Molecular Classification of Psoriasis Disease- Associated Genes

Molecular classiﬁcation of psoriasis disease- associated genes through pharmacogenomic expression proﬁling

JL Oestreicher1 ABSTRACT IB Walters3 Psoriasis is recognized as the most common T cell-mediated inflammatory 3 disease in humans. Genetic linkage to as many as six distinct disease loci has T Kikuchi been established but the molecular etiology and genetics remain unknown. P Gilleaudeau3 To begin to identify psoriasis disease-related genes and construct in vivo path- J Surette2 ways of the inflammatory process, a genome-wide expression screen of mul- U Schwertschlag2 tiple psoriasis patients was undertaken. A comprehensive list of 159 genes 1 that define psoriasis in molecular terms was generated; numerous genes in AJ Dorner this set mapped to six different disease-associated loci. To further interpret 3 JG Krueger the functional role of this gene set in the disease process, a longitudinal phar- WL Trepicchio1 macogenomic study was initiated to understand how expression levels of these transcripts are altered following patient treatment with therapeutic 1Division of Molecular Medicine, Genetics agents that antagonize calcineurin or NF-␬B pathways. Transcript levels for Institute/Wyeth-Ayerst Research, Cambridge, MA, USA; 2Department of Experimental a subset of these 159 genes changed significantly in those patients who Medicine, Genetics Institute/Wyeth-Ayerst responded to therapy and many of the changes preceded clinical improve- Research, Cambridge, MA, USA; 3Department ment. The disease-related gene map provides new insights into the patho- of Investigative Dermatology, Rockefeller genesis of psoriasis, wound healing and cellular-immune reactions occurring University, New York, NY, USA in human skin as well as other T cell-mediated autoimmune diseases. In Correspondence: addition, it provides a set of candidate genes that may serve as novel thera- WL Trepicchio, Genetics Institute/Wyeth- peutic intervention points as well as surrogate and predictive markers of treat- Ayerst Research, 1 Burtt Road, Andover, ment outcome. The Pharmacogenomics Journal (2001) 1, 272–287. MA 01810, USA + Tel: 1 978 247 1332 Keywords: psoriasis; inflammation; gene expression profiling; microarrays; recom- Fax: +1 978 247 1333 binant human Interleukin 11 (rhIL-11); Cyclosporin A E-mail: wtrepicchioȰgenetics.com

JG Krueger, Department of Investigative Dermatology, The Rockefeller University, INTRODUCTION 1230 York Avenue, New York, New York Psoriasis is a chronic cutaneous inflammatory disease characterized by hyperplas- 10021, USA Tel: +1 212 327 7730 tic regenerative epidermal growth, and infiltration of immune cells, particularly Fax: +1 212 327 8232 T cells into the dermis and epidermis resulting in the release of proinflammatory E-mail: [email protected] molecules.1 The therapeutic benefit of immunosuppressive drugs, eg, Cyclosporin A, FK506, anti-CD4 antibody and IL-2 diphtheria toxin conjugate supports the hypothesis that activated immune cells, particularly T cells are pathogenic effec- tors of psoriasis.2–4 More recently, treatment of psoriasis patients with the immu- nomodulatory drug, recombinant human interleukin-11, resulted in a downregulation of type 1 cytokines suggesting a role for Th1-specific cells in disease progression.5 As a result of these studies, there is broad interest among researchers in using psoriasis as a ‘model’ inflammatory disease mediated by Type 1 T-cells and many new anti-inflammatory or immune-modulating drugs are now tested for the first time in humans in this disease. The precise etiology of psoriasis remains elusive, although genetic and cellular factors are being elucidated.6 In addition to environmental factors, there is a Received 6 July 2001 genetic component to the disease. About one-third of patients report a family Revised 28 September 2001 member with the disease and family studies estimate the risk to first-degree rela- Accepted 17 October 2001 tives at between 8–23%. There is a 65–70% concordance in monozygotic twins, Pharmacogenomic expression profiling of psoriasis JL Oestreicher et al 273

compared to 15–20% in dizygotic twins (reviewed in Bethesda, MD, USA). In all instances throughout these stud- Bhalerao and Bowcock7). Several psoriasis susceptibility loci ies, individual patient samples were hybridized to arrays. A have been mapped: PSORS1 on 6p21.3, PSORS2 on 17q, global comparison of gene expression identified 1295–1858 PSORS3 on 4q, PSORS4 on 1cen-q21, PSORS5 on 3q21 and genes (19–26% of sampled genome) expressed in unin- PSORS6 on 19p13.7 volved skin and 1352–2587 genes (19–38% of sampled Despite the identification of disease-associated loci, spe- genome) expressed in lesional skin (Table 1). The number cific genes that play a causative role in disease progression of genes expressed in normal skin varied from 1383–2375 have yet to be identified. Mutations within coding regions (20–22%) indicating no overt differences in expression pro- affecting protein function or levels have traditionally been files between normal and disease tissue at this global level. examined for disease-related genes. However, mutations To determine how similar or dissimilar expression profiles affecting transcriptional regulation or mRNA stability can were between lesional and uninvolved skin types and across also play a role in disease initiation or progression. To begin unrelated individuals, a unsupervised cluster analysis to identify such genes and/or pathways affected at the approach was taken.8 This method, which clusters samples mRNA transcript level, a genome-wide scan of uninvolved based on a hierarchical correlation coefficient, grouped and lesional psoriatic skin as well as inflamed skin from expression patterns of normal and uninvolved skin together non-psoriasis patients was performed using oligonucleotide while expression patterns of lesional skin formed a separate arrays containing over 7000 human genes. A large number cluster (Figure 1a). A dendrogram of these relationships of differentially regulated genes were identified that dis- illustrates greater inter-patient similarities in expression pro- tinctly classified psoriasis lesions from other uninvolved files for like skin types than intra-patient similarities skin types. The functional significance of these expression between uninvolved and lesional skin types (Figure 1a). An patterns was determined by experimentally manipulating illustrative cluster image analysis identified a group of genes human skin in vivo by inducing a cellular immune response whose expression levels were elevated in lesional skin com- or a wound healing response, as well as through therapeutic pared to normal or uninvolved skin (Figure 1a). This result treatment of patients with immunosuppressant or immuno- suggested that significant differences in expression profiles modulatory drugs that impact the calcineurin or NF-␬B between uninvolved and lesional skin could be observed in pathways. These studies indicate that complex genomic a differential comparison of gene expression frequency expression patterns in inflammatory disease can be resolved levels. into immune-associated vs reactive cellular components by functional manipulation coupled with genome-wide assess- Identification of a Psoriasis Disease Gene Classification ments. Furthermore, differentially regulated gene sets and Set pathways identified by these studies may play a causative A differential mRNA expression profile of uninvolved to role in disease progression or may be early markers of clini- lesional skin was performed. Comparisons of gene fre- cal efficacy. Table 1 Summary of gene expression data RESULTS Patient Demographics and Global Expression Analysis Patient Skin type Gene chip data of Uninvolved, Psoriatic and Normal Skin No. To identify disease-specific genes that were differentially No. present No. present Differential regulated in psoriatic lesions, 6-mm full thickness punch nonlesion lesion expression biopsies were obtained from uninvolved and lesional skin of 24 patients (13 females and 11 males) with moderate to 201 psoriasis 1295 1352 519 202 psoriasis 1428 2059 878 severe chronic plaque psoriasis. For comparison, biopsies 203 psoriasis 1671 1847 416 from normal skin of non-psoriatic patients were also 205 psoriasis 1653 2587 1321 obtained. Biopsies were bisected with one half going 212 psoriasis 1776 1868 543 towards histological analysis and one half flash frozen and 222 psoriasis 1743 1812 340 used for RNA expression analysis. All patients enrolled in the 301 psoriasis 1858 2027 454 study had chronic psoriasis vulgaris affecting at least 10% of 304 psoriasis 1680 1773 349 their body surface area as measured by a combination of 401 tape strip 1383 1619 309 clinical and histological parameters.5 On average patients 402 tape strip 1534 1683 62 were affected with psoriasis for 24 years (range 4–44 years) 403 tape strip 1502 1724 283 501 DTH 2187 2336 492 while only one patient had disease for less than 10 years. 503 DTH 2375 2514 925 Patients were primarily of Caucasian descent and had a 504 DTH 1612 2331 782 mean age of 41 years (range of 23–63 years). RNA was prepared from lesional and uninvolved skin RNA was prepared from one-half of a 6-mm punch biopsy of uninvolved skin and from an initial group of eight psoriasis patients and com- lesional skin obtained from patients prior to treatment with rhIL-11 or Cyclosporin pared to normal skin. Samples were analyzed on oligonucle- A or following a DTH or tape-stripping reaction. RNA was hybridized to oligonucleotide arrays containing over 7000 human genes. The number of genes called otide arrays containing approximately 7000 human genes ‘present’ on each array is shown. The number of differentially regulated genes (Unigem collection; National Biotechnology Information, between uninvolved and lesional skin is presented under differential expression.

