Resolved psoriasis lesions retain expression of a subset of disease-related genes

Abstract

Psoriasis is a complex inflammatory disease that usually heals without visible scarring. Histological evaluation often suggests complete resolution, but reversal of genomic disease-associated alterations has not yet been defined. Gene expression profiling was used to determine the extent to which the psoriasis genes were reversed after 3 months of etanercept in patients who responded to treatment. We reviewed the histology, leukocyte counts, and PCR data for inflammatory genes, to compare recovery of these parameters and the genomic studies. Many cellular markers do return close to non-lesional levels, although five inflammatory genes did not improve >75% (IL-12p35, MX1, IL-22, IL-17, IFNγ). Psoriasis-related genes with <75% improvement were defined as comprising a “residual disease genomic profile”, composed of 248 probesets. Genes of interest in psoriasis tissue that did not return to baseline included LYVE-1, WNT5A, RAB31 and AQP9. It appears that even though the epidermal reaction in psoriasis is fully resolved, inflammation, as defined by expression of key cytokines and chemokines, is not completely resolved in treated lesions. We also found that structural cells of the skin continued to express molecular alterations, and that some subtle features of skin structure, e.g., lymphatics, were not fully normalized with treatment.

Introduction

Psoriasis is a complex inflammatory disease with a characteristic clinical and histological phenotype. The non-affected skin of psoriasis patients does not appear to have any distinct clinical features, and effective treatment returns lesional skin (LS) to a non-lesional (NL) state, i.e. skin that appears virtually normal. Our past studies with several different therapies for psoriasis suggest that successful treatment is correlated with reduced epidermal thickness and reductions in inflammatory cellular infiltrates and gene expression (Chamian et al., 2005; Zaba et al., 2007). We have shown that in patients responding to therapy, the cellular infiltrate and expression of selected inflammatory genes in affected LS skin return to NL levels at the end of treatment. The clinical appearance, combined with histological and molecular studies, indicated that psoriasis skin lesions were reversed without obvious differences to NL skin. Thus psoriasis can be viewed as “reversible” when the inflammation is removed from the skin.

Although psoriasis may recur in new areas, it is well recognized that often recurs in the same locations on the body, which may suggest that the skin where psoriasis lesions have been is not fully “healed”. There may be some residual changes in treated lesions that predispose to recurrence of lesions in the same locations. However, histopathology does not give a good indication of cell function, and PCR studies have been conducted on a small group of selected genes. The introduction of microarray profiling has given a more comprehensive platform to study disease-specific genes, and how they respond to treatment.

We wanted to determine the extent to which the genomic psoriasis phenotype, as defined by the differentially expressed genes (DEGs) between LS and NL skin, was reversed by etanercept treatment in responding patients. At the end of the three-month clinical trial period, we analyzed how closely the psoriasis transcriptome returned to a NL state in responders. We were interested to see if psoriasis was fully reversible at the molecular level when the skin was clinically clear of disease. Although many DEGs were normalized in post-treatment biopsies, 248 genes did not return to baseline (NL) after three months of treatment. We have termed this residual genomic expression that did not return to NL levels “residual disease genomic profile (RDGP)”. This gene-set hasimportant implications for the biology and treatment of psoriasis, and it could serve as a means to study progressive alterations in skin structure and function as a consequence of temporal progression or severity of skin lesions.

Results

Resolution of psoriasis lesions

Much of clinical disease pathology in psoriasis is related to altered epidermal growth and differentiation. In this analysis, we sought to relate epidermal improvement to modulation of disease-related genes by etanercept. In the group of patients that respond to etanercept, the epidermis was almost fully recovered (Table 1). The epidermal thickness improved 92% in average, Ki67 (a marker of cellular proliferation) improved 119%, K16 mRNA improved by 100%, and all cases were K16 negative by immunohistochemistry (Zaba et al., 2007). Hence, to a pathologist using an H&E stain or growth-related markers, the epidermis would appear histologically “normal”.

Table 1.

Improvement of Cell Counts and RT-PCR data for 11 responding patients.

