Transcriptional shifts account for divergent resource allocation in feed efficient broiler chickens

Considerable variation in feed efficiency (FE) has been observed in indigenous and selected meat-type chicken populations. Although this variation could be partially linked to extrinsic factors like diet, housing environment and microbiota, it further illustrates the existence of strong molecular mechanisms enabling the differential allocation of resources for various physiological processes. To further deepen the molecular basis of individual allocation capacity in male and female broilers, an RNA-seq experiment was conducted which based on a phenotyped chicken population divergent in FE. Transcriptional differences linked to FE were pronounced in intestinal and muscular tissue sites of male animals. Specifically, signalling pathways of farnesoid X receptor (FXR) and retinoid X receptor (RXR) might contribute to mediate individual FE. The transcriptional profiles suggested ACSBG2 (muscular lipid utilisation), ASBT (intestinal bile salt transport), CLEC2B (natural killer cell activation), HMGCS2 (jejunal, duodenal and muscular ketogenesis), and SCARB1 (jejunal lipid uptake) as potential mediators driving FE. Results indicate that improvements in FE exploit shifts in resource allocation which might occur at the expense of general immune responsiveness in high efficient male chickens. Consequently, to further improve FE traits and to explore causative molecular patterns, effects originating from sex-dimorphism in chickens need to be taken into consideration.

ScIeNtIFIc REPORTs | (2018) 8:12903 | DOI: 10.1038/s41598-018-31072-7 innate immunity is an energetically expensive process which is supposed to be suppressed in meat-type chickens selected for high growth rates 15 . Indeed, selection for high production efficiency might co-evolve behavioral, physiological, and immunological concerns as reviewed elsewhere 16 .
Despite long-term efforts to improve phenotypic traits in chickens, even individuals within highly selected broiler lines still vary considerably in their FE, nitrogen excretion and growth performance under controlled environmental conditions, as previously shown by Metzler-Zebeli et al. 17 Building upon the study of Metzler-Zebeli et al. (2016), extreme male and female chickens with regards to RFI were selected here and analysed via RNA-seq to identify tissue-specific transcriptional responses. Thus, genome-wide transcriptome profiling comprised muscle tissue (i.e. breast muscle; M. pectoralis) as the major consumers of energy and different parts of the small intestine (i.e. duodenum, jejunum, ileum) facilitating digestion and absorption of nutrients.

Results
The study investigates transcriptional patterns of muscular and intestinal tissues in meat-type chickens considerably divergent in their RFI. With regard to the widely unclear molecular mechanisms driving FE and the pronounced sex dimorphism in chickens, the objectives of this study were to identify sex-specific metabolic routes and molecular features contributing to FE.

Measurements of feed efficiency traits.
The analyses of all sampled chickens with divergent RFI (n = 36) revealed differences in mean RFI values for both males (−183 g vs. 303 g; p < 0.001) and females (−195 g vs. 197 g; p < 0.001) 18 . In Table 1, data on RFI and FCR of the twelve animals selected for comprehensive transcriptomic analyses representing the low and high RFI groups of the whole trial population are summarized. Here, significant differences in RFI and FCR values reflect the divergent FE in animals of both sexes (Table 1). Moreover, low FE female and low FE male chickens showed significant differences in their RFI (203 ± 31 g vs. 431 ± 48 g; p = 0.018).