www.nature.com/tpj Pharmacogenomic expression proﬁling of psoriasis JL Oestreicher et al 274

The Pharmacogenomics Journal Pharmacogenomic expression proﬁling of psoriasis JL Oestreicher et al 275

quencies were made between uninvolved and lesional skin tially expressed transcripts between uninvolved and lesional from the same individual in a pairwise fashion in order to skin was observed across patients (Table 1). This finding was reduce the possibility of confounding demographic para- somewhat unexpected especially since the disease severity meters. A large number of differentially regulated transcripts of the affected tissue across patients was relatively homo- were identified. Expression of 340–1321 genes was found to geneous. A likely explanation for these findings is that glo- vary by an average factor of two-fold or greater between bal analysis of gene expression using relatively small sample lesional and uninvolved skin within individual patients sizes has the potential to identify differentially regulated (Table 1). To give an overall picture of gene expression genes that not only correspond to the disease state but also changes in a variety of skin types, off center diagonal plots correspond to demographic differences (age, sex or ethnic of the gene expression measurements from each sample origin) as well as environmental exposure differences (diet, were plotted against all other samples and correlation coef- antigen exposure, temporal gene expression patterns) ficients were calculated (Figure 1b). Comparison of lesional within the starting patient populations. To minimize the to uninvolved skin types from the same individuals indi- inclusion of differentially regulated genes unrelated to the cated a higher degree of unrelatedness (r2 = 0.87) than com- disease-state, a statistical approach to the data analysis was parisons of normal to normal skin (r2 = 0.96), uninvolved to taken. Average frequency gene expression values between uninvolved skin (r2 = 0.94) and lesional to lesional skin types uninvolved and lesional skin were calculated and a paired (r2 = 0.98) between unrelated individuals (Figure 1b). Com- t-test was performed. Four hundred and seventy-six genes parison of normal skin to uninvolved psoriasis skin also with transcripts levels statistically different between these indicated a high degree of similarity (correlation coefficient two defined groups with a confidence level of 95% or greater 0.95) with only 34 genes differing by two-fold or greater were identified. This list was further refined to 159 genes by (Figure 1b). These results validate the consistency of the pro- selecting only those genes whose transcript levels differed filing method and indicate that uninvolved and normal skin on average by two-fold or greater between uninvolved and from different individuals is more similar to each other than lesional skin. To further confirm the significance of these to lesional tissue. findings, we performed 100 random permutations of the 16 A large degree of heterogeneity in the number of differen- samples into two groups. This resulted in only 28–102 genes that were statistically significant between two random groups of eight arrays and only 6–15 genes differed by as Figure 1 Global expression analysis. Total RNA obtained from much as two-fold or greater between the two groups (Figure lesional and uninvolved skin biopsies from eight psoriasis patients 1c). Class prediction analysis using supervised classification and normal skin from three normal patients were analyzed on oligonucleotide arrays containing 7000 full length human genes. In metrics indicated that this 159 gene set can be used to pre- total, 19 samples were analyzed on individual arrays. In no case dict with 100% accuracy expression patterns unique to were RNAs pooled. (a) Unsupervised hierarchical cluster analysis of normal/uninvolved skin vs lesional skin (Figure 1d). These uninvolved and involved psoriasis skin lesions and normal skin. 159 genes therefore comprise a disease classification set for Cluster analysis was performed using the methods of Eisen et al.8 chronic plaque psoriasis. A normal skin sample was used as a common baseline. A dendrogram of sample relatedness is shown. Patient or volunteer identifi- Characterization of Psoriasis Classification Genes cation numbers followed by the designated skin type are shown The 159 differentially regulated genes were categorized into below the dendrogram. Red colors indicate genes that are elevated relative to the baseline and green colors are genes that are functional groups. The expression levels of these genes are decreased relative to the baseline. (b) Representative of diagonal presented as the fold-change of the frequency levels of plots and correlation coefficients (r2) of frequency values of unin- lesional over uninvolved skin (Figure 2). The heterogeneous volved to lesional psoriatic skin from a single patient, normal to nature of the cells within the skin biopsies can result in a normal skin from unrelated donors, uninvolved to uninvolved skin dilutional effect of messages from rare cell types. Our experi- from unrelated patients, lesional to lesional skin from unrelated ence has indicated that small fold-changes in low frequency patients and normal to uninvolved skin from unrelated patients. genes can actually be quite large changes in rare populations Green triangles represent genes whose transcript levels do not of cells such as lymphocytes. Alternatively, they can rep- change in a two-way comparison between the samples. Blue tri- resent small changes in high frequency cells such as kera- angles represent genes whose transcript levels increase in a comparison of one sample vs another. Red triangles represent genes tinocytes. Therefore, to facilitate a clearer biological under- whose transcript levels decrease in comparison of one sample rela- standing of the gene expression changes in the context of tive to another. (c) Supervised cluster analysis. The correlation these different cell populations, genes are further color measurements of the 1000 most statistically significant genes of coded according to their frequency levels (Figure 2). defined classes (uninvolved to lesion) compared to randomly per- Genes involved in functions as diverse as transcriptional muted class distinctions are plotted. The top 1%, 5% and median regulation, metabolic control, protein processing, intracellu- distance measurements of 100 randomly permuted classes com- lar signaling, cell cycle control, lymphocyte regulation and pared to the observed distance measurements for uninvolved and extracellular matrix destruction were identified (Figure 2). lesional classes are plotted. (d) Histogram of class prediction strengths of normal, uninvolved and psoriasis lesions. Scores range Many of these genes were previously reported to be differen- from +1, corresponding to a perfect prediction score of normal or tially regulated in psoriasis lesional skin such as psoriasin uninvolved skin to −1, corresponding to a perfect prediction score (S100A7), fatty acid binding protein (FABP5), elafin (SKALP, of lesional skin. PI3), retinoic acid binding protein (CRAB2), squamous cell

www.nature.com/tpj Pharmacogenomic expression proﬁling of psoriasis JL Oestreicher et al 276

The Pharmacogenomics Journal Pharmacogenomic expression proﬁling of psoriasis JL Oestreicher et al 277