Symbol		LS vs NL	W12 vs 0	% Improvement
EPIDERMIS
Thickness.Original		245.67	−224.51	91.39
Ki67 cell counts		122.13	−145.00	118.7
K16 mRNA		5.06	−5.05	99.86
K16 histology				100% were K16−
CELLS
CD3 (T-cells)	Total	145.12	−138.59	95.5
	Epidermis	51.27	−54.29	105.91
	Dermis	80.87	−84.29	104.24
CD8 (Cytotoxic T-cells)	Total	126.83	−96.50	76.80
	Epidermis	60.33	−54.67	90.80
	Dermis	65.67	−41.83	63.71
CD11c (Myeloid DCs)	Total	216.59	−197.47	88.4
	Epidermis	47.69	−47.41	99.42
	Dermis	168.88	−144.06	85.31
CD83 (Mature DCs)	Total	7.24	−8.76	121.14
	Epidermis	1.76	−2.35	133.33
	Dermis	5.47	−6.41	117.20
CD163 (Macrophages)	Total	68.94	−66.82	96.93
	Epidermis
	Dermis	64.13	−66.82	104.19
CD206 (DCs, macs)	Total	167.12	−140.41	84.02
	Epidermis	9.87	−8.76	88.83
	Dermis	145.07	−131.65	90.75
DCLAMP (Mature DC)	Total	48.59	−48.71	100.24
	Epidermis	5.13	−4.65	90.53
	Dermis	42	−44.06	104.90
GENES
LTA.1		0.37	−2.34	623.75
IL-4		−1.15	1.80	156.66
CCL3		3.17	−4.75	149.87
CCL4		2.94	−3.92	133.26
AREG		2.16	−2.86	132.56
IL-19		5.35	−6.11	114.27
IL-6		2.88	−3.13	108.61
IL-1β		4.04	−4.08	101.03
p40		3.83	−3.55	92.82
IL-8		6.75	−6.07	89.82
iNOS		6.37	−5.54	87.00
IL-20		3.72	−3.13	84.22
Beta.defensin		4.08	−3.34	81.81
IL23p19		3.09	−2.49	80.77
CCL20		3.31	−2.67	80.72
IL12p35		−1.58	1.07	67.43
MX-1		3.67	−2.42	66.06
IL-22		3.60	−2.34	65.09
IL-17		6.16	−3.92	63.66
IFN-γ		2.26	−1.41	62.22

^*cell counts were summarized across patients for the total, epidermal, and dermal counts so the average of the total does not add up to the average of the epidermis and dermis.

In order to determine the extent to which the inflammatory infiltrate and specific inflammatory genes improved with treatment, we calculated the mean improvement in cell counts (Table 1). CD3⁺ T-cells almost completely returned to baseline (95.5% improvement); CD8⁺ T-cells were 77% improved; CD11c⁺ myeloid dendritic cells (DCs) were 88% improved; and CD83⁺ mature DCs were 121% improved; while CD163⁺ macrophages were 97% improved. CD206 is on both DCs and macrophages and improved 84% and DCLAMP⁺ mature DCs were 100% improved. Overall, resolution of inflammatory cell populations in the epidermis was often higher than the dermis, although both were improved greater than 75% for all populations studied, except dermal CD8⁺ cells, which improved only 64%.

In Table 1, improvement of selected genes by RT-PCR expression in responding patients is presented. Five genes that did not improve by greater than 75% were important pro-inflammatory genes, IL-12p35, MX1, IL-22, IL-17, IFN-γ (range of improvement was 62–67%). Hence, while within resolved skin pathologic T-cells were more than 95% reduced based on CD3⁺ cell counts, dermal CD8⁺ cells were not fully resolved, and some T-cell activation may be ongoing within residual infiltrates. Overall, we interpret these findings to indicate that the epidermis can improve almost completely even in the context of some residual inflammatory gene expression.

Resolution of psoriasis disease-related genes

Our first publication of the genomic data from this group of patients (n=15) described the genomic effects of etanercept in responders and non-responders (Zaba et al., 2009b). We have also defined the psoriasis transcriptome using these patients, and compared with other published psoriasis DEG lists (Suárez-Fariñas et al., 2010). Our psoriasis-related DEGs contained 732 up-regulated probesets (579 known unique genes) and 890 down-regulated probesets (703 genes) (Figure 1a). Here, we considered the psoriasis phenotype as genomically defined by this DEG and used them to determine the extent to which the psoriasis phenotype was reversed after 3 months of etanercept in 11 patients who responded to treatment. Indeed, our analysis showed that a large set of disease-related genes was improved by >90%, and some genes had improvement even beyond NL levels. A selection of improved psoriasis genes is presented in Table 2, and full list in Supplementary Table 1. Many inflammatory genes such as granzyme B, IL-8 and S100A8 improved by 91%. Structural genes, such as keratin 6B and 16, as well as epidermal protein lipocalin 2 (LCN2) also improved more than 96%.

