| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Great Lakes WATER Institute, University of Wisconsin–Milwaukee, Milwaukee, Wisconsin, USA 53204
1 Institute for Marine Biosciences, National Research Council, Halifax, Nova Scotia, Canada B3H 3Z1
2 Aquaculture Division, Fisheries and Oceans Canada, 4160 Marine Drive, West Vancouver, British Columbia, Canada V7V 1N6
3 Department of Molecular Biology and Biochemistry, Simon Fraser University, Burnaby, British Columbia, Canada V5A 1S6
4 Centre for Biomedical Research, University of Victoria, Victoria, British Columbia, Canada V8W 3N5
(Requests for offprints should be addressed to R H Devlin; Email: devlinr{at}pac.dfo-mpo.gc.ca)
* Current address is Ocean Sciences Center, Memorial University of Newfoundland, St.John’s, Newfoundland, Canada A1C1S5 
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and methodsResultsDiscussionReferences |
|---|
and ß were R>C in all three studies. Comparisons of informative gene lists within and between studies inferred causes of altered gene expression. For example, ten genes, including 78 kDa glucose-regulated protein, glycerol-3-phosphate dehydrogenase, hemoglobins and ß, and a C-type lectin, were likely induced by GH transgenesis due to their presence in both T>C and R>C gene lists. Eleven genes, including hepcidin, nuclear protein p8, precerebellin-like, transketolase, and fatty acid-binding protein, were present in both T<C and R<C gene lists and were, therefore, likely suppressed by GH transgenesis. A large number of salmonid genes identified in these studies are involved in iron homeostasis, mitochondrial function, carbohydrate metabolism, cellular proliferation, and innate immunity. Pentose phosphate pathway genes phosphogluconate dehydrogenase, transaldolase, and transketolase, were dysregulated in GH transgenic samples relative to control samples. Changes in the expression of genes involved in maintaining hemoglobin levels (heme oxygenase, hemoglobins and ß, Kruppel-like globin gene activator, hepcidin) in R and T fish indicate a need for additional hemoglobin in the transgenic fish, perhaps due to higher metabolic rate required for enhanced growth.
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
GH transgenic salmonids experience extraordinary rates of growth in their 1st year, well before the onset of sexual maturation (Devlin et al. 1994, 2001, 2004a). Feed availability has a profound influence on growth rates of sexually naïve GH transgenic coho salmon (Oncorhynchus kisutch). Our objectives are to study the consequences of growth, GH transgenesis, and ration level, through comparisons of global hepatic gene expression data from three groups of sexually immature coho salmon: GH transgenic on a full ration (group T), GH transgenic on a restricted ration (group R), and control non-transgenic (group C). We have conducted three experiments, each in a different laboratory and using different cDNA microarray platforms containing primarily Atlantic salmon (Salmo salar) sequences. The microarray platforms used in this research were a 3554 gene (3.5K) microarray developed by the Genomic Research on Atlantic Salmon Project (GRASP; Rise et al. 2004b), a 16 006 gene (16K) GRASP microarray (von Schalburg et al. 2005b), and a 4104 gene (4K) microarray developed by the Institute for Marine Biosciences (IMB; Ewart et al. 2005). Heterologous hybridizations indicate that Atlantic salmon cDNA microarrays are effective tools for global gene expression studies involving any salmonid species (Rise et al. 2004b, von Schalburg et al. 2005b). The use of multiple microarray platforms increased the number of different genes, and therefore the proportion of the liver transcriptome, considered in the present study. Cross-platform comparisons confirmed results by identifying genes present in similar informative gene lists between microarray studies, and comparisons within and between studies inferred biological causes of altered gene expression. The functional annotations of informative genes, and the pertinent literature, point to hepatic molecular pathways and biological functions altered by GH transgenesis, growth, and/or ration restriction.
|
Materials and methods |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Coho salmon examined in this study were derived from the Chehalis River, which is a main tributary river of the Fraser River in southwestern British Columbia, Canada. All fish culture was conducted at Fisheries and Oceans Canada’s West Vancouver Laboratory, which is a non-commercial contained research facility specially designed to prevent the escape of genetically modified fish to the natural environment. GH transgenic coho salmon were initially produced by microinjecting a GH-gene construct (OnMTGH1) into the eggs from wild parents from the Chehalis River (Devlin et al. 1994). The strain used in the present study (M77) is stable and shows highly enhanced growth rates relative to wild-type sibling controls (Devlin et al. 2004a).
The objectives of the present study are to examine gene expression changes caused by GH transgenesis and enhanced growth. It is possible that there are both direct effects of GH transgenesis and indirect effects, which arise as a consequence of greatly accelerated growth. To separate these effects, three types of fish were examined in the present study: group T, transgenic fish reared on a full ration (which fully satiates their voluntary intake of food) for 9 months; group C, control non-transgenic coho salmon reared on a full ration for 21 months; and group R, transgenic salmon reared for 21 months on a restricted ration equivalent (per kilogram body weight) to that consumed by control salmon on a full ration. R salmon were pair fed in parallel with a group of C salmon of the same biomass, and were daily provided with the same amount of food consumed by the C group. Thus, group T grew at the very rapid rate typical of this GH transgenic strain, whereas groups C and R grew at similar rates typical of non-transgenic salmon. Use of fish of differing ages of T and C fish was necessary to ensure that fish were at the same developmental stages at the time of analysis. Comparisons of age-matched groups growing at their maximum growth rates are inappropriate due to the very large difference in size, which rapidly occurs between transgenic and non-transgenic individuals resulting in fish of different developmental stages (i.e. after some 4 months of growth, non-transgenic fish are still fry and are adapted for fresh water growth, whereas GH transgenic salmon are smolts, which are adapted for marine conditions). Age-matched samples at the same developmental stage have been generated by restricting the feeding of transgenic fish (group R), but these fish will differ due to effects from ration limitation. Thus, elucidation of the consequences of growth, GH transgenesis, and ration level requires assessment of all three groups (T, C, and R). Comparison of groups Tand C salmon allows examination of effects of both growth and GH transgenesis, comparison of groups T and R salmon reveals effects arising indirectly from growth rate and/or ration restriction, and comparison of groups C and R allows identification of effects of GH transgenesis and/or ration restriction.
