Global Gene Expression Profiling of Hyperkeratotic Skin Lesions from Inner Mongolians Chronically Exposed to Arsenic

Abstract

The skin is an organ that is highly sensitive to chronic arsenic (As) exposure. Skin lesions such as hyperkeratoses (HKs) are common early manifestations of arsenicosis in humans. HKs can be precursor lesions of nonmelanoma skin cancers (NMSCs), but the driving forces behind their formation and how they may ultimately progress to NMSCs are unknown. The goal of this study was to examine the global gene expression profiles of As-related HKs in an effort to better understand gene expression changes that are potentially associated with early stages of As carcinogenesis. HK biopsies were removed from individuals living in an arsenicosis-endemic region in Inner Mongolia who had been exposed to high As levels in their drinking water for >20 years. Gene expression profiling was performed on RNA isolated from 7 individuals in this group and from 4 lesion-free skin samples from healthy individuals. Consistent with the pathological characteristics of the HK lesions, major functional categories and known canonical pathways represented by altered transcripts include those involved in development, differentiation, apoptosis, proliferation, and stress response. The results of this study may help define a signature profile of gene expression changes associated with long-term As exposure in the skin.

Introduction

Chronic exposure to arsenic (As) has been associated with the development of adverse health effects in humans. These effects are often seen in multiple organs in the body, causing diverse diseases such as diabetes mellitus, peripheral vascular disease, cardiovascular disease and cancers of the skin, urinary bladder, liver, lung, and kidney (Centeno et al. 2002; World Health Organization [WHO] 2000).

Most As exposure in humans is related to the consumption of drinking water contaminated from natural, geological sources of inorganic As (Nordstrom 2002). The WHO recently lowered the recommended limit for As in drinking water from 50 ppb to 10 ppb (WHO 2006). However, millions of people worldwide are exposed to significantly higher As concentrations in their drinking water, placing them at risk for developing As-related cancers and other diseases (Nordstrom 2002; Tapio and Grosche 2006). Some of the most severely affected populations are found in Bangladesh, India, China, Taiwan, Mexico, and South America (Rossman, Uddin, and Burns 2004; Tapio and Grosche 2006).

Arsenic accumulates in the skin, most likely due to its binding to the large concentration of thiol groups in the protein-rich environment (Yu, Liao, and Chai 2006). The skin is therefore a sensitive organ to the effects of chronic As exposure, where the first manifestations of exposure often appear (Huang et al. 2004; Rossman et al. 2004). Characteristic early findings in the skin are nonmalignant and include hyperkeratoses (HKs) and pigmentation disorders. The most common HKs are nodular and occur bilaterally on the palms and soles and other areas of friction, but their morphologies can differ greatly and they may occur on other areas of the body as well (Schwartz 1996). Interspersed areas of hyperpigmentation and hypopigmentation may also be present on the trunk or face, occurring in a characteristic “raindrop” pattern (Rossman, Uddin, and Burns 2004; WHO 2000). Bowen’s Disease (BD), classified as squamous cell carcinoma in situ, may also develop, usually forming numerous lesions on the trunk. After a latency period of up to 20 to 30 years, frank nonmelanoma skin cancers (NMSCs), including squamous cell and basal cell carcinomas (SCCs and BCCs, respectively), may appear (Sun et al., 2006). The early dermal changes have been found to occur in dose-response relationship with As exposure. Although these changes are very common, they do not occur in all exposed individuals, and the length of time for their development and for the development of other symptoms related to As exposure can vary greatly (Alain, Tousignant, and Rozenfarb 1993).