Figure 1. Residual Disease Genomic Profile.

a) Treatment response evaluated in the set of psoriasis genes (at fold change greater than 2 and a false discovery rate of 0.05). Genes that improved > 75% after 12 weeks of treatment were considered part of “molecular resolution” while those with < 75% form the “molecular remnant”. b) Example of treatment response of several genes. The x-axis refers to time during the clinical trial, with NL considered at 0, LS levels showing increase or decrease compared to NL levels, and genes that at week 12 attain this line have an improvement of 100%. Up-regulated psoriasis genes WNT5, TCRβ1 and RAB31 improved by < 75% (red lines), while AQP9 and LYVE1 were down-regulated that improved < 75% (green lines). STAT1, showed a complete remission of pathological expression (black line).

Table 2.

Selected Disease Genes that have more than 90% Improvement.

Symbol	Description	LS vs NL	W12 vs 0	Improvement (%)
GZMB	Granzyme B	1.75	−1.83	91.27
IL8*	Interleukin 8	5.85	−6.02	91.87
S100A8	S100 Calcium binding protein A8	1.73	−1.51	91.96
CCL4	Chemokine (C-C motif) ligand 4	1.38	−1.62	92.25
CCL18	Chemokine (C-C motif) ligand 18	1.48	−1.79	92.43
IL19	Interleukin 19	1.43	−2.08	92.50
IL1RN	Interleukin 1 receptor antagonist	1.89	−1.94	94.55
S100A9	S100 Calcium binding protein A9	3.90	−3.59	94.59
IL1B*	Interleukin 1, beta	2.61	−3.22	95.94
IFI16*	Interferon, gamma-inducible protein 16	1.08	−0.90	96.31
KRT6B	Keratin 6B	1.52	−1.74	96.68
KRT16	Keratin 16	3.69	−4.04	96.74
CCL20	Chemokine (C-C motif) ligand 20	2.79	−3.71	96.81
SELE	Selectin E	1.80	−2.39	98.53
LAMP3	Lysosomal associated membrane protein 3	1.97	−2.19	99.95
LCN2	Lipocalin 2	5.65	−5.82	102.35
STAT1	Signal transducer and activator of transcription 1	1.41	−1.35	102.62
TNFSF10	Tumor necrosis factor (ligand) superfamily 10	1.41	−1.14	102.73
CCNF	Cyclin F	1.17	−1.25	104.33
MKi67*	Antigen identified by monoclonal antibody	1.66	−1.78	105.43
ILF9	Interleukin 1 family, member 9	4.73	−5.09	106.97
CXCL9	Chemokine (C-X-C motif) ligand 9	1.85	−2.04	107.97
CDK5R1	Cyclin-dependent kinase 5, regulatory subunit	2.41	−2.55	109.19
TNFRSF21	Tumor necrosis factor receptor superfamily 21	2.00	−2.37	109.55
TK1	Thymidine kinase 1, soluble	1.18	−1.34	110.19
CCNB1	Cyclin B1	2.55	−2.89	110.80
CCNB2	Cyclin B2	1.49	−1.86	111.88
STAT2	Signal transducer and activator of transcription 2	1.59	−1.36	115.37
S100A7	S100 Calcium binding protein A7	1.78	−1.57	115.68
STAT3	Signal transducer and activator of transcription 3	1.40	−1.69	121.52

^*when multiple probesets are present, the smallest improvement is reported.

Resolution of gene-sets representing specific cells (Guttman-Yassky et al., 2009; Haider et al., 2007; Zaba et al., 2010) could also be evaluated using a gene-set approach (Table 3). For example up-regulated keratinocyte genes were improved over 105%, and up-regulated T-cell genes around 88%. Sets representing myeloid cell populations, including immature and mature DCs, inflammatory myeloid DCs, and macrophages, all improved over 89%.