Differentially abundant genes in muscular and intestinal tissues. RNA sequencing yielded an average of 80 million single reads for each sample. The subsequent mapping of reads against the chicken reference genome resulted in a concordant pair alignment rate of 87.6 ± 1.8%. The following statements refer to 44 out of 48 samples that met the quality control criteria. Considering the significance threshold (p < 0.001 and q < 0.25) the following number of genes were found to be differentially abundant between high and low FE groups: 67 (female) and 81 (male) in breast muscle; 60 (female) and 59 (male) in duodenum; 87 (female) and 90 (male) in jejunum; 11 (female) and 11 (male) in ileum (see Supplementary Table S1). Numbers of differentially abundant transcripts as well as their intersections within and between experimental groups are visualized in Fig. 1. Consistent across both sexes the proenkephalin encoding gene PENK was significantly higher abundant in breast muscle of high FE animals. In contrast, the acyl-CoA synthetase bubblegum family member 2 encoding gene ACSBG2 and a novel gene named ENSGALG00000033498 were found to be differentially abundant in males and females. Overlapping results were also present for the transcript abundance of KCNK17 (potassium two pore domain channel subfamily K member 17) and ENSGALG00000045251 in duodenal samples. For the jejunum, genes related to metabolic pathways i.e. 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 3 (PFKFB3), glucose-6-phosphatase catalytic subunit (G6PC), apical sodium-dependent bile salt transporter D (ASBT/SLC10A2) and solute carrier family 10 member 2 (SLC10A2) were revealed as significantly differentially abundant in male and female chickens with opposite expression patterns in both sexes. Interestingly, the transcript encoded by the novel gene ENSGALG00000033116 was consistently less abundant in all examined intestinal parts of high FE male broilers compared to low FE male chickens. Based on the annotation of the human orthologue, this gene is predicted to encode a C-type lectin.
Analyses of pathways. The integration of RNA-seq results revealed significantly enriched canonical pathways (adjusted p < 0.05), which are presented in Table 2. In general, the comparison of high and low FE males showed a higher number of enriched pathways compared to the same comparison in females. In the breast muscle of male chickens, the highest significantly enriched pathways were the 'FXR/RXR activation' (− log 10 [FDR] = 23.3) and 'LXR/RXR activation' (−log 10 [FDR] = 21.1). These pathways are assigned to the 'nuclear receptor signalling' pathway category, which is further represented by 'PXR/RXR activation' and 'LPS/IL-1 The analyses of breast muscle further indicated an enrichment of pathways assigned to the cellular immune response such as 'acute phase response signalling' and 'IL-12 signalling and production in macrophages' in male chickens. In the duodenum, the 'superpathway of cholesterol biosynthesis' was revealed to be altered between high and low FE animals in both sexes. In the ileum, affected pathways in males were based on orthologues genes of the human leukocyte antigen (HLA) system and were assigned to the 'cellular immune response' category ('Crosstalk between Dendritic Cells and Natural Killer Cells' and ' Antigen Presentation Pathway' . It should be noted that pathways identified for duodenum and ileum are represented by only 3 or fewer molecules. No significantly enriched pathways were obtained for the analysis of breast muscle, jejunum and ileum tissue of female chickens.

Analyses of biofunctions.
Biofunctions retrieved from IPA were analysed to deduce consequences as well as tissue-and sex-specific contributions in the context of FE. Biofunctions which were predicted to be significantly activated (z-score > +2) or inactivated (z-score < −2) in high FE chickens are shown in Table 3. Altered biofunctions were exclusively indicated for breast muscle and jejunum samples of male chickens. For breast muscle, the enriched themes were related to cell signalling and interaction, inflammatory response, metabolism and molecular transport. With the exception of the 'inflammation of absolute anatomical region' function, biofunctions were predicted to be inactivated in high FE chickens compared to low FE chickens. Interestingly, all indicated biofunctions rely on AGT and F2, coding for the angiotensinogen precursor and coagulation factor 2, respectively. Consequently, the biofunction related to angiogenesis was predicted to be activated in breast muscle of high FE male chickens, albeit its corresponding z-score was below the significance threshold (z-score = +1.85).
In jejunum samples of FE-divergent male chickens, analysis revealed a predicted activation of molecular themes related to carbohydrate and lipid metabolism as well as endocrine system development. Specifically, IPA results predict a higher quantity of monosaccharides and a higher concentration of lipids in jejunal samples of high FE chickens compared to low FE chickens. Central molecules which contributed to all biofunctions were an orphan member of the nuclear receptor superfamily encoded by NR0B2 and the scavenger receptor class b member 1 encoded by SCARB1. The former was less abundant and the latter was higher abundant in high FE males compared to low FE males.
Verification of RNA-seq results. The validity of the RNA sequencing experiment and subsequent statistical data analysis was verified by comparing results obtained via RNA-seq and qPCR (Fig. 2 Table S3). The magnitude of fold changes in mRNA abundances was reproducible with both approaches. The correlation of results was as significant and revealed a correlation coefficient of 0.868 (see also Supplementary  Table S3). Consequently, the qPCR analyses support the reliability of RNA-seq results.