carcinoma antigen (SCCA), beta defensin-2 (DEFB2), keratin Comparative Expression Profiles with Other Cutaneous 17 (KRT17) and keratin 16 (KRT16).5,9,10,11 This finding vali- Inflammatory Skin Conditions dates our microarray comparison approach (Figure 2). Psoriasis can be viewed as both a focus of T cell-mediated Many genes not previously known as being differentially inflammation and as a local wound-healing reaction. To regulated in psoriatic lesions were also identified. For begin to characterize the role of the differentially regulated example, overexpression of S100A12 (calgizarin C, psoriasis gene set in the pathophysiology of psoriasis, ENRAGE), matrix metalloproteinases (MMP-12) and heparin expression profiles were generated for other cutaneous binding protein 17 (HBP-17) were observed (Figure 2). A inflammatory/hyperproliferative conditions whose mechan- number of genes such as keratin 2, Apolipoprotein E (APOE), istic components comprised a subset of the psoriasis pheno- GATA3, RB1, calponin 1 (CNN1), Cystatin 6 (CST6), TIMP- type. These conditions included antigen-induced inflam- 3 and TNXA were downregulated in the psoriatic lesions vs mation as occurs in the delayed type hypersensitivity uninvolved skin as indicated by their negative fold-change reaction (DTH) and regenerative epidermal hyperplasia values (Figure 2). Other genes involved in inflammation and induced by tape stripping. The DTH reaction was intended immune regulation such as the IL-4R, CD2, CD24, CD47, as a surrogate of a T cell-driven cellular immune response STAT-1, IFI27, IFI56, MX1, MnSOD, and SCYA1 (MCP1) in skin while tape stripping was intended to activate regen- were elevated in lesional vs uninvolved tissue (Figure 2). erative epidermal hyperplasia. From past work it is known The oligonucleotide array findings for a number of differ- that tape-stripping induces K16 expression, keratinocyte entially regulated transcripts were extended to a larger pso- hyperplasia and up-regulation of some inflammatory cyto- riasis patient population using a high throughput quantitat- kines but inflammation as described here is ‘non-specific’ in ive RT-PCR approach. Analysis of 16 additional psoriasis the sense that it is not antigen-driven nor T cell-depen- patients confirmed elevation of S100A12, K16, HBP17, IL- dent.11 To induce a DTH reaction, three volunteers were sen- 4R, CCNF, LAD1, MAPKK3, MMP-12 and DSG3 mRNA levels sitized with DNCB and 72 h later challenged with a second in lesional skin (data not shown). Lower levels of CST6, dose of DNCB. A comparison of a biopsy obtained at the site TNXA, ID4, TIMP-3, GATA3, and ApoE in lesional skin vs of the DTH reaction to an uninvolved region identified a uninvolved skin were also confirmed (data not shown). large number of genes that were differentially regulated Comparable fold changes were observed between the RT- between normal and inflamed skin (Table 1). On average, PCR approach and the oligo arrays for most genes. Better 259 genes differed by two-fold or greater in two out of three correlation was observed for genes expressed at high levels volunteers (Figure 3a). Likewise biopsies were obtained from in both lesional and uninvolved samples. Somewhat normal skin and tape stripped skin from three different sub- reduced fold changes were observed for genes expressed at jects. Significant differences in gene expression levels were lower levels. This was due to the greater sensitivity of the observed between normal vs irritated skin in the three vol- RT-PCR approach that resulted in a more accurate quantit- unteers (Table 1). One hundred and sixty-one genes were ation of gene expression levels in the uninvolved samples. identified that were differentially regulated on average two- Since transcript levels are not always correlative with pro- fold or greater in two out of three tape-stripped samples tein levels, immunohistochemical analysis of a number of (Figure 3a). differentially regulated genes was conducted. Consistent A comparison of the gene sets obtained from the psoriasis with the elevated mRNA levels, protein levels of KRT16, patients, DTH reaction and tape stripping showed some S100A12 and MMP-12 were elevated in lesional vs unin- overlap of the mRNA expression profiles (Figure 3a). Thir- volved skin of these patients (data not shown). teen transcripts including KRT16, S100A2, S100A7, S100A9, PI3 and DEFB2 were found to be differentially regulated in all three inflammatory/hyperproliferative conditions (Figure 3b). Expression of 23 transcripts including STAT-1, IL-4R and SCYA2 were differentially regulated in psoriasis and following a DTH reaction (Figure 3c). mRNA transcript levels Figure 2 Histograms of the 159 statistically significant differen- of 30 genes including KRT17, KRT2A and HBP17 were differ- tially expressed genes between uninvolved and involved psoriasis entially regulated in psoriasis and following tape stripping skin. Average frequency values and standard deviations from eight paired uninvolved and lesional samples were calculated. A paired (Figure 3d). Subtraction of these gene sets from the psoriasis students t-test was performed to identify statistically significant dif- gene set identified 94 transcripts including S100A12 ferences between these samples and P values are presented. P (ENRAGE), RAGE and GATA3 that were differentially regu- values less than 0.05 were considered statistically significant. Fold lated only in psoriasis. Finally, 11 differentially expressed changes in gene expression were calculated as the ratio of the aver- mRNA transcripts were in common between DTH and tape age frequency of lesional skin compared to uninvolved skin change stripped samples but did not change between lesional and and this value is found at the end of the histograms. Increases in uninvolved psoriasis samples (data not shown). transcript levels in lesions relative to uninvolved skin are presented as positive numbers and decreases in transcript levels are presented as negative values. Histograms are color coded according to the Pharmacological Treatment of Patients Identifies average frequency value. Genes are grouped according to func- Pharmacogenomic Classes of Responding Genes tional significance. Gene names and accession numbers are pro- There is a tacit assumption that changes in gene expression vided. are causally related to disease states. However, changes in

www.nature.com/tpj Pharmacogenomic expression proﬁling of psoriasis JL Oestreicher et al 278

The Pharmacogenomics Journal Pharmacogenomic expression proﬁling of psoriasis JL Oestreicher et al 279

mRNA expression levels may also be a consequence of the were treated with Cyclosporin A for 8 weeks. At the end of disease process. We hypothesized that changes in gene this treatment period, patients were monitored for an expression that precede clinical improvement may play a additional 4-week follow-up period. Response rates of these more causal role in disease progression compared with those patients were typical of those previously reported.2,5 Sixty genes whose expression changes mirror clinical improve- per cent of rhIL-11-treated and 90% of Cyclosporin A- ment or do not change despite clinical improvement. To treated patients were considered to have responded to ther- better understand possible functional relationships between apy as defined by changes in several clinical and histopatho- differentially regulated genes and disease, lesion biopsy logical criteria.5 samples were obtained from psoriasis patients before, dur- To identify genes whose expression patterns change over ing, and after pharmacological treatment with therapies pre- the course of drug treatment, comparisons in expression viously shown to be effective. profiles of lesional skin were made between responding and Patients were treated with the experimental immunomod- non-responding patients prior to drug treatment and at ulatory cytokine, rhIL-11, or with the immunosuppressant weeks 1, 4, 8 and 12 following initiation of therapy. Gene drug, Cyclosporin A. Interleukin-11 has shown activity in chip analysis was initially performed on six rhIL-11-treated psoriasis patients to improve clinical and histopathological (five responding and one non-responding) and three Cyclo- scores and these changes correlate with a reduction over sporin-treated (two responding and one nonresponding) time in the expression levels in the skin of type I cytokines patients. The expression levels of the 159 genes that were such as IFN-␥ and IL-12p40.5 In addition, stable and/or over- significantly different between lesional and uninvolved skin expression of type II cytokines such as IL-4 and IL-5 were were monitored over the course of drug treatment in these observed in the same lesion biopsies.5 rhIL-11 has been nine patients. shown to reduce proinflammatory cytokine production in Self-organizing maps (SOMs) were employed to aid in the part through the inhibition of NF-␬B nuclear translocation identification of pharmacogenomic expression patterns.15 in activated macrophages but does not appear to regulate For this analysis, the level of expression of the lesional skin the p38/JNK pathway.12 Cyclosporin A is an immunosup- prior to therapy was normalized to a value of one and the pressant drug active in psoriasis that also inhibits proin- change in expression of the 159 gene set over the course of flammatory cytokine mRNA production in psoriatic treatment was calculated relative to this baseline level. Four lesions.2,5 Mechanistically Cyclosporin A is distinct from patterns of gene expression changes were identified in rhIL-11 with inhibitory effects on T cell activation parti- responding patients over the course of drug treatment cularly through blockade of the calcineurin/NFAT and (Figure 4). In contrast, no significant change in expression p38/JNK signaling pathways.13,14 Comparisons of gene patterns for this gene set was observed over the course of expression levels between uninvolved and lesional skin fol- drug treatment in nonresponding patients. Clusters I, II, and lowing treatment with rhIL-11 or Cyclosporin A cover a III correspond to genes whose levels of expression were elev- diverse spectrum of pathways of therapeutic intervention ated in lesional compared to uninvolved skin. Cluster IV for inflammatory and autoimmune diseases. corresponds to genes whose expression levels were lower in rhIL-11 was administered subcutaneously to patients at lesional compared to uninvolved skin. Of particular note, 27 either doses of 2.5 ␮gkg−1 day−1 and 5 ␮gkg−1 day−1 or a differentially expressed genes mapped to the six identified single 1 mg or 2 mg dose administered once weekly. Cyclo- psoriatic susceptibility loci (Figure 4). The transcript levels sporin A was administered at a dose of 5 ␮gkg−1 day−1.2 Fif- of the 41 genes in cluster I were found to change on average teen patients were treated with rhIL-11 and nine patients from 2–8 fold in the rhIL-11- and/or Cyclosporin A-treated responding patients as early as one week following the initiation of drug therapy (Figure 4, cluster I). Interestingly, Figure 3 Comparison of differentially expressed genes in DTH, these changes in expression levels preceded significant clini- tape strip, and psoriasis samples. Total RNA was prepared from cal improvement as measured by changes in the average PSI affected and unaffected skin of three patients following tapestrip- or PASI scores. The average PSI score declined only modestly ping and three patients following induction of a DTH reaction and from 9.0 to 8.7 after 1 week of therapy (Figure 4). These analyzed on oligonucleotide arrays containing 7000 full length changes were not observed in the non-responding patients human genes. Genes differentially regulated by two-fold or greater between uninvolved and lesional psoriasis skin, normal and DTH examined. Twelve genes in this cluster including S100A12, skin and normal and tape-stripped skin are presented. (a) Venn dia- ID4, MTX and HBP17 mapped to five of the six known pso- gram comparing the overlap in differential gene expression riasis-susceptibility loci with the exception of the PSOR5 loci between the various sample types. (b) Histograms of 13 genes dif- (Figure 4). ferentially regulated following a DTH reaction, tape stripping (TS) Cluster II corresponded to 91 genes whose transcript levels and in psoriasis lesions. (c) Histograms of 23 genes differentially returned to those found in uninvolved skin but only at the regulated following a DTH reaction and in psoriasis lesions. (d) His- end of drug therapy and not prior to clinical improvement tograms of 30 genes differentially regulated following tape strip- (Figure 4, cluster II). Included in this cluster were 11 genes ping and in psoriasis lesions. Data are graphed as fold change and are color coded according to the average frequency value. Average that mapped to all six psoriatic disease loci. Cluster III fold change values are depicted at the end of the histograms. Histo- included 17 genes whose expression levels did not change grams are color coded according to the average frequency value. over the course of therapy despite clinical improvement in Gene names and accession numbers are provided. the patients skin lesions (Figure 4, cluster III). Only one of

