Diet energy change had little effect on production traits but affected feed intake and body composition
Table 1: Means (±SD) and significance for production, feed efficiency and body composition traits, for the effect of the diet, the line and their interaction.
|
{R+,CT}1
|
{R+,LE}1
|
{R-,CT}1
|
{R-,LE}1
|
Diet2
|
Line2
|
Diet × Line2
|
Body weight, week 31 (g)
|
2162.35 (±165.33)
|
2142.46 (±129.28)
|
2089.44 (±216.87)
|
1925.40 (±217.32)
|
*
|
**
|
0.11
|
Laying intensity (%)
|
86.17 (±11.92)
|
87.73 (±7.81)
|
86.87 (±5.44)
|
84.59 (±8.58)
|
0.70
|
0.50
|
0.54
|
Egg number
|
60.94 (±9.33)
|
62.18 (±9.93)
|
61.17 (±6.16)
|
60.47 (±7.43)
|
0.93
|
0.86
|
0.60
|
Egg weight (g)
|
47.91 (±3.11)
|
46.80 (±2.98)
|
48.08 (±2.25)
|
47.61 (±1.82)
|
0.21
|
0.53
|
0.60
|
Egg mass (g)3
|
1166.41 (±181.31)
|
1182.36 (±210.53)
|
1118.36 (±108.85)
|
1055.80 (±126.99)
|
0.43
|
*
|
0.27
|
Static stiffness (N.mm-1)
|
109.68 (±18.75)
|
104.64 (±15.58)
|
126.75 (±18.39)
|
118.95 (±18.76)
|
0.12
|
***
|
0.75
|
Feed intake (g)3
|
4128.47 (±426.94)
|
4398.10 (±551.14)
|
2583.92 (±308.26)
|
2728.73 (±419.65)
|
*
|
***
|
0.50
|
Energy intake (kcal)3
|
11188.16 (±1157.00)
|
10207.97 (±1279.19)
|
7002.41 (±835.38)
|
6333.39 (±974.01)
|
**
|
***
|
0.52
|
RFI (g/21d-1)3
|
868.36 (±329.66)
|
1152.32 (±390.52)
|
-614.35 (±134.93)
|
-196.81 (±211.78)
|
***
|
***
|
0.28
|
Abdominal adipose weight at 31 weeks (g)
|
73.33 (±21.10)
|
57.10 (±18.61)
|
129.83 (±44.23)
|
105.00 (±31.67)
|
*
|
***
|
0.64
|
Ratio of abdominal adipose weight to body weight at 31 weeks (%)
|
3.37 (±0.83)
|
2.65 (±0.78)
|
5.96 (±1.39)
|
5.24 (±1.09)
|
*
|
***
|
1
|
1Values represent the line/treatment group means for each trait (±standard deviation). R+ refers to low feed efficient layers and R- to high feed efficient layers, CT to control group and LE to low energy diet. The number of animals analyzed are: R+,CT n = 34, R+,LE n = 11, R-,CT n = 36, R-,LE n = 15
²***: p < 0.001, **: p < 0.01, *: p < 0.05
3Feed-related traits were measured between 28 and 31 weeks of age.
The line, diet and interaction effects on body weight, egg production and shell strength, feed intake (FI), residual feed intake (RFI) and abdominal adipose weight after 14 weeks of the low-energy diet are summarized in Table 1. The diet energy content difference had no effect on egg production, i.e. on laying rate, egg weight and egg mass. In contrast, we observed a significant decrease in body weight at 31 weeks (on average for both lines, -4.4%, p < 0.05) in the LE group compared to the CT group, despite the fact that at the beginning of the trial (17 weeks of age), the LE group was slightly heavier than the CT group (on average, +3%, p < 0.05, Additional file 1). We also observed a significant (p < 0.05) increase of feed intake in the LE group over 28 to 31th week of age, without significant interaction with the line (p = 0.50). It can however be noted that the increase in feed intake in response to the LE diet is smaller in the R- line (+145g) than in the R+ line (+270g), which can be related to the fact that the R- line generally eats less; the interaction between diet and line remains however not significant. The calculated RFI was significantly higher in the LE group, meaning that the animals were less feed efficient than the CT group. Finally, the LE group had at 31 weeks of age a significantly lower ratio of abdominal adipose tissue weight to body weight compared to the CT group (on average, -0.72, p < 0.05), even if the body weight significantly decreased at the same age (on average -4.4%, p < 0.05) indicating a higher decrease of abdominal tissue (on average, -20.6%, p < 0.05). Concerning the line factor, as expected, we observed significant differences on FI, RFI and abdominal adipose weight. The significant line effect for the body weight at 31 weeks, for which the interaction p-value was the lowest and close to 0.10 is due to the {R-,LE} group, the animals of which are lighter than in the three other groups. However, we observed no significant differences between the body weight of R+ and R- from the control group, as expected since the divergent selection on RFI was performed at constant body weight. Both lines, regardless of their RFI, reacted in a similar way to the energy-depleted diet by increasing their feed intake. However, this increase in feed ingestion was not sufficient to avoid body weight loss in the R- fed with the LE diet and depletion of the energy reserves (body fat). To explore the molecular mechanisms underlying this adaptation, we analyzed gene expression of several tissues of birds from these two lines and diets.
