Short-Term Feed Deprivation Alters Immune Status of Surface Mucosa in Channel Catfish (Ictalurus punctatus)

Short-term feed deprivation (or fasting) is a common occurrence in aquacultured fish species whether due to season, production strategies, or disease. In channel catfish (Ictalurus punctatus) fasting impacts susceptibility to several bacterial pathogens including Flavobacterium columnare, the causative agent of columnaris disease. As columnaris gains entry through the gills and skin of fish, we examined here changes in transcriptional regulation induced in these surface mucosal tissues due to short-term (7 day) fasting. RNA-seq expression analysis revealed a total of 1,545 genes perturbed by fasting. Fasting significantly altered expression of critical innate immune factors in a manner consistent with lower immune fitness as well as dysregulating key genes involved in energy metabolism and cell cycling/proliferation. Downregulation of innate immune actors such as iNOS2b, Lysozyme C, and peptidoglycan recognition protein 6 is predicted to impact the delicate recognition/tolerance balance for commensal and pathogenic bacteria on the skin and gill. The highlighted expression profiles reveal potential mechanistic similarities between gut and surface mucosa and underscore the complex interrelationships between nutrition, mucosal integrity, and immunity in teleost fish.


Introduction
Alterations in feeding regimen are common practice in the modern aquaculture industry, predicated in part on research indicating that fasting is well tolerated by most fish species [1,2]. Food deprivation strategies can be employed as part of a seasonal feeding pattern [3], in response to overproduction, or as a response to disease outbreak [4,5]. When diseases prompt changes in feeding, producers must weigh potential savings on feed and the benefits of limiting fish-fish contact during feeding against reduced growth and heightened stress stemming from nutrient restriction. Traditionally missing in the equation has been any understanding on how short-term fasting may impact the immune status of farmed fish.
Most feed deprivation investigations to-date have focused on changes in intestinal cellular morphology and enzyme activity or on transcriptional changes in critical regulators of protein synthesis or glucose metabolism in tissues such as liver, muscle, and intestine [6][7][8]. Recent studies have documented changes in gene mediators of innate immunity following 4 week starvation studies in rainbow trout intestinal epithelia [9] and Atlantic salmon liver [10]. Comparatively little is known regarding the immune consequences of more common short-term fasting events. However, a small body of previous research indicates that short-term withdrawal of feed, while not visibly impacting growth parameters, can have striking effects, particularly at mucosal surfaces. Vieira et al. [11] reported broad impacts of a one week feed restriction regimen on skin healing and scale regeneration in sea bream utilizing an oligo microarray. Krogdahl and Bakke-McKellep [7] observed rapid 20-50% decreases in intestinal tissue mass and enzyme activities after two days of fasting in Atlantic salmon. Perhaps of greatest relevance here is the recent report that a seven day feed deprivation caused significant changes in microbial density and community composition in the cutaneous mucus of Atlantic salmon [12].
The impact of feed deprivation on mortality rates of channel catfish (Ictalurus punctatus) exposed to the bacterial pathogens Edwardsiella ictaluri and Flavobacterium columnare has previously been established [4,5]. F. columnare, a widespread opportunistic pathogen of freshwater fish, and a leading cause of mortality in the U.S. catfish industry [13], causes higher losses in fish that have been withheld from feed for as little as seven days [4]. Attachment and entry of F. columnare into channel catfish is via surface mucosa, skin and gill. Recently, we identified a rhamnose-binding lectin (RBL) whose expression is strongly induced in channel catfish gill by F. columnare infection [14]. A more detailed subsequent study of RBL activity revealed that expression levels in gill were correlated with columnaris susceptibility, and that saturation of the receptor with its putative ligands resulted in significantly decreased columnaris mortality. Furthermore, RBL expression increased greater than 100-fold in the gill tissue of channel catfish fingerlings fasted for seven days [15]. Given these results and the impact of feeding status on columnaris susceptibility, we wished to examine further the broader molecular effects of short-term fasting of channel catfish on surface mucosal health. Towards that end, here we utilized RNA-seq-based transcriptome profiling of skin and gill homogenates from fed and 7 d fasted channel catfish fingerlings to better understand immune-nutritional regulation in teleost fish.

