Use of a Rainbow Trout Oligonucleotide Microarray to Determine Transcriptional Patterns in Aflatoxin B₁-Induced Hepatocellular Carcinoma Compared to Adjacent Liver

Abstract

Hepatocellular carcinoma (HCC) is one of the most common malignant tumors worldwide, and its occurrence is associated with a number of environmental factors including ingestion of the dietary contaminant aflatoxin B₁ (AFB₁). Research over the last 40 years has revealed rainbow trout (Oncorhynchus mykiss) to be an excellent research model for study of AFB₁-induced hepatocarcinogenesis; however, little is known about changes at the molecular level in trout tumors. We have developed a rainbow trout oligonucleotide array containing 1672 elements representing over 1400 genes of known or probable relevance to toxicology, comparative immunology, carcinogenesis, endocrinology, and stress physiology. In this study, we applied microarray technology to examine gene expression of AFB₁-induced HCC in the rainbow trout tumor model. Carcinogenesis was initiated in trout embryos with 50 ppb AFB₁, and after 13 months control livers, tumors, and tumor-adjacent liver tissues were isolated from juvenile fish. Global gene expression was determined in histologically confirmed HCCs compared to noncancerous adjacent tissue and sham-initiated control liver. We observed distinct gene regulation patterns in HCC compared to noncancerous tissue including upregulation of genes important for cell cycle control, transcription, cytoskeletal formation, and the acute phase response and downregulation of genes involved in drug metabolism, lipid metabolism, and retinol metabolism. Interestingly, the expression profiles observed in trout HCC are similar to the transcriptional signatures found in human and rodent HCC, further supporting the validity of the model. Overall, these findings contribute to a better understanding of the mechanism of AFB₁-induced hepatocarcinogenesis in trout and identify conserved genes important for carcinogenesis in species separated evolutionarily by more than 400 million years.

hepatocellular carcinoma, aflatoxin B₁, rainbow trout, oligonucleotide microarray, gene regulation

INTRODUCTION

Hepatocellular carcinoma (HCC) is one of the most common malignancies in humans worldwide, particularly in Southeast Asia, Japan, and Africa. Even the relatively low incidence of HCC in the United States is rising and exhibits the fastest increase among solid tumors (El-Serag and Mason, 1999). The prevalence of HCC has been correlated with a number of environmental factors including chronic inflammatory liver diseases caused by viral infection and dietary exposure to aflatoxin B₁ (AFB₁), a fungal metabolite that contaminates grain and legume supplies in the same parts of the world. Aflatoxin B₁ is a potent carcinogen in certain animal models, and epidemiological studies support the conclusion that it is also hepatocarcinogenic in humans (Chen et al., 1996; Groopman et al., 1996). Despite the fact that HCC is one of the few human cancers with known etiology, the molecular mechanisms involved in tumorigenesis are poorly understood and have only recently been investigated (Choi et al., 2004; Graveel et al., 2001; Meyer et al., 2003; Okabe et al., 2001).

The carcinogenicity of aflatoxins was first recognized in rainbow trout (Sinnhuber et al., 1968), which have subsequently proven to be an excellent research model for the study of human hepatocarcinogenesis induced by AFB₁ and other environmental carcinogens. The strengths of the trout model include its sensitivity to many classes of carcinogens, low spontaneous background tumor incidence, and low-cost husbandry for large-scale statistically valuable studies. More importantly, certain mechanisms of carcinogenesis have been well characterized in trout, including carcinogen metabolism, DNA adduction and repair, oncogene activation, and tumor pathology (reviewed by Bailey et al., 1996). Results from the study of tumor promotion and chemoprevention by environmental and dietary factors in AFB₁-initiated trout (Breinholt et al., 1995; Oganesian et al., 1999) have also been confirmed and extended in rodent studies (Manson et al., 1998; Stoner et al., 2002), as well as human clinical intervention trials (Egner et al., 2001). At present, however, the molecular mechanisms involved during tumorigenesis have not been described in trout and are not well understood in lower vertebrate models in general.

In this study, we describe construction of a custom rainbow trout 70-mer oligonucleotide array containing genes important for carcinogenesis, immunology, environmental toxicology, stress physiology, and endocrinology. Trout are a well-established and sensitive model for the evaluation of these different processes, and existence of an array targeted to these mechanisms will provide a powerful tool with which to evaluate global molecular changes in the trout model. We applied the arrays to assess molecular changes in trout tumorigenesis by examining transcriptional patterns in AFB₁-induced HCC compared to noncancerous adjacent liver and sham-initiated control liver. Transcriptional changes observed in this study were interpreted within the context of previously compiled data for pathological, physiological, and biochemical changes during carcinogenesis. Recent studies have suggested that comparative analyses of gene profiles across diverse species are more likely to highlight functional gene interactions important in key mechanisms of carcinogenesis (Segal et al., 2005). Global analysis of molecular signatures in trout tumors not only provided information about these processes in an important model for environmental carcinogenesis, but also the analyses were extended to identify mechanisms of tumorigenesis conserved throughout vertebrate evolution by comparing gene profiles in trout HCC with rodent and human HCC. Overall, we observed regulation of genes in a number of distinct functional classes important for carcinogenesis in trout, some of which indicate trout as a relevant model for invasive cancers and chronic inflammatory liver disease.

MATERIALS AND METHODS

Experimental animals and treatments.

Mt. Shasta strain rainbow trout (Oncorhynchus mykiss) were hatched and reared at the Oregon State University Sinnhuber Aquatic Research Laboratory in 14°C carbon-filtered flowing well water on a 12:12 h light:dark cycle. All animal protocols were performed in accordance with Oregon State University Institutional Animal Care and Use Committee guidelines. Approximately 400 embryos were initiated at 21 days post-fertilization with an aqueous exposure of 50 ppb AFB₁ (Sigma, St. Louis, MO) for 30 min. Sham-exposed embryos were exposed to vehicle alone (0.01% ethanol) and served as non-initiated controls. After hatching, fry were fed Oregon Test Diet, a semi-purified casein-based diet, ad libitum (2.8–5.6% body wt) 5 days per week for 50 weeks (Lee et al., 1991). Initiated and sham-exposed fish were maintained separately at a density of 100 fish per 400 l tank under the same conditions as described for rearing. At 13 months of age, trout were sampled for liver tumors over a 2-day period.