Diet energy change leads to transcriptomic modifications, mainly in hypothalamus and blood
To explore the genes involved in the response of birds to the two diets, we analyzed the transcriptomic changes associated with diet changes in the adipose tissue, blood, hypothalamus and liver. A total of 16461 genes were expressed in at least one of the four tissues considered, and represents 66% of the 24881 genes from Ensembl v93 annotation (Fig. 1A and 1B). Of these 16461 genes, 13567, 11440, 15307 and 12873 were expressed in the adipose, blood, hypothalamus and liver, respectively (Fig. 1B), and 10314 (41%) were expressed in all four tissues (Fig. 1A). Some of these genes were tissue-specific, representing 1.34% (adipose) to 10.8% (hypothalamus) of the total number of genes expressed in the tissues (Fig. 1A, Additional file 2). The hypothalamus had markedly higher gene-specificity, with 1653 genes expressed only in this tissue. It also had the greatest number of total expressed genes (15307). Strikingly, diet change had a large effect on the hypothalamic and blood transcriptomes, with 2700 and 1334 differentially expressed genes (DEG), respectively, while the hepatic and adipose tissue transcriptomes were almost unaffected (15 and 2 DEG, respectively) (Fig. 1C and 1D, Additional file 3). The line had a major effect in all tissues, with 3143, 4631, 1874 and 2480 DEG in the adipose, blood, hypothalamus and liver, respectively. As only a very small number of significant interactions (pFDR < 0.05) were observed (Fig. 1C), allowing for an independent analysis of the line and diet factors, the present paper focuses only on the diet effect.
Functional characterization of hypothalamic transcriptome changes upon diet energy challenge
Among the 2700 DEG detected in the hypothalamus in response to the diet energy change, 1438 and 1262 genes were over- and under-expressed, respectively, in the LE group compared to the control. We characterized these two DEG lists using KEGG pathway term enrichment as described in Methods. For the over- and under-expressed gene lists, 26 and 44 pathways (pFDR < 0.05) were significantly enriched (Additional file 4). The 10 top terms with the lowest pFDR for both DEG lists are presented in Table 2.
Table 2: Top 10 (based on pFDR) KEGG pathways associated with under-expressed (A) and over-expressed DEG (B) in the hypothalamus
Term
|
# of genes
|
pFDR
|
A. Under-expressed genes in LE compared to CT
|
Synaptic vesicle cycle
|
22
|
7.36×10-11
|
Glutamatergic synapse
|
26
|
1.79×10-08
|
Dopaminergic synapse
|
26
|
2.37×10-07
|
Axon guidance
|
25
|
5.62×10-07
|
Oxytocin signaling pathway
|
27
|
2.46×10-06
|
Circadian entrainment
|
20
|
2.50×10-06
|
Oocyte meiosis
|
21
|
7.03×10-06
|
Protein processing in endoplasmic reticulum
|
26
|
2.04×10-05
|
Nicotine addiction
|
12
|
2.04×10-05
|
GABAergic synapse
|
17
|
5.18×10-05
|
B. Over-expressed genes in LE compared to CT
|
Ribosome
|
83
|
1.03×10-67
|
Metabolic pathways
|
166
|
2.57×10-25
|
Oxidative phosphorylation
|
46
|
3.26×10-22
|
Glycine, serine and threonine metabolism
|
15
|
7.73×10-08
|
Fatty acid metabolism
|
15
|
1.81×10-06
|
Fatty acid degradation
|
14
|
2.52×10-06
|
Valine, leucine and isoleucine degradation
|
14
|
3.18×10-06
|
PPAR signaling pathway
|
16
|
3.65×10-05
|
Carbon metabolism
|
19
|
1.54×10-04
|
Alanine, aspartate and glutamate metabolism
|
10
|
4.70×10-04
|
Pathways associated with the under-expressed genes (Table 2A) comprised 91 under-expressed genes related to different types of synapses: glutamatergic, dopaminergic and GABAergic synapses, as well as the synaptic vesicle cycle or axon guidance. Among these genes were notably GRIA1, GRIA3 and GRIA4 that code for subunits of the glutamate receptor, the predominant excitatory neurotransmitter in the nervous system; DDC, that code for an enzyme involved in the synthesis of dopamine, a neurotransmitter involved in the reward system, and DRD3 that code for a subunit of the dopamine receptor; GABRQ, GABRG2, GABRR2 that code for subunits of the receptor to the gamma-aminobutyric acid (GABA), the major inhibitory neurotransmitter.