Results
Sequencing of short expressed reads from channel catfish gill and skin A total of 209 million 100 bp high quality reads were generated for the fasted and fed samples. Greater than 26 million reads were generated for each of the six libraries. Raw read data are archived at the NCBI Sequence Read Archive (SRA) under Accession SRP017689.
De novo assembly of channel catfish gill and skin transcriptome Given the importance of assembly of long, accurate contigs to capture channel catfish genes and to correctly identify differential expression, we compared two prominent options for de novo transcriptome assembly: Trans-ABySS and Trinity. We had previously developed an in-house bioinformatics pipeline around Trans-ABySS [14,16] and demonstrated its superior performance in comparison to use of CLC Genome Workbench or Velvet assemblers. However, we sought to determine whether use of Trinity [17] would improve assemblies further.
Trans-ABySS. Use of Trans-ABySS to merge ABySS multi-kassembled contigs, resulted in approximately 301,500 contigs with average length of 1,036.1 bp and N50 size of 1,716 bp, with 24,345 contigs longer than 1,000 bp. After removing redundancy using CD-Hit and CAP-3, about 57.53% contigs were kept, resulting in a final assembly of 173,459 unique contigs with average length 887.3 bp (Table 1).
Trinity. Trinity generated approximately 281,595 contigs in its initial contig assembly with average length of 1,046.9 bp and N50 size of 2,270 bp, with 30,202 contigs longer than 1,000 bp. After removing redundancy, about 96.67% contigs were kept, resulting in a final 272,229 contigs with average length 996.5 (Table 1).
Gene identification and annotation. BLAST-based gene identification was performed to annotate the channel catfish gill/skin transcriptome and inform downstream differential expression analysis. After gene annotation, 60,892 contigs from the Trans-ABySS assembly had significant BLAST hits against 16,610 unigene (unique gene) matches from zebrafish, the closest available reference genome to channel catfish (Table 2). Using a more stringent criteria of a BLAST score $100 and E-value # 1e-20 (quality matches) identified 14,679 zebrafish unigene matches. The same BLAST criteria were used to annotate the Trans-ABySS assembly based on matches against the UniProt and NCBI nr (non-redundant) databases.
In contrast, 82,365 Trinity contigs had a significant BLAST hit against 17,892 unique zebrafish genes (Table 2). 15,821 unigenes were identified based on hits to the zebrafish database with the more stringent criteria of a BLAST score $100 and E-value # 1e-20. As with the Trans-ABySS assembly, the same BLAST criteria were used in comparison of the Trinity reference contigs with the UniProt and nr databases. The largest number of matches was to the nr database with 83,972 contigs with putative gene matches to nr and 20,589 quality unigene matches ( Table 2).
Best assembly selection. In a comparison of the assemblies generated by Trans-ABySS and Trinity (Table 1), it was clear that although Trans-ABySS consistently generated a larger initial number of contigs, redundancy was much higher than that observed with Trinity. CD-HIT/CAP3 removed over 40% of Trans-ABySS contigs due to this redundancy, while almost all Trinity contigs were carried forward after this process. The final Trinity assembly contained almost 100,000 more contigs than Trans-ABySS due to Trinity's superior ability to distinguish splicing isoforms and gene paralogs. Trinity contigs also had larger N50 and average length, 2,270 bp and 996.5 bp, respectively, than Trans-ABySS. These metrics reflect Trinity's superior ability to map paired end reads into the same contig. Trinity mapped 79.9% of reads in pairs versus 66.3% in Trans-ABySS ( Table 1). The superior de novo assembly produced by Trinity was also reflected in the number of quality unigene matches against zebrafish (15,821 vs. 14,679) and nr (20,589 vs. 15,863) databases. Given these results, we utilized the Trinity assembly for subsequent analysis of differential expression.

Identification and analysis of differentially expressed genes
A total of 1,545 genes (unique annotated contigs with significant BLAST identities) were differentially expressed greater than 1.5fold, with 412 up-regulated genes and 1,133 down-regulated genes ( Table 3; Table S2). Short read coverage within differentially expressed contigs is critical for accurate quantification of expression. We obtained good coverage of differentially expressed contigs, with an average of 784.6 reads/contig.

