Feed Restriction Modifies Intestinal Microbiota-Host Mucosal Networking in Chickens Divergent in Residual Feed Intake

The impact of the FE-associated differences in feed intake on intestinal bacterial and host physiological parameters has so far not been clarified. Understanding the underlying principles is essential for the development of cost-effective strategies to improve FE in chicken production. Under conditions of quantitative feed restriction, low- and high-RFI chickens ate the same amount of feed. Therefore, this research helps in distinguishing intestinal bacterial taxa and functions that were highly reliant on feed intake from those that were associated with physiological adaptations to RFI-associated differences in host nutritional needs and intestinal nutrient availability. This work provides a background for further research to assess manipulation of the intestinal microbiota, host physiology, and FE in chickens by dietary intervention.

U nderstanding the mechanisms that underpin individual animal variation is mandatory for the development of effective strategies to improve feed efficiency (FE) to decrease overall production costs, preserve additional edible resources for humans, and reduce the ecological footprint of chicken rearing systems (1,2). Typically, feedefficient chickens have lower feed intake than chickens with low feed efficiency and show differences from those chickens in their intestinal microbiota as well as in intestinal structural and functional characteristics (2)(3)(4)(5). From the data available, it is not clear whether this difference in the feeding behavior is the sole causative factor or whether other host-related factors may underlie the FE-associated variation in the intestinal microbiota and physiology in chickens. The feed intake level (FL) of a chicken modifies the size of the visceral organs, digesta volume, retention time, and nutrient digestion (6) and may alter the intestinal microbiota composition as reported for other livestock species (e.g., sheep) (7), which may have consequences for FE. So far, the impact of the FL on the intestinal microbial community structure and function has not been elucidated in chickens of divergent FE.
In order to study the effect of the FL, it is essential that all birds eat the same amount of feed, irrespective of FE. This may be achieved by limiting the feed intake by means of quantitative feed restriction. Feed restriction in chicken rearing is used to prevent metabolic disorders (e.g., sudden death syndrome and ascites) and to manipulate carcass composition (8). However, feed restriction has also been shown to cause intestinal microbial and digestive-physiological adaptations (9,10) and may impair the intestinal barrier function in chickens (11). The latter finding is interesting in relation to the FE of the birds because we observed greater jejunal permeability in chickens of high FE which had a limited feed intake compared to chickens of low FE in a previous study (5). In general, enhanced mucosal permeability may facilitate the paracellular uptake of nutrients but it may also allow enteric bacteria or their toxic compounds to more easily translocate into the body, thereby triggering a stronger immune response and reducing of the growth performance of the chicken (11).
If the FL is a factor influencing FE-associated variation in the intestine, it may be possible, by imposing the same limitation on all chickens with respect to the amount of feed, to create similar intestinal conditions (i.e., substrate availability, passage rate) regardless of a given bird's FE. Therefore, we hypothesized that restrictive feeding would produce similar intestinal bacterial profiles as well as similar results with respect to development and function of the intestinal epithelium in chickens with low feed efficiency compared to chickens with high feed efficiency. Our objective was to investigate the effect of restrictive versus ad libitum feeding on the ileal and cecal microbiota, visceral organ size, and intestinal morphology and permeability and expression of genes in relation to nutrient transporters, barrier function, and innate immune response in broiler chickens of divergent FE. As a metric for FE, we used the residual feed intake (RFI), with a low RFI value representing good FE and a high (positive) RFI value representing poor FE. The RFI of each chicken was determined for the experimental period from 9 to 30 days posthatch (dph). As the chickens with high and low feed efficiency were assumed to have the most highly contrasting intestinal microbiota and intestinal structure and function, alterations caused by the feed restriction might be more easily detected in those chickens. Therefore, we studied only chickens with the lowest and highest RFI values in the restrictively and ad libitum-fed groups.