Pathways associated with over-expressed genes in LE compared to CT (Table 2B) were related to the “Ribosome” and several metabolic pathways. “Ribosome” comprises 83 ribosomal Protein genes, of which 41 Ribosomal Protein L (RPLx) genes, 27 Ribosomal Protein S (RPSx), as well as 8 Mitochondrial Ribosomal Protein L (MRPLx) and 5 Mitochondrial Ribosomal Protein S (MRPSx). Among the metabolic pathways, energy-related pathways appear to be most affected. Indeed, we found an over-representation of genes associated with oxidative phosphorylation, a process that involves a series of oxidation-reduction reactions in mitochondria, resulting in the phosphorylation of ADP to produce ATP. Among these genes, 31 were related to one of the 5 protein complexes constituting the respiratory chain located in the inner mitochondrial membrane: 15 genes for the complex I (NADH:ubiquinone oxidoreductase), 8 genes the complex II (succinate:ubiquinone oxidoreductase), 3 genes for the complex III (ubiquinol:ferricytochrome C oxidoreductase), 2 genes for the complex IV (cytochrome C oxidase) and 2 genes for the complex V (FoF1-ATP synthetase), in addition to SLC25A4, the ADP/ATP translocase 1. More than 21 of them are located in the mitochondrial genome. In addition, genes involved in fatty acid transport (ACSBG1, APOA1, APOC3, DBI, SLC27A1, FABP4, FABP7, SCP2), the fatty acid β-oxidation in the mitochondria (CPT2, CACT, ACADL, ACADS, ECHS1, ECI1, HADH, HADHB, ACAA2), and to a lesser extent, in the peroxisomes (ACAA1, ACOX, ECI2) were also over-expressed. On the contrary, genes involved in the de novo lipogenesis were significantly under-expressed, in particular FASN, that codes for a key enzyme of the saturated fatty acid synthesis, and ACLY that codes for the primary enzyme involved in the synthesis of cytosolic acetyl-CoA from citrate. Similarly genes involved in the cholesterol synthesis such as HMGCR, FDFT1, SQLE, CYP51A1, DHCR7, and DHCR24 were also under-expressed. Interestingly, we observed an over-expression of genes involved in the biosynthesis of ω3 and ω6 polyunsaturated fatty acids, with FADS2, ELOVL5, FADS1, ELOVL2 and (see top 5 and 19 KEGG term). It is noteworthy that one of the products of this pathway, the arachidonic acid, can be used by the enzyme coded by NAPEPLD, which is over-expressed (FC = 1.93, pFDR = 6.86×10-11) as a substrate for the synthesis of anandamide. Since the lipid metabolism was largely impacted (Fig. 2A), we studied the transcription factors related to this metabolism (Fig. 2B). The expressions of PPARA, SREBF2 and SREBF1 genes were not affected (FC = 1; 0.88 and 1.08 respectively, with pFDR = 0.99; 0.44 and 0.79, respectively). On the other hand, NR1H3 (alias LXRA) was significantly over-expressed (FC = 1.55, pFDR = 2×10-6). The 30 genes most correlated (r > 0.8) to NR1H3 are showed in Fig. 2C in which can be found FADS2 and NAPE-PLD (r = 0.81 and r = 0.84, pFDR < 2.24×10-5 and pFDR < 5.4×10-6, respectively, Fig. 2D).
Functional characterization of blood transcriptomic changes upon diet energy change
Among the 1334 DEG detected in the blood in response to the dietary change, 719 and 615 genes were over- and under-expressed, respectively, in the LE compared to the CT group. KEGG characterization of the over- and under-expressed DEG lists reveals 2 and 8 significantly enriched pathways, respectively (pFDR < 0.05) (Additional file 5). The terms for both DEG lists are presented in Table 3.