Enrichment and pathway analysis
A total of 3,482 GO terms including 970 (27.86%) cellular component terms, 1,012 (29.06%) molecular functions terms and 1,491 (42.82%) biological process terms were assigned to 1,545 unique gene matches. The percentages of annotated channel catfish sequences assigned to GO terms are shown in Fig. S1. The differently expressed unique genes were then used as inputs to perform enrichment analysis using Ontologizer. A total of 125 terms with p-value (FDR-corrected) ,0.05 were considered significantly overrepresented. Ten higher level GO terms were retained as informative for further pathway analysis (Table S3). The GO terms included response to stress, regulation of cell death, and cell cycle regulation.
Based on enrichment analysis and manual annotation and literature searches, representative key genes were arranged into three broad categories, including immune response, energy

Validation of RNA-seq profiles by QPCR
We selected 10 genes for QPCR confirmation, choosing from those with differing expression patterns and from genes of interest based on functional enrichment and pathway results. Samples from fasted and fed channel catfish (with three replicate sample pools per timepoint) were used for QPCR. Melting-curve analysis revealed a single product for all tested genes. Fold changes from QPCR were compared with the RNA-seq expression analysis results. As shown in Fig. 1, QPCR results were significantly correlated with the RNA-seq results (average correlation coefficient 0.91, p-value ,0.001; Figure 1). With the exception of anterior gradient protein 2 (AGR2), all examined genes had the same direction of differential expression by both methods indicating the reliability and accuracy of the Trinity reference assembly and RNA-seq-based transcriptome expression analysis.

Discussion
While it is well understood that nutritional perturbations can modulate resistance to infectious diseases in agricultural animals [18], the molecular mechanisms by which dietary changes alter fish host immunity are largely unknown. The impact of these changes would be expected to differ depending on duration of feed withdrawal (fasting vs. starvation), species-specific evolutionary adaptation to natural levels of feed availability, season, and pathogen dynamics (prevalence, routes of infection, etc). Short (one week) periods of feed deprivation in channel catfish can decrease mortality rates to a Gram-negative bacterial pathogen, E. ictaluri [5], but, conversely, appear to lead to higher mortality to F. columnare [4]. Given that F. columnare gains entry through surface mucosa, we examined here transcriptomic changes in channel catfish gill and skin following a 7 d feed deprivation challenge. RNA-seq analysis revealed a total of 1,545 genes with expression perturbed by fasting. Fasting significantly altered expression of critical innate immune factors in a manner consistent with lower immune fitness as well as dysregulating key genes involved in energy metabolism and cell cycling/proliferation.
Studies of feed deprivation have traditionally focused on direct impacts on the digestive tract [19] or on systemic alterations in energy homeostasis [20]. Our understanding of the importance of mucosal surfaces as the front line in host immunity and the primary target for pathogen invasion continues to grow [21,22]. However, the sensitivity of mucosal defensive barriers to nutritional changes is only now being explored in fish [12,23]. Advances in next-generation sequencing now allow rapid, comprehensive analysis of the molecular underpinnings of these phenomena through RNA-seq approaches [24]. In order to gain a perspective on the molecular actors in the surface mucosa responding to fasting we pooled equal amounts of skin and gill from each fish contributing to the replicated tissue pools utilized for RNA-seq analysis. Clearly, by pooling heterogeneous cell populations stemming from the two tissues the potential exists for masking or confusing gene expression patterns from particular cellular subsets. However, we accepted this compromise in this initial study to more broadly capture novel genes and patterns in the surface mucosa. Future research will seek to ascertain which tissues and constituent cell types are contributing to key transcript profiles using laser capture micro-dissection and antibody-based cell sorting.
A total of 1,545 differentially expressed contigs could be annotated based on BLAST analysis (Table S2). We attempted to classify key differentially expressed genes into broad functional categories based on GO annotation and manual imputation of putative function via literature search of studies in vertebrate model organisms (Table 4). Below we highlight several important pathways likely mediating the channel catfish mucosal response to short-term (7 d) fasting.