Necropsy and histopathology.

At termination, fish were euthanized by deep anesthesia with 250 ppm tricaine methanesulfonate. Fish body weights were recorded and livers were removed, weighed, and inspected for neoplasms under a dissecting microscope. After the size and location of all surface tumors was marked, a portion of each tumor and 100–200 mg of adjacent liver tissue were collected, placed separately in TRIzol Reagent (Invitrogen, Carlsbad, CA), and quick frozen in liquid nitrogen for gene expression analysis. Precautions were taken to avoid contamination of adjacent liver samples with tumor tissue, and samples were frozen within minutes of removal. The remaining liver was fixed in Bouin's solution for 2–7 days for histological analysis. Fixed livers were then cut into 1-mm slices with a razor blade to retrieve previously marked tumors. At least one piece of liver from each tumor-bearing fish was then processed by routine histological evaluation and stained with hematoxylin and eosin (H&E). Neoplasms were classified by the criteria of Hendricks et al. (1984) as either HCC or mixed carcinoma, and only HCC tumors were used for microarray analysis.

Microarray construction and quality control.

A rainbow trout 70-mer oligonucleotide array (OSUrbt ver. 2.0) containing 1672 elements representing approximately 1400 genes was created at Oregon State University (see Supplementary Data of the Supplementary Data online; also see http://www.science.oregonstate.edu/mfbsc/facility/micro.htm). Human sequences known to be involved in the physiological processes of carcinogenesis, immunology, endocrinology, toxicology, and other stress responses were chosen for the array based on several methods, including interrogation of literature, use of commercially available targeted microarray gene lists, and consultation of experts in respective fields for additional genes relevant for each process. The human full-length amino acid sequences for each gene were then blasted (tBLASTn) into GenBank and The Institute for Genome Research (TIGR) rainbow trout database (http://www.tigr.org/tdb/tgi/rtgi). Trout sequences with an E-value <10⁻¹⁰ were chosen as putative homologs. Oligonucleotides of 70 bases were designed for unique regions of each gene using ProbeSelect (Li and Stormo, 2001). Other oligonucleotide design considerations to minimize cross-hybridizations included proximity to the 3′ end of the gene, avoidance of internal self-annealing structures, and a narrow T_m range across the entire oligonucleotide set. For each gene, at least one 70-mer oligonucleotide was selected; however, for several genes, more than one distinct 70-mer oligonucleotide was included in the array. Oligonucleotides were synthesized by Operon Technologies (Alameda, CA) and Sigma Genosys (The Woodlands, TX).

Oligonucleotides were resuspended in spotting buffer (3× SSC and 1.5 M betaine) at 25 μM final concentration and were printed onto Corning UltraGap slides (Acton, MA) using a Biorobotics Microgrid II arrayer (Genomic Solutions, Ann Arbor, MI) with 16 pins. The trout 70-mer oligonucleotides were each printed in duplicate, and a set of 10 SpotReport Alien Oligos (Stratagene, La Jolla, CA) were printed 16 times (once per block) on the array. Buffer-only spots were included as negative controls. Test hybridizations of samples without alien spiking controls indicated that experimental cDNA does not cross-hybridize to alien spots on the array. After printing, slides were desiccated for 24–48 h and cross-linked with UV Stratalinker (Stratagene) at 600 mJ. Desiccation was chosen over baking as a post-processing method for drying slides, based on resulting spot quality as determined by morphology and spot criteria in test hybridizations (data not shown). To assess printing quality and oligonucleotide retention, all slides were scanned for red reflectance using a ScanArray 4000 scanner (PerkinElmer, Boston, MA), and at least one slide from each printing was stained with Syto 61 (Molecular Probes, Eugene, OR) and scanned at 633 nm. Slides were stored under desiccation for no longer than 6 months prior to use.

RNA isolation.

Total RNA was isolated from three individual trout HCC tumors and from corresponding adjacent liver tissue using TRIzol Reagent, followed by clean-up with RNeasy Mini Kits (Qiagen, Valencia, CA) according to the manufacturer's instructions. RNA was isolated from only those tumors that were histologically identified as HCC and that were of adequate size to yield at least 20 μg total RNA. RNA was also isolated from individual livers of 10 sham-initiated trout, and aliquots were pooled in equal amounts (μg) for use as a reference sample. RNA quality and quantity were assessed by agarose gel electrophoresis and spectrophotometric absorbance at 260/280 nm.

Microarray hybridization and analysis.

Hybridizations were performed with the Genisphere Array 350 kit and instructions (Hatfield, PA) using standard reference design with dye-swapping. Briefly, 7 μg total RNA were reverse-transcribed with Superscript II (Invitrogen) using the Genisphere oligo d(T) primer containing a capture sequence for the Cy3 or Cy5 labelling reagents. Each reaction was spiked with a range of concentrations (0.0049 – 2.5 ng/μl) of the 10 SpotReport Alien Oligo controls (Stratagene). Each cDNA sample containing the capture sequence for the Cy3 or Cy5 label was combined with equal amounts of reference cDNA (pooled from sham-initiated controls) containing the capture sequence for the opposite label. cDNA from two of the three biological replicates were dye-swapped and hybridized to two slides as technical replicates. cDNA from the reference sample was also hybridized to dye-swapped slides (against itself) following the same protocol as experimental samples for use as a negative control. Prior to hybridization, microarrays were processed post-printing by washing twice in 0.1% SDS for 5 min, 2× SSC, 0.1% SDS at 47°C for 20 min, 0.1× SSC for 5 min, water for 3 min, then dried by centrifugation. The cDNAs (25 μl) were hybridized to arrays in formamide buffer (50% formamide, 8× SSC, 1% SDS, 4× Denhardt's solution) for 16 h at 47°C with 22 × 25 mm Lifterslips (Erie Scientific, Portsmouth, NH). Arrays were then washed once in 2× SSC, 0.1% SDS at 47°C for 10 min, twice in 2× SSC, 0.1% SDS for 5 min, twice in 1× SSC for 5 min, twice in 0.1× SSC for 5 min and dried by centrifugation. Shaded from light, the Cy3 and Cy5 fluorescent molecules (3DNA capture reagent, Genisphere) were hybridized in formamide buffer for 3 h at 49°C to the corresponding capture sequences on cDNAs bound to the arrays. Arrays were washed in the dark with SSC containing 0.1 M DTT and dried as described earlier.