Discussion
In the context of divergent RFI as an indicator for FE, the study revealed pronounced differences between male and female broilers at the phenotypic and transcriptional level. According to the data of the whole trial population, the phenotypic differences comprised a higher final body weight, a better FCR and higher daily dry matter excretion of males compared to females 18 . Similarly, sex-specific differences were observed for various production traits across different meat-type chicken lines 19,20 . In the studied population, the range of RFI values was higher in males than in females 18 . Accordingly, the subset of birds with most extreme phenotypes regarding RFI used for transcriptome analyses showed significant differences of RFI values between low FE males and low FE females. Transcriptomic profiles observed for the analysed tissues of high and low FE chickens differed considerably between the two sexes. It is also worth mentioning that a lower number of differentially abundant transcripts were identified in ileum samples compared to the other tissues analysed. This might represent the major importance of the proximal parts of the small intestine for the FE of pigs in concentrate-based diet. These sex and tissue-specific peculiarities were also reflected in the number of significantly affected canonical pathways and biofunctions in the comparisons between the different groups. Hence, despite extensive performance testing and novel insights into quantitative genetics contributing to improved FE, results implicate that certain sex-specific molecular routes and biofunctions might be targeted in the context of FE in broilers.
The most prominent category of enriched pathways identified in the jejunum and breast muscle of male broilers was related to the nuclear receptor signalling. This category comprised pathways triggered by retinoid X receptor (RXR), farnesoid X receptor (FXR or NR1H4) and liver X receptor (LXR) activation. These receptors are present in numerous tissues with highest expression levels of FXR and LXR in liver, intestine and kidney 21 . Signalling pathways downstream of these receptors influence different developmental and metabolic processes including cholesterol and bile acid homeostasis 22 , which might contribute to improve FE in broilers. Indeed, an altered lipid, cholesterol and bile acid metabolism including effects mediated via FXR/RXR signalling were previously presented as one of the key molecular mechanisms driving differences of FE in the duodenum of male meat-type chickens divergently selected for RFI 12 . Similar pathways were found to be affected at the level of the muscle and jejunum in the current study. Based on the altered biofunctions in muscle tissue, these differences were suggested to result in the decrease of molecular transport processes and fatty acid metabolism in high FE male chickens. In this context, the ACSBG2 encoded acyl-CoA synthetase plays an important role for the regulation of lipid metabolism in chickens by initiating the intracellular decomposition or conversion of long-chain fatty acids (mainly C18:1 and C18:2) 23 . Although  expression of human and murine ACSBG2 was reported to be specific for testis and brainstem 24 , the gene appeared to be among the highest expressed genes in chicken muscle of both sexes (see Supplementary Table S1). Interestingly, the FC of ACSBG2 in muscle tissue of males and females was shown to be in opposite directions. Therefore, this gene might contribute to known sex-specific differences in FE and fat traits in chickens 19 . The lower abundance of transcripts associated with apolipoproteins such as APOA4, APOB and APOH in muscle tissue of high FE male chickens indicates for a reduced muscular presence of lipoproteins linked to a putative decrease in intramuscular fat, as previously described 25 . Consequently, serum cholesterol levels tended to be lower in high FE male broilers and were proposed as a valuable indicator for FE by Metzler-Zebeli et al. 18 .
In the jejunum of male broilers, pathways dedicated to nuclear receptor signalling (FXR, RXR, PXR) seem to mediate effects on the intestinal absorption of bile salts as well as dietary cholesterol and triglycerides. Specifically, genes like SCARB1 and ASBT were found to be more abundant in the jejunum of high FE male birds, while NR0B2, which mediates the FXR effects 26 was less abundant. SCARB1 encodes a plasma membrane receptor for lipoproteins which is involved in the absorption of cholesterol from the diet and, thus, is known to differ in abundance depending on the dietary level of lipids in humans 27,28 . ASBT is the primary intestinal bile salt transporter in the intestine 29 and is negatively regulated by FXR 26 . Interestingly, the analyses of jejunal samples of female broilers revealed that ASBT was significantly less abundant in high FE animals. This might implicate sex-specific differences regarding the jejunal bile acid metabolism. Based on these results, it is suggested that high FE male chickens had an increased absorption of dietary cholesterol and a more efficient recovery of bile acid compared to low FE broilers. Moreover, it is known that an intestinal FXR activation is linked to a reduced de novo synthesis of fatty acids in other tissues 30 , which is in line with the predicted inactivation of fatty acid synthesis in muscle tissue of high FE male broilers.