Immune Function
Recent work by our group examined how the gill transcriptome differed between F. columnare resistant and susceptible channel catfish both basally and at early timepoints following infection     [25]. That dataset, combined with recent analysis of pathogenmediated changes in the catfish skin and intestinal transcriptomes [16,26,27], provided a foundational reference of genes likely modulating immunity in mucosal tissues useful for comparison here. Initial analysis revealed enrichment of genes with immune functions among the mostly highly differentially expressed genes following fasting. Of the top 15 upregulated genes following fasting, 11 had immune functions according to a combination of GO annotation and manual literature searches. Similarly, 6 of the 15 most highly downregulated genes also had known immune functions. Among immune genes perturbed by fasting were acidic chitinase precursor genes which were not sequenced in fed fish (0 reads), but were present with between ,30-125 reads in the fasted replicates. Previously, we examined zero read data and concluded that although fold changes are difficult to predict, these data reflect true differential expression [25]. Chitinases have received little attention in fish to-date and have not been reported in the skin or gill of any species. However, in mammals, chitinases have the ability to cleave inhaled or ingested chitins from fungi, parasitic worms, crustaceans, and insects and are known to participate in the pathogenesis of allergic inflammation, particularly in the lung [28]. The acidic mammalian chitinase (AMCase) in particular has been found to be causative in lung inflammation and induced by IL-13 [29]. Also induced by fasting was an intestinal-type alkaline phosphatase (IAP; 90.69-fold). Known as a critical brush-border protein in mammals, its functions and expression patterns in fish have not been previously characterized. In mammals, IAP has the ability to detoxify lipopolysaccharide (LPS) and prevents infection across the gut mucosal barrier. Additionally, as here, IAP is reported to be sensitive to nutritional changes [30,31]. However, IAP was silenced in the gut after 2 days of fasting in mice [30], contrasting with the induced expression seen here in skin and gill samples at 7 d. Clearly, additional study is required to delineate novel roles of IAP in modulating mucosal immune events in teleost fish. Also induced following fasting were several genes whose expression had been perturbed in fish mucosa in previous studies in channel catfish. These included beta-2-glycoprotein 1, a novel mediator of innate immunity recently discovered to interact with LPS and activate macrophages [32]. We previously observed  sharp upregulation of beta-2-glycoprotein 1 at 2 h after Aeromonas hydrophila infection of channel catfish skin [27]. Interestingly, beta-2-glycoprotein 1 was also reported to be upregulated following 7 d fasting in sea bream skin. Other examples such as SAM domain and HD domain-containing 1 (SAMHD1) and serum amyloid-P precursor (SAP) were induced here by fasting and upregulated in the gill following columnaris challenge [14,25]. A suite of immune genes were also strongly downregulated following fasting (Table 4). Immunoresponsive gene 1 (IRG1), for example, was downregulated over 53-fold after the 7 d fasting period. IRG1 is a LPS-inducible gene linked to susceptibility to Marek's Disease in chickens [33]. Additionally, we found basally lower levels of IRG1 in F. columnare susceptible channel catfish gill when compared with resistant channel catfish [25]. Also among downregulated immune genes following fasting was inducible nitric oxide synthase 2 b (iNOS2b), previously established as the most likely orthologue of mammalian iNOS [34]. iNOS generates nitric oxide (NO) from L-arginine. NO is a potent cytotoxic agent in immune defenses which can have beneficial antimicrobial activity at mucosal surfaces including respiratory epithelium [35]. Previously, we reported high constitutive expression of iNOS2b in healthy channel catfish gill and higher iNOS2b levels in F. columnare resistant catfish than susceptible channel catfish [25]. Here, following 7 d fasting, iNOS2b transcript levels plunged greater than 17-fold. Based on our previous studies, we would predict that these fasted levels of iNOS2b may open up channel catfish to heightened levels of pathogen colonization.
Another noteworthy alteration in immune status induced by food deprivation was the downregulation of peptidoglycan recognition protein 6 (PGRP6; -9.13-fold). Peptidoglycan recognition proteins are poorly-studied pathogen recognition molecules which can bind and kill commensal and pathogenic bacteria, often upstream of the better characterized Toll pathway [36]. Host species from insects to mammals rely on a diverse array of PGRP molecules, some mucus-secreted, to shape and coordinate responses to a wide range of microorganisms. They appear particularly adapted to local responses at mucosal epithelial interfaces rather than more systemic immune responses [37]. For example, in Drosophila PGRP proteins mediate tolerance of gut epithelia toward endogenous microbiota [38]. PGRP6 is one of four PGRP genes in zebrafish. Suppression of PGRP6 by RNAi in zebrafish downregulated the Toll pathway and resulted in markedly increased susceptibility to F. columnare [39]. Taken together, these studies point to the potential importance of PGRP6 as a nutrition-sensitive mediator of mucosal health. Its presence at the mucosal surface in fish may be important for the establishment of tolerance to commensal microbial populations through regulation of release of appropriate levels of antibacterial factors such as mucins and iNOS [37]. Future work should further characterize PGRP6 activity in catfish in the context of nutrition and disease.
In our previous study of patterns of gene expression linked to F. columnare resistance, levels of lysozyme C in channel catfish gill were markedly higher both pre-challenge and following infection in resistant fish [25]. Here, fasting for 7 d decreased lysozyme C levels greater than 6-fold. In mammals, lysozymes are critical mucosal enzymes secreted from the epithelium as well as a major component of granules of professional phagocytes. They help to kill bacterial pathogens through enzymatic and antimicrobial activity [40]. Previous research in zebrafish, in which a transgenic zebrafish strain expressing a chicken lysozyme gene under the control of a keratin promoter resulted in a 65% survival rate against F. columnare compared to 0% survival in wild-type fish, also points to the importance of lysozyme in fish mucosal immunity [41]. Additional genes exhibiting co-regulation in both the context of feed deprivation and F. columnare immunity included autoimmune-regulator like, CCL19, CD83, IgFc-binding protein-like, IL-17A, MFAP4, polymeric immunoglobulin receptor (pIgR), and toxin-1 precursor (Table 4; [25]). The fasting-induced modulation of genes previously indicated to be critical in resistance to F. columnare may explain in part the heightened sensitivity to F. columnare observed in fasted fish [4]. Fasting appears to decrease levels of innate immune mediators including iNOS2b, Lysozyme C, and PGRP6 at the surface barriers where F. columnare gains entry.