Table 3.

Resolution of cellular gene-sets and canonical pathways at the end of treatment.

CELLULAR GENE-SETS
Name	Mean	n	Description	Reference
Keratinocytes	105.14	85	Up-regulated in Keratinocytes Haider’s Cell Maps	(Haider et al., 2007).
Terminal Differentiation Genes	93.78	17	Genes representing epidermal differentiation complex and cornified envelope	(Guttman-Yassky et al., 2009)
Keratinocytes + TNF	98.37	179	Up-regulated in TNF-treated keratinocytes at 1.5-fold-change and FDR<0.1	(Zaba et al., 2010)
Tcells	87.73	25	Up-regulated in Tcells Haider’s Cell Maps	(Haider et al., 2007)
Immature DCs	89.36	8	Up-regulated in immature DCs Haider’s Cell Maps	(Haider et al., 2007)
Mature DCs
	97.61	19	Up-regulated in mature DCs Haider’s Cell Maps	(Haider et al., 2007)
Inflammatory myeloid DCs	97.30	41	Up-regulated in BDCA1− cells (vs BDCA1+) from psoriasis lesions at 2-fold-change and FDR<0.1	(Zaba et al., 2010)
Macrophages	91.10	14	Up-regulated in macrophages Haider’s Cell Maps	(Haider et al., 2007)
CANONICAL PATHWAYS (C2 CP collection from MSigDB)
Name	Mean	n	Description	Genes in pathway (%improvement)
ST_MYOCYTE_AD_PATHWAY	51.70	6	Cardiac myocytes have a variety of adrenergic receptors that induce subtype-specific signaling effects.	TPR3(41.37), ITPR3(65.9), ITPR2(49.19), ITPR2(70.8), ITPR2(38.44), EPHB2(44.51)
LEPTIN PATHWAY	55.62	5	Leptin is secreted by adipose tissue that in skeletal muscle promotes fatty acid oxidation and insulin sensitivity, decreases lipid content.	LEP (110.95), LEPR (69.68), LEPR (26.49), LEPR (43.29), LEPR (27.71),
ST_WNT_CA2_CYCLIC_GMP_PATHWAY	65.61	7	Some Wnt glycoprotein/Frizzled receptor interactions increase intracellular calcium and decrease cGMP.	ITPR3(41.37), ITPR3(65.9), ITPR2(49.19), ITPR2(70.8), ITPR2(38.44), ITPKB(100.93), NFAT5(92.66)
TCYTOTOXIC PATHWAY	76.22	4	Cytotoxic T-cells release perforin and granzyme to lyse foreign cell targets and express Fas ligand to promote Fas-induced apoptosis.	CD2(48.72), PTPRC(87.43), PTPRC(59.08), PTPRC(109.66)

Defining “residual disease genomic profile” in treated psoriasis

Genes with less than 75% improvement were defined as comprising the RDGP. From the psoriasis up-regulated genes, 76 probesets (64 genes) did not improve above 75% with treatment, and among disease down-regulated genes, 206 probesets (184 genes) were below 75% improvement (Figure 1a). A selection of potentially interesting genes that did not return to NL levels is presented in Table 4, for both up and down-regulated psoriasis-genes (See Supplementary Table 1 for complete RDGP). These genes identify a set of disease-related genes with incomplete reversal even after effective therapy. Unfortunately, many primary cytokines such as IFNγ, IL-17, IL-22, are not reliably detected by Affymetrix chips due to low mRNA abundance (Suárez-Fariñas et al., 2010). Since our definition of RDGP is based on Affymetrix expression data, those genes are not in Supplementary Table 1. However, since we have assessed their improvement by PCR (Table 1), they could be considered as residual genes in a broader definition of RDGP.

Table 4.

Selected Disease Genes that have less than 75% Improvement.