At the same time, altered transcript abundances in muscle (e.g. G6PC, HMGCS2) and jejunum (e.g. G6PC, PFKFB3) of male chickens affect metabolic routes of carbohydrate metabolism. Specifically, effects observed in the muscle of male broilers indicate a reduced ketogenesis in high FE animals which were likely mediated via mitochondrial HMG-CoA synthase (encoded by HMGCS2), as a central regulator of ketogenesis 31 . This might imply for a metabolic demand to allocate sufficient amounts of carbohydrates into muscle tissue of high FE broilers. Biofunctions pointing to an increased availability of monosaccharides and lipids in the jejunum derived by a lower jejunal abundance of HMGCS2 in high FE males support this suggestion. Interestingly, chickens selected for reduced RFI values showed an increased abundance of HGMCS2 in the duodenum 12 . The same transcriptional alterations were found in the comparison of duodenal samples between high and low FE chickens in the current study, albeit reaching significance exclusively in females. In conclusion, the findings promote a central role of nuclear receptor signalling for FE in different tissues affecting the utilization of lipids and carbohydrates.
Although considerable differences were observed between high and low FE chickens of both sexes, there were no signs for major differences in muscle development and cellular energy homeostasis at the transcriptional level.  Especially, mitochondrial efficiency and AMPK signalling were recently proposed as important biological processes involved in improving FE 9,10 , but were unaffected in the analysed FE-divergent broilers. However, commercial broiler lines are under selection for improved body weight gain and FE for many generations leading to a favoured genetic pattern regarding these traits 3 . Thus, minor alterations of different biological pathways might culminate in the observed differences between high and low FE broilers. As recently described in livestock species, trade-offs between performance traits and immune response have to be considered as one important factor contributing to individual differences in productivity 32,33 . With this regards, all tissues examined in the current study are known to contribute to divergent FE via molecular factors involved in the modulation of immune responses 34 . Thus, biofunctions derived from the comparison of breast muscle expression profiles of high and low FE male chickens as well as pathways enriched for the comparisons of breast muscle and ileum indicated for modest differences in immune system functions. Additionally, alterations observed at the level of the lipid metabolism might also have implications on cell signalling and inflammatory responses, for instance, via the predicted decrease in eicosanoid synthesis and release in muscle tissue of high FE male chickens 35 . This is in line with ongoing discussions regarding the energy demand to maintain immune system functions [14][15][16] and possibilities to modulate the immune system in chickens 36 . Specifically, the results of the current study argue for shifts in favour of a disease resistance strategy of low FE broilers which might employ the innate immune response providing a broad-spectrum protection while being energetically expensive 14,37 . In contrast, high FE chickens seem to rely on tolerance mechanisms promoting specific immune reactions 38 . This is supported by the predicted decrease in the activation of macrophages in the muscle of high FE male chickens, which at the same time showed activated biological functions of inflammatory response. Moreover, Metzler-Zebeli et al. 18 reported a significant increase in the lymphocyte fraction and a trend towards a decreased heterophile fraction of white blood cells in high FE male broilers compared to low FE male birds of the same population. In this context, ENSGALG00000033116 (orthologue of human CLEC2B) was consistently found to be lower abundant in all three analysed parts of the small intestine of high FE chickens, albeit reaching statistical significance exclusively for the comparison of male broilers. In humans, CLEC2B is located in the natural killer (NK) cell gene complex and is known to encode a cell activation antigen presented by NK cells, which are an important component of the innate immune system 39,40 . Hence, the intestinal lower abundance of CLEC2B in high FE chickens might represent an indicator for a reduced activation of the innate immune system. However, due to the complex contributions to differences in FE, other aspects like FE-related differences in the intestinal microbiota composition 41 need to be considered modulating the host's immune capacity at the transcriptional level (e.g. 42 ).

