Responses to selection on production and carcass traits

The selection index in the divergent lines had a heritability of 0.13±0.05 (Table 1), just the same as the heritability estimated for RFI computed from the realized phenotypic correlations in the population (i.e. 0.13±0.05). These figures are lower than generally reported for growing pigs (0.20 to 0.40, Saintilan et al., Reference Saintilan, Brossard, Vautier, Sellier, Bidanel, van Milgen and Gilbert2013). The genetic correlation between RFI and FCR was 0.39±0.12. The responses to selection were significant since G1 on RFI and DFI (Figure 1), reaching −165 g/day (LRFI line – HRFI line) for RFI (3.84 genetic standard deviation (σ _g )), and −270 g/day for DFI in generation G9 (Table 1). The line difference for FCR was −0.32 kg feed/kg BW gain. Correlated responses to selection on ADG were slightly significant (−8 g/day) in generation G9, but no clear increase over successive generations was observed. Responses on carcBFT were not significant, but LMCcalc was 1.30 σ _g higher in the LRFI compared with the HRFI line in G9, indicating that predicting RFI from BFT does not fully constrain changes in body composition to achieve better feed efficiency. This was associated with significant increases in dressing percentage, loin, ham and shoulder weights and reductions in backfat and belly weights. In the ISU lines, a significant decrease of ADG was reported in the low RFI line compared with the high RFI line, and similar to the INRA lines, the LRFI pigs had more muscle and less fat (Young and Dekkers, Reference Young and Dekkers2012). In addition to differences in DFI, a reduced water intake as g/kg BW^0.60 per day (−33%, P=0.062) was reported in LRFI pigs (Renaudeau et al., Reference Renaudeau, Gourdine, Fleury, Ferchaud, Billon, Noblet and Gilbert2013).

Figure 1 Genetic trends in the divergent selection experiment for residual feed intake (RFI) on component traits and meat quality expressed in genetic standard deviations of the traits (σ _g ), obtained from a linear mixed model including an animal random effect structured by the pedigree relationship matrix. HRFI=high RFI line; LRFI=low RFI line; Index=selection index; DFI=daily feed intake; ADG=average daily gain; FCR=feed conversion ratio; carcBFT=backfat thickness measured on the carcass; MQI=meat quality index; LMCcalc=lean meat content of the carcass computed from a linear combination of cut weights.

Table 1 Genetic parameters (h ²=heritability; ρ _g =genetic correlation with RFI; σ _g =genetic SD; σ _p =phenotypic SD), responses to selectionFootnote ¹ in generation G9 at the genetic level in the low RFI (LRFI) and high RFI (HRFI) lines and significance level of the difference (P)

Index=selection index; FCR=feed conversion ratio; RFI=residual feed intake; DFI=daily feed intake; ADG=average daily gain; LMCcalc=lean meat content of the carcass computed from a linear combination of cut weights; carcBFT=backfat thickness measured on carcass; DP=dressing percentage of cold carcass; pH 24 h: determined 24 h after slaughter; AD=adductor; SM=semimembranosus; GS=gluteus superficialis; LM=longissimus muscle; GM=gluteus medius; L*=lightness, a*=redness, b*=yellowness, measured 24 h after slaughter; WHC=water holding capacity; MQI=meat quality index.

¹ Least square means (SE) of a linear model including the fixed effects of line, generation and the interaction line×generation applied to the estimated breeding values of the pigs tested in the two lines from G0 to G9 (n=1451 candidates to selection and 1629 slaughtered sibs).

² P value for the difference between least square means of the LRFI and LRFI lines in generation G9, ***=P<0.001, *=0.01<P<0.05, t=0.05<P<0.10.

Meat quality traits

The substantial progress made since the 1960s to increase growth rate and carcass LMC and to improve feed efficiency in swine has resulted in unfavourable effects on meat quality (Supplementary Material S1, Lebret, 2004). From the third generation of divergent selection for RFI, technological quality traits and sensorial meat quality indicators were impaired in the LRFI line in both loin (longissimus) and ham (semimembranosus, gluteus, adductor) muscles (Tables 1 and 2) (Gilbert et al., Reference Gilbert, Alain, Sellier, Lagant, Billon, Bidanel, Guillouet, Noblet, van Milgen and Brossard2007; Lefaucheur et al., Reference Lefaucheur, Lebret, Ecolan, Louveau, Damon, Prunier, Billon, Sellier and Gilbert2011; Faure et al., Reference Faure, Lefaucheur, Bonhomme, Ecolan, Meteau, Coustard, Kouba, Gilbert and Lebret2013). In generation G9, these differences reached −3.41 points (−2.12 σ _g ) for a meat quality index based on a combination of ultimate pH, L* and water holding capacity. These responses were related to changes in tissue deposition efficiency and in muscle metabolic properties (see Biological components of residual feed intake section). Indeed, a higher percentage of fast-twitch glycolytic myofibres and a hypertrophy of all fast-twitch myofibres were found in the longissimus muscle (LM) of LRFI pigs, resulting in greater muscle glycogen stores, especially in the glycolytic muscles (Lefaucheur et al., Reference Lefaucheur, Lebret, Ecolan, Louveau, Damon, Prunier, Billon, Sellier and Gilbert2011). Besides, the intra-muscular fat content was lower in the glycolytic and oxidative muscles of the LRFI line (Lefaucheur et al., Reference Lefaucheur, Lebret, Ecolan, Louveau, Damon, Prunier, Billon, Sellier and Gilbert2011; Faure et al., Reference Faure, Lefaucheur, Bonhomme, Ecolan, Meteau, Coustard, Kouba, Gilbert and Lebret2013), without any impact on oxidation traits of lipids and proteins in meat after ageing (Supplementary Material S1, Gilbert et al., 2012a). These unfavourable genetic relationships between low RFI and meat quality traits were confirmed in French commercial populations (Saintilan et al., Reference Saintilan, Brossard, Vautier, Sellier, Bidanel, van Milgen and Gilbert2013), but were not always found in other studies: in the ISU RFI lines, Smith et al. (Reference Smith, Gabler, Young, Cai, Boddicker, Anderson, Huff-Lonergan, Dekkers and Lonergan2011) reported no line difference in ultimate pH, drip loss or colour coordinates of loin, but Arkfled et al. (Reference Arkfled, Young, Johnson, Fedler, Prusa, Patience, Dekkers, Gabler, Lonergan and Huff-Lonergan2015) reported lower drip loss, colour scores, lean tissue a* and lipid content and greater moisture content in LRFI pigs in later generations of the same selection experiment.