Symbol	Description	LS vs NL	W12 vs LS	Improvement (%)
UP Genes that have less than 75% improvement
RAB31*	Member RAS oncogene family	1.14	−0.31	33.80
TRCB1*	T-cell receptor beta constant 1	1.82	−0.86	42.62
MMP 9	Matrix metallopeptidase 9 (gelatinase B, 92kDa type IV collagenase)	1.54	−0.83	46.42
CD2	CD2 molecule	1.51	−0.75	48.72
LCK	Lymphocyte-specific tyrosine kinase	1.23	−0.77	52.59
WNT5A*	Wingless-type MMTV integration site family	2.83	−1.55	54.24
CD48	CD48 molecule	1.04	−0.69	57.69
IF144*	Interferon-induced protein 44	2.86	−1.46	61.06
SYK	Spleen tyrosine kinase	1.09	−0.79	62.04
GZMA	Granzyme A (granzyme 1, cytotoxic T-lymphocyte	1.73	−0.97	64.34
LTB	Lymphotoxin beta (TNF superfamily, member)	1.30	−1.01	66.30
CD36*	CD36 molecule (thrombospondin receptor)	1.89	−0.67	68.20
IRF1	Interferon regulatory factor 1	1.13	−0.90	68.66
HLA-DPA1	Major histocompatibility comples, class II	1.40	−0.79	68.73
IFIT3	Interferon-induced protein with tetratricope	1.80	−1.20	70.86
CCL2	Chemokine (C-C motif) ligand 2	1.36	−1.33	73.06
IFI27	Interferon, alpha-inducible protein 27	2.24	−1.73	73.42
CXCR4	Chemokine (C-X-C motif) receptor 4	3.05	−2.40	73.54
CCL18	Chemokine (C-C motif) ligand 18	2.35	−2.13	73.86
ISG20	Interferon stimulated exonuclease gene 20	1.59	−1.05	74.87
DOWN Genes that have less than 75% improvement
MUC7	Mucin 7, secreted	−2.27	0.37	15.51
LYVE1*	Lymphatic vessel endothelial hyaluronan receptor 1	−1.38	0.4	21.45
AQP9	Aquaporin 9	−1.85	0.43	25.85
CLDN10	Claudin 10	−1.38	0.43	26.01
LEPR*	Leptin receptor	−1.26	0.28	26.49
CLDN8	Claudin 8	−3.44	1.52	44.95
AQP5	Aquaporin 5	−1.52	0.85	52.72
MMP7	Matrix metallopeptidase 7 (matrilysin, uterine)	−1.17	0.68	54.83
DSG2	Desmoglein 2	−1.23	0.93	56.53

^*when multiple probesets are present, the smallest improvement is reported.

Within this RDGP, we consider two sets of effects. First, there was continued expression of many inflammation-related genes, possibly indicating incomplete suppression of inflammatory circuits by etanercept. Genes such as lymphotoxin and TCRβ1 were up-regulated in psoriasis and improved less than 75% (66% and 43% improvement, respectively). This is in keeping with our observations described above at the PCR level, that epidermal improvement can occur even in the setting of some continued inflammatory gene expression.

Structural changes in the skin

The second major effect that we considered within the RDGP was the altered expression of genes associated with “structural” cell types of the skin. This suggests altered physiology or tissue structure in resolved lesions, in some sense, a “molecular scar”. Many of the genes that did not return to NL levels were for structural cutaneous components. There were genes that were down-regulated during psoriasis inflammation (LS<NL) that were not completely resolved by week 12 (e.g. LYVE-1 and AQP9). There were up-regulated genes that did not return to NL (e.g. RAB31 and WNT5A). Gene expression for these four genes is shown in Figure 1b, and compared to STAT1, an up-regulated gene that does return to NL. We have also performed immunohistochemstry for these four gene products from both down and up-regulated groups over the treatment period (Figure 2).

Figure 2. Immunohistochemistry for select genes in residual disease genomic profile.

Representative staining in normal, NL, LS and after 12 weeks of etanecept treatment. a) LYVE-1 showed discernible lymphatic lumen in normal and NL skin compared to collapsed lymphatic vessels in LS skin and after 12 weeks of treatment. b) WNT5A showed increased expression in LS skin and even after 12 weeks of treatment compared to NL and normal skin. c) RAB31 identified many positive dermal cells in LS skin, which decreased after 12 weeks of treatment but did not completely return to NL levels. d) AQP9 showed decreased expression in LS skin compared to NL and normal skin and after 12 weeks of treatment. Bar= 100μm.