Conclusion
Phenotype differences in RFI values were reflected by pronounced sex-dimorphic transcriptional patterns. Indeed, the RNA-seq analyses of muscle and intestine tissues in broiler chickens selected for divergent RFI revealed cholesterol and bile acid metabolism as well as immunomodulatory aspects as molecular drivers of FE in male chickens. Moreover, results indicate that improvements of FE exploit shifts in resource allocation which might occur at the expense of general immune responsiveness in high efficient male chickens. Hence, data suggest that breeding strategies have also to focus beyond improvements in FE traits towards maintenance of phenotypes balancing weight, performance and immune competence.

Material and Methods
Animals, housing, feed efficiency assessment and tissue sampling. The conducted chicken trial was approved by the animal ethics committees of the University of Veterinary Medicine (Vienna, Austria) and the national authority according to paragraph 26 of Law for Animal Experiments, Tierversuchsgesetz 2012 (GZ 68.205/0131-II/3b/2013). Animal trials were performed in three independent batches with at least 52 birds per batch. Chickens were obtained from commercial hatcheries. The experimental design was balanced for sex and comprised 157 birds (79 males and 78 females). Housing and feeding conditions have been previously described 17,18 . In brief, one-day-old Cobb500FF birds (Cobb-Vantress, Siloam Springs, AR, USA) were housed in groups of four birds for 7 days. Afterwards, they were individually penned in cages until the end of the experimental period. Chickens were checked daily for health status and welfare issues; signs of illness led to the exclusion of animals for further analysis. Common diets based on corn and soybean meal were fed during starter (days 1-10), grower (days [11][12][13][14][15][16][17][18][19][20][21], and finisher periods (days 22-42) as previously described 17,18 . Ad libitum access to fresh water was provided. Individual FI was recorded daily. Body weight of all chickens was recorded weekly (d 1, 7, 14, 21, 28 and 35). To calculate overall RFI values, body weight was further measured at d 36. For RFI calculation, data for net total FI, metabolic midweight and total BWG between day 7 and 36 were analysed using a nonlinear mixed model (SAS Stat Inc., version 9.2, Cary, NC) as previously described 17,18  The model included the interaction of sex (male, female) and RFI group (high, low) as fixed effect and the day of slaughter as random effect. P-values were adjusted for multiple comparisons using the Tukey-Kramer procedure (R, lsmeans package). After DNA extraction following a conventional phenol/chloroform procedure, the sex of selected animals was verified by PCR using primers specific for chicken sex chromosome W 43 . Individual samples of duodenum (n = 12), jejunum (n = 12), ileum (n = 12), and breast muscle (n = 12) were used for RNA isolation. Extraction of total RNA was performed according to TRIzol Reagent protocol (Invitrogen, Darmstadt, Germany). Therefore, samples were disrupted using either mortar and pestle (duodenum, jejunum, and ileum) or a Precellys cell disruption device (breast muscle; 12 sec at 5000U; Peqlab, Erlangen, Germany). Subsequently, total RNA was treated with Baseline-ZERO DNase (Biozym, Hessisch Oldendorf, Germany). Purification of RNA was achieved using the NucleoSpin RNAII kit (Macherey-Nagel, Düren, Germany) according to manufacturer's protocols. Total RNA was checked on agarose gel and measured by NanoDrop ND-1000 (NanoDrop, Peqlab, Erlangen, Germany). The quality of RNA was assessed employing an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). All samples had RNA integrity numbers (RIN) values ≥ 7.8. Subsequent extraction of mRNA was supported by a magnetic mRNA Isolation Kit (New England Biolabs, Frankfurt, Germany). Libraries were prepared using the TruSeq Stranded mRNA preparation kit (Illumina, San Diego, CA, USA). Quantity of libraries were analysed using both a DNA 1000 chip on a Bioanalyzer instrument and the Illumina Library Quantification Kit (Kapa Biosystems, Cape Town, South Africa) performed on a LightCycler 480 PCR System (Roche, Mannheim, Germany). Quality controlled libraries were sequenced on an Illumina HiSeq. 2500 instrument (Illumina) by generating 125-bp paired-end reads in high-output run mode as described by the manufacturer.