Arsenic-induced HKs can be precursor lesions of NMSCs (Centeno et al. 2000; National Research Council [NRC] 1999), which are generally indistinguishable microscopically from those related to other causes (i.e., sun exposure), except that they tend to occur in clusters and in sun-protected areas of the body (Centeno et al. 2002; WHO 2000). When As-related HKs undergo malignant transformation, they most often develop into SCCs, which tend to be more aggressive than SCCs arising from actinic keratoses (AKs) related to sun exposure and may more readily metastasize to internal sites (Schwartz 1996; Southwick and Schwartz 1979). Although BCCs do not develop from AKs, they infrequently arise from As-related HKs (Cabrera and Gomez 2003). As-related HKs may also develop de novo or from BD lesions (Alain, Tousignant, and Rozenfarb 1993). The underlying mechanisms of any these processes are not well understood. The results of numerous in vitro and in vivo studies have led to several proposed mechanisms for As carcinogenesis in the skin, including the overproduction of growth promoting cytokines and growth factor overproduction (Germolec et al. 1996; Vega et al. 2001), generation of reactive oxygen species (ROS) (Yu, Liao, and Chai 2006), deficiencies in DNA repair and oxidative stress defense mechanisms (Ahsan, Chen, Wang, et al. 2003; Andrew, Karagas, and Hamilton 2003; Banerjee et al. 2007; Hamadeh et al. 2002), suppression of terminal epidermal differentiation (Jessen et al. 2001; Kachinskas et al. 1997), and alterations in cell signaling and cell cycling (Huang, Costa, and Shi 2004; Hughes 2002; Kitchin 2001; Yu, Liao, and Chai 2006). More than one of these mechanisms is likely to be involved, acting independently or synergistically. Although As can produce some genotoxic effects (reviewed by Andrewes, Kitchin, and Wallace 2003; Basu et al. 2001; Gebel 2001; Rossman, Uddin, and Burns 2004), it does not cause point mutations. It is therefore unlikely to act as an initiator of carcinogenesis but may act as a tumor promoter, progressor, or cocarcinogen (Rossman, Uddin, and Burns 2004).

To better understand the mechanism of action of As carcinogenesis in the skin, we performed global gene expression profiling on arsenic-related HKs isolated from individuals from arsenicosis-endemic populations in Inner Mongolia. These gene expression profiles were compared to profiles obtained from normal skin samples of healthy individuals living in a nearby area. Arsenic exposure is a significant problem in some regions of mainland China, where there are ∼3 million people who have been chronically exposed to >50 ppb As in their drinking water and at least 30,000 patients diagnosed with arsenicosis (Sun et al. 2006). Subjects included in this study resided in the two major arsenicosis endemic sites located south of the Daqing mountains along the northern coast of the Yellow River in Inner Mongolia (Luo et al. 1997; Ma et al. 1999). The residents of these endemic sites have good nutritional status but have been exposed to As for >20 years through their usage of artesian wells for drinking water, their only significant source of As exposure. The high arsenic concentrations in these wells (up to ∼1,800 ppb) is due to natural geologic sources and is a mixture of soluble, inorganic trivalent As (arsenite) (54%), soluble pentavalent inorganic As (arsenate) (30%), and an insoluble As fraction (16%) (Gong et al. 2006). Arsenicosis prevalence was related to As well water concentration in a dose-response manner, with common symptoms including skin lesions, gastroenteritis, and cardiovascular and peripheral neurological effects (Ma et al. 1999).

Ours is one of the few microarray studies that has examined samples from humans chronically exposed to arsenic (Andrew et al. 2008; Argos et al. 2006; Lu et al. 2001; Wu et al. 2003; Yih, Peck, and Lee 2002), and to our knowledge, this is the first study of As-related human lesions from a major target organ of As toxicity. Because extremely limited amounts of biological materials were available, this study is observational in nature and provides an overview of the functions of the 2,824 transcripts differentially expressed between the HK lesions and normal skin. The most significant categories of altered genes include those involved in differentiation, cellular organization, apoptosis, proliferation, and stress response, gene categories known to play important roles in the carcinogenesis process and that have been shown to be transcriptionally altered after As exposure in a variety of other systems. In particular, several of the most significantly altered pathways (i.e., mitogen activated protein kinase [MAPK] pathways, Wnt/β-catenin and calcium signaling pathways) are known to play important roles in skin development and homeostasis and may play important roles in As-related HK formation.

Materials and Methods

Study Subjects and Skin Sample Collection

This study was conducted according to the recommendations of the World Medical Association Declaration of Helsinki (WMA; 1989) for international health research. The research protocol met the requirements for protection of human subject certification as approved by the U.S. Environmental Protection Agency. All subjects gave written, informed consent to participate in this study and completed questionnaires that provided information regarding sociodemographic characteristics, tobacco use, medical history, and arsenic exposure information (see Table 1 for some of these characteristics).

