The mapping population consisted of four F2 families that originated from a cross between two divergent populations of Atlantic salmon in Norway, the 'ancient', landlocked Byglands Bleke population and a commercial selected line, a resource population known as 'SALBANK'. These two populations are highly divergent for a number of traits, including flesh colour and growth, and the F2 population was created as a mapping resource with an increased likelihood of segregation at QTL affecting these traits. The details of the pedigree, phenotypic recording and sampling procedures can be found in Baranski et al. . Fifty progeny from each extreme of the flesh colour distribution, measured visually using the Roche SalmoFan, were selected from three F2 families (8B, 9B and 10B), and all 76 progeny from a fourth family (10A) were selected for genotyping. Due to differences in progeny numbers between the families, this represented selective genotyping fractions (both extremes) of 44%, 35%, 35% and 100% respectively for the four families.
Atlantic salmon tissues (NIVA fish)
The NIVA-fish served as a reference population (normal breeding population) since we did not have tissues for RNA from the actual QTL-mapping population. Tissues for total RNA-isolation were collected from Atlantic salmon with an average weight of 1000 grams (before sexual maturation), kept in tanks at 12°C at NIVAs Research station, Drøbak, Norway. Muscle, liver and mid gut were collected from ten fish, and immediately stored on liquid nitrogen and subsequently transferred to -80°C until RNA-extraction.
Genomic DNA was isolated from about 20 mg muscle of Atlantic salmon, using either MagAttract DNA M48 Tissue kit on the Bio-Robot M48 (Qiagen, Hilden, Germany) or DNeasy 96 protocol (Qiagen, Hilden, Germany). Prior to isolation, muscle samples were lysed in Proteinase K at 56°C over night according to the manufacturer's protocol. Insoluble materials after lysis were removed by centrifugation (300 × g, 1 minute). DNA was extracted according to manufacturer's instruction with the inclusion of RNaseA (0.1 mg/sample RNase A R5503 (Sigma-Aldrich, St. Louis, MO, USA)), for 30 minutes at room temperature.
Oligonucleotides used for PCR-amplification of intron sequences were designed based on Atlantic salmon SCARB1 [Genbank:DQ914655.1] using Vector NTI Advance (Invitrogen, Carlsbad, CA, USA). Putative exon-intron junctions for intron three, four, five and ten were identified by comparative sequence alignment to the human SCARB1 reference sequence [Genbank: NM_005505]. For real-time PCR analysis oligonucleotides were designed to span one intronic sequence in order to keep control of any potential genomic contamination in the RNA samples.
PCR-amplification was carried out using 0.4 μM oligo-nucleotides, 0.05 U/μl Taq Gold, 200 μM dNTPs and 1 × reaction buffer containing 15 mM MgCl2 (Applied Biosystems, Foster City, CA, USA). After 10 minutes of initial denaturation at 95°C, 40 cycles of amplification at 95°C for 30 seconds, 54-60°C for 30 seconds, 72°C for 30-60 seconds was followed by a final extension of 7 minutes at 72°C. The PCR-product was fractionated on a 2% agarose gel containing ethidium-bromide and visualised by UV-transiluminator 2000 (BIORAD).
DNA sequence analysis
Prior to sequencing, the PCR-products were purified using Montage PCR Genomics cleanup kits (Millipore, Billerica, MA, USA), according to the manufacturer's protocol. DNA-sequencing was performed using the BigDye Terminator version 3.1 cycle sequencing kit according to the manufacturer's protocol with the following thermocycling conditions: 96°C 1 minute, 25 cycles at 96°C for 45 seconds, 50°C for 45 seconds and 60°C for 4 minutes. The products were purified with Montage-SEQ96 cleanup kits (Millipore, Billerica, MA, USA) and sequencing was performed on a 3730 DNA Analyser (Applied Biosystems, Foster City, CA, USA).