Group T transgenic salmon used in this study were F5 generation progeny (year 2001 brood) produced by crossing a father homozygous for the transgene insert with five wild females derived from the Chehalis River. These crosses stably produce 100% transgenic hemizygous progeny (Devlin et al. 2004a). Group C were non-transgenic salmon (year 2000 brood) produced by crossing wild salmon sires with five wild females. Group R transgenic salmon were F4 progeny (year 2000 brood) derived as for group T, except using the same females as in group C, but homozygous fathers from a previous year class. Group R salmon were reared on a restricted, control level, ration throughout the pre-experiment period. The variance associated with intrafamily growth rates is very small relative to the difference in growth observed between transgenic and non-transgenic salmon. Nevertheless, transgenic strain M77 has been maintained in an otherwise wild genetic background by continuous backcrossing to wild fish. Thus, the transgenic and non-transgenic strains should on average differ genetically only at the transgenic locus. All fish groups were initially incubated in Heath trays during embryonic development followed by culture under parallel conditions in oxygen-saturated fresh well water at 10 °C with a flow rate in excess of 1 l/min/kg and a fish density <5 kg/m3. Stage-specific feeds were obtained from Skretting Canada (Vancouver, BC, Canada).
During the pre-sample period (August 21, 2002), 11 control non-transgenic (C), 11 ration-fed GH transgenic (R), and 11 full ration GH transgenic (T) post-smolt coho salmon were identified with unique passive integrated transponder tags implanted in their i.p. space, and placed in separate 200 l tanks and fed twice daily. C fish were fed to satiation, R fish were fed the same amount of food as C salmon (one R fish died and ration was adjusted accordingly), and T fish were fed to satiation. Length-specific growth rates (SGRL) and weight-specific growth rates (SGRW) were calculated using the equation, SGR=[ln(x2/x1)/(t2–t1)]x100, with x1 as length or weight at time t1, and x2 as length or weight at time t2 (Fig. 1
; see Table S1 in Supplementary data at http://jme.endocrinology-journals.org/content/vol37/issue2/). One-way ANOVA was used to assess differences in SGRL and SGRW. Tukey post-hoc tests were used for pairwise comparisons of growth rates for the C, R, and T treatment groups, and a significance level of P=0.05 was utilized.
|
Microarray analysis
Three different Atlantic salmon microarray platforms were utilized in this study: a 3.5K GRASP microarray, 16K GRASP microarray, and 4K IMB microarray. Differential gene expression in T, R, and C liver samples was determined using the 3.5K and 16K GRASP microarrays, and R and C samples were compared on the IMB microarray. Construction and initial testing of the GRASP microarrays described previously (Rise et al. 2004a,b, von Schalburg et al. 2005a,b). Briefly, clones were stringently selected from an expressed sequence tag (EST) database containing approximately 80 000 sequences derived from over 30 high-complexity salmonid cDNA libraries at the time the 3.5K microarray was designed and built, and over 300 000 sequences from over 175 high-complexity salmonid cDNA libraries at the time the 16K microarray was constructed. Inserts were amplified by PCR, cleaned, and printed as double, side-by-side spots on ez-rays aminosilane slides (Matrix) in 3x SSC. Arabidopsis thaliana cDNAs were spotted on each micro-array (Rise et al. 2004b, von Schalburg et al. 2005b) and used for thresholding (see Supplemental data at http://jme.endocrinology-journals.org/content/vol37/issue2/).
Construction of the IMB microarray has been described previously (Ewart et al. 2005). Briefly, 4104 clones were selected from four non-normalized salmonid cDNA libraries from liver, head kidney, macrophage, and spleen, as well as cDNA libraries constructed by suppression subtraction hybridization from the same tissues infected by the pathogen Aeromonas salmonicida. cDNAs were amplified by PCR, purified, resuspended in dH2O, and spotted in duplicate on Gap II glass slides (Corning, Corning, NY, USA) in 50% DMSO. The entire array of 8208 spots was reprinted on the lower half of the slide. Twofold dilutions of a plasmid encoding chlorophyll synthetase G4 from A. thaliana were included at the bottom of each sub-array to serve as internal controls.
RNA isolation, and microarray hybridization and analysis
Microarray experiments were designed to comply with Minimum Information About a Microarray Experiment (MIAME) guidelines (Brazma et al. 2001). For GRASP microarrays, all scanned microarray TIF images and quantified raw data files are available as online supplemental data at http://web.uvic.ca/cbr/grasp. Corresponding data for IMB microarrays are available as online supplemental data at http://www.imb.nrc.gc.ca/projects/salmonlivergeneexp/index_e.php. In addition, microarray TIF images and raw data (ImaGene or QuantArray files) have been deposited in gene expression omnibus (GEO; http://www.ncbi.nlm.nih.gov/geo/). For the 3.5K GRASP microarray experiment, the GEO platform accession number is GPL966 [NCBI GEO] ; the GEO sample accession numbers are GSM44847 [NCBI GEO] (slide DD035.038), GSM44848 [NCBI GEO] (DD035.040), GSM44849 [NCBI GEO] (DD035.041), GSM44988 [NCBI GEO] (DD035.042), GSM44989 [NCBI GEO] (DD035.043), GSM44990 [NCBI GEO] (DD035.044), GSM44991 [NCBI GEO] (DD035.046), GSM44992 [NCBI GEO] (DD035.047), and GSM44993 [NCBI GEO] (DD035.048); the GEO series accession number is GSE2388 [NCBI GEO] . For the 16K GRASP microarray experiment, the GEO platform accession number is GPL2989 [NCBI GEO] ; the GEO sample accession numbers are GSM78638 [NCBI GEO] (slide EB017.061), GSM78641 [NCBI GEO] (EB017.062), GSM78664 [NCBI GEO] (EB017.063), GSM78665 [NCBI GEO] (EB017.064), GSM78666 [NCBI GEO] (EB017.065), and GSM78667 [NCBI GEO] (EB017.066); the GEO series accession number is GSE3477 [NCBI GEO] . For the 4K IMB microarray experiment, the GEO platform accession number is GPL2844 [NCBI GEO] ; the GEO sample accession numbers are GSM78628 [NCBI GEO] (slide 12605187-L-n-bottom), GSM78629 [NCBI GEO] (slide 12605533-L-n-top), GSM78630 [NCBI GEO] (slide 12605532-L-n-bottom), GSM78631 [NCBI GEO] (slide 12605532-L-n-top), GSM78632 [NCBI GEO] (slide 12841776-L-n-bottom), GSM78633 [NCBI GEO] (slide 12841776-L-n-top), GSM78634 [NCBI GEO] (slide 12841777-L-n-bottom), and GSM78635 [NCBI GEO] (slide 12841777-L-n-top); the GEO series accession number is GSE3475 [NCBI GEO] .