Table 2 Meat quality traits of the longissimus muscle of pigs from low (n=60) and high (n=57) residual feed intake lines of generation G6

HRFI=high residual feed intake; LRFI=low residual feed intake; p.m.=post mortem.

Adapted from Faure et al. (Reference Faure, Lefaucheur, Bonhomme, Ecolan, Meteau, Coustard, Kouba, Gilbert and Lebret2013).

¹ Ns=non-significant at P>0.05, *P<0.05, **P<0.01, ***P<0.001.

Sensorial analyses in the INRA lines showed that the appearance of raw meat was significantly modified, but eating quality of loin was lowly affected by selection (Figure 2) as reported for the ISU lines (Arkfled et al., Reference Arkfled, Young, Johnson, Fedler, Prusa, Patience, Dekkers, Gabler, Lonergan and Huff-Lonergan2015), despite muscle metabolic responses to selection. Besides, a multidimensional analysis allowed the identification within the LRFI line of a sub-group of efficient pigs that exhibited both lean carcasses and satisfactory levels of technological and sensory meat quality (Supplementary Material S1, Faure et al., 2012).

Figure 2 Sensorial meat quality in the loin evaluated on a 0 to 10 scale (none to very high) in pigs from a low residual feed intake (RFI) line (LRFI, n=60) and a high RFI line (HRFI, n=57) after seven generations of selection. Red colour intensity and marbling intensity were appreciated on raw meat, the other traits were appreciated on cooked meat (dry heat for 10 min at 250°C and then humid heat at 100°C up to a core temperature of 80°C). †0.05<P<0.1, *P<0.05, **P<0.01, ***P<0.001. Adapted from Faure et al. (Reference Faure, Lefaucheur, Bonhomme, Ecolan, Meteau, Coustard, Kouba, Gilbert and Lebret2013).

Growth and feed intake curves and nutrient requirements

Correlated responses to selection for RFI on growth and DFI curves were investigated. Gilbert et al. (Reference Gilbert, Bidanel, Gruand, Caritez, Billon, Guillouet, Lagant, Noblet and Sellier2009) analyzed individual curves for growth and DFI on the first six generations of selection of the RFI lines to obtain individual lysine requirement curves from the InraPorc^® model (Figure 3). Growth rate was slightly lower in LRFI animals throughout the growing period (−7.2% on average, data not shown). The DFI was also lower in LRFI animals (−6% to −13% from early growth to late growth), resulting in higher digestible lysine requirements (in g lysine/MJ NE) during the whole growing period, ranging from +7% at 10 weeks of age up to +20% at slaughter weight. Estimation of genetic parameters with RFI, DFI and shape descriptors of curves describing DFI (e.g. estimated DFI at 50 kg BW, DFI₅₀) and growth showed that DFI₅₀ could be an interesting early predictor for RFI, having a heritability of 0.41 and a genetic correlation of 0.61 with RFI. Line differences in amino acid requirements were later validated experimentally by feeding all pigs a control diet covering requirements of all pigs or by feeding a diet covering the requirements of the lowest 25% of the HRFI pigs. Feeding the low-lysine diet reduced ADG in the LRFI line to a larger extent than in the HRFI line (−25% v. −5%, respectively) (Supplementary Material S1, Brossard et al., 2012). In the ISU RFI lines Cai et al. (Reference Cai, Kaiser and Dekkers2011, Reference Cai, Kaiser and Dekkers2012) also reported that the LRFI line had lower DFI, especially in the later part of the growth period. These results indicate that the dynamics of DFI and possibly growth can respond to selection on RFI over a rather long period (30 to 95 kg). The relationships between RFI and model parameters characterizing growth, DFI and lysine requirements curves were also analyzed in commercial pigs (Saintilan et al., Reference Saintilan, Mérour, Brossard, Tribout, Dourmad, Sellier, Bidanel, van Milgen and Gilbert2015), confirming different dynamics depending on feed efficiency. This indicates that the amino acid content of the diet (and probably other nutrients) must be considered when selecting for efficient pigs to avoid poor performance due to mismatch between nutrient supplies and requirements. This will be more important for entire males, which are leaner and more efficient than barrows.