A total of 11 study subjects were recruited for the study: 7 individuals living in the arsenicosis endemic areas in the Ba Men and Huhhot regions of Inner Mongolia and 4 unexposed individuals living in Huhhot (the capital of Inner Mongolia). Of the As-exposed subjects, 6 lived in Ba Men (2 in the village of Tie Men Geng and 4 in the village of Zhi Ji Liang), and 1 lived in the Gu Cheng village of Huhhot (Table 1). These individuals had skin characteristics (e.g., HKs and pigmentation changes) consistent with chronic arsenicosis and been exposed to 212 to 950 ppb As in their drinking water for >20 years, with the Tie Men Geng individuals having slightly longer periods of exposure (close to 30 years). Physical examinations and questionnaire results indicated that all of the study subjects had good nutritional status. The As-exposed individuals had similar levels of skin disease with the exception of subject 342, who had no skin pigmentation disorder and fewer HKs compared to the other As-exposed individuals.

Skin HKs were diagnosed by a dermatologist according to the China National Standards for Diagnosis of Arsenicosis (People’s Republic of China [PRC]), and all of those removed for analysis in this study had similar pathological characteristics. A skin sample from each study subject (hyperkeratotic lesions from As-exposed individuals; lesion-free skin from unexposed individuals) was collected by a dermatologist and immediately placed in RNAlater ^®(Ambion, Austin, Texas, USA) for RNA stabilization. The samples were transported on ice packs via air to the United States and stored at −20 °C.

Drinking water samples collected from the subjects’ homes were analyzed for As content using hydride generation atomic fluorescence spectrometry (HGAFS) (Le and Ma 1998) or a standard colorimetric method using silver diethydithiocarbamate (Zhang et al. 1994).

RNA Extraction

For total RNA isolation, approximately half of each ∼0.2 cm × 0.3 cm skin sample was pulverized using a mortar and pestle under liquid nitrogen. Immediately after the liquid nitrogen had evaporated, 1 ml of buffer RLT from the RNeasy Mini Kit (QIA-GEN, Valencia, CA, USA) was added to the sample. The sample was transferred to a tube and further homogenized for 45 seconds in an electric homogenizer at low speed (Tissue Tearor model 985-370, Biospec Products, Bartlesville, OK, USA). The RNeasy Mini Kit (QIAGEN) protocol was used to purify total RNA, yielding between 225 ng and 8.5 μg of total RNA per sample. RNA concentrations and quality were determined spectrophotometrically using A₂₆₀ and A₂₆₀/A₂₈₀, respectively, and RNA quality was verified using the RNA 6000 Pico kit and the Bioanalyzer 2100 (Agilent, Palo Alto, CA, USA). Only samples with A₂₆₀/A₂₈₀ values between 1.9 to 2.1 and the presence of two distinct, intact 18S and 28S ribosomal RNA populations identified using the Bioanalyzer 2100 were used in the study.

Microarray Analyses

Total RNA from each sample (4 control and 7 HK samples) was used for global gene expression profiling performed by Expression Analysis, Inc. (Durham, NC, USA). Biotinylated cRNA was obtained from 50 ng of total RNA from each sample using Affymetrix^® Two-Cycle Target Labeling Assay and was hybridized to GeneChip^® Human U133 Plus 2.0 Arrays (containing 54,675 probe sets) according to the manufacturer’s instructions (Affymetrix^®, Santa Clara, CA, USA). The output files (.cel files) were subsequently analyzed for differential gene expression between control and HK samples using methods described by Chloe et al. (2005). Several sequential steps were performed: (1) background correction was performed using MAS v5 (Affymetrix^®), (2) probe-level quantile normalization using Robust Multiarray Average (RMA) (Irizarry et al. 2003), (3) perfect match adjustment for nonspecific hybidization using MAS v5, (4) expression summary (median polish) using RMA, (5) probe set-level normalization using Loess, and (6) a Bayesian t-test using CyberT (Baldi and Long 2001).

Pathway and functional analyses of the differentially expressed transcripts were performed using the following software: GeneSpring^® (v. 7.2) (Silicon Genetics, Redwood City, CA, USA), DAVID 2.0 (NIAID), Ingenuity Pathway Analysis (v. 5.5) (Ingenuity Systems^®, Redwood City, CA, USA), BioRag (Bio Resource for Array Genes) at www.biorag.org and MetaCore^™(GeneGo; San Diego, CA, USA). Principal components analysis (PCA) was performed using Cluster 3.0 (http://rana.lbl.gov/EisenSoftware.htm).