RESULTS
Restrictive feeding improves feed efficiency. Chickens with extremely low and high RFI values were selected in both FL groups to discriminate between their intestinal microbiota, structure, and function characteristics. Consequently, chickens in both FL groups had largely contrasting RFI values (see Table S1 in the supplemental material). Nevertheless, as indicated by the FL ϫ RFI interaction (P Ͻ 0.05), restrictive feeding decreased the RFI value of high-RFI chickens by 153 g compared to ad libitum-fed high-RFI chickens. Although post hoc comparisons showed similar trends (P Ͻ 0.10) for the 2 groups of high-RFI males, the FL ϫ RFI interaction did not reach significance (FL ϫ RFI, P ϭ 0.171). Overall, restrictive feeding lowered the total feed intake (TFI) of chickens by 338 g on average between 9 and 30 dph compared to ad libitum-fed birds (P Ͻ 0.05). Because low-RFI chickens commonly ate less than high-RFI chickens (12), the feed restriction was less severe among the low-RFI chickens (92% of ad libitum group) than among the high-RFI chickens (80% of ad libitum group).
Feed restriction alters bacterial microbiome in ileal and cecal digesta. Samples were rarefied to 2,079 reads to account for unequal numbers of sequences between samples before calculating ␣-diversity indices. Restrictive feeding increased the ileal bacterial diversity (P Ͻ 0.05) as well as the ileal species abundances (P Ͻ 0.05) and, as a trend, cecal species abundances (P Ͻ 0.10), whereas the RFI rank did not alter the ileal and cecal ␣-diversity among chicken groups (Table S2). Nonparametric multidimensional scaling (NMDS) was used to represent separation of the taxonomic composition data among the 4 groups. A significant separation was detected in Bray-Curtis-derived dissimilarity matrices (permutational multivariate analysis of variance [PERMANOVA]) for bacterial composition in ileal digesta (P ϭ 0.003; Fig. 1A) and cecal digesta (P ϭ 0.001; Fig. 1B) between feeding-level groups, whereas the RFI rank did not show a significant effect (P Ͼ 0.10) on the ␤-diversity structure irrespective of the FL group examined. Moreover, it was discernible that, in particular, the taxonomic composition in the ceca of the restrictively fed high-RFI group differed from that of the ad libitum-fed high-RFI group, as those groups clustered separately in the NMDS plot.
Taxonomic assessment of the bacterial microbiota in ileal and cecal digesta showed that restrictive feeding enriched the Firmicutes population numbers and decreased the Proteobacteria population by 11% at both intestinal sites (P Ͻ 0.05; Table S3), with family-level analysis indicating certain intestinal site-specific differences in the FLrelated abundances in taxa, though. In the ileal digesta, the most drastic changes in population numbers within the Firmicutes were among the high-abundance taxa, including a trend showing a 16.5% decrease in the population of Turicibacteraceae (P ϭ 0.058) and an 18.5% increase in the population of Lactobacillaceae (P ϭ 0.046), as well as trends for increases of 2.9% and 0.5% in the populations of Ruminococcaceae (P ϭ 0.084) and Lachnospiraceae (P ϭ 0.096), respectively, in restrictively fed chickens compared to ad libitum-fed chickens (Table 1). With regard to Proteobacteria, restrictive feeding decreased the level of Enterobacteriaceae by 15.7% in ileal digesta compared to ad libitum feeding (P ϭ 0.033). Similar shifts within the Firmicutes and Proteobacteria were seen in cecal digesta. In addition to the aforementioned families, populations of an unclassified family of the order Clostridiales, the predominant Firmicutes family in cecal digesta, were enriched by 13.9% in the ceca of restrictively fed birds compared to ad libitum-fed birds (P Ͻ 0.001). The numbers of the second and third most dominant Firmicutes families in the ceca, Ruminococcaceae and Lachnospiraceae, and the dominant Proteobacteria family (i.e., Enterobacteriaceae) declined by 5.6, 3.1, and 5.0%, respectively, with restrictive compared to ad libitum feeding (P Ͻ 0.05). Low-abundance Firmicutes families that further decreased in number with restrictive feeding included Peptostreptococcaceae (P Ͻ 0.001) (and results showed similar trends for Clostridiaceae and Enterococcaceae [P Ͻ 0.10]) in cecal digesta, whereas Dehalobacteriaceae, as a trend in ileal (P ϭ 0.054) and significantly in cecal digesta (P ϭ 0.007), and Christensenellaceae, as a trend in cecal digesta (P Ͻ 0.10), were enriched with restrictive compared to ad libitum feeding. RFI-associated changes were found in the ceca only at the family level for low-abundant Christensenellaceae (P Ͻ 0.05) and, as a trend, Bacillaceae (P Ͻ 0.10), which were more and less abundant in low-RFI and high-RFI chickens, respectively.