www.nature.com/tpj Pharmacogenomic expression proﬁling of psoriasis JL Oestreicher et al 280 7) and =

n al, 45 biopsy d and lesional (two responding onder ( fferentially regulated . # Average expression values in lesional skin at baseline, week 1, week 4, week 8 and week 5 4-cluster (I-IV) SOM was generated using the 45 biopsy samples. Genes contained in each SOM are × 2) categories based on clinical and histopathological criteria. =

n 12 for respondershown and in nonresponder alphabetical groups order werespecifically by to calculated. rhIL-11 the A are right 1 indicated of by the a maps. *. Genes Transcripts that that are correspond differentially to regulated psoriasis specifically susceptibility by loci Cyclosporin are A color-coded. are Transcripts indicated that by are di nonresponder ( Figure 4 Self-organizingskin biopsies map (SOM) obtained from analysisand one of six non-responding). psoriasis patients Biopsies 159 treatedsamples were gene with were obtained set rhIL-11 hybridized at baseline, to following (five week individual rhIL-11 responding 1, oligonucleotide and and week arrays Cyclosporin one 4, containing A non-responding) week 7000 8 treatment. and human of RNA three genes. treatment patients was Samples and were prepared treated 4 not from with weeks pooled. following uninvolve Cyclosporin Patients end A were of treatment grouped (week into 12). resp In tot

The Pharmacogenomics Journal Pharmacogenomic expression proﬁling of psoriasis JL Oestreicher et al 281

these genes, KPNB1, mapped to a known psoriasis locus observed in either rhIL-11- or Cyclosporin A-treated nonre- (PSORS2). Finally, cluster IV contained 10 genes whose tran- sponding patients (data not shown). script levels were lower in lesional compared to uninvolved skin. The transcript levels of these genes increased over the DISCUSSION course of drug treatment and approached the levels found The pathophysiology of psoriasis is complex involving mul- in uninvolved skin. Changes in the expression levels of tiple cell types and is characterized by epidermal hyperplasia some of these genes such as TNXA, RAGE, ID4 and GATA3 with differentiation of keratinocytes along a regenerative were elevated as early as 1 week following the start of ther- pathway.1 The underlying disease mechanism involves an apy and these changes also preceded clinical improvement immune response with epidermal and dermal infiltration of (Figure 4, cluster IV). Both RAGE and TNXA map to the activated CD4 and CD8 T cells.16 We have used a pharmaco- PSORS1 locus and CNN1 maps to the PSORS6 locus. Interest- genomic mRNA expression analysis approach to identify ingly, a ligand for the RAGE receptor, S100A12, mapped to genes linked to these pathogenic processes and cell types in the PSORS4 locus at 1q21. order to understand better the molecular basis of the disease. Quantitative RT-PCR was used to analyze gene expression A comparison of gene expression profiles from psoriatic changes for the early responding genes in clusters I and IV lesions and uninvolved skin from multiple patients ident- in a larger patient population. Expression levels of genes for ified over 159 differentially regulated genes that comprise a S100A12, KRT16, SCYA2, PRKMK3, CST6, TNXA, CCNF and predictor set of the disease-state as determined by nearest GATA3 were analyzed in 15 patients treated with rhIL-11 neighbor analysis. As validation of our experimental and nine patients treated with Cyclosporin A (Figure 5a–b approach, most genes previously identified by orthogonal and data not shown). Consistent with gene chip analysis in methods as being differentially regulated in psoriasis lesions, the smaller patient population, similar changes in gene such as S100A7 (psoriasin), PI3 (elafin), FABP5, DEFB2 (beta expression in response to Cyclosporin A or rhIL-11 were defensin-2), KRT16 and SCCA1, were also identified in the 5,9,10,11 observed as early as 1 week following drug therapy in current studies. In addition to these known differentially expressed genes, responding patients but not in nonresponding patients. a large number of genes were identified with no previous Changes in gene expression continued to decline over the link to psoriasis and/or inflammatory conditions. Many of course of drug treatment. This change in gene expression these genes fell into clusters that define a number of clinical preceded clinical improvement in these patients. Following and histological components of psoriasis such as hyperkera- discontinuation of drug treatment at week 8, mRNA tosis, metabolic dysfunction and immune deregulaton. For expression levels for some of these genes such as KRT16, example, altered expression of a number of genes involved S100A12 and SCYA2 began to rebound at week 12 towards in protein synthesis, degradation and metabolism were those found in untreated lesions (Figure 5a). observed. Members of this functional cluster include a group rhIL-11 and Cyclosporin A treatment had differential of genes involved in lipid and sterol metabolism (FABP5, effects on the transcript levels of 11 genes. Interestingly, fatty acid binding protein 5; APOE, apolipoprotein E; SQLE, many of these differences were observed in genes in cluster squalene epoxidase; ALDH10, fatty aldehyde dehydrogen- IV. For example, only rhIL-11-treatment resulted in signifi- ase; LDLR, low density lipoprotein receptor). Some of these cant elevation of GATA3, CRIP1 and TNXA. Cyclosporin A metabolic enzymes are associated with skin diseases. For treatment did not significantly modulate the transcript lev- example, deficiency in fatty aldehyde dehydrogenase is asso- els of these genes. Alternatively, Cyclosporin A significantly ciated with Sjogren–Larsson syndrome that is characterized reduced MMP-12, CCNF and HBP17 levels following treat- by ichthyosis and hyperkeratosis.17 In addition, the LDLR ment while rhIL-11 had no significant effect on these tran- gene maps to the PSORS5 locus supporting a potential link script levels. These differential effects of rhIL-11 and Cyclo- in deregulated lipid metabolism with disease development. sporin A were examined in a larger treatment group using In support of this, engulfment of excess lipid molecules by quantitative RT-PCR (Figure 5b). Nine rhIL-11-treated macrophages can result in the subsequent activation of this patients who responded to therapy were compared to eight cell type resulting in the release of proinflammatory cyto- Cyclosporin A-treated patients who responded to therapy. kines.18 Elevation of GATA3 and TNXA were observed as early as 1 Another functional cluster of genes can be associated with week following rhIL-11 treatment but these changes were the regenerative phenotype and hyperkeratosis observed in not observed in the Cyclosporin A treated patients (Figure psoriatic plaques. Genes in this cluster include KLK7 5b). Following termination of rhIL-11 administration these (stratum corneum chymotryptic enzyme), SCCA2 transcripts returned to levels observed prior to treatment (squamous cell carcinoma antigen-2), SPRR2A and SPRR1B and these changes in gene expression preceded a clinical (small proline-rich proteins), TGM1 (keratinocyte worsening of disease. Inhibition of HBP17 and CCNF (cyclin transglutaminase), TGM3 (transglutaminase E3) and mul- F) was observed in Cyclosporin A-treated responding tiple keratins (Keratin 16, 17, 2 and 6E), PI3 (elafin) and patients but was not observed in rhIL-11-treated responders S100A7 (psoriasin). Many of these genes are components of (Figure 5b). Again, upon termination of Cyclosporin A the stratified squamous epithelium, which are cross-linked administration these transcripts returned to levels observed by transglutaminases or involved in associated proteolytic in the lesions prior to treatment. These changes were not processes.19

www.nature.com/tpj Pharmacogenomic expression proﬁling of psoriasis JL Oestreicher et al 282

