantibody production. J 389–394. Clin Invest 1996; 97 (9): 2063–2073. 40 Lowes MA, Chamian F, Abello MV, Fuentes-Duculan J, Lin SL, 36 Harigai M, Hara M, Nakazawa S, Fukasawa C, Ohta S, Nussbaum R et al. Increase in TNF-alpha and inducible nitric Sugiura T et al. Ligation of CD40 induced tumor necrosis oxide synthase-expressing dendritic cells in psoriasis and factor-alpha in rheumatoid arthritis: a novel mechanism of reduction with efalizumab (anti-CD11a). Proc Natl Acad Sci activation of synoviocytes. J Rheumatol 1999; 26 (5): 1035–1043. USA 2005; 102 (52): 19057–19062. 37 Pasch MC, Timar KK, van Meurs M, Heydendael VM, Bos JD, 41 Ohta Y, Hamada Y. In situ Expression of CD40 and CD40 Laman JD et al. In situ demonstration of CD40- and CD154- ligand in psoriasis. Dermatology 2004; 209 (1): 21–28. positive cells in psoriatic lesions and keratinocyte production 42 Eisen MB, Spellman PT, Brown PO, Botstein D. Cluster of chemokines by CD40 ligation in vitro. J Pathol 2004; 203 (3): analysis and display of genome-wide expression patterns. 839–848. Proc Natl Acad Sci USA 1998; 95 (25): 14863–14868.

Genes and Immunity