Figure 3 Curves for daily feed intake and digestible lysine requirements during the growing–finishing period of low residual feed intake (LRFI) and high RFI (HRFI) lines in generations G1 and G5 of selection as predicted using the INRAPorc^® model (n=1370). First, recorded BW were used to fit a Gompertz model to the repeated BW records. Next, the daily feed intake (DFI) records were modelled for each individual with a non-linear exponential model DFI=a×BW ^b (Gilbert et al., Reference Gilbert, Bidanel, Gruand, Caritez, Billon, Guillouet, Lagant, Noblet and Sellier2009). Finally the daily digestible lysine requirements were calculated individually with InraPorc^® on the basis of modelled protein deposition and observed growth and DFI curves as described in Saintilan et al. (Reference Saintilan, Mérour, Brossard, Tribout, Dourmad, Sellier, Bidanel, van Milgen and Gilbert2015).

N and P excretion

Selection for improvement of feed efficiency is expected to reduce the environmental impact through a decrease of N and P excretion, at least by decreasing the total amount of excreta. In the sixth generation of selection for RFI, N and P excretion, evaluated by a modelling approach, was shown to be slightly reduced in LRFI pigs (4.0% and 2.4%, respectively; Supplementary Material S1, Faure et al., 2012). However, some of these differences were lower or not observed in experimental trials during short periods of time where animals were individually housed (Barea et al., Reference Barea, Dubois, Gilbert, Sellier, van Milgen and Noblet2010) or exposed to stressors (Renaudeau et al., Reference Renaudeau, Gourdine, Fleury, Ferchaud, Billon, Noblet and Gilbert2013; Labussière et al., Reference Labussière, Dubois, Gilbert, Thibault, Le Floc’h, Noblet and van Milgen2015). Recent studies on the genetic relationships between feed efficiency and excretion in commercial populations allowed to quantify moderate to high genetic correlations between N excretion and feed efficiency using predictive equations (Table 3; Saintilan et al., Reference Saintilan, Brossard, Vautier, Sellier, Bidanel, van Milgen and Gilbert2013) or direct measurement (Shirali et al., Reference Shirali, Doeschl-Wilson, Duthie, Knap, Kanis, van Arendonk and Roehe2012, Reference Shirali, Doeschl-Wilson, Knap, Duthie, Kanis, van Arendonk and Roehe2014). Also, Saintilan et al. (Reference Saintilan, Brossard, Vautier, Sellier, Bidanel, van Milgen and Gilbert2013) reported moderate to high correlations between P excretion and feed efficiency in these populations. However, these studies were performed with only one diet fed to the animals, and provide no evidence of differences in efficiency of N or P deposition in relation with RFI. The correlations confirm the potential of the selection for reduced RFI or FCR to decrease excretion when the diet is not adjusted to the pig requirements. Taking into account the differences in nutrient requirements between animals to supply adequate nutrient levels should enhance the effects of RFI improvement. However, possible interactions with feed composition and stress must also be considered when reasoning the impact of RFI on excretion levels.

Table 3 Genetic correlations (SE) between feed efficiency traits (feed conversion ratio (FCR); residual feed intake (RFI)) and excretion traits recorded in performance test station in three commercial populations

Adapted from Saintilan et al. (Reference Saintilan, Brossard, Vautier, Sellier, Bidanel, van Milgen and Gilbert2013).

DFI=average daily feed intake; ADG=average daily gain; Ne=N quantity excreted during the test period; Pe=P quantity excreted during the test period; Nr=proportion of retained nitrogen relative to feed intake; Pr=proportion of retained phosphorus relative to feed intake.

Genomic dissection of residual feed intake

Even though feed intake can be measured accurately, it is an expensive trait to record. The advances in genomics now allow the identification of genomic regions affecting the variability of quantitative traits (QTL) to consider selection schemes incorporating molecular information as predictors of feed efficiency and related traits. A genome wide association study to detect QTL for RFI and production traits was run using a single-step approach in the selected lines (Riquet et al., Reference Riquet, Labrune, Feve, Billon and Gilbert2014). All sires from generations G0 to G6 and dams from generations G0, G3 and G6 were genotyped with 60 K single nucleotid polymorphisms (SNPs) (149 LRFI and 121 HRFI). Only few significant regions were identified (Figure 4) for all traits given the limited power of the study and only suggestive regions were identified for RFI: the most significant one was located on sus scrofa chromosome (SSC) 16 (Figure 4). Three, two and one regions were detected for DFI, ADG and FCR, respectively. Some of these regions were already detected in other experiments: the QTL detected on SSC7 affecting DFI and ADG could be similar to the QTL influencing the same traits in the ISU lines at position 125 Mb (Onteru et al., Reference Onteru, Gorbach, Young, Garrick, Dekkers and Rothschild2013). The two QTL influencing ADG (SSC7, 130.5 Mb; SSC18, 32.5 Mb) have also been reported by Fontanesi et al. (Reference Fontanesi, Schiavo, Galimberti, Calò and Russo2014). On the opposite, the regions detected on SSC10 and SSC14 for DFI and SSC1 for FCR have not been reported before, neither the suggestive region for RFI on SSC16. A comparative analysis of these candidate regions with bovine results highlighted two suggestive regions on SSC7 and SSC8 orthologous to QTL regions affecting RFI in cattle (bos taurus chromosome (BTA)BTA21, 70 Mb: Santana et al., Reference Santana, Utsunomiya, Neves, Gomes, Garcia, Fukumasu, Silva, Oliveira Junior, Alexandre, Leme, Brassaloti, Coutinho, Lopes, Meirelles, Eler and Ferraz2014 and BTA6, 45 Mb: Rolf et al., Reference Rolf, Taylor, Schnabel, McKay, McClure, Northcutt, Kerley and Weaber2012). These first results are promising, but require confirmation. The absence of a significant region for RFI suggests that biological strategies for improving this trait are diverse. The identification of the sub-traits contributing to RFI could also contribute to the dissection of the trait.