Restrictive feeding and RFI-associated effects on ileal and cecal fermentation acids. Restrictive feeding of chickens mainly decreased the ileal and cecal concentrations of acetate, resulting in reductions in total short-chain fatty acid (SCFA) concentrations compared to ad libitum-fed chickens (P Ͻ 0.05) ( Table 2). In addition to the measured concentrations, FL and RFI rank affected the molar SCFA concentrations, showing that not only the acetate concentration but also its molar proportion decreased with restrictive feeding at both intestinal sites (P Ͻ 0.05). Although RFI rank effects were observed for several single SCFAs, the FL ϫ RFI interactions (P Ͻ 0.05) for concentrations and molar proportions of these SCFAs showed that restrictively fed high-RFI chickens contained more valerate, isovalerate, and caproate in ileal digesta and more isobutyrate and isovalerate (only with respect to the molar proportion) in cecal digesta than the other 3 groups. Concurrently, the cecal digesta of restrictively fed low-RFI chickens was more concentrated with respect to butyrate than was seen with the other groups (P Ͻ 0.05). ANOVA also showed higher (P Ͻ 0.001) molar proportions of valerate in cecal digesta of restrictively fed chickens than in that of ad libitum-fed chickens. Moreover, the FL ϫ RFI interaction (P ϭ 0.020) for propionate indicated that restrictive feeding enhanced its cecal proportion in high-RFI chickens compared to low-RFI chickens, whereas the opposite was found for ad libitum-fed birds.
Restrictive feeding and RFI-associated differences in visceral organ size and intestinal structure and function. Although the body weight (BW) values measured at sampling between 33 and 37 dph were not different among the 4 groups, restrictively fed chickens had heavier (P Ͻ 0.05) crop and duodenum than ad libitum-fed birds and a Data are presented as least-squares means and pooled SEM. n ϭ 7 per FL group, residual-feed-intake (RFI) rank, and sex except for n ϭ 8 high-RFI ad libitum females. RFI was calculated for the experimental period from 9 to 30 days posthatch. FL, feed intake level; SCFA, short-chain fatty acids. b Different letters within a row indicate significant differences (P Յ 0.05).
tended (P Ͻ 0.10) to have heavier ceca (Table S4). Along with the higher weight, the duodenum was 0.9 cm longer in restrictively fed chickens than in ad libitum-fed chickens (P Ͻ 0.05). Low-RFI chickens were further characterized by trends (P Ͻ 0.10) toward lighter liver and pancreas but heavier ceca than were measured for high-RFI chickens. The FL ϫ RFI interaction (P Ͻ 0.01) for the pancreas weight data indicated that the RFI-associated difference existed only in restrictively fed chickens. Most histomorphological parameters in the jejunum and ileum were unaffected by FL and RFI rank (Table S5). The feed restriction tended (P Ͻ 0.10) to decrease the ileal and cecal lymphocyte counts by 17.5% and 21.4%, respectively, compared to ad libitum feeding. In addition, low-RFI chickens had shallower crypts and fewer goblet cells (P Ͻ 0.01) than high-RFI chickens and tended to have thinner longitudinal muscles (P Ͻ 0.10) in the cecum.
In the mid-jejunum, restrictive feeding tended (P ϭ 0.086) to increase the level of expression of jejunal MCT1 by 16.8% (Table S6). The SMCT gene, in turn, was differently expressed in the jejunum of high-RFI compared to low-RFI chickens, depending on the FL, as indicated by the FL ϫ RFI interaction (P ϭ 0.035). With ad libitum feeding, the level of jejunal SMCT expression in low-RFI chickens was approximately one-third that in high-RFI chickens, whereas with restrictive feeding, the jejunal SMCT expression was 41.6% greater in low-RFI chickens than in high-RFI chickens. A similar trend (P Ͻ 0.10) for a FL ϫ RFI interaction with regard to the expression levels of SMCT and SGLT1 was found for the cecal mucosa. Regardless of the FL, in contrast to the jejunal results, the levels of expression of cecal MCT1 were higher (P ϭ 0.018) in both high-RFI chicken groups than in the low-RFI chicken group. With respect to the mucosal cytokine gene expression in the ceca, the expression levels of NFKB tended to be higher with low-RFI chickens than with high-RFI chickens as well as with restrictive feeding compared to ad libitum feeding (P Ͻ 0.10), whereas the cecal expression of TNFA was higher only in restrictively fed high-RFI chickens and not in the other 3 groups (P Ͻ 0.05).