Figure 5 Quantitative RT-PCR analysis of treatment-responsive genes in a larger psoriasis patient population. Skin biopsies were obtained from 24 patients treated with rhIL-11 (n = 15) or Cyclosporin A (n = 9) at baseline, week 1, week 4, week 8 of treatment and 4 weeks following the end of treatment (week 12). In addition, a single biopsy of uninvolved skin was obtained at baseline. (a) Patients were divided into treatment responders (R; n = 17) and nonresponders (NR; n = 7) based on a combination of clinical and histopathological criteria.5 Quantitative RT-PCR was performed on the selected genes. Average fold change and standard deviations were calculated from baseline (lesion) for each group at the indicated time points and are presented as line graphs. Statistically signiﬁcant differences (P Ͻ 0.05) are represented by *. (b) Patients were divided into rhIL-11 responding (n = 9) and Cyclosporin A responding (n = 8) groups. Quantitative RT-PCR was performed on the indicated genes. Average fold change and standard deviations from baseline (lesion) for each group at the indicated time points were calculated and are presented as line graphs.

Finally, gene sets have also been identified related to an ogy of psoriasis and inflammation in general, the differen- activated immune response and inflammation. Genes in this tial gene expression profiles allow for the formation of new set include RAGE and the S100 family of RAGE ligands hypotheses to further explain the underlying pathology of which activate and are activated in part by NF-␬B.20 Several this multi-genic disease. For example, parallels can be drawn interferon inducible genes such as IFI56 and IFI27SEP also between genes that are differentially expressed in tumor fall into this category and are consistent with our previous cells as well as in psoriasis lesions such as CTSB (cathepsin findings that IFN-␥ and IL-12 are elevated in psoriatic B), CST6 (cystatin 6), and HBP17 (heparin binding protein lesions.5 In addition, reduced levels of the TH2-specific tran- 17). CTSB is a lysosomal cysteine protease that plays a role scription factor, GATA3, and elevated levels of IL-4R and the in antigen processing. In the neoplastic state, CTSB is over- T cell surface molecule, CD2, as well as molecules involved expressed in multiple tumor types and its localization corre- in leukocyte trafficking such as CD24 and CD47 were lates with areas of angiogenesis, inflammation and observed in this study. GATA3 is a transcription factor, necrosis.23 It is thought to play a role in the loss of contact abundantly found in T lymphocytes, that positively regu- inhibition of tumor cells. Expression of cystatin 6, an anti- lates CD4 cells to a type-2 phenotype.21 We have previously protease inhibitor of cathepsin B, is downregulated in tumor identified a relative deficiency of type-2 T cells in psoriatic samples and this downregulation has been correlated with patients.22 tumor progression.24 Overexpression of cathepsin B and In addition to supporting and enhancing the known etiol- downregulation of cystatin 6 in the epidermis of psoriasis

The Pharmacogenomics Journal Pharmacogenomic expression proﬁling of psoriasis JL Oestreicher et al 283

patients may contribute to deregulated keratinocyte growth reduction in apoptosis.32 Reduction of ID4 expression in the and differentiation. Likewise, HBP17, a heparin-binding pro- epidermis, through either inherited mutations or following tein for basic fibroblast growth factor (bFGF), is over- insult, may lead to deregulation of a number of downstream expressed in squamous cell carcinoma.25 Ribozyme-specific bHLH transcription factors that control keratinocyte targeting of HBP17 resulted in the decreased growth and growth. Another gene, S100A12 (calgranulin C) located on angiogenesis of xenograft tumors in mice. Interestingly, 1q21, is a soluble ligand for the RAGE receptor which has HBP17 is upregulated by p38 MAP kinase and inhibition of been implicated in activation of the proinflammatory NF- p38 results in a loss of HBP17 production, further demon- ␬B pathway.20 The RAGE gene maps to the PSORS1 locus on strating a role for the p38 pathway in the progression of 6p21.3. Finally, PRKMK3 is mitogen-activated protein kinase disease.26 A deregulated p38 MAP kinase pathway in pso- kinase 3 (MKK3), an upstream regulator of p38 MAPK.33 p38 riasis in addition to effects on the inflammatory process MAPK is activated by proinflammatory molecules such as would result in upregulation of HBP17 in the epidermis, TNF-␣ and plays a role in the inflammatory process. Block- resulting in increased angiogenesis and aberrant kera- ing p38 activity results in a reduction in the inflammatory tinocyte growth. Interestingly, despite these similarities, state.34 conversion of a psoriatic plaque to squamous cell carcinoma is rare, suggesting the presence of additional tumor sup- pressor pathways that are not deregulated in the lesions.27 Comparison of the Psoriasis Gene Set to Other This hypothesis can be tested with further functional Cutaneous Inflammatory Conditions analysis. A comparison of the 159 gene set to expression profiles from inflamed skin following a DTH response or tape stripping Association of Differentially Regulated Genes to has allowed us to construct genomic maps that resolve pso- Psoriasis-susceptibility Loci riasis into immune-associated vs reactive cellular compo- Understanding the multifactorial contribution of an nents. Some of the genes that comprise these maps are pso- unknown number of genes interacting with each other and riasis-specific while others are shared in common with with the environment has complicated the identification of general skin inflammatory conditions. Furthermore, some of individual risk-associated genes in multi-genic diseases in these genes are only found differentially regulated in either general and in psoriasis in particular. Interestingly, 27 genes DTH or tape-stripped samples but not both. Functional in the psoriasis gene set identified in the current studies map characterization of the genes shared and/or specific to these to one of the six known psoriasis-susceptibility loci. We and experimental conditions was consistent with the known others have hypothesized that deregulated genes that reside biology of the systems. For example, many of the differen- in chromosomal disease loci may play a fundamental role tially regulated transcripts shared between psoriasis and in the etiology of the disease. Autosomal or somatic DTH samples are immune-specific genes expressed in vari- mutations in regions of these 27 genes that affect transcrip- ous cell types of the immune system such as monocytes, B tional regulation or message stability may contribute to cells and T cells or are induced by proinflammatory immun- increased risk of developing psoriasis. In addition, some of omodulatory cytokines such as IL-12 or IFN-␥. Genes shared these genes code for transcription factors whose aberrant between psoriasis and tape stripped samples are expressed expression may serve to deregulate other downstream path- predominantly in cells of epithelial origin, such as kera- ways that further contribute to development of the disease. tinocytes, consistent with the shared wound healing compo- This may contribute to observed epistatic interactions nent of both tissue types. Psoriasis-specific genes crossed the between different psoriasis disease loci.28 In addition, spectrum of cell types such as neutrophils, leukocytes and environmental factors that impact regulation of these tran- keratinocytes. These genes also crossed the spectrum of scription factors may also contribute to disease development cellular functions from roles in metabolism (ATP1AL1, and progression. These findings could explain the difficulty HAL), transcriptional regulation (ID1, ID4), immune modu- in identifying disease alleles in the coding region of candi- lation and inflammation (S100A12, IFI56) and cornified date genes within psoriasis disease loci.29 Furthermore, a envelope development and keratinocyte growth regulation nonrandom cluster of autoimmune susceptibility loci (TGM1, GJB2, SPRR2A). resides on 6p21 and suggests that the genes identified here Genes activated in an antigen-specific manner during a may play a role in other autoimmune diseases as well.30 DTH response may play a role in a variety of inflammatory Expression analysis of other complex inflammatory/ conditions such as rheumatoid arthritis and Crohn’s disease autoimmune diseases will validate this approach. while genes activated during a tape-strip reaction may play These 27 genes have a variety of cellular functions that a more fundamental role in generalized wounding of the support a role for the protein products of these transcripts skin. In support of this, a differentially regulated gene ident- in the etiology of the disease. For example, ID4, located in ified in the current studies as overexpressed during the DTH the HLA locus on 6p21.3, is a dominant/negative regulator response, MMP-12, has previously been shown through of the basic helix-loop-helix (bHLH) family of transcription microarray analysis to be deregulated in Crohn’s and RA factors and serves as a general antagonist of cellular differen- tissue.35 Other genes overexpressed in tape-stripped samples, tiation and proliferation in a variety of cell lineages.31 Cellu- such as HBP17, have been shown to play a role in angiogen- lar injury results in a downregulation of ID4 leading to a esis and wound healing.25

www.nature.com/tpj Pharmacogenomic expression proﬁling of psoriasis JL Oestreicher et al 284