Quantitative Reverse Transcriptase-PCR (qRT-PCR)

There was sufficient total RNA from 3 of the 4 control samples (C-5, C-6, C-8) and the 7 hyperkeratosis subjects (332, 333, 334, 336, 338, 342, 343) to perform limited real-time quantitative RT-PCR analyses. For each sample, 10 ng of total RNA was converted to cDNA using the High-Capacity cDNA Reverse Transcription kit (Applied Biosystems Catalog #4368814) in 20 μl reactions according to the manufacturer’s instructions. Two reactions were performed per sample and pooled. The TaqMan^® PreAmp Master Mix Kit (Applied Biosystems, #4364130) was used for cDNA preamplification prior to qRT-PCR according to the manufacturer’s instructions. Preamplification reactions from the three control samples (C5, C6, C8) were pooled and used as a standard sample. A dilution series from the amplified standard was prepared and qPCR reactions using TaqMan^® Gene Expression Assays were run to determine preliminary expression levels, stability of potential of reference targets between HK and control sample groups, and appropriate dilutions to use for each target to be interrogated. Quantitative PCR reactions were run in triplicate for each target using FastStart Universal Probe Master mix (Roche Applied Science Cat #01 −14 066 001) and cycled as follows: 95°C for 10 min followed by 40 cycles of 95°C for 30 sec and 60°C for 60 sec. Standard curves were generated for each target and data were normalized to the concentration of input RNA in each cDNA amplification reaction. Significant differences in gene expression were determined using Tukey’s HSD (alpha = .05), and fold-change values were calculated for each target in the HK samples relative to controls.

Results

Skin Sample Pathology

Representative histology slides of normal (control) and HK skin samples used in microarray analyses are shown in Figures 1A and 1B, respectively. Each of the control skin samples exhibited characteristics of healthy skin with no evidence of disease (Figure 1A). In the control abdomen sample, the four layers of the normal stratified squamous epithelium (basal, spinous, granular, and cornified) are visible (Figure 1A). Keratinocytes are the dominant cell type in each layer and exhibit layer-specific biochemical and morphological characteristics (Fuchs and Weber 1994).

Each of the HK samples in this study had similar pathological characteristics. Epidermal ridges and cornified layer hyperkeratinization are all observed in the HK lesions in this study (Figure 1). Vacuolization, a particular characteristic of arsenic-related HKs (Schwartz 1996), was also found during pathology analyses.

Overview of Differentially Expressed Transcripts

In each of the samples, transcripts corresponding to approximately 39% of the 54,675 probe sets on the Human 133 Plus 2.0 arrays were expressed. Of the expressed transcripts, 2,824 different transcripts (representing ∼1,180 different known genes and >700 unannotated transcripts) were differentially expressed between the HK and control sample groups (Table 2, Supplemental Material). There were 1,013 transcripts upregulated in HK compared to controls and 1,811 downregulated transcripts (ranging from 1.3-to 59.5-fold and from −1.2-to −69.9-fold change in expression in HKs vs. controls, respectively). All but 33 of the differentially expressed transcripts (DETs) had an expression value that differed at least 1.5-fold in the HK samples compared to controls. Each of the 9 genes chosen for qRT-PCR analysis demonstrated the same direction of modulation in the HK samples relative to controls as observed in microarrays. The expression values of these genes were also significantly modulated in HK lesions compared to controls in the qRT-PCR analyses (Tukey’s HSD; alpha = .05) (Supplemental Figure 1, Supplemental Material).

The 11 samples used in the microarray analysis were subjected to a PCA to determine how similar the expression values of the 2,824 DETs are between each sample. PCA is a data reduction tool that allows multidimensional expression data to be viewed in three-dimensional space, and the degree to which samples cluster together reveals how similar the DETs’ expression profiles are between them (Aittokallio et al. 2003; Quakenbush 2001). As seen in Figure 2, PCA using the 2,824 DETs clearly separates the control and HK samples into two distinct groups. The control samples cluster more tightly together than the HK samples, revealing that there is a larger disparity in the expression values between individual samples within the HK group than observed among the control samples. Visualizations of PCA data can reveal that diseased samples are more scattered or form subgroups relative to normal tissue, which may reflect some subtle pathological differences in the diseased samples (O’Driscoll et al. 2006).