GRASP microarrays
GRASP microarray studies were conducted using liver samples from fish of year class 1 (Fig. 2A and B). To minimize technical variability within a study, all targets were synthesized in one round, and all hybridizations were conducted simultaneously on slides from a single batch (DD035 for 3.5K study and EB017 for 16K study). Total RNA was individually prepared from flash-frozen adult liver tissues using TRIzol reagent and methods (Invitrogen). RNA from each fish was quantified and quality-checked by spectrophotometer and agarose gel respectively. The microarray experiments used pooled RNA templates (pooled C, pooled R, and pooled T; Fig. 2A and B), and each individual fish contributed an equal quantity of high-quality total RNA to a pool. Since the microarray studies involved pooled RNA samples, they did not provide information on biological variability of expression of informative genes. However, individual fish total RNA samples contributing to pools were archived at –80 °C, and used as templates in the quantitative reverse transcription-PCR (QPCR) validation of a selection of microarray-identified genes (Fig. 3). The QPCR data, therefore, reveal biological variability of expression levels for genes validated in this manner. Microarray experimental design (Fig. 2A and B) involved three comparisons (C versus R, C versus T, and R versus T), each run in triplicate (two replicates and one dye flip) for the 3.5K study, or duplicate (one microarray and a dye flip) for the 16K study. The microarray experiments included only technical replicates (i.e. with averaged biological variation contained within technical replicates), while QPCR experiments included both technical and biological replicates. Microarray hybridizations were performed using the 3DNA Array 50 Expression Array Detection Kit and instructions (Genisphere, Inc., Hatfield, PA, USA). The Array 50 instruction manual, which includes an explanation of the chemistry involved in the Array 50 labeling system, is available online at http://www.genisphere.com/pdf/array50v2_10_19_04.pdf. Briefly, 20 µg total RNA (pooled C, pooled R, or pooled T) were reverse transcribed using oligo d(T) primers with unique 5-prime sequence overhangs for the Cy3 or Cy5 labeling reactions. Microarrays were prepared for hybridization by washing 2x5 min in 0.1% SDS, washing 5x1 min in MilliQ H2O, immersing 3 min in 95 °C MilliQ H2O, and drying by centrifugation (5 min, 2000 r.p.m. in 50 ml conical tube). Microarray hybridizations were run in the dark under HybriSlips hybridization covers (Grace Biolabs, Bend, OR, USA) in slide hybridization chambers (Corning, Corning, NY, USA) submerged in a 48 °C water bath. The coho salmon liver cDNAs were hybridized to the salmonid cDNA microarray in a formamide-based buffer (25% formamide, 4x SSC, 0.5% SDS, 2x Denhardt’s solution) for 16 h at 48 °C. Coverslips were floated off at 48 °C in 2xSSC, 0.1% SDS buffer, and arrays were washed once for 10 min in 2x SSC, 0.1% SDS at 48 °C, twice for 5 min in 2x SSC, 0.1% SDS at room temperature (RT), twice for 5 min in 1xSSC at RT, and twice for 5 min in 0.1xSSC at RT, and dried by centrifugation as before. The Cy3 and Cy5 three-dimensional fluorescent molecules (3DNA capture reagent, Genisphere) were hybridized to the bound cDNA on the microarray in a formamide-based buffer (25% formamide, 4x SSC, 0.5% SDS, 2x Denhardt’s solution) for 3 h at 48 °C, and washed and dried as before.
|
|
IMB microarrays
The IMB microarray study was conducted using liver samples from fish of year class 2 (Fig. 2C
). mRNA, prepared from adult liver tissues preserved in RNALater (Ambion, Austin, TX, USA) using the FastTrack kit (Invitrogen), was quantified and quality-checked by spectrophotometer and agarose gel. Microarray target synthesis used 100 ng pooled C or pooled R mRNA template, and each pool contained approximately equal quantities from five individual fish. All experiments were conducted in quadruplicate (two replicates and two dye flips). The top array from each slide was scanned and analyzed separately from the bottom array, yielding eight data sets and 16 possible spots for each gene. Preparation of cDNA using the SuperSmart PCR cDNA Synthesis Kit (BD Biosciences, San Jose, CA, USA), amplification using the Advantage 2 PCR Kit (BD Biosciences), labeling of cDNA with Cy3- and Cy5-dCTP (Amersham), hybridization, and scanning were performed as previously described (Ewart et al. 2005). Normalization and data analysis were carried out using the GeneTraffic software (Iobion Informatics, La Jolla, CA, USA). Background fluorescence was subtracted from the spot intensities and spots were flagged and removed from the data set, if the signal to background ratio was <1, the signal to average background ratio was <1, or the signal intensity was <1000. Data were normalized on a sub-grid basis using both locally weighted non-linear regression (Lowess) and intensity-based methods. Differentially regulated genes were identified using the following filter: normalized log 2 ratios of spots >0.9 or <–0.9 in the arrays using RNAs from R versus C fish, eight or more (out of 16) unflagged spots contributing to this assessment, and an overall coefficient of variation for that spot of <0.3. Since biologically significant changes in gene expression may not be greater than twofold or <0.5-fold, the non-parametric significance analysis of microarrays (SAM; Tusher et al. 2001) method was carried out on the normalized data to identify differentially regulated genes based on their changes in expression in relation to the S.D. for multiple measurements. The false discovery rate was set at 1%, meaning that, on average, 1% of the genes would be classified incorrectly using this method. Genes were identified as up- or downregulated only if they were identified using both the filter and SAM methods.