Transcript data analyses. After demultiplexing, adapter-trimming, and base-quality filtering, TopHat (version 2.0.12) and Bowtie (version 1.1.1.0) mapping tools were used to align the sequences to the chicken reference genome (Gallus gallus genome build 5, March 2017). Read counts per gene were calculated via htseq-count version 0.6.1p1 44 . In order to exclude genes with very low read counts, features with less than 5 reads in at least 2 samples were discarded from further analyses. Quality of sequencing data was assessed using the arrayQualityMetrics package for R (version 3.32.0). The package uses three different outlier detection approaches, including the distance between individual data sets, the distribution of signal intensities and the quality of the individual data set. Samples were judged as outliers which were conspicuous in at least two of these criteria. In total, 4 out of 48 samples (4 tissues, n = 12) were removed from further analyses (see Supplementary Fig. S1).
Analysis of differentially expressed genes was performed using the DESeq. 2 package of the R environment 45 . For each tissue and sex, the negative binomial generalised linear model included day of slaughter and RFI group as fixed effects. Corresponding contrasts were assessed via Wald tests. False discovery rates (FDR, q-values) were calculated using the qvalue R package following the Benjamini-Hochberg (BH) procedure (version 2.8.0). The significance threshold was set at p < 0.001 and q < 0.25. Fold changes were calculated to display the relative transcript abundance between high FE and low FE groups (high FE > low FE indicated by positive FC; high FE < low FE indicated by negative FC). No FC cut off was applied. The number of differentially abundant transcripts and the size of the intersections between experimentally groups were displayed using the UpSetR package 46 . Analysis of canonical pathways and biofunctions. Lists of significantly altered transcripts were evaluated with Ingenuity Pathway Analysis (IPA; Ingenuity Systems, Redwood City, CA). Human gene identifiers were preferentially used if available. Therefore, orthologous gene identifiers were assigned using the BiomaRt R package (Gallus_gallus-5.0 and GRCh38.p7, version 88, accessed on March 2017). Results related to canonical pathways and biofunctions were considered significant at a Benjamini-Hochberg adjusted p-value < 0.05. In addition, molecular themes related to biofunctions had to pass a second cut-off criteria (absolute value of z-scores > 2.0 were considered as significant) stating a prediction of its activation status 47 . Quantitative real-time PCR (qPCR). First-strand cDNA was synthesized from 1.5 µg of total RNA using random primers and oligo d(T) 13VN in the presence of Superscript III reverse transcriptase (Invitrogen, Karlsruhe, Germany) and RNAsinPlus RNase Inhibitor (Promega, Heidelberg, Germany). The cDNA samples were diluted in 200 µL Aqua dest. and stored at −20 °C until further analyses. Transcript levels of selected target (ACSBG2, AMY2A, AQP4, CA7, FABP4, FNDC5, G6PC, GATM, GBP, HPX, PFKFB, THRSP) and reference genes (ACTB, GAPDH, DCAF7) were quantified by qPCR using transcript-specific primers. Primer specifications and tissues analysed are shown in Supplementary Table S2. Individual mRNA samples (n = 12 per tissue) were analysed in duplicate on a LightCycler 480 system using the LightCycler 480 SYBR Green I Master (Roche, Mannheim, Germany) according to manufacturer's instructions. Amplified products were subjected to melting curve analyses and gel electrophoresis to verify the absence of non-specific products. For all assays, threshold cycles were converted to copy numbers using a standard curve generated by amplifying serial dilutions of a corresponding PCR standard (10 7 -10 2 copies). Values were calculated by factorial normalisation on ACTB, GAPDH, and DCAF7 expression values. Correlation of normalized expression values was calculated using Pearson test and considered significant at p < 0.05. Data availability. The data generated during the current study have been deposited in the ArrayExpress database at EMBL-EBI (www.ebi.ac.uk/arrayexpress) under accession number E-MTAB-6169.