Overview of Major Altered Cellular Functions and Signaling Pathways

Figure 3 is a summary of the most significant (p < .00001) molecular, cellular, and physiological system functions and diseases (3A) and known canonical signaling and metabolic pathways (p < .01, 3B) associated with the DETs as determined using Ingenuity Pathway Analysis (version 5.5). (For complete lists of genes associated with significant functions/diseases and known canonical pathways, see Tables 3A and 3B, respectively, Supplemental Material.) Similar gene ontology and pathway results were obtained from analyses using MetaCore^™ (results not shown). Many of the high-order physiological functions and diseases altered in the HK lesions compared to controls include genes associated with skin development, function, and disease (Figure 3A).

At the cellular and molecular levels, the most significant functions of the modulated genes can be broadly classified into categories involving cell death, cellular development/organization, and cell proliferation. These genes can be further subdivided into several functional categories, such as signaling/regulation, cell differentiation/cytoskeleton/adhesion, cell proliferation/cell cycle checkpoint/apoptosis, and stress response/DNA damage and repair (Table 4A-4E, respectively, Supplemental Material).

Several common functions are present among the most significantly altered pathways (Figure 3B). These include pathways that are indicators of cellular stress (e.g., extracellular signal-regulated kinase [ERK], c-jun N-terminal kinase [JNK], and p38 MAPK pathways; protein ubiquitination pathway, xenobiotic metabolism signaling), those involved in cytoskeletal rearrangements (axonal guidance, ephrin receptor, and actin cytoskeletal signaling), and pathways broadly classified as involved in skin differentiation or development (MAPKs, calcium signaling, Wnt/β-catenin signaling). Many of the altered pathways include genes that are effectors, components, or targets of pathways involving signaling through guanine nucleotide binding proteins (G proteins), including MAPK pathways. In particular, many of these altered genes either influence or are controlled by the Ras-Raf-ERK in a majority of the most significantly altered pathways as indicated by asterisks (*) in Figure 3B.

Summary of Major Modulated Gene Categories

Signaling/Regulatory Genes:

Gene ontology analyses reveal that more than 500 of the DETs are predominantly involved in regulatory roles, and several of the most significantly altered pathways are signaling pathways that play regulatory roles (i.e., MAPKs, Wnt/β-catenin, and calcium signaling pathways). The most commonly altered regulatory genes among the significantly altered pathways (Table 3B) are listed in Table 4A. Figure 4 illustrates the most predominant of the regulatory pathway components, which, as mentioned above, includes components, regulators and effectors of P38, JNK, and, in particular, ERK MAPK pathways. For simplicity, not all altered regulatory genes are displayed in Figure 4. The altered regulatory genes include G proteins, kinases, integral membrane proteins such as integrins and receptor tyrosine kinases (RTK) (e.g., epidermal and fibroblast growth factor receptors and EFNB3), genes for extracellular proteins such as growth factors (e.g., HGEBF) and transcription factors such as members of the ETS oncogene family (ETS1, ETS2, ELF5), ternary complex ETS subfamily (TCF) ELK1 and ELK4, and AP-1 components such as MAFB, MAFF, and FOSL2 (Shaulian and Karin 2001) (Table 4A).

Cell Differentiation/Cytoskeleton/Adhesion Genes:

A large number of genes directly involved in keratinocyte terminal differentiation are also modulated (Table 4B). Many of these genes are upregulated in the HK lesions, including those that code for precursor components of cornified envelope such as LOR, IVL, multiple small proline rich proteins (SPR genes), and the enzymes that crosslink these components, such as transglutaminases (TGM genes) (Fuchs 1990; Fuchs and Weber 1994). Genes coding for components of the extracellular matrix (collagens, cartilage oligomeric matrix proteins); matrix remodeling enzymes such as those belonging to the matrix metalloproteinase (MMP) and a disintegrin and metalloprotease domain (ADAM) families, and various adhesion molecules, such as DSG1, DSC1, DSC2, JUP, and CDSN, are also modulated (Table 4B). Cytoskeletal component genes are also modulated, including tubulins, actins, myosins, and a large number of keratins (Table 4B).

Cell Proliferation/Cell Cycle/Apoptosis:

These genes include growth or survival factors, receptors or downstream effectors of these factors (many of these genes described above and listed in Table 4A), cell cycle checkpoint genes, genes involved in genome replication and cell division, and genes that promote or inhibit apoptosis (Table 4C).