Scanned images (5 μm) were acquired with ScanArray Express (PerkinElmer) at an excitation of 543 nm for Cy3 and 633 nm for Cy5 and at 90% power. The photomultiplier tube (PMT) settings for each fluor were set based on intensity of spiked internal alien controls to normalize among all slides in the experiment. Image files were quantified in QuantArray (PerkinElmer; see Supplementary Data of the Supplementary Data online), and raw median signal and background values were exported to BioArray Software Environment (BASE; Saal et al., 2002) for analysis. Data were background subtracted and normalized by LOWESS, which is recommended for two-color experiments to eliminate dye-related artifacts and produce ratios that are not affected by signal intensity values (see Supplementary Data of the Supplementary Data online). Stringent criteria were used to filter for genes that were regulated at least twofold consistently in all features from biological replicates (n = 3 per treatment) and had a p value <0.05 by Welch's t-test (GeneSpring v.6, Silicon Genetics, Redwood City, CA). The genes that met these criteria were minimally categorized based on function using the Gene Ontology and Online Mendelian Inheritance in Man (OMIM) databases for putative homolog descriptions. Hierarchical clustering of gene expression profiles was performed in BASE; further filtering to identify similarities was performed with Array File Maker 4.0 (AFM; Breitkreutz et al., 2001), and comparisons of microarray and real time PCR gene regulation were performed with GraphPad Prism (GraphPad Software, San Diego, CA).

Real time RT-PCR.

To assess the authenticity of results from the microarray analyses, mRNAs for select genes were also analyzed by real-time reverse transcriptase polymerase chain reaction (RT-PCR). Genes for confirmation were chosen from each functional category, as determined by Gene Ontology, and were differentially upregulated or downregulated or resulted in no change. Total RNA was isolated as described previously and was treated with DNase (Invitrogen) according to the manufacturer's protocol. cDNA was synthesized from 2 μg RNA with an oligo (dT)₁₈ primer using SuperScript II (Invitrogen), following the manufacturer's instructions, with a final volume of 100 μl. Synthesized cDNAs (1 μl) were used as templates for amplification of specific gene products in total volumes of 20 μl containing 1× SYBR Green master mix (DyNAmo qPCR kit, Finnzymes, Finland) and 0.3 μM of each primer. Primer sequences were as follows: 5′-GCTGCCTCCTCTTCCTCTCT-3′ and 5′-GTGTTGGCGTACAGGTCTT-3′ for β-actin; 5′-GGATCACTTCTCACGTCCAC-3′ and 5′-TTAAACACAGTAAGCCCATC-3′ for chemotaxin; 5′-CTTTGTTTGACTCCGACACG-3′ and 5′-GAGAAATTTGCTTTTTGTGC-3′ for CD63; 5′-CAGCCACCTGTGGAATGCAC-3′ and 5′-AAAAATGGGATTCAATAGCC-3′ for urokinase plasminogen activator receptor (UPAR); 5′-CAAATACAGACGCGTTGGAC-3′ and 5′-GGCTGGTTCGTGACGGATGG-3′ for retinol binding protein; 5′-TTGCCTTTGCCAACATCGAC-3′ and 5′-CGGACATTGACGTATGCTTT-3′ for vitellogenin; 5′-GATGTCTTTCTCACTGCAACCT-3′ and 5′-GCTGTCTTTTTCCTGGTCACT-3′ for hepcidin; 5′-CCTGCGGCACGGTCTT-3′ and 5′-CTGACATCTTCATGCATCTCTTG-3′ for differentially regulated trout protein (DRTP). Primer sequences were chosen so that the product contained the 70-mer array oligonucleotide sequence to ensure validation of the microarray experiment, except for β-actin which was used only for normalization purposes. Polymerase chain reaction was performed with a DNA Engine Cycler and Opticon 2 Detector (MJ Research, Waltham, MA). Polymerase chain reaction was carried out for 35 cycles with denaturation at 94°C for 10 s, annealing at optimum temperature for primers (54°–60°C) for 20 s, and extension at 72°C for 12 s. DNA amplification was quantified (pg) from the C(T) value based on standard curves to ensure that quantification was within a linear range. Standards were created from gel-purified PCR products (QIAX II, Qiagen, Valencia, CA) for each primer set after quantification with a PicoGreen dsDNA Quantification Kit (Molecular Probes, Eugene, OR) and serial dilutions ranging from 0.25 to100 ng DNA. All signals were normalized against β-actin, and ratios were calculated for treated samples compared to sham-initiated control as for the microarray analysis. Expression of β-actin was not altered by treatment based on either microarray analysis or RT-PCR and so was found to be an appropriate housekeeping gene for normalization in this study.