Real-time quantitative reverse transcription (RT)-PCR (QPCR)
Validating QPCRs were run using RNA templates from the livers of individual fish in year class 1. Total RNA was prepared from flash-frozen C, R, and T coho livers using TRIzol reagent and methods (Invitrogen). For each individual, 250 ng RNA were reverse transcribed using 31.25 ng oligo d(T)16 primer. Reverse transcription reactions included 5 U RNase inhibitor, 500 µM dNTPs, 5.5 mM MgCl2, and 15.625 U MultiScribe Reverse Transcriptase with 1x TaqMan RT buffer (TaqMan Reverse Transcription Reagents Kit, Applied Biosystems, Foster City, CA, USA). Thermal cycling parameters for reverse transcription reactions were as follows: 25 °C for 10 min, 48 °C for 30 min, 95 °C for 5 min. First-strand cDNAs were diluted 1:3 and used as templates for QPCR analysis.
QPCR used two PCR primers per gene and SYBR Green I dye. Transcripts of interest and a normalizer gene were selected from the results of microarray experiments. For both the 3.5K and 16K GRASP studies, five different microarray features with ESTs identified as ubiquitin (top BLASTX hit: Oncorhynchus mykiss ubiquitin, accession number BAA88568 [GenBank] ) had normalized Cy5/Cy3 ratios between 0.5 and 2.0 on all slides. Ubiquitin was not dysregulated by GH transgenesis and/or ration restriction, and was therefore an appropriate normalizer gene for QPCR validation of microarray results. Primer pairs for transcripts of interest and ubiquitin were designed from alignments of the transcripts of interest and other available homologous sequences in GenBank using Primer Express software (Applied Biosystems) and the following guidelines: product size 150–250 bp, Tm 60± 1 °C, and at least two of the 3'terminal six bases G/C. For each gene, sequences of the forward and reverse QPCR primers respectively, are as follows: ubiquitin, 5'-CAA-CAGCGTCTGATCTTCGC-3', 5'-TTTGTCACAGTTG-TACTTCTGGGC-3'; iron–sulfur cluster assembly U2 (ISCU2), 5'-AGCTTCATCACATCTCCACAAGC-3', 5'-CAAGAAGGTAGTGGACCACTATGAGAA-3'; hepci-din antibacterial peptide, 5'-GTTGGAAGCTTTGA CAGTCCAGTT-3', 5'-AGAAGCCACAGCCAATGTTGT-3'; phosphogluconate dehydrogenase, 5'-CACCGTGG-CCTGTCCAGT-3', 5'-ACCACCGCCTTGTCCTTCTA-3'; and transaldolase 1, 5'-GCCGTCAGAGAGCTTCTCCA-3', 5'-AGAGCTGAGCAAAGACCACAGC-3'. Reactions (20 µl total volume) containing 4 µl diluted template, 300 nM each primer (with the exception of transaldolase at 50 nM), and 1xSYBR Green PCR Master Mix (Applied Biosystems), were run in duplicate using an ABI Prism 7000 Sequence Detection System (Applied Biosystems) and the following cycling parameters: 95 °C for 10 min, then 40 cycles of (95 °C for 15 s, 60 °C for 1 min). Controls (no template) were run for all primer pairs. Standard curves were generated by template dilution, with > 98.95% efficiency, for all primer pairs for relative quantitation by the ABI Prism 7000 Sequence Detection System.
Duplicate data (threshold cycle, or Ct, values) for each transcript of interest and for each individual were first normalized to ubiquitin (individual Ct value minus average ubiquitin Ct value for the same template). Data (normalized Ct values) from control non-transgenic, ration-fed transgenic, and full ration transgenic templates were compared (converted to fold differences) in Excel using the relative quantification method (Pfaffl 2001) and assuming 100% efficiencies. All Ct values and calculations are available online (Supplementary data, Tables S18–S22; http://jme.endocrinology-journals.org/content/vol37/issue2/). Melting curves for the QPCR products for all transcripts of interest and ubiquitin showed single peaks. Endpoint analysis on agarose gels showed a strong, single band of the expected size for the iron–sulfur cluster assembly U2 primers, but the primers for ubiquitin and the remaining transcripts of interest showed two bands of the appropriate size, as may be anticipated for transcripts derived from a tetraploid genome such as salmon. The two bands were extracted from the gel and sequenced. The resulting sequences matched one another and the EST sequence used for primer design.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
The following growth data pertain to the year class analyzed using the 3.5K and 16K GRASP microarrays (Figs 1 and 2; See Supplementary data, Tables S1 and S2 at http://jme.endocrinology-journals.org/content/vol37/issue2/). At the start of the pre-experiment period (t1=August 21, 2002), for both mean weights and length, C and R salmon did not differ but both were larger than T salmon (Fig. 1A and B; See Supplementary data, Table S1). The salmon in all treatment groups were re-measured for length and weight (Fig. 1A and B; See Supplementary data, Table S1) on September 13, 2002 (t2), at which time T salmon had grown to the same weight as C and R salmon and the same length as C salmon, but were slightly shorter than R salmon (Fig. 1A and B; See Supplementary data, Table S1). Specific growth rates over the period for both weight (SGRW) and length (SGRL) were significantly different among all groups, with SGRW in T salmon greater than eightfold higher than in C salmon and fourfold higher than R salmon (Fig. 1C and D; See Supplementary data, Table S1). Samples for the IMB microarray analysis were from a 2nd year class of fish, which were treated and grew in a similar fashion; however, growth rates were not specifically calculated.