Pharmacogenomic Profiling Identifies Disease Gene cular fragility and poor wound healing.38 Cyclosporin A Candidates and Novel Pathways for Pharmacological does not impact the expression of these transcript levels; Intervention rather it specifically upregulates transcript levels for several Changes in gene expression can be causally related to the genes such as CCNF and HBP17. Induction of HBP17 is regu- disease state or may be a consequence of the disease process. lated by the p38/JNK pathway that is blocked by Cyclospo- A general problem with genome-wide differential expression rin A.39 Understanding the impact of different therapeutic analysis is the difficulty in differentiating between these two treatments on these pathways may help differentiate possibilities. For example, the gene DEFB2, a defensin, is an between desirable (specific and efficacious) and undesirable antimicrobial protein induced during inflammation whose (toxic) drug activities. expression levels are increased as much as 200 fold in psori- These studies present for the first time a genome-wide atic lesions.36 DEFB2 may be upregulated as an effect of the expression response to pharmacological intervention in a wound healing process. However, recent studies indicate longitudinal study and indicate that this method has utility that defensins are also chemoattractive for lymphocytes and in prioritizing disease-associated genes for further functional may contribute to adaptive immunity by mobilizing T cells genomic analysis. Pharmacogenomic gene expression analy- and dendritic cells.37 Thus, descriptions of the biological sis of a complex multi-genic disease such as psoriasis has the role of these differentially expressed genes are not sufficient potential to identify novel regulators of the disease-state. to determine candidates for further functional analysis. This One hundred and fifty-nine genes were identified that are is particularly the case when the differentially expressed differentially expressed in the disease state. Of these, 41 gene is relatively uncharacterized. To address this issue in transcripts are modulated rapidly by inhibitors of the cal- an in vivo human setting, we analyzed psoriasis biopsy tissue cineurin and NF-␬B pathways and 12 of these genes map to from patients following pharmacological intervention. known psoriasis susceptibility loci. Further analysis of these Drugs such as rhIL-11 and Cyclosporin A are known to genes will aid in elucidating the etiology of psoriasis and block the NF-␬B and calcineurin/p38 pathways, respectively. perhaps other inflammatory diseases such as rheumatoid We have hypothesized that changes in gene expression that arthritis, multiple sclerosis and Crohn’s disease where NF- precede clinical improvement may play a more causal role ␬B and Type I cytokines also play a predominant role. They in disease progression as opposed to genes whose expression may also serve to identify novel therapeutic intervention changes mirror clinical improvement or do not change points or clinical markers of drug efficacy. despite clinical improvement. This approach identified a subset of 41 differentially regulated genes that returned to METHODS normal or uninvolved levels following therapeutic inter- Study Design and Psoriasis Patient Entry Criteria vention with rhIL-11 or Cyclosporin A at time points that These studies were approved by The Rockefeller University preceded clinical improvement. Members of this group Hospital Institutional Review Board, New York, NY. Thirty included 12 genes such as ID4, HBP-17, KRT16, S100A2, adult patients (16 female, 14 male) with chronic psoriasis S100A9, S100A12, GNA15, MTX, PRKMK3, and SCYA2 that vulgaris affecting at least 10% of their body surface area, as all localized to psoriasis susceptibility loci. Sequence analysis measured by a combination of clinical and histological para- of these genes in families with a history of psoriasis may meters as previously described, signed informed consent help to identify specific affected alleles at susceptibility loci. forms and were enrolled in this study.5 Patients had stable Pharmacogenomic analysis also can serve to identify path- chronic disease as evidenced by a median duration of disease ways suitable for pharmacological intervention. Of parti- of 24 years (range 4–44 years). Patients were primarily of cular interest is the finding that 11 IFN-␥-regulated genes are Caucasian descent with a mean age of 41 years (range of 23– aberrantly expressed in psoriasis lesions. Thus, agents that 63 years). target the IFN-␥ or Type I cytokine pathway may have future Global clinical assessment of each patient before, during therapeutic potential. For example, neutralization of the and after therapy was based on the Psoriasis Area and Sever- IFN-␥ inducer, IL-12, or inhibition of IL-12 production from ity Index (PASI) as described previously.2 The grading system mature Langerhans cells may provide effective treatment. In used to measure local clinical disease activity was the Pso- addition, blockade of Type I cytokine production from acti- riasis Severity Index (PSI). The PSI index grades individual vated CD8+ Tc1 cells or CD4+ Th1 cells may also be effective. psoriasis lesions for scale, erythema, and induration on a 0– This approach has also allowed us to begin to identify in 6 point scale for each parameter (0, absent; 1, trace; 2, mild; vivo pathways that are modified in a drug-specific manner. 3, mild to moderate; 4, moderate; 5, moderate to severe; and For example, 11 genes were differentially regulated between 6, severe). The final score sums the individual parameters rhIL-11- and Cyclosporin A-treated patients. Blockade of the for a range of 0–18.40 NF-␬B pathway by rhIL-11 resulted in specific upregulation Patients were treated on an outpatient basis with rhIL-11 of a number of genes including the type II helper T cell tran- (n = 21) or Cyclosporin A (n = 9) for 8 weeks. The dose and scriptional regulator GATA3 and the lymphocyte differen- schedule selection of rhIL-11 (2.5 ␮gkg−1 day−1,5␮gkg−1 tiation and maturation molecule CRIP1. Also, rhIL-11 upreg- day−1, and 1 mg once weekly and 2 mg once weekly) and ulates the extracellular matrix protein, TNXA. TNXA Cyclosporin A (5 ␮gkg−1 day−1) was as previously described.5 deficiency has been associated with Ehlers–Danlos syn- At the start of enrollment, a psoriatic plaque was chosen for drome that is characterized by skin hyperextensibility, vas- weekly assessment of lesion severity scores by the study

The Pharmacogenomics Journal Pharmacogenomic expression proﬁling of psoriasis JL Oestreicher et al 285

coordinator. A 6-mm punch biopsy was taken of this included in vitro synthesized transcripts of 11 bacterial genes lesional skin prior to rhIL-11 or Cyclosporin A treatment in each hybridization reaction as described by Hill et al.42 and at weeks 1, 4, 8 and 12 during rhIL-11 or Cyclosporin The abundance of these transcripts ranged from 1:300000 A treatment. In addition, prior to initiation of treatment, a (3 ppm) to 1:1000 (1000 ppm) stated in terms of the number 6 mm punch biopsy was taken from uninvolved skin at a of control transcripts per total transcripts. As determined by location of the patient’s choosing. Biopsies were equally div- the signal response from these control transcripts, the sensi- ided for immunohistochemical analysis and for RNA prep- tivity of detection of the arrays ranged between ෂ1:300000 aration. Normal skin was obtained from healthy volunteers. and 1:100000 copies per million. Labeled probes were Six-micrometer cryostat sections were made from frozen denatured at 99°C for 5 min and then 45°C for 5 min and biopsies, were reacted with antibodies to CD3, CD8, keratin hybridized to oligonucleotide arrays comprised of over 7000 16, Ki67, ICAM-1, or HLA-DR and processed for immunohis- human genes (HuGeneFL, Affymetrix, Santa Clara, CA, tochemistry as previously described.5 Computer-assisted USA). Arrays were hybridized for 16 h at 45°C with rotation image analysis was used to quantify epidermal thickness and at 60 rpm. After hybridization, probes were removed and the the number of CD3+, CD8+, or Ki67+ cells in tissue sections cartridges washed extensively with 6 × SSPET and stained as described.5 with phycoerythrin coupled to streptavidin as described.41