RESULTS

Liver Histopathology

Exposure of trout embryos to 50 ppb AFB₁ resulted in 5% liver tumor incidence with only one tumor per animal at 13 months of age, whereas no tumors were observed in sham-initiated fish. This frequency is similar to what would be expected from historical AFB₁ exposure data (Oganesian et al., 1999). Tumors were determined to be approximately 30% HCC and 70% mixed carcinoma by histopathological examination, and only HCC tumors were further evaluated for gene expression profiles. Histological examination showed distinct structural differences between HCC and non-tumor tissues (Fig. 1). The trout liver from both non-initiated controls and noncancerous adjacent liver showed hepatocytes oriented in tubules with only two hepatocytes between adjacent sinusoids. However, HCC samples showed both increased basophilia and cellularity between adjacent sinusoids. These structural differences provided distinct borders between the HCC tissue and the surrounding liver.

FIG. 1.

Representative H&E-stained histological sections of (A) sham-initiated control liver, (B) noncancerous adjacent liver (NAL), (C) hepatocellular carcinoma (HCC), and (D) both HCC and NAL. (A, B) Hepatocytes oriented in tubules (shown in brackets) around a central bile canaliculus (arrow) with only two hepatocytes, end-to-end, between adjacent sinusoids. (C) Increased basophilia and cellularity between adjacent sinusoids (shown in brackets). (D) Border between HCC on the left and NAL on the right, with characteristics similar to those described for panels (A, B, C). Magnification 32×.

Gene Expression Analysis

The rainbow trout 70-mer spotted oligonucleotide array, OSUrbt v2.0, has been developed in a collaboration between the Marine and Freshwater Biomedical Sciences Center and the Center for Gene Research and Biotechnology at Oregon State University. The OSUrbt v2.0 array contains 1672 elements representing over 1400 genes important for processes of carcinogenesis, environmental toxicology, comparative immunology, endocrinology, and stress physiology. Existence of an array targeted to these mechanisms provides a powerful tool with which to evaluate global molecular changes in the trout model. The 70-mer oligonucleotides were designed for trout using multiple BLAST searches against annotated databases in GenBank and TIGR, and the resulting oligos are annotated by sequence homology based on the top BLAST hit (E-value <10⁻¹⁰; Supplementary Data and Supplementary Data of the Supplementary Data online). We report methods for spotting, post-processing, and hybridization of trout arrays for strong sensitivity and technical reproducibility (Fig. 2). Analyses of technical controls indicate that dye-swapped replicates have a high level of reproducibility, as do the duplicate spots printed on each slide. The correlation coefficients, r-values, were 0.93 (r² = 0.86) and 0.96 (r² = 0.93) for the duplicate slides and duplicate spots within an array, respectively, p < 0.001.

FIG. 2.

Pairwise analysis of gene profiles from (A) dye-swapped replicate slides and (B) replicate spots printed on each array. Values are fold change (log₂) compared to reference sample and were plotted to generate correlation coefficients (r) among the replicates.

Hepatic gene expression was analyzed using the OSUrbt v2.0 array to characterize the mRNA profiles for AFB₁-initiated HCC compared to noncancerous adjacent liver and sham-initiated control liver. As described in Materials and Methods, RNA from individual HCC and adjacent liver tissue was hybridized to arrays with dye-swapping. Gene expression was examined in tumors and liver samples, and genes were considered to be differentially expressed if their mRNA levels were consistently changed ≥ 2.0- or ≤ 0.5-fold and there was a p value <0.05 (Welch's t-test) among biological replicates (n = 3). Hierarchical clustering was used as a visualization tool to identify similarities among biological replicates within a treatment and differences in gene expression between treatments. Bidirectional hierarchical clustering of all genes on the array (Fig. 3A) and a filtered subset of genes that were differentially regulated within a treatment (Fig. 3B) revealed distinct separations among the treatments by Pearson correlation. The HCC samples and noncancerous adjacent liver samples were clustered to separate nodes in both instances, indicating that there were distinct gene expression patterns between the treatment groups.

FIG. 3.

Clustering of gene expression in trout liver by Pearson correlation in aflatoxinB₁-initiated HCC and noncancerous adjacent liver (NAL). Results are shown as fold change compared to sham-initiated control liver for biological replicates (n = 3 per treatment). Two of the three biological replicates were dye-swapped, and these data are also included. Red, upregulation; green, downregulation; black, unchanged expression; gray, missing values. (A) Gene expression profiles for all genes on the array. (B) Subgroup reflects gene expression profiles for genes differentially regulated twofold up or down in at least one treatment group.

Overall, 55 elements were differentially regulated between HCC and the noncancerous adjacent liver tissue (Table 1). Similarly, 40 elements were found to be regulated in HCC compared to sham-initiated control liver, whereas only 6 elements were expressed differentially in the noncancerous adjacent liver samples compared to sham-initiated controls. Most elements on the array represent different genes; however, a few elements represent distinct oligonucleotide sequences for the same gene. For example, vitellogenin, prostaglandin D synthase, UPAR, retinol binding protein, and DRTP are represented by multiple elements on the array and can serve to provide internal validation of our microarray results. In our negative control arrays, where the reference sample was hybridized against itself and dye-swapped, no genes were found to be differentially regulated. These arrays served to provide a technical validation of our microarray results.

TABLE 1

Select Genes Differentially Regulated in AFB₁-Initiated HCC Compared to Adjacent Liver Tissue