Effects of GH transgenesis on global liver gene expression with and without rationing
Coho salmon transcripts responding reproducibly (as defined in Materials and methods) to GH transgenesis with and without feed rationing are reported in Tables S3–S16 (See Supplementary data at http://jme.endocrinology-journals.org/content/vol37/issue2/). Minimum microarray feature signal requirements (see Supplementary data, Table footnotes) were imposed to prevent inclusion of false positives in candidate informative gene lists due to random variation close to background levels. Changes in transcription levels, and degrees of similarity between salmonid cDNA ESTs and their top (most negative E-value) BLASTX or BLASTN hits against GenBank nr. or nt databases, are shown to identify candidate informative transcripts (See Supplementary data, Tables S3–S16). Informative unknown transcripts (cDNAs with no significant BLAST hits) are also reported. Informative transcripts in Tables S3–S14 (See Supplementary data) are listed by descending signal values in the appropriate channel (e.g. T for upregulated in T relative to C, or C for downregulated in T relative to C) of one representative slide in the study. This allows rapid assessment of relative expression level, with highly abundant transcripts towards the top of the list and rare transcripts towards the bottom. For the purposes of intra- and interplatform comparisons, informative salmonid features (or spots) having significant BLAST hits with the same gene name are called by that name. These transcripts may represent identical salmon genes, orthologous sequences (for GRASP microarrays, which are predominantly Atlantic salmon, but contain some rainbow trout sequences), or paralogous sequences.
Many of the microarray-identified hepatic transcripts responsive to GH transgenesis with and without rationing are involved in iron homeostasis, energy metabolism, mitochondrial function, cellular proliferation and stress, and innate immunity. We, therefore, present cross-platform microarray data relevant to these biological processes in Tables 1–5. In cross-platform comparisons, reproducibly informative (see Materials and methods) gene names and accession numbers from individual microarray studies were used to query the normalized data from other microarray studies (See Supplementary data, Tables S3–S17). In Tables 1–5, the reproducibly informative feature data are shown in regular font, and the cross-platform data are shown in italicized font. When cross-platform scans identified multiple same-named microarray features, data were taken from features with signal intensities above threshold and regulation (average fold change between samples and direction of change) most closely resembling that observed in the reproducibly informative platform. With a few exceptions, there is a good level of cross-platform confirmation of informative genes in Tables 1–5. For example, four genes in Table 1 (hepcidin, heme oxygenase, and hemoglobins and ß) receive cross-platform confirmation, while one gene (GABARAP) does not. In Table 2, data from four genes (phosphogluconate dehydrogenase, glycerol-3-phosphate dehydrogenase, prostaglandin D-synthase, and 
–6 fatty acyl desaturase) were confirmed by cross-platform comparison. In some instances, features assigned the same gene name between microarray platforms (or even within a platform) may represent distinct but related genes. This could explain apparent disagreements between platforms (i.e. GABARAP in Table 1, and barrier-to-autointegration factor in Table 5).
|
|
|
|
|
). Comparisons of informative gene lists within and between microarray studies allow identification of suites of genes with the following inferred causes of dysregulation (followed by criteria for inclusion in the suites): genes induced by transgenesis (gene name must appear in at least one T>C list and at least one R>C list, so Tand R> C); genes suppressed by transgenesis (gene name must appear in at least one T<C list andatleastoneR<C list, so T and R<C); genes indirectly induced by growth (gene name must appear in at least one T>C list and at least one R<T list, so T>C and R); genes indirectly suppressed by growth (gene name must appear in at least one T<C list and at least one R>T list, so T<C and R); genes induced by ration restriction (gene name must appear in at least one R>C list and at least one R>T list, so R>C and T); and genes suppressed by ration restriction (gene name must appear in at least one R<C list and at least one R<T list, so R<C and T; Fig. 4B).
|
. It is important to note that the mean fold change and S.E.M. values presented in Tables S3–S16 (See Supplementary data) pertain to technical replicates (i.e. data from replicate microarrays comparing gene expression in pooled samples), while statistics associated with QPCR experiments (See Supplementary data, Tables S18–S23) pertain to data from biological replicates.
3.5K GRASP T<C
In the 3.5K GRASP T versus C study on fish from year class 1 (Fig. 2A), 28 microarray features representing 26 different transcripts were reproducibly greater than twofold downregulated in T liver samples relative to C liver samples (See Supplementary data, Tables S3a and S3b; ‘a’ tables show mean fold change values for a non-redundant set of informative transcripts as well as the presence of same-named genes in other informative gene lists, and ‘b’ tables show data for each reproducibly informative feature from each contributing microarray). The most abundant downregulated transcript in T samples relative to C samples was alcohol dehydrogenase. Different microarray features with identical top BLAST hits in an informative gene list represent single genes or closely related paralogs. For example, there are two CCAAT/enhancer-binding protein delta microarray features and two iron–sulfur cluster assembly U2 (ISCU2) microarray features in Table S3 (see Supplementary data at http://jme.endocrinology-journals.org/content/vol37/issue2). The presence of multiple entries of genes in an informative gene list provides internal validation of microarray results. From 3.5K GRASP microarray data, transcripts most highly suppressed (>4.5-fold) in T relative to C liver were alcohol dehydrogenase (8.2-fold), metalloproteinase inhibitor 2 (8.2-fold), differentially regulated trout protein 1 (6.1-fold), and hepcidin antibacterial peptide (4.6-fold). QPCR validation of microarray results using tissue samples from individual fish of year class 1 showed a 2.3-fold suppression of ISCU2 (S.E.M. 0.2) and a 3.2-fold suppression of hepcidin antibacterial peptide (S.E.M. 0.5) in T relative to C liver samples (Fig. 3A and B; See Supplementary data, Tables S18–S20 and S23). QPCR with fish from year class 1 also showed that ISCU2 was 2.2-fold suppressed (S.E.M. 0.2) in R relative to C samples (Fig. 3A; See Supplementary data, Tables S19 and S23).
16K GRASP T<C
In the 16K GRASP T versus C study on fish from year class 1 (Fig. 2B), 28 microarray features representing 20 different transcripts were reproducibly greater than twofold downregulated in T liver relative to C liver (See Supplementary data, Table S9a and S9b). The most abundant downregulated transcript in Tsamples relative to C samples was an unknown EST (GenBank nt accession number CA038313
[GenBank]
). There are five nuclear protein p8, three differentially regulated trout protein 1, and three fatty acid-binding protein microarray features in Table S9 (See Supplementary data). From 16K GRASP microarray data, transcripts most highly suppressed (>4.4-fold) in T relative to C liver were nuclear protein p8 (7.3-fold), precerebellin-like protein (5.3-fold), alcohol dehydrogenase (4.7-fold), and fatty acid-binding protein (4.5-fold).