Modulated genes involved in G1/S phase transition include cyclins CCND2 and CCNE2 and the p14^ARF transcript of the CDK2NA locus) (Table 4C). Genes that play roles in the regulation of the G2 to M phase transition of the cell cycle are upregulated in the HK lesions, such as CCNB2 and CDK7. Several genes that control the G2 to M checkpoint in response to DNA damage are also modulated in HK lesions, such as CHEK1, WEE1, FOXN3/CHES1, GADD45A, and the alternate transcript p19^ARF from the CDK2NA locus (Pomerantz et al. 1998; Tao and Levine 1999; Zhang, Xiong, and Yarbrough 1998).

Several of the modulated genes have direct roles in apoptosis. These include pro-apoptotic genes of the B-cell lymphoma/leukemia 2 (BCL-2) family of proteins BNIP3, BNIP3L, and BCL2L14 (Table 4C). Other modulated pro-apoptotic genes include several members of the baculoviral repeat inhibitor of apoptosis (IAP) proteins such as BIRC-4, BIRC-5 (survivin), and BIRC7 (livin). Modulated anti-apoptotic genes include FAIM2, CFLAR, and serine proteinase inhibitors (serpins) SCCA1 and SCCA2 (Table 4C).

Stress-Induced Genes/DNA Damage Recognition and Repair:

Modulated stress genes aside from the DNA damage/cell cycle checkpoint genes previously mentioned (Table 4C) include those in the protein ubiquitination pathway (e.g., ubiquitinconjugating enzymes; Table 3B), xenobiotic metabolism pathway (e.g., cytochrome P450 genes, glutathione metabolism-related genes; Table 3B), and components of various MAPK pathways including those involved in signaling through P38 via interleukin 1 as illustrated in Figure 4. Other stress-related genes include S100A16, PPARβ/δ, and heat shock protein genes (HSPs) such as heme oxygenase 1 (HMOX1). Modulated DNA repair genes include RAD51L3, REV3L, and XRCC5 (Table 4E).

Discussion

There are several noteworthy outcomes of this study: (1) the HK microarray results are consistent with the observed histological features, (2) there is concordance between the microarray and qRT-PCR data, (3) gene expression changes observed in the HK lesions in this study are consistent with data obtained from As-exposed keratinocytes and/or As-related skin lesions in other studies, and (4) this study provides the first comprehensive microarray analysis of DETs of As-associated human skin lesions compared to normal skin. As these HKs may be precursors of NMSCs, their gene expression profiles may provide insight into the driving forces behind both benign and malignant As-related skin diseases.

Aside from histological observations, few studies have characterized As-related HKs (Centeno et al. 2002, 2000). Arsenic-related HKs are believed to be similar to HKs related to other causes, in which alterations in cell proliferation and differentiation (normally restricted to the basal and suprabasal epidermal layers, respectively), occur (Fuchs and Weber 1994; Schwartz 1996). HK lesions are characterized by keratinocyte hyperproliferation in the suprabasal prickle layer, which may extend into the dermis to form epidermal ridges (Centeno et al. 2000), and an abnormally large deposition of terminal differentiation proteins (hyperkeratinization) in suprabasal layers, particularly the cornified layer (Lee et al. 2006; Schwartz, 1996). Epidermal ridges, cornified layer hyperkeratinization, and vacuolization (vacuolization a particular characteristic of arsenic-related HKs [Schwartz 1996]) are all observed in the HK lesions in this study (Figure 1). Stress-related characteristics that have been reported in As-related skin lesions or keratinocytes are also supported by the HK transcriptional profiles, including G2/M cell cycle arrest (several genes that promote G2/M cell cycle arrest are upregulated, e.g., GADD45A, CHEK1, WEE1) (Chen and Shi 2002; Yu, Liao, and Chai 2006) and stress-related KRT expression found in advanced As-related skin lesions (i.e., upregulation of KRT6, -16, and -17 and downregulation of KRT15) (Yu et al. 1993).