DTH and Tapestrip Induction Gene Expression Data Reduction Patients were sensitized with 2 mg of dinitrochlorobenzene Data analysis was performed on raw fluorescent intensity (DNCB, 0.2 ml of 1% solution) to an area of normal skin. values using GENECHIP 3.2 software (Affymetrix). Gene- The DNCB was placed into a 2-cm hole in a Teflon template Chip 3.2 software uses an algorithm to calculate the likeli- and slowly evaporated under a steady stream of nitrogen hood as to whether a gene is ‘absent’ or ‘present’ as well as gas. The area was covered with a sterile bandage. Patients a specific hybridization intensity value or ‘average differ- returned in 2 weeks for challenge with 50 ␮g DNCB (0.1 ml ence’ for each transcript represented on the array. The algor- of 0.05% solution) in the same fashion as described above. ithms used in these calculations are described in the Affyme- If no reaction occurred within 96 h, a second challenge was trix GeneChip Analysis Suite User Guide (Affymetrix). The placed with 100 ␮g (0.2 ml of 0.05% solution). A 6-mm full ‘average difference’ for each transcript was normalized to thickness punch biopsy was taken from the middle of the ‘frequency’ values according to the procedures of Hill et al.42 erythematous plaque (DTH-reaction) for histological and This was accomplished by referring the average difference RNA preparation 72 h following challenge. For tape strip- values on each chip to a calibration curve constructed from ping, a 1 × 3 inch area of normal skin was repetitively the average difference values for the 11 control transcripts stripped with adhesive tape as previously described.11 When with known abundance that were spiked into each hybridiz- a smooth glistening erythematous epidermis was reached, ation solution. This process also served to normalize the stripping was stopped; the wound was covered with a between arrays.42 sterile bandage. After 24 h, a full thickness 6 mm punch Specific transcripts were evaluated further if they met the biopsy was taken for histological and RNA preparation. following criteria. First, genes that were designated ‘absent’ by the GENECHIP 3.2 software in all samples were excluded Isolation of RNA and Preparation of Labeled Microarray from the analysis; 3320 (49%) of the transcripts met this Targets criteria. Second, in comparisons of transcript levels between Total RNA was isolated from one-half of a 6-mm full thick- arrays, a gene was required to be present in at least 50% of ness skin biopsy using the RNeasy mini kit (Qiagen, Valen- the arrays comprising one or more groups. Third, for com- cia, CA, USA). Labeled target for oligonucleotide arrays was parisons of transcript levels between groups, a t-test was prepared using a modification of the procedure described by applied to identify a subset of transcripts that had a signifi- Lockhart et al.41 Two micrograms total RNA were converted cant (P Ͻ 0.05) difference in frequency values. Fourth, aver- to cDNA by priming with an oligo-dT primer containing a age fold changes in frequency values across this statistically T7 DNA polymerase promoter at the 5Ј end. The cDNA was significant subset of genes was required to be 2-fold or used as the template for in vitro transcription using a T7 greater. These criteria were established based on numerous DNA polymerase kit (Ambion, Woodlands, TX, USA) and replicate experiments performed in our lab that estimated biotinylated CTP and UTP (Enzo, Farmingdale, NY, USA). the intra-assay reproducibility (data not shown). Quantitat- Labeled cRNA was fragmented in 40 mM Tris-acetate pH 8.0, ive RT-PCR was also performed on a select subset of genes 100 mM KOAc, 30 mM MgOAc for 35 min at 94°Cinafinal to further validate these criteria (data not shown). volume of 40 ␮l. Quantitative Reverse Transcriptase-polymerase Chain Hybridization to Affymetrix Microarrays and Detection Reaction (RT-PCR) of Fluorescence Quantitative RT-PCR was performed as previously Ten micrograms of labeled target were diluted in 1 × MES described.5 RNA samples from patient biopsies were treated buffer with 100 ␮gml−1 herring sperm DNA and 50 ␮gml−1 with 10 U of RQ1 DNase I (Promega, Madison, WI, USA) for acetylated BSA. To normalize arrays to each other and to 30 min at 37°C. rTth DNA polymerase was used to reverse estimate the sensitivity of the oligonucleotide arrays, we transcribe and amplify 10 ng of total RNA in a single tube

www.nature.com/tpj Pharmacogenomic expression proﬁling of psoriasis JL Oestreicher et al 286

assay using the TaqMan EZ RT-PCR kit (ABI, Foster City, CA, ance in data analysis and experimental design. This work was supported USA) with gene specific sense and anti-sense primers and in part by grant GM-42461, from the National Institutes of Health. This a probe fluorescently labeled at the 5Ј end with 6-carboxy- work has previously been presented in abstract form at the 61st Annual Meeting of the Society for Investigative Dermatology, Chicago, IL, May fluorescein (6-FAM). Amplification was performed using the 10–14, 2000. ABI Prism 7700 sequence detection system as described by the manufacturer (ABI). Primers and fluorescently labeled probes were generated using Primer Express software (ABI). DUALITY OF INTEREST None declared. Sequence-specific amplification was detected as an increased fluorescent signal of 6-FAM during the amplification cycle. Quantitation of gene-specific message levels was based on a REFERENCES comparison of the fluorescent intensity in the unknown 1 Greaves MW, Weinstein GD. Treatment of psoriasis. NEJ Med 1995; 332; 581–588. mRNA sample to the fluorescent intensity from a standard 2 Gottlieb AB et al. Studies of the effect of cyclosporine in psoriasis in curve of known mRNA levels. Amplification of the gene for vivo: combined effects on activated T lymphocytes and epidermal human acidic ribosomal protein (HARP) was performed on regenerative maturation. J Invest Dermatol 1992; 98: 302–309. all samples tested to control for variations in RNA 3 Gottlieb SL et al. Response of psoriasis to a lymphocyte-selective toxin (DAB IL-2) suggests a primary immune but not keratinocyte patho- 43 389 amounts. All genes were subsequently normalized to HARP genic basis. Nature Med 1995; 1: 442–447. mRNA levels. Levels of gene-specific messages were graphed 4 Bachelez H et al. Treatment of recalcitrant plaque psoriasis with a as normalized message units as determined from the stan- humanized non-depleting antibody to CD4. J Autoimmun 1998; 11: dard curve. A no template control was included in each 53–62. 5 Trepicchio WL et al. Interleukin-11 therapy selectively downregulates amplification reaction to control for contaminating tem- type I cytokine proinflammatory pathways in psoriasis lesions. J Clin plates. Invest 1999; 104: 1527–1537. 6 Nickoloff BJ. The immunologic and genetic basis of psoriasis. Arch Dermatol 1999; 135: 1104–1110. Statistical Analysis 7 Bhalerao J, Bowcock AM. The genetics of psoriasis: a complex disorder Quantitative measures of gene expression changes between of the skin and immune system. Hum Molec Genet 1998; 7: 1537– groups were statistically evaluated using the JMP 3.2.6 stat- 1545. istical discovery software package (SAS Institute, Cary, NC, 8 Eisen MB, Spellman PT, Brown PO, Botstein D. Cluster analysis and display of genome-wide expression patterns. Proc Natl Acad Sci USA USA). Unsupervised hierarchical clustering of samples into 1998; 95: 14863–14868. groups on the basis of similarity of their expression profiles 9 Schalkwijk J, Chang A, Janssen P, de Jongh GJ, Mier PD. Skin-derived was performed using the procedure of Eisen et al.8 Self antileucoproteases (SKALPs): characterization of two new elastase organizing maps (SOM) were constructed using methods inhibitors from psoriatic epidermis. Brit J Dermatol 1990; 122: 631– 15 641. described by Tamayo et al. Nearest neighbor prediction 10 Madsen P, Rasmussen HH, Leffers H, Honore B, Celis JE. Molecular analysis and supervised cluster analysis was performed using cloning and expression of a novel keratinocyte protein (psoriasis-asso- metrics defined by Golub et al.44 The Pearson correlation ciated fatty acid-binding protein [PA-FABP]) that is highly up-regulated coefficient was chosen to measure the degree of overall simi- in psoriatic skin and that shares similarity to fatty acid-binding proteins. J Invest Dermatol 1992; 99: 299–305. larity between expression profiles. For hierarchical clus- 11 Rivas MV et al. Identification of aberrantly regulated genes in diseased tering, SOM analysis, and nearest neighbor prediction analy- skin using the cDNA differential display technique. J Invest Dermatol sis, data were log transformed and normalized to have a 1997; 108: 188–194. mean value of zero and a variance of one. Differences 12 Trepicchio WL, Wang L, Bozza M, Dorner AJ. IL-11 regulates macrophage effector function through inhibition of Nuclear Factor-kB. J between paired sites were analyzed using 2-tailed paired t- Immunol 1997; 159: 5661–5670. test. An unpaired t-test was used to compare normal to unin- 13 Matsuda S, Koyasu S. Mechanisms of action of cyclosporine. Immuno- volved skin. In all comparisons a P value Ͻ0.05 was used to pharmacology 2000; 47: 119–125. indicate statistical significance. 14 Kiani A, Rao A, Aramburu J. Manipulating immune responses with immunosuppressive agents that target NFAT. Immunity 2000; 12: 359–372. Note Added in Proof 15 Tamayo P et al. Interpreting patterns of gene expression with self- Since submission of this manuscript for publication, Bow- organizing maps: methods and application to hematopoietic differen- cock et al have also described differential expression profile tiation. Proc Natl Acad Sci USA 1999; 96: 2907–2912. 16 Weinstein GD, Kreuger JG. Overview of psoriasis. In Weinstein GD, analysis of psoriasis lesions (Bowcock AM, Shannon W, Du Gottlieb AB (eds). Therapy of Moderate to Severe Psoriasis. National Pso- F, Duncan J, Cao K, Aftergut K, Catier J, Fernandez-Vina MA riasis Foundation, Portland, OR, 1994, pp 1–22. and Merter A. Insights into psoriasis and other inflamma- 17 De Laurenzi V et al. Sjogren-Larsson syndrome is caused by mutations tory diseases from large-scale gene expression studies. Hum in the fatty aldehyde dehydrogenase gene. Nature Genet 1996; 12: 52–57. Mol Genet 2001; 10: 1793–1805. 18 Jackson SK. Role of lipid metabolites in the signalling and activation of macrophage cells by lipopolysaccharide. Progr Lipid Res 1997; 36: ACKNOWLEDGEMENTS 227–244. 19 Kim I-G et al. Structure and organization of the human transglutamin- The authors would like to thank the many patients and volunteers who ase 3 gene: evolutionary relationship to the transglutaminase family. donated biopsy samples for these studies. Expert technical assistance J Invest Dermatol 1994; 103: 137–142. was provided by Mary Bozza, Yared Tekabe, John P Kim and Lori Casci- 20 Tanaka N et al. The receptor for advanced glycation end products is otti. We would also like to thank Maryann Whitley, William Mounts, induced by the glycation products themselves and tumor necrosis fac- John Ryan and Gene Brown for many thoughtful discussions and assist- tor-alpha through nuclear factor-kappa B, and by 17beta-estradiol