Figure 4 (a) Manhattan plot for detection of associations between SNP and residual feed intake (RFI) (n=149 low RFI pigs, n=121 high RFI pigs), using a genomic relationship matrix to account for the pedigree structure. The y-axis corresponds to the ratio between the variance explained by successive 0.3 Mb windows as estimated from a single-step genomic selection approach and the empirical 5% significance threshold at the genome level computed by simulation under the null hypothesis. (b) Chromosomal location of SNPs with suggestive (small italic) or significant (large bold) P for each group of traits (growth rate, feed intake and feed efficiency in black, carcass traits in dark grey and meat quality traits in light grey). Adapted from Riquet et al. (Reference Riquet, Labrune, Feve, Billon and Gilbert2014). FCR=feed conversion ratio; RFI=residual feed intake; DFI=daily feed intake; ADG=average daily gain; LMC=lean meat content of the carcass computed from a linear combination of cut weights; backfat=backfat thickness measured on carcass or backfat weight; Length=carcass length; Yield=carcass yield; pHu=pH determined 24h after slaughter on adductor, semimembranosus, gluteus superficialis or longissimus muscles; L=lightness, a=redness, b=yellowness, measured 24 h after slaughter on gluteus superficialis or gluteus medius muscles; WHC=water holding capacity; MQI=meat quality index.

Digestion and fibrous diets

The digestibility was measured at multiple sites of the gastro intestinal tract (GIT) to compare the digestive efficiency of the lines at each step of the digestion process. However, there was no line difference in these digestibility measurements for nutrients and energy during the post-weaning and growing stages in pigs fed conventional European diets (Barea et al., Reference Barea, Dubois, Gilbert, Sellier, van Milgen and Noblet2010; Montagne et al., Reference Montagne, Loisel, Le Naou, Gondret, Gilbert and Le Gall2014; Labussière et al., Reference Labussière, Dubois, Gilbert, Thibault, Le Floc’h, Noblet and van Milgen2015). Such minor role of digestion in explaining RFI differences has been reported in laying hens and broilers (Supplementary Material S1, Luiting et al., 1994) and in mice (Supplementary Material S1, Bunger et al., 1998), whereas digestibility has been shown to contribute very significantly to RFI in beef cattle (Herd and Arthur, Reference Herd and Arthur2009).

Compared with feeding a control diet, feeding the LRFI and HRFI pigs with a low-energy high-fibre diet supplied by a mixture of wheat bran, dehydrated sugar beet pulp and soya bean hulls during 3 weeks decreased digestibility similarly in both lines, despite lower weights of the digestive tract in LRFI pigs (Montagne et al., Reference Montagne, Loisel, Le Naou, Gondret, Gilbert and Le Gall2014). The faecal NDF digestibility and concentrations of volatile fatty acids in the caecum were lower in LRFI pigs compared with HRFI pigs when fed the high-fibre diet (Montagne et al., Reference Montagne, Loisel, Le Naou, Gondret, Gilbert and Le Gall2014). The increase in eating time of pigs fed with a low-energy high-fibre diet was 22% for the HRFI line, whereas it was up to 30% for the LRFI line (Supplementary Material S1, Hauptli et al., 2013). The dietary fibre significantly increased the ratio of acetate to propionate concentrations in the distal part of the GIT for the HRFI line only, which might affect the metabolism of peripheral tissues. The visceral mass is a major source of utilization of nutrients and energy (Supplementary Material S1, Yen et al., 1989), and fermentation in the hindgut diverts nutrients from the animal metabolism. Both were reduced in LRFI pigs, which may contribute to their better efficiency. The reduction of ADG observed when feeding a high-fibre diet during 3 weeks was smaller in the LRFI pigs than in the HRFI pigs (Montagne et al., Reference Montagne, Loisel, Le Naou, Gondret, Gilbert and Le Gall2014). However, other studies found a similar reduction in ADG in both lines when a high-energy high-fibre diet was given to pigs during 10 weeks (Gondret et al., Reference Gondret, Louveau, Mourot, Duclos, Lagarrigue, Gilbert and van Milgen2014). The LRFI pigs from the ISU lines fed a conventional US diet had a greater digestive efficiency in early generations (Harris et al., Reference Harris, Patience, Lonergan, Dekkers and Gabler2012), which was only found in animals fed a low-energy high-fibre diet in the later generations (Supplementary Material S1, Mauch et al., 2015). Altogether, the digestive efficiency does not seem to explain the variation in RFI between lines. The two lines also show some differences in gut microbiota composition, in particular in the phylotypes of the Prevotella and Lactobacillus genus present in each line (Zemb et al., submitted) that could contribute to their different use of the feed.