Electrophysiological data for the distal jejunum showed that the restrictive feeding decreased the short-circuit current (I sc ) compared to ad libitum feeding, whereas the tissue resistance was higher in low compared to high-RFI chickens (P Ͻ 0.05) ( Table 3). When glucose was added mucosally, only the jejunal mucosa of ad libitum-fed chickens responded differently, as indicated by the FL ϫ RFI interaction (P Ͻ 0.05), with low-RFI chickens showing a 100% lower change in the tissue resistance than high-RFI chickens. Restrictively fed chickens, however, had similar tissue resistance responses to the added glucose. c Data represent responses to glucose addition to reach a final chamber concentration of 5 mmol/liter. ΔI sc , difference between the maximal I sc value obtained 2 min after glucose addition and the basal value determined 1 min before glucose addition; ΔR T , difference between the basal glucose transport (GT) value determined 1 min before glucose addition and the R T value obtained 2 min after glucose addition.
FL-and RFI-related microbiota effects on the host. Relevance network analysis provided us with a predictive model to identify the most discriminant bacterial taxa (operational taxonomic units [OTUs]) in ileal and cecal digesta on chicken RFI and TFI and on performance. Thresholds were set for each network to present only the strongest pairwise associations as the most influential (Fig. 2). Ileal OTUs were discriminative for TFI and total body weight gain (TBWG), whereas only weak relationships with RFI were found that did not reach the threshold correlation level (|r| ϭ 0.25). Only Enterobacteriaceae OTU151 was positively correlated with both TFI and TBWG. Lactobacillus OTU8, OTU85, and OTU222 were negatively correlated with TFI and TBWG, whereas Ruminococcus OTU143 was negatively associated only with TFI as well as Lactobacillus OTU70 and Lactobacillus OTU129 and Turicibacter OTU138 were negatively related only to TBWG (Fig. 2). In cecal digesta, 8 OTUs were found to be similarly discriminative for TFI and TBWG with the correlation threshold set to |r| ϭ 0.40. These included several Enterobacteriaceae OTUs (OTU2, OTU22, OTU28, OTU151, and OTU230), Clostridiales OTU112, Turicibacter OTU83, and Blautia OTU260. In the ceca, 5 OTUs (3 Anaerotruncus OTUs [OTU20, OTU27, and OTU267], Ruminococcus OTU80, and Clostridiales OTU120) could be distinguished that were influential with respect to the RFI.
Relationships between the bacterial presence and the host response in the ceca were further investigated with multigroup supervised sparse partial least-squaresdiscriminant analysis (sPLS-DA) (13) to identify OTUs, genes, and SCFAs that contributed best to the discrimination of component 1 and component 2 data between FL and RFI groups (Fig. 3). The sPLS-DA identified 14 OTUs within the unclassified Clostridiales family in the cecal digesta as most discriminant for the first component (Fig. 3A), some of which were positively associated with the host mucosal expression of TNFA, TLR2, MCT1, TGFB1, and IL1B and with the cecal isobutyrate concentration. By contrast, the acetate concentration was negatively correlated with 4 Clostridiales OTUs (OTU4, OTU74, OTU92, and OTU137). For component 2, the expression level of NFKB in the cecal mucosa was negatively associated with 3 Anaerotruncus OTUs (OTU20, OTU27, and OTU267) and Clostridiales OTU213 as well as with the concentrations of total SCFA, FIG 2 Determination of discriminant operational taxonomic units (OTUs) in ileal and cecal digesta for total feed intake (TFI), total body weight gain (TBWG), and residual feed intake (RFI) in low-and high-RFI chickens fed either ad libitum or restrictively. Covariations between the relative abundances of bacterial OTUs (relative abundances Ͼ 0.01) and TFI, TBWG, and RFI levels were assessed using sparse-partial-least-squares regression. The network is displayed graphically as nodes (OTUs and performance traits) and edges (biological relationship between nodes), with the edge color intensity indicating the level of the association as follows: red, positive; green, negative. Only the strongest pairwise associations were projected. Relevance networks for (A) relationships between OTUs in ileal digesta, TFI, and TBWG (r Ͼ 0.25) and (B) relationships between OTUs in cecal digesta, RFI, TFI, and TBWG (r Ͼ 0.40) are shown. acetate, isovalerate, and caproate (Fig. 3B). The same Anaerotruncus and Clostridiales OTUs appeared to enhance total SCFA isovalerate and caproate levels in cecal digesta for component 2.