Table 3: KEGG pathways associated with over-expressed (A) and under-expressed DEG (B) in the blood
Term
|
# of genes
|
pFDR
|
A. Under-expressed genes in LE compared to CT
|
Metabolic pathways
|
61
|
7.92×10-05
|
Biosynthesis of amino acids
|
10
|
2.18×10-03
|
Carbon metabolism
|
11
|
8.02×10-03
|
Fructose and mannose metabolism
|
6
|
9.32×10-03
|
Steroid biosynthesis
|
5
|
9.32×10-03
|
Amino sugar and nucleotide sugar metabolism
|
7
|
9.32×10-03
|
Pentose phosphate pathway
|
5
|
2.20×10-02
|
Galactose metabolism
|
5
|
3.82×10-02
|
B. Over-expressed genes in LE compared to CT
|
Ribosome
|
13
|
2.95×10-02
|
RNA degradation
|
9
|
3.24×10-02
|
The pathways associated with under-expressed genes in blood are related to “Metabolic pathways”, in particular amino acids biosynthesis (ACO2, ALDH7A1, CPS1, CTH, ENO2, GOT1, PFKP, TALDO1, TKT, TPI1), fructose and mannose metabolism (AKR1B1, AKR1B10, PFKFB4, PFKP, PMM2, TPI1) or galactose metabolism (AKR1B1, AKR1B10, GALK2, PFKP, PGM2). Genes involved in cholesterol biosynthesis were under-expressed in blood (FDFT1, SQLE, CYP51A1, NSDHL and DHCR24) as in hypothalamus. The two pathways associated with over-expressed genes are “RNA degradation”, with EDC3, EXOSC5, PABPC1, PAN2, PAN3, PATL1, RQCD1, SKIV2L and TOB2, and “Ribosome”, which contains 3 RPL, 3 MRPL, 3 Ribosomal Protein Lateral Stalk Subunit P (RPLPx) and 4 RPS genes, 11 out of these 13 genes were also over-expressed in hypothalamus.
Detection of co-expressed genes with WGCNA within hypothalamus and blood DEG lists
To detect gene subsets in our DEG lists, we used the R package WGCNA to identify and cluster co-expressed gene modules (see Methods). As shown in Figure 3, WGCNA separated for hypothalamus (Fig. 3A) and blood (Fig. 3C) different co-expression groups (noted by a color) for both “LE > CT” (in red) and “LE < CT” (in blue) DEG lists. Interestingly, 2 modules of the same DEG list were not positively correlated in the blood (Fig. 3D, pink color in the correlation matrix) with the blue and purple modules for the red “LE > CT” DEG list and the red and turquoise modules of the blue “LE < CT” DEG list, while all modules were positively correlated in the hypothalamus (Fig. 3B). The plots of module eigengenes of these two pairs can be found in Figure 3E. We can clearly distinguish in the two plots, two distinct parallel series of points that correspond to the R+ and R- lines. This parallelism reveals two facts: first, a difference of expression between the lines with a positive “R- / R+” expression ratio for the purple module (i.e., the x-axis of the plot in Fig. 3E top) whereas it is negative for the blue module (i.e., the y-axis). Second, the eigengene expression differential between the LE and CT groups (symbolized by a Δdiet in Fig 3E) is similar for both lines confirming the absence of a diet × line interaction. We found the same characteristics for the red vs. turquoise modules (Fig. 3E bottom). This illustrates again that this difference is independent of the line effect, and the absence of interactions at the gene expression level, as already seen in Fig. 1C.
The functional analysis of each co-expressed gene module in the hypothalamus revealed KEGG terms similar to the full list of over- and under-expressed genes for the turquoise and blue modules, respectively, and no KEGG term enrichment for the green, red and yellow modules. In the pink module, three genes were associated with “N-Glycan biosynthesis”, while the brown module was enriched in genes related to vesicles and organelles. Finally, the black module was enriched in terms associated with immunological functions (see Additional file 6). This last module, composed of 134 genes, is associated with 10 immunological-related pathways, supported by 22 genes in total, such as C1QA, C1QB and C1QC, C3AR1, CD14, IRF1 and TLR4. In the blood, we found seven modules in the list of over-expressed genes and five modules in the list of under-expressed genes. Functional analysis revealed KEGG terms similar to the full list of under-expressed genes for the black module. No KEGG term enrichment were found for the purple, magenta, green, blue, pink, turquoise, brown, and red modules. The greenyellow module was enriched with genes associated to “Ribosome” and “Protein processing in endoplasmic reticulum”, while the salmon module was enriched with 3 genes associated with the “Estrogen signaling pathway” (See Additional file 7).
Focus on genomic regions concentrating differentially expressed genes
We searched for groups of three or more DEG in close physical proximity (i.e., side by side) along the genome that had a significant pairwise expression correlation (|r| > 0.7 & pFDR < 10-4), hypothesizing that such genes might be co-regulated by a local common mechanism. We found two such proximal co-expressed gene groups in the hypothalamus (Fig.4 A and B), composed of RPS6KA2, MPC1 and SFT2D1 for the first one (Fig. 4A) and C1QA, C1QB and C1QC for the second (Fig. 4B), genes that belong to the black WGCNA module, which was enriched in immunity-related genes.