Energy and protein metabolism in healthy animals

Both nutrient use and metabolism in tissues may participate to the biological basis of feed efficiency. Different observations suggest that LRFI pigs have a reduced nutrient catabolism for energy production in skeletal muscles. Indeed, lower glycolytic and lower oxidative enzyme activities have been reported in their muscles (Le Naou et al., Reference Le Naou, Le Floc’h, Louveau, Gilbert and Gondret2012; Faure et al., Reference Faure, Lefaucheur, Bonhomme, Ecolan, Meteau, Coustard, Kouba, Gilbert and Lebret2013). These changes were associated with a reduced activity of the adenosine monophosphate-activated protein kinase (AMPK), a sensor of the cellular energy deficit that generally acts as a master switch to stimulate oxidation. Importantly, a reduced oxidation of nutrients to provide ATP later released as heat was found in LRFI pigs (Barea et al., Reference Barea, Dubois, Gilbert, Sellier, van Milgen and Noblet2010; Renaudeau et al., Reference Renaudeau, Gourdine, Fleury, Ferchaud, Billon, Noblet and Gilbert2013; Labussière et al., Reference Labussière, Dubois, Gilbert, Thibault, Le Floc’h, Noblet and van Milgen2015). In the LM of LRFI pigs, lower expression levels of genes associated with mitochondrial metabolism (Vincent et al., Reference Vincent, Louveau, Gondret, Tréfeu, Gilbert and Lefaucheur2015) have been reported, including several antioxidant proteins. This could indicate a lower oxidative stress: a positive correlation between RFI and reactive oxygen species was also reported in LM mitochondria in ISU lines (Grubbs et al., Reference Grubbs, Fritchen, Huff-Lonergan, Dekkers, Gabler and Lonergan2013). However, when evaluated in bundles of 20 to 30 fresh permeabilized myofibres, no difference in the potential of mitochondrial respiration or in the expression of genes coding for uncoupling proteins were found between LRFI and HRFI lines in earlier generations of selection (Lefaucheur et al., Reference Lefaucheur, Lebret, Ecolan, Louveau, Damon, Prunier, Billon, Sellier and Gilbert2011). Because efficiency is directly linked to energy use, muscles of LRFI pigs may have fewer but more efficient mitochondria. Decreased rates of the Cori cycle (involving glucose and lactate turnovers between muscle and liver) may also participate to limit energy losses in the LRFI pigs (Le Naou et al., Reference Le Naou, Le Floc’h, Louveau, Gilbert and Gondret2012). Paradoxically, in adipose tissues increased mitochondrial oxidative enzyme activities had been reported in LRFI pigs (Gondret et al., Reference Gondret, Louveau, Mourot, Duclos, Lagarrigue, Gilbert and van Milgen2014), a metabolic change that can account for their leaner phenotype. Other molecular changes in adipose tissues concerned many genes involved in the regulation of apoptosis and cell death and immunity pathways (Louveau et al., Reference Louveau, Vincent, Tacher, Gilbert and Gondret2016).

It is also reasonable to suspect changes in protein metabolism to account for differences in efficiency between RFI lines. However, due to differences in methodologies, age/weights of studied animals and experimental designs, it is difficult to render definitive conclusions on this aspect. First, there was no difference in nitrogen utilization (as % of absorbed N) between LRFI and HRFI lines, resulting in similar rates of protein deposition in animals from the 6^th generation (Barea et al., Reference Barea, Dubois, Gilbert, Sellier, van Milgen and Noblet2010). Nitrogen utilization and protein deposition were also similar between lines in the 7^th generation during post-weaning (Labussière et al., Reference Labussière, Dubois, Gilbert, Thibault, Le Floc’h, Noblet and van Milgen2015), but were lower in LRFI animals during the growing period (Renaudeau et al., Reference Renaudeau, Gourdine, Fleury, Ferchaud, Billon, Noblet and Gilbert2013). Second, similar rate of protein synthesis and expressions of protein synthesis markers have been reported in LM in the two lines both at INRA (Le Naou et al., Reference Le Naou, Le Floc’h, Louveau, Gilbert and Gondret2012) and at ISU (Cruzen et al., Reference Cruzen, Harris, Hollinger, Punt, Grubbs, Selsby, Dekkers, Gabler, Lonergan and Huff-Lonergan2013). On the other hand, Vincent et al. (Reference Vincent, Louveau, Gondret, Tréfeu, Gilbert and Lefaucheur2015) have shown an over-expression of various genes encoding initiation and elongation translation factor subunits in LM of LRFI pigs at market weight. Regarding protein catabolism, the overall proteasome and calpain activities in muscle did not differ between RFI lines when pigs were considered early in the post-weaning period as well as at market weight (Le Naou et al., Reference Le Naou, Le Floc’h, Louveau, Gilbert and Gondret2012). However, Cruzen et al. (Reference Cruzen, Harris, Hollinger, Punt, Grubbs, Selsby, Dekkers, Gabler, Lonergan and Huff-Lonergan2013) have reported that muscles of ISU LRFI pigs at 68 kg BW exhibited lower activities of the 20S proteasome, a key catalytic subunit responsible for proteolysis of ubiquitin-tagged proteins, and of calpains involved in Ca²⁺-dependent proteolysis. Further investigations are thus needed to clarify the possible involvement of protein metabolism in the muscle and the viscera in divergence for RFI.