Overlap of T<C from studies 1 and 2 (Fig. 2A and B)
Four genes (alcohol dehydrogenase, differentially regulated trout protein 1, second mitochondria-derived activator of caspase protein, and hepcidin antibacterial peptide) were greater than twofold suppressed in T relative to C in both the 3.5K and 16K GRASP microarray studies (Fig. 4A).
3.5K GRASP T>C
In the 3.5K GRASP T versus C study on fish from year class 1 (Fig. 2A), 36 microarray features representing 27 different transcripts were reproducibly greater than twofold upregulated in T liver relative to C liver (See Supplementary data, Tables S4a and S4b). The most abundant upregulated transcript in T relative to C liver was retinoid-binding protein 7. There are five prosta-glandin D-synthase, three cytochrome c, three proliferating cell nuclear antigen (PCNA), and two H2A histone family member Z microarray features in Table S4 (See Supplementary data). From 3.5K GRASP microarray data, transcripts most highly induced (>4.1-fold) in T relative to C liver were cytochrome c (4.5-fold), replication factor C 4 (4.5-fold), H2A histone family member Z (4.3-fold), and prostaglandin D-synthase (4.2-fold). Since there was a high level of redundancy (providing internal validation) in this informative gene list, transcripts of interest for QPCR validation were selected from informative gene lists with less redundancy.
16K GRASP T>C
In the 16K GRASP T versus C study on fish from year class 1 (Fig. 2B), 31 microarray features representing 18 different transcripts were reproducibly greater than twofold upregulated in T liver relative to C liver (See Supplementary data, Tables S10a and S10b). The most abundant upregulated transcript in Tsamples relative to C samples was ubiquitin carboxy-terminal hydrolase L1. There are two C-type lectin, seven ß hemoglobin, two hemoglobin, and six prostaglandin D-synthase micro-array features in Table S10 (See Supplementary data). From 16K GRASP microarray data, transcripts most highly induced (>3.7-fold) in T relative to C liver were C-type lectin (8.9-fold) and prostaglandin D-synthase (3.8-fold).
Overlap of T>C from studies 1 and 2 (Fig. 2A and B)
Three genes (prostaglandin D-synthase, 20ß-hydroxy-steroid dehydrogenase B, and glycerol-3-phosphate dehydrogenase) were greater than twofold induced in T relative to C in both the 3.5K and 16K GRASP studies (Fig. 4A).
3.5K GRASP R<C
In the 3.5K GRASP R versus C study on fish from year class 1 (Fig. 2A), ten microarray features representing ten different transcripts were reproducibly greater than twofold downregulated in R liver relative to C liver (See Supplementary data, Tables S5a and S5b). The most abundant downregulated transcript in R relative to C liver was related to antifreeze protein LS-12 and is probably an apolipoprotein. From 3.5K GRASP micro-array data the transcripts most highly suppressed (>3.2-fold) in R relative to C liver were differentially regulated trout protein 1 (4.6-fold), cysteine protease CP14 (3.6-fold), guanylate-binding protein 5 (3.4-fold), and antifreeze protein LS-12 (3.3-fold). From this informative transcript list, hepcidin antibacterial peptide and transaldolase 1 were selected for QPCR validation, because their functions relate to immunity and carbohydrate metabolism respectively. QPCR validation of microarray results using tissue samples from individual fish of year class 1 showed a 2.8-fold suppression of hepcidin antibacterial peptide (S.E.M. 0.2) and a 2.3-fold suppression of transaldolase 1 (S.E.M. 0.2) in R relative to C liver samples (Fig. 3B and C; See Supplementary data, Tables S18, S20, S21, and S23). QPCR with year class 1 liver template samples and transaldolase 1 primers also showed 1.5-fold higher expression in T relative to C samples (S.E.M. 0.2), and 3.1-fold higher expression in T relative to R samples (S.E.M. 0.3; Fig. 3C; See Supplementary data, Tables S21 and S23).
16K GRASP R<C
In the 16K GRASP R versus C study on fish from year class 1 (Fig. 2B), 20 microarray features representing 16 different transcripts were reproducibly greater than twofold downregulated in R liver relative to C liver (See Supplementary data, Tables S11a and S11b). The most abundant downregulated transcript in R samples relative to C samples was thyroid hormone-inducible hepatic protein. There are three antifreeze protein LS-12, two nuclear protein p8, and two fatty acid-binding protein microarray features in Table S11 (See Supplementary data online). From 16K GRASP microarray data, transcripts most highly suppressed (>3.1-fold) in R relative to C liver were antifreeze protein LS-12 (3.9-fold), thyroid hormone-inducible hepatic protein (3.7-fold), transketolase (3.5-fold), and nuclear protein p8 (3.2-fold).
4K IMB R<C
In the IMB R versus C study on fish from year class 2 (Fig. 2C), the reproducibly downregulated transcripts in R liver relative to C liver encoded precerebellin-like protein, tachylectin, heme oxygenase, two complement C3 components, and four unknown products (See Supplementary data, Tables S15a and S15b).
Overlap of R<C between microarray studies
Two genes (antifreeze protein LS-12 and differentially regulated trout protein 1) were greater than twofold suppressed in R relative to C in both the 3.5K and 16K GRASP microarray studies (Fig. 4A). Two other genes (precerebellin-like protein and complement C3-3) were present in the R<C gene lists of both the 16K GRASP and the 4K IMB studies (Fig. 4A). Interestingly, the IMB microarray study found that differentially regulated trout protein 1 was upregulated in R relative to C and this was found to be significant using SAM analysis (see Materials and methods). None of the remainder of the down-regulated transcripts uncovered using the 3.5K GRASP microarray was confirmed using the IMB microarray, either because the intensities fell below threshold (e.g. hepcidin, GABAA receptor associated protein), because SAM analysis did not find them to be statistically significant (transaldolase, amoebocyte aggregation factor, antifreeze protein LS-12), or because they were not present on the IMB microarray (metalloproteinase inhibitor 2, retinoid X receptor , guanylate-binding protein 5, cysteine protease CP14). Using the IMB microarray with year class 2 fish samples, antifreeze protein LS-12 homolog was significantly downregulated in R relative to C head kidney, but not in liver (data not shown).