LYVE-1, a well-recognized marker of lymphatic endothelial cells, showed only 21% maximal improvement at week 12. To determine the significance of persistent down-modulation of LYVE-1 at the protein level, we performed standard immunohistochemistry of LYVE-1 in normal skin, NL and LS skin, and at 12 weeks of treatment (Figure 2a). In normal skin the lymphatic vessels identified by this antibody were mostly in the upper reticular dermis and had a discernible (wide) lumen. In LS skin the lymphatic vessels were more collapsed and were observed to be closer to the dermo-epidermal junction compared to NL skin and week 12. Hence LYVE-1⁺ cells showed a different pattern in normal skin compared to psoriatic lesions, and even after treatment, it did not return to the pattern of expression of normal skin.

WNT5A (wingless-related MMTV integration site 5A) regulates epidermal differentiation in adult skin and acts in synergistic induction of type-I IFN target genes (Romanowska et al., 2009). This gene has been previously reported to show increased expression in psoriatic plaques (Reischl et al., 2007). WNT5A was only improved 54% at the genomic level. By immunohistochemistry, WNT5A showed the greatest expression in the epidermis of LS skin, with some dermal staining (Figure 2b). There was a persistent expression by week 12, greater than NL and minimal staining of WNT5A in normal skin.

RAB31 was another gene of interest that is increased in psoriasis tissue but did not return to NL levels. RAB31, a member of RAS oncogene family that plays an essential role in vesicle and granule targeting (Bao et al., 2002), showed 34% improvement at week 12. Immunohistochemistry was performed in normal skin, NL and LS skin, as well as after 12 weeks of treatment (Figure 2c). There were many RAB31 positive cells in the papillary dermis of LS skin close to the dermo-epidermal junction, with fewer positive cells in NL skin and at week 12. There was minimal RAB31 expression in normal skin. In LS skin, RAB31 was expressed on CD45⁺ cells, and also some CD11c⁺ myeloid DCs mainly in the papillary dermis and few CD163⁺ macrophages, (Figure S1).

Aquaporin 9 (AQP9) was another interesting down-regulated gene with maximal improvement of 26%. Mammalian aquaporins are a family of 13 membrane proteins that form water channels across cell membranes, and may transport small solutes such as glycerol (Boury-Jamot et al., 2006). Aquaporin 9 has been described in human skin, although its role there is not fully understood. In normal and NL skin there was some epidermal expression, with subtle loss of expression in LS skin, and some recovery at week 12, although not to NL levels (Figure 2d).

Canonical pathways do not fully resolve

To obtain a global view of the RDGP in terms of pathways rather than gene by gene, we quantified the average improvement for a number of canonical pathways. Of 358 canonical pathways that were represented in the transcriptome several of them showed less than 75% improvement (selected pathways shown in Table 4). For example disease-genes in the Leptin pathway improved only 56%. This could be important as leptin is considered to be partially responsible for obesity (Kelesidis et al.), and recently there has been an association with psoriasis and metabolic syndrome (Azfar and Gelfand, 2008). The failure of this pathway to improve 100% may be related to ongoing risk for obesity and the metabolic syndrome, even with clinical improvement in the skin. Myocyte AD pathway only improved 52%, and Wnt-pathway was 66% improved, both of which could relate to structural components in the skin or other organs.

Discussion

Psoriasis vulgaris tends to be active in focal regions of skin for long periods of time, unless treated with an effective therapeutic. However, lesions that appear to be resolved by therapy also tend to recur within weeks to months after ceasing treatment and recurrence often involves the same regions as prior lesions. Hence, an unresolved issue is whether successfully treated lesions continue to express factors that favor lesion development as compared to background skin in that patient. From the clinical perspective, psoriasis plaques often resolve completely, except for some pigmentary change. From histopathologic assessment, there is often complete reversal of acanthosis, disordered terminal differentiation, and psoriasiform pattern along with elimination of excess infiltrating leukocytes (Abrams et al., 2000; Gottlieb et al., 1995; Haider et al., 2008; Johnson-Huang et al., 2010; Zaba et al., 2007).