Basal metabolism

An important fraction of the difference in energy requirements between LRFI and HRFI lines is related to differences in basal metabolic rate. In growing pigs, basal metabolic rate is generally estimated through the determination of fasting heat production (FHP). In 60 kg BW pigs kept in thermoneutral conditions (24°C) and fed close to ad libitum, a modelling approach used to partition total heat production (HP) between its different components indicated that FHP was lower in LRFI pigs than in HRFI pigs (−10% on average, Barea et al., Reference Barea, Dubois, Gilbert, Sellier, van Milgen and Noblet2010), which is consistent with results reported by Boddicker et al. (2011) in the ISU lines. Similarly, basal metabolic rate estimated by HP measurements in feed-deprived animals has also been shown to be numerically lower in LRFI laying hens (−15%, Supplementary material S1, Gabarrou et al., 1997) and cockerels (Swennen et al., Reference Swennen, Verhulst, Collin, Bordas, Verbeke, Vansant, Decuypere and Buyse2007). Little information is available in the literature on the physiological mechanisms underlying the lower FHP in LRFI pigs. As in others species (Supplementary Material S1, Bottje et al., 2006; Herd and Arthur, Reference Herd and Arthur2009; Murphy et al., Reference Murphy, McDonald and Nielsen2013), differences in mitochondria number or activity in tissues could contribute to explain changes in basal metabolic rate between the two lines (see Energy and protein metabolism in healthy animals section). In addition, the variation in energy expenditure and especially the FHP component due to line differences in the size of visceral organs can contribute to the differences in basal metabolic rate between animals with different RFI (see Digestion and fibrous diets section). However the direct effect of a decreased FI only explained 25% of the maintenance difference between the lines (Labussière et al., Reference Labussière, Dubois, Gilbert, Thibault, Le Floc’h, Noblet and van Milgen2015).

Activity and feeding behaviour

Activity is one of the main non-productive functions contributing to energy use in the pig. The LRFI pigs spent less time standing (Table 4), leading to reduced physical activity (−2.5% of a 24 h scan for time standing, i.e. −35 min, representing a 21% difference in generation G6; Meunier-Salaün et al., Reference Meunier-Salaün, Guérin, Billon, Sellier, Noblet and Gilbert2014). This difference represented 14% of the line difference in metabolizable energy (ME) intake (Meunier-Salaün et al., Reference Meunier-Salaün, Guérin, Billon, Sellier, Noblet and Gilbert2014), which agrees with calorimetric observations (17%; Barea et al., Reference Barea, Dubois, Gilbert, Sellier, van Milgen and Noblet2010). A similar line effect in physical activity was described in the ISU lines (Sadler et al., Reference Sadler, Johnson, Lonergan, Nettleton and Dekkers2011). Investigations conducted in other species also reported differences in physical activity related to RFI (Supplementary Material S1, Luiting et al., 1994; Supplementary Material S1, Bunger et al., 1998; Herd and Arthur, Reference Herd and Arthur2009). These differences were not due to a higher prevalence of lameness in LRFI pigs (Meunier-Salaün et al., Reference Meunier-Salaün, Guérin, Billon, Sellier, Noblet and Gilbert2014): in fact, higher scores were observed in HRFI pigs, potentially related to their higher physical activity. Social and pen investigations were slightly reduced in LRFI pigs, but feeding activity significantly contributed to differences in physical activity (Table 4): LRFI pigs showed shorter daily eating time, lower number of visits and increased feeding rate (Gilbert et al., Reference Gilbert, Bidanel, Gruand, Caritez, Billon, Guillouet, Lagant, Noblet and Sellier2009; Meunier-Salaün et al., Reference Meunier-Salaün, Guérin, Billon, Sellier, Noblet and Gilbert2014). In ISU lines, Young et al. (Reference Young, Cai and Dekkers2011) reported a trend for LRFI to visit the feeders fewer times and to spend less time eating per visit. These line differences could be modified if the pen composition was modified, for example, by mixing lines or sexes. Investigations of the impact of the selection on the social interactions between individuals have to be explored to complete the evaluation of the activity component.

Table 4 Behavioural activity in pigs from the low residual feed intake (LRFI) and high residual feed intake (HRFI) lines, on pigs from generation G6

Adapted from Meunier-Salaün et al. (Reference Meunier-Salaün, Guérin, Billon, Sellier, Noblet and Gilbert2014).

¹ Traits in reference to the animal trait ontology for livestock ATOL: http://www.atol-ontology.com/index.php/en/.

² Ns=non-significant at P>0.10, ***P<0.001.

³ Video recordings during 24 h at 17 weeks of age (mean BW 77±8 kg); % on total scan (n=96 LRFI, n=96 HRFI). Least square means, RMSE and P from linear mixed models including the batch, line and sex as fixed effects and interactions between these factors. Analysis was done on the ratio value after arcsin racin transformation.

⁴ Feeding patterns were determined during a 3 to 4 weeks-period surrounding 12 (P1), 17 (P2) and 22 (P3) weeks of age (n=140 LRFI; n=125 HRFI). Least square means, RMSE and P from linear mixed models including the batch, line, sex and growth stage (P1=54±7 kg; P2=77±8 kg; P3=96±9 kg) as fixed effects, interactions between these factors and the animal as repeated random effect. Only line effects for period P2 are presented.