			Average fold change (p value)c,d
Array ID	TIGR IDa	Gene name (accession number, species)b	HCC/C	NAL/C	HCC/NAL	Reported in HCC or other cancers (OC)
Liver-specific proteins (vitellogenesis)
OmyOSU1540	TC65780	Vitelline envelope protein alpha (AF231706; Oncorhynchus mykiss)	0.22 (0.007)	1.15	0.19 (0.002)
OmyOSU1552	TC55460	Vitelline envelope protein gamma (AF231708; O. mykiss)	0.27 (0.001)	1.12	0.26 (0.004)
OmyOSU222	TC47576	Vitellogenin precursor (X92804; O. mykiss)	0.21 (0.015)	0.44	0.64
OmyOSU248	TC47577	Vitellogenin (X92804; O. mykiss)	0.34 (0.02)	0.63	0.57
OmyOSU203	TC47576	Vitellogenin precursor (X92804; O. mykiss)	0.33 (0.004)	1.31	0.32
Cell proliferation (signal transduction, transcription factors, and cytokines)
OmyOSU1615	TC50701	Transmembrane 4 superfamily member (AF281357; O. mykiss)	3.84 (0.003)	0.81	5.44 (0.008)	HCCe
OmyOSU39	TC61920	CD63 tetraspan protein (AY593998; O. mykiss)	3.60 (0.028)	0.95	3.79 (0.036)	HCCe,f
OmyOSU165	TC57084	Interferon inducible transmembrane protein 2 (AJ291989; O. mykiss)	3.48 (0.029)	0.77	5.34 (0.017)	OCg
OmyOSU983	TC86968	Interferon inducible transmembrane protein 1 (AJ291989; O. mykiss)	2.50 (0.012)	0.88	2.99 (0.006)	OCg
OmyOSU565	TC85655	Interluekin 8-like chemokine CK2 (AF418561; O. mykiss)	1.30	0.52	2.50 (0.009)	HCCh
OmyOSU804	TC65531	Transcription factor JunB (Q800B3; Fugu rubripes)	2.49	0.76	3.52 (0.005)	HCCi
OmyOSU435	TC54298	Calmodulin (J04046; Homo sapiens)	2.09	0.66	3.11 (0.035)	HCCj
OmyOSU227	NP543968	Estrogen receptor beta (AJ289883; O. mykiss)	2.11	0.72	2.94 (0.0009)	HCCk
OmyOSU347	TC54777	B-cell translocation gene 2 (AB036784; Danio rerio)	2.02	0.66	3.05 (0.013)	OCl
Protein/Ion stability and transport
OmyOSU471	TC71770	Procathepsin B (Q90WC3; O. mykiss)	2.39 (0.031)	1.02	2.41	HCCf,m
OmyOSU497	TC78415	Cathepsin L-like cysteine peptidase (AY332270; Tenebrio molitor)	2.35	0.61	3.82 (0.027)	HCCe,f,m
OmyOSU228	TC55313	Hepcidin (AF281354; O. mykiss)	3.26 (0.015)	0.40	8.33 (0.0006)	HCCn,o
OmyOSU685	TC54275	Ferritin middle subunit (S77386; Salmo salar)	0.36 (0.025)	0.49	0.77	OCp
OmyOSU707	TC86600	Ferritin H-3 subunit (D86627; O. mykiss)	0.32 (0.004)	0.57	0.60	OCp
Extracellular matrix and vascularization factors
OmyOSU1263	TC62562	Plasminogen activator, urokinase receptor (AF007789; Rattus norvegicus)	13.64d	0.95	13.26 (0.041)	HCCq
OmyOSU380	TC62562	Plasminogen activator, urokinase receptor (AF007789; R. norvegicus)	3.57 (0.028)	0.92	4.02 (0.027)	HCCq
OmyOSU562	TC62077	Collagen alpha 2(VIII) C1q (AF394686; Salvelinus fontinalis)	3.53 (0.03)	0.56	6.90 (0.006)	HCCe
OmyOSU744	TC51973	Putative interlectin, fibrinogen (AF281350; O. mykiss)	3.58	0.50	7.22 (0.017)	HCCe
OmyOSU1502	TC50691	Tissue factor, blood coagulation (AJ295167; O. mykiss)	0.10 (0.011)	0.34	0.87	HCCr
OmyOSU1551	TC54343	CD9 antigen, platelet aggregation (NM_212619; D. rerio)	0.42 (0.012)	1.31	0.32 (0.0006)	HCCs
OmyOSU864	TC72628	Hemagglutinin aggregation factor (M96983; Limulus polyphemus)	0.33 (0.036)	2.16	0.24d
Redox regulation
OmyOSU37	TC58221	Catalase (AAF89686; D. rerio)	0.63	2.11	0.30 (0.003)	HCCt
OmyOSU1422	TC47183	Thioredoxin (AAH49031; D. rerio)	2.54	0.54	4.83 (0.002)	HCCu
OmyOSU134	TC62472	Glutathione peroxidase 3 (AAH61950; Mus musculus)	2.24	0.94	2.46 (0.027)	OCv
OmyOSU238	TC56389	Glutathione peroxidase 4 (AAO86704; D. rerio)	2.06	0.81	2.59 (0.0009)	HCCe
Drug, lipid, glucose and retinol metabolism/homeostasis
OmyOSU1393	TC69788	Prostaglandin D synthase (AF281353; O. mykiss)	0.05 (0.001)	0.05 (0.0001)	1.06	OCw
OmyOSU1395	TC62304	Prostaglandin D synthase (AF281353; O. mykiss)	0.13 (0.019)	0.08 (0.0008)	0.62	OCw
OmyOSU1140	TC62082	Retinol binding protein (AF257326; O. mykiss)	0.36 (0.007)	1.19	0.30 (0.007)	HCCx
OmyOSU1142	TC62082	Retinol binding protein (AF257326; O. mykiss)	0.37 (0.016)	1.18	0.32 (0.014)	HCCx
OmyOSU352	TC72158	Cytochrome P450 2K1v2 (L11528; O. mykiss)	0.26d	0.84	0.30 (0.046)
OmyOSU354	TC72158	Cytochrome P450 2K3 (AF043551; O. mykiss)	0.23d	0.79	0.27d
OmyOSU356	TC63105	Cytochrome P450 2P2 (AF117342; Fundulus heteroclitus)	0.25d	0.90	0.26 (0.041)
OmyOSU1380	TC94217	Cytochrome P450 2K1v3 (AF045053; O. mykiss)	0.19 (0.005)	0.74	0.26 (0.005)
OmyOSU146	TC63282	Cytochrome P450 1A (AF059711; O. mykiss)	0.25 (0.044)	0.84	0.33 (0.049)	HCCe
OmyOSU460	TC62817	Aldo-keto reductase 1D1 (Z28339; H. sapiens)	0.17 (0.002)	0.89	0.19 (0.0002)	HCCe
OmyOSU829	TC46928	20β-hydroxysteroid dehydrogenase (AF100931; O. mykiss)	0.26 (0.022)	0.40	0.62	HCCe
OmyOSU164	TC58328	Insulin-like growth factor II (M95184; O. mykiss)	0.28 (0.018)	0.85	0.34 (0.002)	HCCy
OmyOSU1194	TC50776	Tyrosine aminotransferase (Q8QZR1; M. musculus)	0.56	1.71	0.35 (0.013)	HCCz
OmyOSU1451	TC69925	Cytosolic malate dehydrogenase (Q801E7; Oryzias latipes)	0.44	1.31	0.34 (0.006)	HCCaa
OmyOSU86	TC63296	Glucose-6-phosphatase (XM_702785; D. rerio)	0.55	2.01	0.28 (0.0.04)	HCCe
Potential immunoregulators and acute phase response proteins
OmyOSU268	TC71098	Chemotaxin (AF271114; O. mykiss)	3.73	0.24 (0.049)	16.90 (0.041)	HCCbb
OmyOSU232	TC91273	Differentially regulated trout protein (AF281355; O. mykiss)	4.49 (0.007)	0.30	23.94 (0.017)
OmyOSU148	TC91273	Differentially regulated trout protein (AF281355; O. mykiss)	3.91 (0.022)	0.30	19.82 (0.014)
OmyOSU1512	TC87050	Apolipoprotein A1 (AB183290; F. rubripes)	6.69 (0.044)	1.11	6.93 (0.039)	HCCf
OmyOSU766	TC78877	MHC class I heavy chain (AF296366; O. mykiss)	3.02	0.54	6.16 (0.025)	HCCcc
OmyOSU756	TC80478	MHC class IA heavy chain precursor (AF115518; O. mykiss)	3.22 (0.037)	0.87	3.86 (0.012)	HCCcc
OmyOSU527	TC87038	Precerebellin-like protein (AF192969; O. mykiss)	2.29	0.69	3.24 (0.022)
OmyOSU29	TC87038	Precerebellin-like protein (AF192969; O. mykiss)	1.78	0.74	2.41 (0.015)
OmyOSU533	TC71048	Novel protein similar to gliacolin, C1Q (AL627248; D. rerio)	2.69 (0.032)	0.98	2.83 (0.016)	OCdd
OmyOSU375	TC73041	Complement component C1q (XM_417653; Gallus gallus)	2.11	0.70	2.95 (0.041)	OCdd
OmyOSU407	TC87086	Complement component C7 (AJ566190; O. mykiss)	2.00	0.76	2.61 (0.018)	OCee
OmyOSU405	TC87533	Complement component 6 (BAD02321; H. sapiens)	1.93	0.72	2.62 (0.03)	OCee
OmyOSU219	TC55520	Complement factor Bf-1 (Q9YGE7; O. mykiss)	0.44 (0.003)	1.85	0.28 (0.029)
OmyOSU1477	TC55253	Trout C-polysaccharide binding protein 1 (AF281345; O. mykiss)	0.14 (0.033)	0.44	0.58
OmyOSU1478	TC55253	Trout C-polysaccharide binding protein 1 (AF281345; O. mykiss)	0.10 (0.031)	0.39	0.59