The histological characteristics of the HK lesions are consistent with the most significant transcriptional alterations occurring in genes involved in cell death, proliferation/cell cycle control, differentiation, and stress response in these lesions (Figure 3; Table 4). In general, perturbations in these processes (i.e., proliferation/cell cycle, differentiation, and apoptosis) are believed to be important driving forces behind the development of As-related NMSCs (Yu, Liao, and Chai 2006), other preneoplastic and malignant skin diseases (Batinac et al. 2007), and many other known cancers (Hanahan and Weinberg 2000). Increasingly, ROS generation is believed to play a major role in the formation of benign and malignant skin diseases via its disturbance of these processes (Bickers and Athar 2006). Arsenic is a potent generator of ROS in the skin, and ROS production is believed to be a major factor in As-mediated skin toxicity both by generating direct damage to intracellular components such as DNA and by activating signaling pathways (particularly those such as MAPKs that converge on stress response transcription factors AP-1 and NF-κB) that regulate apoptosis, proliferation, and differentiation (Cooper, Liu, and Hudson 2007; Ding, Hudson, and Liu 2005; Pi et al. 2005; Shi et al. 2004). Several of the most significantly altered canonical pathways in this study (i.e. MAPKs, calcium signaling, and Wnt/β-catenin signaling), can regulate these processes and can be modulated by oxidative stress (Hidalgo and Donoso 2008; Korswagen 2006; Torres and Forman 2003). Importantly, each of these pathways also plays a crucial role in epidermal development and/or homeostasis (Dong 2002; Yu, Liao, and Chai 2006).

One of the most predominant features of the significantly altered pathways includes components, regulators, and effectors of the Ras-Raf-ERK cascade (Figures 3B and 4). This cascade plays a crucial role in maintaining the normal balance of differentiation and proliferation in the epidermis, and its disruption may lead to aberrant differentiation and/or proliferation and ultimately tumorigenesis (Khavari and Rinn 2007). Alterations in the Ras-Raf-ERK pathway have been correlated with pre-neoplastic and malignant skin disease in human and murine models and has been shown to be essential for As-mediated keratinocyte transformation (Huang et al. 1999). Recent micro-array work has demonstrated that this pathway, along with the P38 MAPK pathway, plays an essential role in the expression of differentiation genes in keratinocytes (Gazel et al. 2008).

Along with ERK, the SAPK/JNK and P38MAPK pathways have also been shown to be activated by As exposure in keratinocytes (Tanaka-Kagawa et al. 2003). The exact effect each of these pathways has on apoptosis, differentiation, and proliferation in keratinocytes are complex and depend on the dose and duration of As exposure (Dong 2002). In this study, upstream components of SAPK/JNK and P38 MAPK pathways are also modulated in the HK lesions. Importantly, modulated upstream components of the P38 pathway may also activate transcription factor NF-κB (Figure 4), which regulates processes such as survival and apoptosis in response to stress and is implicated as an important player in As-mediated NMSC formation (Yu, Liao, and Chai 2006). Few NF-κB-dependent genes have been identified in the epidermis, but at least one of them, the anti-apoptotic gene CFLAR, is upregulated in the HK lesions (Banno et al. 2005).

Transcriptional alterations in calcium and Wnt/β-catenin pathways may also have profound effects on epidermal homeostasis. Disturbances in intracellular calcium levels disrupt the balance between proliferation and differentiation in keratinocytes (Hennings et al. 1980) and cause nuclear DNA damage (Dopp et al. 1999) and disruptions in mitosis and DNA repair (Florea, Yamoah, and Dopp 2005; Xu, Luo, and Chang 2003). Arsenic can disrupt cellular calcium homeostasis (Florea, Yamoah, and Dopp 2005), and this disruption has been implicated as a mechanism of As trioxide–induced DNA damage and apoptosis induction (Florea and Busselberg 2008). Wnt/β-catenin signaling is a highly conserved process involved in development, regeneration, and homeostasis in many tissues (Gordon and Nusse 2006). In the skin, it is involved in the development and homeostasis of the epidermis and dermis, including structures such as sweat glands and hair follicles (Narhi et al. 2008). Increased Wnt signaling is seen in SCCs and plays an important role in the maintenance of tumorlike properties in epidermal cancer stem cells (Malanchi et al. 2008). Perturbations in Wnt signaling in response to arsenic exposure in the skin have not been reported in the literature until recently, when Wnt signaling was among the statistically significant altered pathways in the skin of arsenic-exposed K6/ODC transgenic mice (Ahlborn et al. 2008).