Biological markers

Identifying early RFI biomarkers to easily measure large numbers of animals is highly desirable. Among the potential sources of biomarkers, blood is the most frequently targeted biological fluid: it stands as a surrogate tissue that can be repeatedly sampled by minimally invasive procedures. First, circulating concentrations of IGF-I, a growth factor synthesized by the liver and most tissues and known to play a major role in growth and metabolism (Supplementary Material S1, Le Roith et al., 2001) have been analyzed. Juvenile IGF-I, that is circulating IGF-I measured soon after weaning, has been shown to have significant genetic correlations with production traits in commercial populations (Supplementary Material S1, Bunter et al., 2005), and to respond to selection for RFI in the ISU lines (Bunter et al., Reference Bunter, Cai, Johnston and Dekkers2010). Investigations in the INRA RFI lines confirmed lower circulating juvenile IGF-I concentrations in LRFI pigs compared with HRFI pigs (Bunter K. and Louveau I., unpublished results). This difference was not found at market weight (Lefaucheur et al., Reference Lefaucheur, Lebret, Ecolan, Louveau, Damon, Prunier, Billon, Sellier and Gilbert2011; Le Naou et al., Reference Le Naou, Le Floc’h, Louveau, Gilbert and Gondret2012). Second, a lower plasma leptin level in the fed state has been observed in LRFI pigs at 100 kg BW (Lefaucheur et al., Reference Lefaucheur, Lebret, Ecolan, Louveau, Damon, Prunier, Billon, Sellier and Gilbert2011) and a positive genetic correlation between RFI and serum leptin concentrations has been reported in Duroc pigs (Hoque et al., Reference Hoque, Katoh and Suzuki2009). However, the precise relationships between these blood measurements and RFI deserve further studies.

Advances in high-throughput technologies offer new opportunities to identify novel physiological markers of feed efficiency. Plasma metabolomic profiles did not clearly discriminate 132-day-old pigs from the RFI lines (Jégou et al., Reference Jégou, Gondret, Vincent, Tréfeu, Gilbert and Louveau2015). By contrast, transcriptomic profiles of whole blood examined between 35 and 132 days of age showed line differences both at INRA (Jégou et al., Reference Jégou, Gondret, Martin-Lalande, Tea, Baéza and Louveau2016) and at ISU (Supplementary Material S1, Liu et al., 2015). The main differences were found in the expression of genes associated with the immune system, despite limited line differences in response to inflammatory and immune challenges (see Stress and other functions: robustness and residual feed intake section). Proteomic analyses may also be a relevant approach to find serum biomarkers of RFI in young pigs (Grubbs et al., Reference Grubbs, Lonergan, Dekkers and Tuggle2014). Altogether, these putative candidates need to be further validated as relevant biomarkers of feed efficiency.

Response to an inflammatory challenge

The inflammatory response is caused by sanitary events the pig has to face during his lifespan and is a typical situation inducing trade-offs between productive and non-productive functions. To investigate the line differences in metabolic and immune responses to an inflammatory challenge, weaned piglets were injected with Complete Freund’s Adjuvant (CFA) to induce a non-infectious pneumonia (Supplementary Material S1, Melchior et al., 2004). Both lines showed a similar transient depression in FI during the first 24 h and hyperthermia during the first 2 days. They also displayed similar increases in blood levels of haptoglobin, a hepatic inflammatory protein, and interferon (IFN)-γ, an inflammatory cytokine (Merlot et al., Reference Merlot, Evrard, Vincent, Gilbert and Le Floc’h2011). Blood transcriptome analysis indicated that few genes related to inflammation and immunity as well as to other functions, including antioxidant defences, proteasome and lysosome functions were differently expressed between lines during the acute phase of the inflammatory challenge. Together with the transcriptome data from whole blood and adipose tissue of older healthy animals from the INRA lines presented above, these results suggest that the differences in the gene profiling could account for baseline differences between the LRFI and HRFI lines rather than for differences in the line capacity to respond to inflammation (Rogel-Gaillard C. and Merlot E., unpublished results). In the tissues where the immune response develops, for example in lungs and their draining lymph nodes, the expression of inflammatory cytokines was lower in LRFI pigs 1 week after the CFA challenge.

The first 2 days after CFA administration, contrary to HRFI piglets, LRFI piglets exhibited a marked reorientation of nutrients from anabolic to catabolic pathways (Labussière et al., Reference Labussière, Dubois, Gilbert, Thibault, Le Floc’h, Noblet and van Milgen2015). On the 8^th day, during the recovery phase, differences in the potential for protein accretion were observed between lines (Labussière et al., Reference Labussière, Dubois, Gilbert, Thibault, Le Floc’h, Noblet and van Milgen2015; Merlot et al., Reference Merlot, Gilbert and Le Floc’h2016). The tendency for a higher muscle protein synthesis rate and lower indicators of protein catabolism (muscle calpain activity and plasma hydroxyproline levels), and the higher blood clearance of dietary amino acid (AA) immediately after the meal in LRFI compared with HRFI pigs supports the hypothesis that selection for low RFI favoured the preservation of muscular protein accretion during inflammation (Merlot et al., Reference Merlot, Gilbert and Le Floc’h2016). Divergent selection might have generated preferences for different energetic pathways in LRFI and HRFI pigs to respond to inflammation. Indeed, individual AA plasma concentrations on the week after the CFA challenge suggested that the alanine cycle, producing pyruvate in the liver, might be more reduced in LRFI than in HRFI pigs. To conclude, in young pigs the lines differed moderately in their immune and metabolic responses to an inflammatory challenge, and no disadvantage was observed for the LRFI line. Further studies of chronic immunity and sanitary challenges should provide a complementary understanding of these line responses.