a

TIGR ID number of the tentative consensus or singleton expressed sequence tag (EST) sequence corresponding to OSUrbt ver. 2 microarray feature.

b

The most significant BLASTX is shown. If an EST has no significant (E-value <10⁻⁶) BLASTX hit, then the most significant BLASTN hit is shown (see Supplementary Data in the Supplementary Data online). Genes have been categorized by function based on putative trout homolog using the Gene Ontology and Online Mendeliam Inheritance in Man (OMIM) databases.

c

Average fold change values represent background corrected, Lowess normalized signal ratios. Stringent criteria were used to filter for genes that were regulated at least twofold consistently in all features from biological replicates and had a p value <0.05 by Welch's t-test. HCC = hepatocellular carcinoma; NAL = noncancerous adjacent liver; C = sham-initiated control liver. Fold change values for genes that passed stringency criteria are in bold and those that did not pass are shown in italics.

d

A few genes were consistently dysregulated >2.0- or <0.5-fold in all biological replicates, but had a p value >0.50 by Welch's t-test due to individual variability.

e

Meyer et al., 2003

f

Graveel et al., 2001

g

Jazaeri et al., 2002

h

Akiba et al., 2001

i

Borlak et al., 2005

j

Neo et al., 2004

k

Carruba et al., 2004

l

Ficazzola et al., 2001

m

Tumminello et al., 1996

n

Leong and Lonnerdal, 2004

o

Weinstein et al., 2002

p

Kakhlon et al., 2001

q

Zheng et al., 2000

r

Poon et al., 2003

s

Kanetake et al., 2001

t

Ding et al., 2004

u

Choi et al., 2004

v

Lodygin et al., 2005

w

Su et al., 2003

x

Okabe et al., 2001

y

Lu et al., 2005

z

Ouyang et al., 2002

aa

Thirunavukkarasu et al., 2001

bb

Ovejero et al., 2004

cc

Chen et al., 2005

dd

Bjorge et al., 2005

ee

Oka et al., 2001.

The genes most highly upregulated in HCC included those known to be involved in regulation of cell growth, formation and maintenance of the extracellular matrix, immunoregulation, and the acute phase response. Some immune-relevant genes, such as chemotaxin, were upregulated in HCC, but downregulated in the noncancerous adjacent liver, resulting in strong expression of these genes in HCC, 5- to 30-fold above the adjacent tissue. Genes commonly downregulated in HCC included those involved in drug, lipid, glucose, and retinol metabolism, immunoregulation, and vitellogenesis (estrogen-responsive liver proteins). The few genes differentially expressed in the noncancerous adjacent tissue compared to sham controls included prostaglandin D synthase and chemotaxin. Thus, expression of genes from distinct functional classes are altered during AFB₁-initiated HCC in rainbow trout compared to noncancerous adjacent or non-initiated liver and may be characteristic of molecular pathways important for tumorigenesis.