RNA-isolation and cDNA-synthesis
Total RNA was isolated using a combined protocol with Trizol and RNeasy mini kit, and DNAse-treated using the RNase-free DNase set as described by the manufacturer (Invitrogen, Carlsbad, CA, USA). RNA quantity was measured using the NanoDrop ND-1000 Spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA) and quality was examined by the 28S:18S rRNA ratio using the RNA 6000 Nano LabChip® Kit on 2100 Bioanalyser (Agilent Technologies, Santa Clara, CA, USA). First strand cDNA synthesis was performed using SuperScript™-II Rnase H-Reverse Transcriptase (Invitrogen, Carlsbad, CA, USA) and oligo-dT oligonucleotide T270 (5'-GACTCGAGTCGACATCGATTTTTTTTTT-TTTTTTT-3'). 0.25 μg of total RNA was used as template for cDNA-synthesis, all RNA was standardised to the same concentration prior to cDNA-synthesis.
Quantitative Real-Time PCR
Real-Time PCR was conducted using LC480 (Roche) and gene-specific oligonucleotides using oligonucleotide combination 2 and 5, for SCARB1 and SCARB1-2, respectively (table 1). Oligonucleotide and cDNA-concentration were optimised to obtain the lowest possible Ct-value. 0.4-1.25 μM of each oligonucleotide, 2 μl of 2 × SYBR Green PCR Mastermix (Applied Biosystems, Foster City, CA, USA) and the cDNA-template were mixed in a total volume of 12 μl. A two-step PCR was run for 45 cycles (10 s. at 95°C, 30 s. at 65-67°C) with an initial denaturation of 10 min. at 95°C. Specificity of the PCR-products was verified by agarose gel-electrophoresis and amplicon sequencing. Standard curves for each oligonucleotide pair were generated by serial dilution (1, 1:2, 1:10) of cDNA consisting of a pool of the representative samples, and PCR efficiencies (E) were calculated according to the formula E = 10(-1/slope(a)) . E-values was 0.7 for SCARB1 and for SCARB1-2, and comprise Ct-values ranging from 21 to 27, and from 30 to 34, respectively. Relative mRNA expression was calculated by the 2-ΔΔCt method  adjusted for PCR efficiencies. 18S was used as a reference gene. Two independent cDNA-syntheses of each sample were performed. Two independent PCR reactions were run for each cDNA synthesis with duplicate samples within each PCR.
Statistical analyses of the qPCR results were carried out using a GenEx software package (MultiD Analyses AB, Sweden). The samples were already normalized to reference gene (18S) and PCR efficiency when imported into GenEx. The samples were normalized for technical replicates (and missing data), and calculated using the average of three independent qPCR runs. One-way ANOVA, with post test all pairwise comparisons (Tukey-Kramer's) was used to compare the gene expression in mid gut, muscle and liver, for each of the two paralogs.
Linkage mapping and QTL analysis
Genotyping of the SCARB1-2 locus was carried out on the 'SALBANK' founder parents and F2 progeny with thermocycling conditions 95°C 10 min., 40 cycles of 95°C 30 sec., 60°C 30 sec., 72°C 30 sec., and 72°C 10 min., using oligonucleotide combination 5 shown in table 1. The lengths of the fluorescent PCR products were determined relative to the LIZ1200 size standard (Applied Biosystems, Foster City, CA, USA) on a 3730 DNA Analyzer (Applied Biosystems), using GeneMapper 4.0 (Applied Biosystems) software for allele calls. Genotypes of the F1 parents were inferred from the grandparent and offspring genotypes. The SCARB1-2 genotype data was added to the microsatellite data previously genotyped for these individuals , and linkage analysis was performed in Joinmap 3.0 using a minimum LOD score of 3.0 in order to map the SCARB1-2 locus to the Atlantic salmon linkage map. In order to evaluate QTL linkage with the hypothesis that a QTL peak is located at this marker position, half-sib regression interval mapping for the linkage group containing the SCARB1-2 locus was carried out in GridQTL at 1 cM intervals , with sex fitted as a fixed effect and body weight as a covariate due to a strong positive phenotypic correlation between body weight and colour. Separate male and female analyses were performed due to the map differences observed in the sexes. 50 cM distance was inserted between the unlinked pair of markers in the female map and marker Albumin1 in the first group. Significance thresholds were estimated after 10,000 chromosome-wide permutation tests. Parents segregating for the QTL were identified based on their individual t-values in the GridQTL analysis. Confidence intervals (CI) were estimated for the QTL using the bootstrap method and 10,000 iterations . The proportion of phenotypic variance explained by the QTL using the half-sib model was calculated as (2*(1-MSfull /MSreduced)sire) + (1-MSfull/MSreduced)dam) .