3.5K GRASP R>C
In the 3.5K GRASP R versus C study on fish from year class 1 (Fig. 2A), 22 microarray features representing 17 different transcripts were reproducibly greater than twofold upregulated in R liver relative to C liver (See Supplementary data, Tables S6a and S6b). The most abundant and fourth-most abundant upregulated transcripts in R relative to C liver were the and ß chains of hemoglobin respectively. There are three prostaglandin D-synthase, three cytochrome c, and two H2A histone family member Z microarray features in Table S6 (See Supplementary data online). From 3.5K GRASP micro-array data, transcripts most highly induced (>3.5-fold) in R relative to C liver were 3-phosphoglycerate dehydrogenase (8.2-fold), 78 kDa glucose-regulated protein (4.1-fold), nucleolar protein NOP5 (3.8-fold), and cytochrome c (3.6-fold). Due to the high level of internal validation within this informative gene list, transcripts of interest for QPCR validation were selected from informative gene lists with less redundancy.
16K GRASP R>C
In the 16K GRASP R versus C study on fish from year class 1 (Fig. 2B), three microarray features representing three different transcripts were reproducibly greater than twofold induced in R liver relative to C liver (See Supplementary data, Tables S12a and S12b). The most abundant upregulated transcript in R samples relative to C samples was hemoglobin . From 16K GRASP microarray data, transcripts reproducibly induced in R relative to C liver were hemoglobin (2.2-fold), C-type lectin (2.4-fold), and phospholipid hydroperoxide glutathione peroxidase (2.3-fold).
4K IMB R>C
In the IMB R versus C study on fish from year class 2 (Fig. 2C), the reproducibly upregulated transcripts in R relative to C samples encoded liver-expressed antimicrobial peptide 2, C-type lectin, barrier-to-autointegra-tion factor (also called breakpoint cluster protein), histone H5, a novel protein containing a ChaC-like domain, hemoglobins and ß, and two unknown products (See Supplementary data, Tables S16a and S16b). There are two barrier-to-autointegration factor microarray features in Table S16. One additional feature in this informative gene list came from a clone (BG935738
[GenBank]
) containing two inserts, one for hemoglobin and the other for monocyte chemoattractant protein-1 (See Supplementary data, Table S16b). Given the positive result from another spot containing hemoglobin , the signal from BG935738
[GenBank]
likely arises from hemoglobin :; however, there may also be a contribution from monocyte chemoattractant protein. Again, several of the transcripts identified using the 3.5K GRASP microarray were not confirmed using the IMB microarray due to low intensities, non-significance using SAM, or absence of the spot.
Overlap of R>C between microarray studies
Hemoglobin chain was greater than twofold induced in R relative to C in all three microarray studies (Fig. 4A). Hemoglobin ß chain was present in the R> C gene lists of both the 3.5K GRASP and the 4K IMB studies, and C-type lectin was present in the R>C gene lists of both the 16K GRASP and the 4K IMB studies (Fig. 4A).
3.5K GRASP R<T
In the 3.5K GRASP T versus R study on fish from year class 1 (Fig. 2A), 11 microarray features representing 11 different transcripts had greater than twofold lower expression in R relative to T liver in all three slides (See Supplementary data, Tables S7a and S7b). The most abundant transcript in this list was protein disulfide isomerase related. From 3.5K GRASP microarray data, transcripts with the highest T/R fold differences (> 3.4-fold) were an unknown (6.3-fold), phosphogluconate dehydrogenase (4.5-fold), GABAA receptor associated protein (4.2-fold), kinesin-like protein 2 (3.7-fold), -6 fatty acyl desaturase (3.6-fold), and apoptosis inhibitor survivin 1 (3.5-fold). The unknown, GABAA receptor associated protein, and -6 fatty acyl desaturase microarray features in this informative transcript list are also present in Table S4 (See Supplementary data online) (3.5K GRASP T>C; Fig. 4B), suggesting that these genes are induced in T relative to R rather than suppressed in R relative to T. QPCR validation of microarray results, using templates from fish of year class 1 and primers for phosphogluconate dehydrogenase, showed a 4.2-fold T/R difference (S.E.M. 0.2; Fig. 3D; See Supplementary data, Tables S18, S22, and S23). QPCR with year class 1 templates also showed that phosphogluconate dehydrogenase was 1.7-fold induced in T relative to C (S.E.M. 0.1), and 2.8-fold suppressed in R relative to C (S.E.M. 0.3; Fig. 3D; See Supplementary data, Tables S22 and S23).
16K GRASP R<T
In the 16K GRASP T versus R study on fish from year class 1 (Fig. 2B), eight microarray features representing seven different transcripts had greater than twofold lower expression in R liver relative to T liver in both slides (See Supplementary data, Tables S13a and S13b). The most abundant transcript in this list is C-type lectin, and there are two C-type lectin features in Table S13 (See Supplementary data online). From 16K GRASP microarray data, transcripts with the highest T/R fold differences (>4.0-fold) were C-type lectin (6.2-fold), an unknown (EST accession number CA038900
[GenBank]
; 5.9-fold), and TSC domain family protein 4 (4.1-fold).
Overlap of R<T from studies 1 and 2 (Fig. 2A and B)
One gene, phosphogluconate dehydrogenase, was greater than twofold induced in T relative to R liver in both the 3.5K and 16K GRASP studies (Fig. 4A).
3.5K GRASP R>T
In the 3.5K GRASP T versus R study on fish from year class 1 (Fig. 2A), nine microarray features representing nine different transcripts had greater than twofold higher expression in R relative to T liver in all three slides (See Supplementary data, Tables S8a and S8b). The most abundant transcript in this list was alcohol dehydrogenase. From 3.5K GRASP microarray data, transcripts with the highest R/T fold differences (> 3.8-fold) were 3-phosphoglycerate dehydrogenase (10.4-fold), immunoglobulin H chain variable region (4.3-fold), and alcohol dehydrogenase (3.9-fold). Since 3-phosphoglycerate dehydrogenase and translation initiation factor SUI1 are also found in Table S6 (See Supplementary data online) (3.5K GRASP R>C; Fig. 4B), these genes are likely induced in R relative to T rather than suppressed in T relative to R.