The Pharmacogenomics Journal Pharmacogenomic expression proﬁling of psoriasis JL Oestreicher et al 287

through Sp-1 in human vascular endothelial cells. J Biol Chem 2000; 32 Andres-Barquin PJ, Hernandez MC, Israel MA. Id4 expression induces 275: 25781–25790. apoptotis in astrocytic cultures and is down-regulated by activation of 21 Rincon M, Flavell RA. Reprogramming transcription during the differ- the cAMP-dependent signal transduction pathway. Exp Cell Res 1999; entiation of precursor CD4+ T cells into effector Th1 and Th2 cells. 247: 347–355. Microbes & Infection 1999; 1:43–50. 33 Enslen H, Raingeaud J, Davis RJ. Selective activation of p38 mitogen- 22 Austin LM, Ozawa M, Kikuchi T, Walters IB, Krueger JG. The majority activated protein (MAP) kinase isoforms by the MAP kinase kinases of epidermal T cells in Psoriasis vulgaris lesions can produce type 1 MKK3 and MKK6. J Biol Chem 1998; 273; 1741–1748. cytokines, interferon-gamma, interleukin-2, and tumor necrosis factor- 34 Herlaar E, Brown Z. p38 MAPK signalling cascades in inflammatory alpha, defining TC1 (cytotoxic T lymphocyte) and TH1 effector popu- disease. Mol Med Today 1999; 5: 439–447. lations: a type 1 differentiation bias is also measured in circulating 35 Heller RA et al. Discovery and analysis of inflammatory disease-related blood T cells in psoriatic patients. J Invest Dermatol 1999; 113: 752– genes using cDNA microarrays. Proc Natl Acad Sci USA 1997; 94: 759. 2150–2155. 23 Hughes SJ et al. A novel amplicon at 8p22–23 results in overexpression 36 Liu L et al. Structure and mapping of the human beta-defensin HBD- of cathepsin B in esophageal adenocarcinoma. Proc Nat Acad Sci USA 2 gene and its expression at sites of inflammation. Gene 1998; 222: 1998; 95: 12410–12415. 237–244. 24 Sotiropoulou G, Anisowicz A, Sager R. Identification, cloning, and 37 Yang D et al. Beta-defensins: linking innate and adaptive immunity characterization of cystatin M, a novel cysteine proteinase inhibitor, through dendritic and T cell CCR6. Science 1999; 286: 525–528. down-regulated in breast cancer. J Biol Chem 1997; 272: 903–910. 38 Burch G et al. Tenascin-X deficiency is associated with Ehlers-Danlos 25 Czubayko F et al. A secreted FGF-binding protein can serve as the angi- syndrome. Nat Genet 1997; 17:5–7. ogenic switch in human cancer. Nature Med 1997; 3: 1137–1140. 39 Harris V et al. Induction of the angiogenic modulator fibroblast growth 26 Harris VK et al. Induction of the angiogenic modulator fibroblast factor-binding protein by epidermal growth factor is mediated growth factor-binding protein by epidermal growth factor is mediated through both MEK/ERK and p38 signal transduction pathways. J Biol through both MEK/ERK and p38 signal transduction pathways. J Biol Chem 2000; 275: 10802–10811. Chem 2000; 275: 10802–10811. 40 Coven TR et al. Narowband UV-B produces superior clinical and histo- 27 Nickoloff B. Creation of psoriatic plaques: the ultimate tumor sup- pathological resolution of moderate-to-severe psoriasis in patients pressor pathway. A new model for an ancient T-cell-mediated skin dis- compared with broadband UV-B. Arch Dermatol 1997; 133: 1514– ease. Viewpoint. J Cutan Pathol 2001; 28:57–64. 1522. 28 Veal C et al. Identification of a novel psoriasis susceptibility locus at 1p 41 Lockhart DJ et al. Expression monitoring by hybridization to high-den- and evidence of epistasis between PSORS1 and candidate loci. J Med sity oligonucleotide arrays. Nature Biotechnol 1996; 14: 1675–1680. Genet 2001; 38:7–13. 42 Hill AA, Hunter CP, Tsung BT, Tucker-Kellogg G, Brown EL. Genomic 29 Nair RP et al. Localization of psoriasis-susceptibility locus PSORS1 to a analysis of gene expression in C-elegans. Science 2000; 290: 809–812. 60-kb interval telomeric to HLA-C. Am J Hum Genet 2000; 66: 1833– 43 van Ruissen F, Le M, Carroll JM, van der Valk PG, Schalkwijk J. Differen- 1844. tial effects of detergents on keratinocyte gene expression. J Invest 30 Becker KG et al. Clustering of non-major histocompatibility complex Dermatol 1998; 110: 358–363. susceptibility candidate loci in human autoimmune diseases. Proc Natl 44 Golub TR et al. Molecular classification of cancer: class discovery and Acad Sci USA 1998; 95: 9979–9984. class prediction by gene expression monitoring. Science 1999; 286: 31 Pagliuca A, Bartoli PC, Saccone S, Della Valle G, Lania L. Molecular 531–537. cloning of ID4, a novel dominant negative helix-loop-helix human gene on chromosome 6p21.3-p22. Genomics 1994; 27: 200–203.