We present a method to more accurately define the pathological activity at the end of effective treatment, using microarray expression. We have defined the RDGP, as a set of genes that did not resolve by 3 months of etanercept treatment. This study shows four important features of clinically resolved lesions: 1) inflammation, as defined by expression of key cytokines and chemokines, is not completely resolved in treated lesions, although the epidermal reaction is fully resolved; 2) there are residual dermal CD8⁺ T-cells; 3) structural cells of the skin continue to express molecular alterations; and 4) some subtle features of skin structure, e.g., lymphatics, are not fully normalized with treatment. A similar pattern of collapsed lymphatics was seen in psoriasis lesions (Henno et al., 2009), but normalization after treatment was not studied.

TNF blockade with etanercept in psoriasis leads to two sets of early effects (Zaba et al., 2007; Zaba et al., 2009b). First, the classic “sepsis cascade” pathway of TNF leading to IL-1 production, then IL-6 and IL-8 cytokines, is suppressed within one week of starting treatment in all patients (Zaba et al., 2009b). Second, DC activation (presumably TNF-stimulated) is rapidly decreased in responding patients, resulting in reduced production of IL-23 and other pro-inflammatory molecules (Zaba et al., 2007). Th17 and Th22 cells are also rapidly de-activated, as evidenced by reduced IL-17, IL-22 and down-stream genes (Zaba et al., 2007). In this study, we showed that overall, CD3⁺ T-cells were reduced by 96%, and CD11c⁺ DCs were reduced by 88% in skin lesions, with nearly complete elimination of leukocytes infiltrating the epidermis. Pathogenic psoriasis DEGs were also highly suppressed by etanercept treatment, such that greater than 80% of DEGs were effectively normalized.

However, TNF-targeted therapeutics modulate adaptive immunity only indirectly, so MHC-antigen-TCR and co-stimulatory interactions between DCs and T-cells resident in skin should still occur even with full TNF blockade present. We postulate that ongoing interactions between residual DCs and T-cells in healed psoriasis plaques may lead to low levels of T-cell cytokines and products, but below the threshold that triggers epidermal activation and visible disease phenotype. Given that some inflammatory circuits appear to persist at low levels in treated lesions, it is not yet clear what are the specific molecules and their level of expression required for an active lesion to become recognizable by histopathology or clinical criteria. We hypothesize that persistent IL-17 and IL-22 expression may reach a threshold where sufficient cytokine is available to bind to IL-17 and IL-22 receptors on keratinocytes and trigger known psoriasis genes, such as S100A7 (psoriasin), β defensins (DEFB4), and cathelicidin (LL37). In particular, production of LL37 and binding of self-RNA could trigger DC activation with accumulation of DC-LAMP⁺ DCs in the skin (Lande et al., 2007). Collapsed lymphatics could lead to poor lymphatic drainage and may increase edema in healing lesions, while also restricting T-cell and DC trafficking to draining lymph nodes, effectively trapping activated leukocytes in the skin.

Recently, it has been shown that effector memory T-cells (T_EM) are found as resident T-cells within normal appearing skin and act as a first line of defense when skin is re-challenged with antigen (Clark; Liu et al., 2010). Hence, it is possible that healed psoriasis could also contain residual “psoriasis-specific” T_EM, and thus be “primed” to expand when given the right signal, and lead to site-specific inflammation. We observed that while CD3⁺ cells appear to resolve almost completely in the epidermis and dermis, bulk CD8⁺ cells in the dermis only showed improvements of 64%, which could include a subpopulation of CD8⁺ T_EM that could be part of the residual inflammatory signature in psoriasis. As T_EM frequency is best determined by FACS of single cell suspensions (T_EM are CCR7⁻ CD45RA⁺), further studies with appropriate tissue samples will be needed to answer this question.

The clearance of epidermal T-cells, but persistence of some dermal CD8⁺ T-cells is consistent with the proposed pathogenic role of CD8⁺ T-cells in psoriasis (Kryczek et al., 2008). This is supported by murine xenotransplant studies of human NL skin, which showed that blockade of T-cells into the epidermis by inhibiting α1β1 integrin, prevented epidermal hyperplasia and the development of psoriasis lesions (Conrad et al., 2007).