16K GRASP R>T
In the 16K GRASP T versus R study on fish from year class 1 (Fig. 2B), 11 microarray features representing ten different transcripts reproducibly had greater than twofold higher expression in R relative to T liver (See Supplementary data, Tables S14a and S14b). The most abundant transcript in this list was precerebellin-like protein. There are two nuclear protein p8 microarray features in Table S14 (See Supplementary data online). From 16K GRASP microarray data, transcripts with the highest R/T fold differences (>2.8-fold) were precerebellin-like protein (6.8-fold), nuclear protein p8 (3.0-fold), and alcohol dehydrogenase (2.9-fold).
Overlap of R>T from studies 1 and 2 (Fig. 2A and B)
Two genes, RNA terminal phosphate cyclase-like 1 and alcohol dehydrogenase, had greater than twofold higher expression in R relative to T liver in both the 3.5K and the 16K GRASP studies (Fig. 4A). Since alcohol dehydrogenase is also present in the T<C gene lists from both the 3.5K GRASP study (See Supplementary data, Table S3) and the 16K GRASP study (See Supplementary data, Table S9; Fig. 4A and B), this gene is likely suppressed in T relative to R rather than induced in R relative to T.
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
The functional annotations of informative salmonid liver transcripts responding to GH transgenesis with and without food rationing point to molecular pathways altered under these different genotypic and nutritional conditions. A large number of salmonid genes ident-ified in these studies are involved in iron homeostasis, mitochondrial function, carbohydrate metabolism, apoptosis, cellular proliferation, and innate immunity. The discussion is, therefore, focused primarily on these biological processes, with particular attention paid to genes (or gene names) with expression profiles that were confirmed within and/or across platforms, and informative transcripts with growth-relevant functional annotations.
Iron homeostasis and signal transduction
GABARAP, hepcidin, and hepatic transcription factors
In the 3.5K GRASP study, GABAA receptor associated protein (GABARAP) was upregulated in T relative to C and T relative to R liver, and downregulated in R relative to C liver (Table 1; T>C>R). Thus, in this study, GABARAP was found to be both indirectly induced by growth (T>C and R) and suppressed by ration restriction (R<C and T; Fig. 4B). A transcript identified as GABARAP was not informative in the 16K GRASP study (Table 1), and potential causes of such apparent disagreement between platforms were previously presented. GABARAP is extremely highly conserved in the metazoa, with amino acid identity values ranging from 93 to 98% in such organisms as fruit fly (Drosophila melanogaster), amphioxus (Branchiostoma belcheri tsing-taunese), and zebrafish (Danio rerio). Human GABARAP, first identified by a yeast two-hybrid system using parts of the GABAA receptor as bait, bears significant amino acid sequence similarity to light chain-3 of microtubule-associated proteins (MAPs) 1A and 1B (Wang et al. 1999). GABARAP is ubiquitously expressed, suggesting that it has likely generalized or multiple functions in addition to its putative role in moving, sorting, and clustering GABAA receptors (Wang et al. 1999, Kneussel & Betz 2000, Stangler et al. 2002). GABARAP binds transferrin receptor (TfR) and potentially plays a role in trafficking or cycling of TfR (Green et al. 2002). Liver TfR plays a key role in iron metabolism, delivering iron-bound transferrin to the endosomal compartment and stimulating hepatocytes to release hepcidin (Green et al. 2002, Sharma et al. 2005). In addition to GABARAP, other genes involved in iron homeostasis (e.g. hepcidin and heme oxygenase) were also dysregulated in the GH transgenic coho salmon experiments giving rise to this report.
Microarray (3.5K and 16K GRASP) and QPCR data on tissues from fish of year class 1 (Fig. 2A and B) indicated that a coho salmon transcript encoding a putative polypeptide related to the acute phase protein hepcidin was downregulated in both T and R relative to C liver (Table 1; Figs 3B and 4A and B). Hepcidin was 1 of 11 genes found to be suppressed by GH transgenesis (T and R<C; Fig. 4B). Mammalian hepcidin (or liver-expressed antimicrobial peptide-1) genes are expressed primarily in liver, and hepcidin is believed to be among the most abundant transcripts in adult mammalian liver (Pigeon et al. 2001, Courselaud et al. 2002). Similarly, coho salmon hepcidin was the fourth-most abundant transcript in the 3.5K GRASP T<C list (See Table S3 at http://jme.endocrinology-journals.org/content/vol37/issue2), and the third-most abundant transcript in both the 3.5K GRASP R<C list (See Table S5 in Supplementary data) and the 16K GRASP T<C list (See Supplementary data, Table S9).
Hepcidin is thought to play an important role in vertebrate innate immune responses to microbial challenge (Shike et al. 2004, Lauth et al. 2005). Mammalian hepcidin proteins exhibit antibacterial and antifungal activity (Krause et al. 2000, Park et al. 2001). Hepcidin expression is upregulated in human (Nemeth et al. 2003) and mouse (Pigeon et al. 2001) hepatocytes exposed to lipopolysaccharide, and also in Atlantic salmon (Douglas et al. 2003), zebrafish (Shike et al. 2004), and striped bass (Lauth et al. 2005) liver tissues in response to bacterial infections. In addition to its conserved function in innate immunity, hepcidin also serves as a modulator of iron homeostasis (Courselaud et al. 2002, Lee et al. 2005). Hepcidin expression in mammalian liver is strongly induced by iron excess (Pigeon et al. 2001), and strongly suppressed by anemia and hypoxia (Nicolas et al. 2002). Therefore, the suppressed hepcidin expression observed in GH transgenic coho salmon liver regardless of nutritional status suggests that these fish should have compromised innate immunity, as well as altered iron homeostasis, specifically increased iron availability for hemoglobin biosynthesis. Challenge studies have indicated that the strain of GH transgenic salmon used in the present study do indeed have reduced resistance to bacterial infections (Jhingan et al. 2003). Hemoglobin an