Energy Metabolism
While our primary focus here was to examine direct impacts of short-term feed deprivation on immune regulation in channel catfish surface mucosa, we also examined broader physiological impacts on cellular health and function. Changes in metabolic or stress parameters may be important indicators of reallocation of energy reserves and/or changes in barrier permeability which may, in turn, impact host pathogen susceptibility [18,42]. Among the most highly induced genes following fasting were several apolipoproteins. Apoliproteins are proteins which classically bind and transport lipids through the blood but are also recognized for roles in immunity and acute phase responses [43]. Greatest changes were seen in the apolipoprotein ApoBb (Table 4). ApoB serum levels in humans have been reported to rise following fasting [44]. ApoE and ApoL1 levels also rose in channel catfish skin/gill following fasting, contrasting with starvation-induced decreases of apolipoproteins in the liver of rainbow trout and salmon [10,45].
Expression of several fatty-acid binding proteins (FABP) were perturbed by fasting (Table 4), indicating changes in lipid metabolism. Largest changes were observed in the intestinal FABP, FABP2, which was induced greater than 29-fold. FABPs are well studied as intracellular fatty acid transporters in vertebrates, with limited studies to-date in fish [46]. Better studied in fish are genes linked to proteolytic gene expression during muscle atrophy [47,48]. The ubiquitin ligases F-box only protein 32 (Fbx32) and muscle RING finger 1 (MuRF1) were upregulated following fasting, 9.07-fold and 2.72-fold respectively.
The absorption of oligopeptides following protein digestion in vertebrates is facilitated by a member of the solute carrier 15 family, SLC15A1 or PepT1. In our results, the transporter was upregulated 5.36-fold in channel catfish following fasting. Given that most luminal products of protein digestion in fish are di-and tripeptides rather than individual free amino acids, PepT1 is thought to be the major transporter of ingested protein across the gut mucosa [49]. Considerable research has been directed at understanding PepT1 function in a range of vertebrates including fish [50]. Studies in several teleost species have revealed that intestinal expression levels of PepT1 respond to fasting and refeeding, with expression levels rising in some species after shortterm fasting, followed by a decline in long-term fasting/starvation. Re-feeding stimulates upregulation of PepT1 above pre-fasted levels, indicating sensitive responsiveness to food availability and a potential role for the transporter in the phenomenon known as compensatory growth [51][52][53][54]. Aquaculture researchers are keenly interested in use of PepT1 as an indicator of protein uptake and as a direct predictor of animal growth [51,55]. PepT1 may also serve as another bridge between nutritional and immune regulation, given its roles in binding bacterial peptides and stimulation of gut innate immune activation and inflammation [56]. However, all the aforementioned studies focus on PepT1 function in the intestine. While Terova et al. [54] reported second highest tissue expression of PepT1 in sea bass gills (behind some but not all intestinal segments), no previous reports of a PepT1 response to fasting in teleost surface mucosa exist. Given the importance of skin and/or gill epithelia for nutrient absorption in invertebrates and at least one ancient vertebrate, the hagfish [57,58], one is tempted to speculate that catfish PepT1 fasting-induced expression in the skin/gill may reflect conserved mechanisms of nutrient sensing and/or uptake previously unknown in fish. Further research is needed to pinpoint cellular sources of PepT1 expression in surface mucosa and examine responses to longer-term fasting and refeeding.