We acknowledge several limitations of this study. First, biopsies were available only until 12 weeks of treatment. Potentially, longer treatment could lead to additional genomic improvements, but at some point, continued expression of DEGs might represent a “molecular scar” of inflammation in the skin. Second, some genomic effects produced by etanercept might be agent-specific and future studies will be required to compare treatment outcomes with different therapeutic agents. It will be interesting to ascertain the relationship between quantitative or qualitative suppression of DEGs, and remission periods upon therapeutic discontinuation. It is important to determine if the psoriasis transcriptome could be fully reversed by any treatment, should a durable response or “cure” of psoriasis be possible. Third, to the extent that psoriasis NL skin has abnormal gene expression (Gudjonsson et al., 2009) it would be interesting to study whether anti-inflammatory treatment affects NL skin gene-sets. Fourth, inflammatory products in the skin might have “systemic” inflammatory effects that modulate psoriasis, such as recently described co-morbidities (Davidovici et al., 2010), and hence it will be important to study other tissues that might be impacted by targeted immune inhibitor.

In conclusion, we suggest that histological resolution of psoriasis skin lesions is not accompanied by full resolution of molecular alterations as assessed by gene profiling. We present a set of genes that represent the RDGP, genes that do not go back to NL levels. Other inflammatory diseases such as rheumatoid arthritis, psoriatic arthritis, or inflammatory bowel disease are destructive or scarring at the primary sites of involvement, while psoriasis is not. Even so, many of the questions that we raise for psoriasis - the extent to which a disease process can be reversed or progression prevented and whether a real cure is possible, are shared across common inflammatory diseases in humans.

Materials and Methods

Patients

Twenty adult patients with moderate to severe psoriasis were treated with etanercept 50 mg subcutaneously twice weekly for 12 weeks (clinical trial no. NCT00116181). Written informed consent was obtained from all patients. The clinical and histological response of these 20 patients was previously published (Zaba et al., 2007). Patients were defined histologically as “responders” if the serial biopsies of an index plaque showed return of epidermal thickness and keratin 16 (K16) immunohistochemistry to NL levels.

Gene Expression Data

RNA from skin biopsies of 15 patients was hybridized to Affymetrix HGU133a2 chips as described (Zaba et al., 2009b). Microarray is deposited at the Gene Expression Omnibus repository (GSE11903). Probeset annotation was obtained with Bioconductor’s hgu133a2.db package (version 2.3.5).

Statistical Analysis

In a prior publication (Zaba et al., 2009b), a mixed-effect linear model with “patient” as random effect was used to model gene expression using interaction TimexResponse. From this model, the treatment effect at 12 weeks was estimated for responders using limma package form Bioconductor.

Psoriasis-related genes (Suárez-Fariñas et al., 2010) were analyzed at the end of treatment to evaluate return to NL. For each disease-gene, we quantified improvement after 12 weeks of treatment as:

$$\text{Improvement}=100\times \frac{{log}_2(X_{12}/X_{LS})}{{log}_2(X_{LS}/X_{NL})}$$

where X_LS X_NL X₁₂ are the expression values at LS, NL and 12 weeks of treatment respectively. RDGP is defined as the genes with improvement below 75%.

We used Gene-Set approach to quantify the average improvement for a collection of pathways (Table 3). We included the canonical pathways (C2 CP) from MDigDB (http://www.broadinstitute.org/gsea/msigdb) and several gene-sets developed by our group (Haider et al., 2007); (Guttman-Yassky et al., 2009).

RT-PCR, Immunohistochemistry and Immunofluorescence

Skin biopsies for leukocyte markers were stained and counted, and PCR conducted, both in a standard manner (Zaba et al., 2007). Most of these results had been published (Zaba et al., 2007), and are re-analyzed here to determine mean improvement after 12 weeks of treatment. Standard procedures were followed for immunohistochemistry (n=6), immunofluorescence (n=3) and CD8⁺ cell counts (n=6) as previously described (Zaba et al., 2009a). Antibodies used for immunohistochemistry and immunofluorescence are listed in Supplementary Table 2. Images were acquired using appropriate filters of a Zeiss Axioplan 2 widefield fluorescence microscope with Plan Neofluar 20×0.7 numerical aperture lens and Hamamatsu Orca ER-cooled charge-coupled device camera, controlled by METAVUE software (MDS Analytical Technologies, Downington, PA). Immunohistochemistry was conducted in batches for paired samples and representative staining is shown.

Abbreviations

DEG

differentially expressed genes

FCH

fold change

FDR

false discovery rate

LS

lesional

NL

non-lesional

RDGP

residual disease genomic profile