Real-Time PCR Confirmation

The expression profiles of selected genes that were found to be differentially increased or decreased in the microarray analysis, including chemotaxin, CD63, retinol binding protein, vitellogenin, UPAR, hepcidin, and DRTP, were confirmed by real-time RT-PCR using SYBR Green (Fig. 4). The same RNA preparations were used for each technique, and the mean expression ratios for all samples in each treatment (n = 3 biological replicates) were compared. Values for duplicate spots and dye-swapped slides were averaged prior to analysis of biological replicates for microarray data. Reverse transcriptase PCR was more sensitive in several cases than microarray analysis and detected greater changes, but it also resulted in higher variability among biological replicates in some of our assays. Overall, we were able to confirm gene expression profiles measured by oligonucleotide microarray analysis by RT-PCR. These data indicate that our strict criteria for determining differential gene regulation by array, including the twofold change in all biological replicates with p < 0.05, resulted in detection of meaningful changes that could be validated by other methods.

FIG. 4.

Hepatic gene expression in trout AFB₁-initiated HCC and noncancerous adjacent liver (NAL) measured by microarray and real-time reverse transcriptase polymerase chain reaction (RT-PCR.) HCC = hepatocellular carcinoma; NAL = noncancerous adjacent liver; C = sham-initiated control liver. Values are expressed as average fold change (log₂) with standard deviation (n = 3 biological replicates) compared to non-initiated control liver or noncancerous adjacent liver as indicated for select genes including (A) chemotaxin, (B) vitellogenin, (C) urokinase plasminogen activator receptor (UPAR), (D) hepcidin, (E) differentially regulated trout protein (DRTP), (F) CD63, and (G) retinol binding protein.

DISCUSSION

Evaluation of Gene Expression Profiles in Trout HCC

Rainbow trout have been used as a research model for environmental carcinogenesis for over 40 years. The strengths of this model have been described previously and include sensitivity to a number of carcinogens, low-cost husbandry, and well-described tumor pathology, carcinogen metabolism, DNA adduction, and mutational oncogene activation (Bailey et al., 1996). Recently, a number of studies supporting the strength of this model in biomedical research have been published; those studies have employed microarray technology in trout and other salmonid models to examine gene expression patterns and functional classes important for stress responses, chemical toxicology, and immune function (Krasnov et al., 2005a, 2005b; Rise et al., 2004). In this study, we applied the OSUrbt v2.0 array to further evaluate trout as a model for carcinogenesis based on molecular changes in AFB₁-induced HCC compared to noncancerous adjacent liver and sham-initiated control liver. Tumorigenesis is a multistage process that involves a number of genetic alterations during initiation, promotion, and progression of the disease. We detected distinct gene expression patterns in HCC compared to noncancerous tissue in rainbow trout; these are summarized in Table 1 and Figure 3. In our study, the noncancerous tissue surrounding HCC showed few transcriptional differences compared to non-initiated control liver, indicating that it still maintained a relatively “normal” phenotype. This observation is further supported by the structural similarities observed between the two noncancerous tissue types by histopathology (Fig. 1). Consequently, the genes regulated in HCC compared to both types of noncancerous tissue were very similar and included genes relevant for cell proliferation, synthesis of the extracellular matrix, vascularization, oxidative stress, immune response, and drug, lipid, and retinol metabolism and homeostasis.

Genes encoding some proteins important for cell signaling and proliferation were upregulated in trout HCC. JunB, a proto-oncogene involved in cell cycle regulation, was the only oncogene transcriptionally upregulated in our study. Previous studies have found that hepatic tumors in trout initiated by AFB₁ and other carcinogens have a high incidence (71–100%) of mutationally activated Ki-ras oncogenic alleles (Bailey et al., 1996). However, in this study, we did not observe transcriptional upregulation of the Ki-ras oncogene in HCC tumors. Also of interest was upregulation of the B-cell translocation gene 2 (BTG2), which is an effector protein of tumor suppressor genes important for regulation of the cell cycle. The B-cell translocation gene is an antiproliferative factor whose expression is upregulated in response to growth signals during tumorigenesis to promote cell cycle arrest, thereby helping to provide a “growth brake” (Ficazzola et al., 2001).

A number of genes associated with metabolism and homeostasis of drugs, lipids, glucose, and retinol were downregulated in trout HCC. Also downregulated were the liver-specific proteins involved in vitellogenesis. These transcriptional profiles suggest changes in normal liver function and differentiation in trout HCCs. Dedifferentiation in neoplastic development is supported by the step-wise progression from foci to benign to malignant tumors and has been well documented in rodents (Pitot et al., 1996). Histopathological evidence exists for a similar type of progression in fish models; for example, HCC has been observed to develop within benign adenomas in trout liver (Bailey et al., 1996; Okihiro and Hinton, 1999). The present data suggest that transcriptional profiles also support the process of dedifferentiation in trout neoplasia. Of particular interest is the decrease in expression of retinol-binding protein, which provides cellular storage for vitamin A in the liver. Low serum retinol, retinol binding protein, or vitamin A has been correlated with neoplastic HCC (Okabe et al., 2001; Schmitt-Graff et al., 2003). Retinol and vitamin A therapy have been applied to patients with HCC or chronic liver disease (Moriwaki et al., 2000), suggesting that reduced expression of these genes may play a crucial role in hepatocarcinogenesis.