Response to heat stress

In hot conditions, the capacity to maintain homoeostasis is driven by the animal’s ability to lose heat and/or to reduce its metabolic HP. Selection for low RFI reduced the total amount of heat produced by unit of ME intake (see Basal metabolism section), which could favour LRFI pigs to cope with high temperatures. In a first approach, thermoregulatory responses and energy utilization in HRFI and LRFI lines were compared during a standardized thermal challenge (Renaudeau et al., Reference Renaudeau, Frances, Dubois, Gilbert and Noblet2008). Although not significant, LRFI pigs had a numerically lower reduction of FI in hot conditions than HRFI pigs (Campos et al., Reference Campos, Noblet, Jaguelin-Peyraud, Gilbert, Mormede, de Oliveira Donzele, Donzele and Renaudeau2014). In addition, the time required to initiate acclimation responses was shorter in LRFI compared with HRFI pigs (Figure 5). However, this favourable low RFI line effect was not accompanied by changes in blood hormones concentration and/or in the ability to lose heat (Renaudeau et al., Reference Renaudeau, Gourdine, Fleury, Ferchaud, Billon, Noblet and Gilbert2013; Campos et al., Reference Campos, Noblet, Jaguelin-Peyraud, Gilbert, Mormede, de Oliveira Donzele, Donzele and Renaudeau2014). In the former study, changes in energy metabolism during the thermal acclimation period at 32°C were similar in both lines. In connection with their higher water intake (see Responses to selection on production and carcass traits section), HRFI pigs tended to have a greater capacity to lose heat by evaporation in hot and thermoneutral conditions that could compensate their greater HP. Similar results were reported in HRFI laying hens (Supplementary Material S1, Bordas and Minvielle, 1997). In a second approach, the RFI lines have been evaluated in a tropical environment. From 11 to 23 weeks of age, the ADG was slightly lower in the LRFI than in the HRFI. This effect was mainly explained by a reduced growth of the LRFI line during the post-weaning period without a compensation during the growing–finishing period (Supplementary Material S1, Gilbert et al., 2012b).

Figure 5 Effects of line and high ambient temperature on rectal temperature and respiratory rate in growing pigs from the low residual feed intake (LRFI) and high residual feed intake (HRFI) lines. The thermoregulatory responses were modelled using the following equation: Y=y ₀+v ₁ h−r ₁(v ₁−v ₂) ln{1+exp[(d−th₁)/r ₁]}−r ₂(v ₂−v ₃) ln{1+exp[(d−th₂)/r ₂]} where Y is the response variable; d the day of exposure, y ₀ the value of Y on day 0 L; th₁ and th₂ the threshold days of the first and second phase of response, respectively; v ₁, v ₂ and v ₃ are the linear variations of Y before and after th₁ and th₂, respectively. Only the th₁ parameter for rectal temperature was significantly affected by the line (0.85 v. 1.88 day, for the LRFI and the HRFI line, respectively). Adapted from Campos et al. (Reference Campos, Noblet, Jaguelin-Peyraud, Gilbert, Mormede, de Oliveira Donzele, Donzele and Renaudeau2014).

Lactation

Decreased voluntary FI during growth combined with increased growth rate, leanness and prolificacy limits the availability of sow resources to face lactation, which is the only period of negative energy balance in healthy pigs, with potential deleterious impacts on reproductive performance and longevity (Prunier et al., Reference Prunier, Heinonen and Quesnel2010). Despite low genetic correlations between RFI and sow reproductive traits (Gilbert et al., Reference Gilbert, Bidanel, Billon, Lagant, Guillouet, Sellier, J Noblet and Hermesch2012), the correlated responses to selection after seven generations of selection (Figure 6) showed significantly reduced daily FI during lactation (−280 g/day) in the LRFI line compared with the HRFI line, increased loss of BW and BFT (+5.6 kg BW and +1.3 mm BFT), higher litter weight at 21 days of age (+2 kg), and +0.6 weaned piglet, but no change in piglet BW at weaning (28 days of age) despite numerically heavier piglets in most studies in the LRFI line. Given the low number of unsuccessful inseminations, no significant line difference was reported for rebreeding. Young et al. (Supplementary Material S1, 2010) obtained similar results at the phenotypic level in sows of the ISU lines after seven generations of selection. Applying the RFI concept to lactating females (Supplementary Material S1, Veerkamp et al., 1995) on the INRA sows showed a reduction of 110 g/day of the lactation RFI after seven generations of selection in LRFI sows (Gilbert et al., Reference Gilbert, Bidanel, Billon, Lagant, Guillouet, Sellier, J Noblet and Hermesch2012), indicating that LRFI sows are more resource-efficient for a given production level. When assessed in breeding sows from the eighth generation in tropical conditions (Renaudeau et al., Reference Renaudeau, Kerdoncuff, Anaïs and Gourdine2014), the line difference in voluntary FI during lactation was further increased (−590 g/day in the LRFI line compared with the HRFI line), as were BW and BFT losses, probably essentially due to the combination of heat and humidity. Altogether, the two experiments show no deleterious impact of selection for low RFI during growth on sow reproduction traits, even when FI is restricted by heat stress. This can be interpreted as an improved capacity of LRFI sows to produce milk and ensure litter survival compared with HRFI sows, suggesting a higher robustness of the LRFI sows. The dynamics of body resources and FI during lactation and gestation over multiple parities in LRFI pigs could provide indicators of life production efficiency of the sows.

Figure 6 Correlated responses to selection for residual feed intake (RFI) after seven generations of divergent selection for sow reproductive traits. LRFI=low RFI line; HRFI=high RFI line; SDFI=sow daily feed intake; loss-BW=body weight loss during lactation; loss-BFT=backfat thickness loss during lactation; D-LW21=litter weight gain from farrowing to 21 days of age; SRFI=sow residual feed intake during lactation. All line differences were significant at P<0.001. Adapted from Gilbert et al. (Reference Gilbert, Bidanel, Billon, Lagant, Guillouet, Sellier, J Noblet and Hermesch2012).