Cell Cycling/Proliferation
A final notable category of genes perturbed in channel catfish skin/gill following short-term fasting were those involved in cell cycling and proliferation. Again here, while no previous study has examined how fasting may impact cell division in fish surface mucosa, our understanding of these processes in the mammalian gastrointestinal mucosa is fairly well refined [59]. There, fasting/ starvation has been found to be one of the best models of a hypoproliferative response [60]. Subsequent studies have established a consistent pattern of feed deprivation reducing proliferation rates across multiple cell types, such that fasting and calorie restriction are now recommended as approaches to impede tumor growth [61]. We observed consistent downregulation of a large number of genes involved in cell cycling, proliferation, and differentiation (Table 4 and Table S2). These included Ki67 and PCNA-associated factors from channel catfish, cyclins, DNA replication licensing factors and checkpoint proteins ( Table 4). The alteration of cell cycling indicated by these transcript changes may disrupt the integrity of catfish mucosal barriers with consequences for immune defenses at these surfaces [62].

Conclusion
RNA-seq-based transcriptome profiling in channel catfish revealed that short-term feed deprivation altered immune status in the surface mucosa. Changes in innate immune actors such as iNOS2b, Lysozyme C, and PGRP6 may impact the delicate recognition/tolerance balance for commensal and pathogenic bacteria on the skin and gill. Furthermore, our analysis identified critical regulators of metabolism, cell cycling, and transport previously unstudied and/or unreported in these tissues whose expression was perturbed by fasting. The highlighted expression profiles reveal potential mechanistic similarities between gut and surface mucosa and underscore the complex interrelationships between nutrition, mucosal integrity, and immunity in teleost fish. Channel catfish were subjected to two treatments with two replicate tanks per treatment. Fish in treatment group 1 were fed to satiation three times daily with a standard catfish ration (35% protein, 2.5% fat). In treatment 2, fish were withheld feed for 7 d. All fish were sacrificed on day 7, euthanized with tricaine methanesulfonate (MS 222) at 300 mg/L before tissues were collected.

Experimental animals and tissue collection
Equivalent portions of the gill and skin were isolated from matching locations on each fish and stored in RNALater (Ambion, Life Technologies, Grand Island, New York) at 280uC until RNA extraction. Three pools (five fish each) of tissue were generated from each condition. Equal amounts of tissue were collected from each fish within a pool. Samples were immediately placed in RNAlater and stored at 280uC until extraction. Samples were homogenized with a mortar and pestle.

RNA extraction, library construction and sequencing
Extractions were performed according to the manufacturer's directions using an RNeasy Kit (Qiagen, Valencia, California). RNA concentration and integrity of each sample was measured on an Agilent 2100 Bioanalyzer. For each condition, equal amounts of RNA from skin and gill were pooled to generate three pooled replicates for RNA-seq library construction.
RNA-seq library preparation and sequencing was carried out by HudsonAlpha Genomic Services Lab (Huntsville, AL, USA) following the standard TruSeq protocols with 100 bp PE read chemistry on an Illumina HiSeq 2500 instrument [14].

De novo assembly of sequencing read
The de novo assembly of short reads was performed on channel catfish using both ABySS and Trinity [17,63], versions 1.3.2 and the 2012-10-05 editions, respectively. Before assembly, raw reads were trimmed by removing adaptor sequences and ambiguous nucleotides. Reads with quality scores less than 20 and length below 30 bp were all trimmed. The resulting high-quality sequences were used in the subsequent assembly.
In ABySS, briefly, the clean reads were first hashed according to a predefined k-mer length, the 'k-mers'. After capturing overlaps of length k-1 between these k-mers, the short reads were assembled into contigs. The k-mer size was set from 50 to 96, assemblies from all k-mers were merged into one assembly by Trans-ABySS (version 1.4.4).
In Trinity, briefly, the raw reads were assembled into the unique sequences of transcripts in Inchworm via greedy k-mer extension (k-mer 25). After mapping of reads to Inchworm contigs, Chrysalis incorporated reads into deBruijn graphs and the Butterfly module processed the individual graphs to generate full-length transcripts.
In order to reduce redundancy, the assembly results from different assemblers were passed to CD-Hit version 4.5.4 [64] and CAP3 [65] for multiple alignments and consensus building. The threshold was set as identity equal to 1 in CD-Hit, the minimal overlap length and identity equal to 100 bp and 99% in CAP3.