Molecular Indicators of Metastatic Potential

A number of genes altered in trout HCC are involved in matrix-membrane associations, cell migration, and metastasis and may be indicative of tumor invasive potential. Urokinase plasminogen activator receptor, which was strongly upregulated, is a serine protease and part of the plasminogen activation system responsible for cell migration and matrix–membrane interactions important for cancer progression. Urokinase plasminogen activator receptor is on the cell membrane and promotes degradation of the extracellular matrix, which is considered to be an early step in the invasion and metastasis of cancer (Zheng et al., 2000). Two cysteine proteases, cathepsin L and pro-cathepsin B, were also upregulated in trout HCC. Like UPAR, these proteases function in protein stability, and their activity is attributed to the ability of tumor cells to invade the extracellular matrix (Graveel et al., 2001; Tumminello et al., 1996). A member of the transmembrane 4 superfamily, CD63, which regulates cell growth and motility, was strongly upregulated, and it also appears to influence tumor progression and metastasis in other models, serving as a good prognostic marker for disease (Graveel et al., 2001; Meyer et al., 2003). Other factors involved in coagulation and leukocyte aggregation were downregulated in trout HCC and have been indicated in tumor metastasis. For example, CD9, which is also a member of the transmembrane superfamily 4, is important for platelet aggregation. Downregulation of CD9 has been associated with tumor invasiveness and metastatic potential in a number of cancers (Kanetaka et al., 2001; Sauer et al., 2003). Also, the pro-inflammatory cytokine, IL-8, which was upregulated in trout HCC, has been found to promote angiogenesis in human HCC and may be important for vascularization required in both invasion and metastasis (Akiba et al., 2001).

These transcriptional profiles indicate that AFB₁-induced HCC in trout is of an invasive and potentially aggressive nature, which further enhances the merits of the trout as a model for human HCC. Trout HCCs are composed of broad tubules with many basophilic hepatocytes between adjacent sinusoids (Fig. 1) that are capable of distant metastasis and direct growth into surrounding tissues (Bailey et al., 1996). However, invasior is more common in trout over 2 years of age, and histopathological analysis of HCC tumors in this study does not indicate invasive growth. Therefore, the genes expressed in malignant trout tumors at this stage may be early predictors of metastasis and could provide molecular targets for treatment.

Trout as a Model for Chronic Liver Disease and Inflammation

Also of interest was the regulation of genes found to be important in chronic liver disease and inflammation, including cytokines, chemokines, and acute phase response proteins. Teleosts have a robust acute phase response, and several of their acute phase proteins are known to exist in mammals, where they function in innate immunity and inflammatory responses. Acute phase proteins are generally synthesized as an initial stress response to tissue injury, including such processes as malignant growth, and they were upregulated in trout HCC. Hepcidin, for example, is a liver hormone induced during inflammation that controls iron homeostasis by negatively regulating intestinal iron absorption; it was recently observed to be important in anemia of chronic liver disease (Leong and Lönnerdal, 2004). Iron-refractory anemia has been observed in liver adenomas of patients with glycogen storage disease caused by a deficiency in glucose-6-phosphatase, which results in an inability to maintain glucose homeostasis (Weinstein et al., 2002). We have previously observed disruption of iron homeostasis in trout neoplasms in studies where hepatocytes from carcinogen altered-foci and tumors were resistant to iron loading and were deficient in glucose-6-phosphatase (Hendricks et al., 1984; Lee et al., 1989). In this study, hepcidin upregulation was also correlated with transcriptional downregulation of glucose-6-phosphatase along with downregulation of ferritin subunits in HCC tumors. Although the molecular basis of anemia in chronic liver disease is not well understood, production of cytokines and interferons during inflammation has been correlated with anemia in patients and animal models. In our study a number of chemokines and cytokines were upregulated in trout HCC, which can regulate cell growth, differentiation, and inflammation during carcinogenesis. These data indicate that rainbow trout may also provide a novel and relevant model for the study of anemia and inflammation in chronic liver disease.

Trout as a Model for Human HCC

One of the inherent strengths of microarray platforms is the ability to extrapolate data across multiple species. Such comparative analyses can highlight mechanisms that have a key role in processes such as carcinogenesis. Studies that have examined the relationship of gene profiles across diverse species found that transcriptional responses conserved across evolution were more likely to correspond to true functional interactions (Segal et al., 2005). In this study, we found that most genes or gene classes differentially expressed in the trout tumor samples are conserved in human and rodent HCC or in other cancers (Table 1; Choi et al., 2004; Graveel et al., 2001; Meyer et al., 2003; Okabe et al., 2001). The fact that these studies used HCC of different etiologies (many viral) supports the likelihood that some processes of HCC pathogenesis have been conserved during long periods of evolutionary time.

In summary, we applied a novel rainbow trout oligonucleotide array to examine gene expression profiles in trout HCC compared to adjacent noncancerous tissue and identified distinct gene classes important in hepatocarcinogenesis in trout. Genes whose altered expression were identified through our microarray studies were typical of those observed in HCC in humans and other mammalian models. This finding is consistent with the notion that HCC pathogenesis has been conserved during vertebrate evolution. The genes that were regulated seem to indicate that AFB₁-induced HCC in trout is of an invasive and aggressive nature and may also be indicative of changes observed during chronic inflammatory liver diseases, which further enhances the merits of the trout as a model for human HCC. Future work will evaluate gene expression during progression over time of HCC in trout and examine how gene expression profiles are altered in the presence of dietary tumor modulators.

SUPPLEMENTAL DATA

Supplementary data are available online at Supplementary Data.

The authors thank Eric Johnson and Greg Gonnerman for care and maintenance of fish and Sheila Cleveland for histological preparation. This work was supported by National Institutes of Health (NIH) grants ES07060, ES03850, ES00210, ES11267, and CA90890.

Author notes

*Department of Environmental & Molecular Toxicology, Oregon State University, Corvallis, Oregon 97331; †Marine & Freshwater Biomedical Sciences Center, Oregon State University, Corvallis, Oregon 97331; ‡The Linus Pauling Institute, Oregon State University, Corvallis, Oregon 97331; §Department of Zoology, Oregon State University, Corvallis, Oregon 97331; ¶Center for Gene Research and Biotechnology, Oregon State University, Corvallis, Oregon 97331