The transcriptional disturbances of these regulatory pathways support theories in which several mechanisms are involved in As-mediated skin disease. In addition to disturbances in regulatory pathways, several aspects of the DETs in this work support other factors believed to play important roles such as the upregulation of mitogenic growth factors (e.g., HBEGF, EREG), cytokines (e.g., IL1F9), and cyclins (e.g., CCND2, CCNE2) in the HK lesions (Table 4). Several DNA repair genes (e.g., MSH5, REV3L, RAD51L3, XRCC5) and oxidative stress response genes (e.g., GSR, M1TM, glutathione s-transferase genes) are downregulated in the HK lesions (Table 4E). These expression profiles may be significant as epidemiological studies have indicated that arsenic-exposed individuals with mutations/defects in DNA repair genes (Ahsan, Chen, Kibriya, et al. 2003; Banerjee et al. 2007; Breton et al. 2007) or oxidative stress response genes (Ahsan, Chen, Wang, et al. 2003; Lin et al. 2006, 2007) have increased susceptibilities of developing arsenic-related HKs and/or NMSCs.

The importance of any of these transcriptional alterations in As-mediated HK formation and their potential progression to NMSCs is unknown and must be addressed in future work. The limited amounts of biological materials available precluded performing experiments beyond gene expression analyses and limited histological examinations. Ideally, additional analyses would also involve comparisons between normal skin, perilesional skin and As-related HKs, BD lesions, and NMSCs to gain a better knowledge of the gene expression changes involved in As-related skin disease. Samples from these more advanced forms of As-related lesions were not available for this study, nor are gene expression studies of them reported in the literature. Similarities in the processes behind As-related HK and NMSC formation, and those involved in the formation of other benign lesions and NMSCs, are unknown. The major functional classes represented in the DETs (including multiple significantly altered pathways playing major roles in skin homeostasis and development) and the HK histology data suggest that As-mediated HKs exhibit similarities to other preneoplastic skin lesions. In particular, hyperkeratinization and the large number of DETs involved in terminal epidermal differentiation modulated in the HK lesions indicate the presence of an aberrant differentiation phenotype (Figure 1, Table 4B). These phenotypes are often displayed in preneoplastic skin lesions as opposed to highly proliferative ones; progression to NMSCs generally involves a shift in the balance between differentiation and proliferation to favor proliferation (Caldwell, Hobbs, and McKee 1997; Nickoloff et al. 2002). Microarray studies of As-unrelated AKs reveal that differentiation-related genes are among those that are most consistently modulated in these lesions (van Ruissen et al. 2002). In particular, these include several that belong to the epidermal differentiation complex (EDC) at chromosome 1q21 such as KRTs, SPR genes, and those that belong to the S100 family of calcium-binding proteins (van Ruissen et al. 2002; Torres et al. 2007). These genes are predominantly upregulated in AKs and have a similar expression pattern in the HKs in this study (Table 2). Other modulated genes in both types of lesions that may play important roles in maintaining epidermal homeostasis include components of the Wnt/β-catenin pathway (i.e., upregulation of WNT5A and downregulation of CTNNB1, WIF1); the upregulation of cytokine IL1F9, downregulation of the immune modulator CCL27, and upregulation of anti-apoptotic genes SCCA1 and SCCA2 (Torres et al. 2007). As seen in the HK lesions in this study (Table 4C), apoptosis-related genes are often modulated in preneoplastic lesions; their progression to NMSCs involves a shift from the dominance of pro-apoptotic signals in pre-neoplastic lesions to favor anti-apoptotic events in malignancies (Nickoloff et al. 2002). Future work is necessary to determine the importance of particular genes and/or processes in the formation and progression of As-related HKs as well.

At least some differences are known between As-related and As-unrelated NMSC formation. As in As-mediated disease, oxidative stress is believed to play a role in UV-mediated NMSCs (Bickers and Athar 2006), but direct UV-mediated mutations in DNA also play a large role, creating so-called “UV signature” mutations that tend to occur in particular regions in genes and differ from the major DNA lesions caused by As (Bau et al. 2002; Soehnge, Ouhtit, and Ananthaswamy 1997). The formation of the two major forms of non-As related NMSCs, SCC and BCC, are both linked to UV exposure but the processes behind their development are quite different (Boukamp 2005; Green and Khavari 2004). For instance, SCCs arise from preneoplastic, interfollicular lesions in a multistep process most likely due to multiple factors, whereas BCCs are cancers of the hair follicle that do not arise from precursor lesions and the activation of genes involved in sonic hedgehog signaling have been identified as important, if not causative, events (Boukamp 2005; Tilli et al. 2005). Arsenic exposure has been linked to both SCC and BCC formation in humans (Cabrera and Gomez 2003), and future comparisons between the gene expression profiles of As-related skin lesions and As-unrelated NMSCs, of which there are growing numbers represented in the literature, may provide new insight into any commonalities between them.