Gene Annotation and Ontology
The selected assembly contigs were used as queries against the NCBI zebrafish protein database, the UniProtKB/SwissProt database and the non-redundant (nr) protein database using the BLASTX program. The cutoff E-value was set at 1e-5 and only the top gene id and name were initially assigned to each contig. Gene ontology (GO) annotation analysis was performed using the zebrafish BLAST results in Blast2GO version 2.6.3 [66]. The zebrafish BLAST result or the nr BLAST result was imported to BLAST2GO. The final annotation file was categorized with respect to Biological Process, Molecular Function, and Cellular Component at level 2.

Identification of differentially expressed contigs
The high quality reads from each sample were mapped onto the Trinity reference assembly using CLC Genomics Workbench software. During mapping, at least 95% of the bases were required to align to the reference and a maximum of two mismatches were allowed. The total mapped reads number for each transcript was determined and then normalized to detect RPKM (Reads Per Kilobase of exon model per Million mapped reads). The proportions-based test was used to identify the differently expressed genes between fed and fasted group with three replicates in each group with corrected p-value ,0.05 [67]. After scaling normalization of the RPKM values, fold changes were calculated. Analysis was performed using the RNA-seq module and the expression analysis module in CLC Genomics Workbench [68]. Transcripts with absolute fold change values of larger than 1.5 were included in analysis as differently expressed genes.
Contigs with previously identified gene matches were carried forward for further analysis. Functional groups and pathways encompassing the differently expressed genes were identified based on GO analysis, pathway analysis based on the Kyoto Encyclopedia of Genes and Genomes (KEGG) database, and manual literature review.

Enrichment Analysis
Enrichment analysis of significantly expressed GO terms was performed using Ontologizer 2.0 using the Parent-Child-Intersection method with a Benjamini-Hochberg multiple testing correction [69,70]. GO terms for each gene were obtained by utilizing zebrafish annotations for the unigene set. The threshold was set as FDR value ,0.1.

Experimental validation-QPCR
Ten significantly expressed genes with different expression patterns were selected for validation using real time QPCR with gene specific primers designed using Primer3 software. Primers were designed based on contig sequences (Table S1). Total RNA was extracted using the RNeasy Plus kit (Qiagen) following manufacturer's instructions. All the cDNA products were diluted to 250 ng/ml and utilized for the quantitative real-time PCR reaction using the SsoFast TM EvaGreenH Supermix on a CFX96 real-time PCR Detection System (Bio-Rad Laboratories, Hercules, CA). The thermal cycling profile consisted of an initial denaturation at 95uC (for 30 s), followed by 40 cycles of denaturation at 94uC (5 s), an appropriate annealing/extension temperature (58uC, 5 s). Results were expressed relative to the expression levels of 18S rRNA in each sample using the Relative Expression Software Tool (REST) version 2009 [71] as described in (14). Figure S1 Gene ontology (GO) term categorization and distribution of differently expressed genes in channel catfish. GO-terms were processed by Blast2GO and categorized at level 2 under three main categories (cellular component, molecular function and biological process). (TIF)

Table S2
Differentially expressed genes in the gill and skin between fasted and fed channel catfish. Positive/negative values indicate upregulation and downregulation, respectively, in fasted fish relative to the fed control. Included genes showed fold changes of 1.5-fold or greater and corrected p-value ,0.05. Annotation is based on the NCBI zebrafish and nr databases. (XLSX) Table S3 Summary of GO term enrichment result of GO-based functional categories of significantly expressed genes in channel catfish after fasting. The 1,545 differentially expressed genes were analyzed as the study set in comparison to all the catfish unigenes. P-value #0.1 was considered significant. Population count is the number of genes associated with the term in the population set. Study count is the number of genes associated with the term in the study set. GO names were retained only from GO terms of levels .2. (DOCX)