Fish under intensive culture conditions are exposed to a variety of acute and chronic stressors, including elevated rearing densities, sub-optimal water quality including decreased dissolved oxygen (DO) and high carbon dioxide (CO2), and thermal fluctuations [1, 2]. During transportation and sorting, several factors may also lead to stress including handling and tank confinement [3, 4]. Osmoregulatory disturbance may be an outcome of stress, for which altered salinity is routinely used to mitigate negative effects, however high salinity such as in adaptation to seawater induces a stress response . These and other stressors which are common to aquaculture production can induce physiological responses which may have adverse effects on traits important to producers and consumers, including those associated with growth, nutrition, reproduction, immune response, fillet quality, and environmental impacts [1, 2, 6–13]. Understanding and monitoring the biological mechanisms underlying stress responses will facilitate alleviating their negative effects through selective breeding and changes in management practices, resulting in improved animal welfare and production efficiency.
Studies which aim to characterize global gene expression in response to stress often use hybridization-based approaches (i.e. microarrays) to identify differences between challenged and control fish [14–16]. Hybridization techniques including those that employ microarrays can be high-throughput and are relatively inexpensive, however they present some limitations . For instance: 1) the preparation of custom-made arrays requires previous knowledge of the target transcriptome expected to be represented in a sample; 2) hybridization data often present high background levels due to cross-hybridization; and 3) microarrays are not suitable for genes expressed at very low or very high levels because there is a limited range of detection due to background and saturation . Sequencing-based methodologies have also been used to characterize gene expression, full-length cDNA and EST (expressed sequence tags) libraries have been sequenced using Sanger technology, however, this method is low-throughput, expensive and usually not quantitative [17, 18]. Tag-based short sequence reads methodologies like SAGE (serial analysis of gene expression) or MPSS (massive parallel signature sequence) are also high-throughput and quantitative, however the sequence reads cannot be unambiguously mapped to the reference genome; the transcriptome is still partially
analyzed; and it is usually not possible to distinguish different gene isoforms [18, 19]. Recently developed high-throughput sequencing technologies offer a new approach (RNASeq) for characterizing transcriptomes through high-throughput sequencing, mapping and quantification [20–31]. This approach is especially advantageous for non-model species, because it is not restricted by the unavailability of a genome reference sequence. Additionally, RNASeq has very low background and sequences can be unambiguously mapped to reference sequences. The quantification is based on the number of sequences obtained, therefore it offers the detection of a higher expression range and it is possible to identify low or highly expressed genes. Furthermore, this approach permits additional analysis such as the identification of putative single nucleotide polymorphisms (SNPs), which can be identified in transcribed regions and have the potential to affect gene function . RNASeq has previously been used for transcriptome characterization of non-model species, including butterfly , silkworm , garter snake , coral , pearl oyster  and several fish species [25–30], including rainbow trout .
In the absence of a genome reference sequence for rainbow trout, we used Roche 454 pyrosequencing technology to develop a reference transcriptome sequence to be used specifically in gene expression analyses associated with stress, including RNASeq based approaches. The library construction included a normalization process [22, 23, 33] such that the reference transcriptome would constitute a qualitative resource attempting to represent the maximum number of unique transcripts from each of the tissues/treatment combinations as possible . Although the majority of stress gene expression research in rainbow trout has concentrated on evaluating the effect of only one stressor on one specific tissue [34–36], our goal was to capture and characterize a collection of transcripts from multiple tissues from fish exposed to one of five stressors. As a result we aim to establish a comprehensive stress transcriptome resource that will facilitate understanding stress responses in this species. To this end we selected stressors that are among those commonly experienced by rainbow trout during aquaculture production and transport including high temperature, low temperature, low DO/high CO2, seawater transfer, and handling/confinement.
Previous studies have evaluated the combined effects of temperature and salinity on physiological condition of rainbow trout. Niu et al.  transferred juvenile rainbow trout from 13.5°C to 25.5°C, held them for two hours, and then transferred them to a 32‰ water at 13.5°C to observe the effects osmotic stress during transfer from freshwater to saltwater. Also, low temperature combined with sea water tolerance was analyzed by Findstad et al  who transferred fresh water acclimated rainbow trout to sea water at 1 and 8°C, and Saunders et al  who first acclimated the fish to salt water and placed them in floating cages until temperatures fell below 0°C. In this project we used similar parameters for high and low temperature and salinity treatments however we were conducted treatments independently. Previous studies have suggested that oxygen concentrations above 6 mg/l are not limiting factors for growth  and carbon dioxide in culture tanks should range between 10 and 20 mg/l  therefore we combined effects of low DO and high CO2 by supplying fish with re-use water, a common practice in aquaculture production which uses water from rearing sites "upstream" to increase production capacity . Finally, fish were subjected to a handling and confinement stress routinely used for rainbow trout stress response research [15, 43–46].
We demonstrate that each stressor produced a physiological response by reporting changes in plasma variables indicative of various phases of a stress response (cortisol, glucose, lactate, and chloride concentrations and lysozyme activity [47–49]) as well as changes in the expression of genes in gill tissue related to apoptosis [50–52] and Na/K transport. Our reference transcriptome created by sequencing RNA from multiple tissues of control and stressed fish is inclusive of transcripts expressed within the context of these unique physiological states and typical basal expression from control fish. We characterize the reference transcriptome by conducting an assembly of sequencing reads, assigning Gene Ontology annotation  and identifying putative SNPs.