DISCUSSION
In the present study, we used restrictive feeding to impose the same feed intake on all chickens in order to clarify whether the RFI-associated variation in chicken's feed intake is a factor vital to RFI-related differences in intestinal microbial composition, physiology, and function. In doing so, we observed a strong impact of restrictive feeding on the intestinal physiology and bacterial communities, mainly influencing the predominant families, including Turicibacteraceae, Ruminococcaceae, and Enterobacteriaceae in ileal and cecal digesta. Relevance network analysis further supported the idea that, in particular, the abundance of Lactobacillus (negative) in ileal digesta, Turicibacter (positive) in cecal digesta, and Enterobacteriaceae (positive) in both intestinal segments depended on chicken's feed intake. By contrast, only a few RFI-associated taxa were detectable; in particular, Anaerotruncus OTUs were discriminative for high RFI. Overall, more RFI-associated differences were found for the host physiology, i.e., shallower crypts and fewer goblet cells in ceca and a trend toward a lower-weight liver in low-RFI chickens, which may be explained by host-related energy-saving mechanisms associated with low RFI (5). In contrast to our hypothesis, the restrictively fed high-RFI chickens often showed intestinal profiles (e.g., SCFA profile and size of the pancreas) that differed from those shown by the other three chicken groups as indicated by the FL ϫ RFI interactions. These findings suggest the presence of host-related physiological adaptations, presumably in relation to the RFI-associated greater energy and nutrient needs of high-RFI chickens, to compensate for the restricted feeding and hence reduced intestinal nutrient availability and may help to explain the improved RFI values in the restrictively fed high-RFI birds described here.
By restriction of the total amount of feed available, chickens received lower levels of energy sources as well as of macro-and micronutrients than their ad libitum-fed counterparts. Because low-RFI chickens typically eat less feed than high-RFI chickens (12), the feed restriction was more severe for the high-RFI chickens than for the low-RFI chickens. The two characteristic physiological adaptations in the restrictively fed chickens were a bigger crop and a bigger duodenum, possibly to enhance the utilization of the feed, irrespective of the RFI rank. Further in-depth experimentation is required to determine whether the enlargement of the crop was related to increased retention time or to bacterial activity with respect to the feed in this segment. Intriguingly, only the restrictively fed high-RFI chickens had a larger pancreas and not their low RFI counterparts, which may explain the greater nutrient retention in these birds than in the birds in the other 3 groups (4). This finding may imply higher energy and nutrient needs for basal metabolism in high-RFI chickens than in low-RFI chickens (1) which would be met by the increased feed intake in high-RFI chickens under ad libitum feeding conditions (12). We did not perform histomorphological measurements for the duodenum to evaluate changes in the villus surface as we mainly focused on the jejunum and ileum. There, the absorptive and secretory surfaces were similar among the chicken groups.
The lower (negative) jejunal I sc in restrictively fed chickens indicated a greater net absorption of anions (i.e., chloride and bicarbonate) than of cations compared to ad libitum-fed chickens (14). The similar I sc responses seen after stimulation of active Na ϩ -coupled glucose transport (SGLT1) via mucosal glucose addition in the Ussing chambers and similar expression levels of SGLT1 and GLUT2 among chicken groups indicated, however, that neither restrictive feeding nor RFI rank influenced chicken's jejunal capacity to actively absorb glucose. By contrast, high-RFI chickens appeared to take up more glucose in a paracellular manner as indicated by the FL ϫ RFI interaction for the delta transepithelial resistance (R T ). As this effect was not detected with restrictive feeding, it is likely that the RFI-associated differences in the feed intake of the chickens represented the driving force behind this RFI-related physiological adaptation. Concurrently, low-RFI chickens had a higher barrier function as indicated by the higher levels of jejunal tissue resistance than were seen with high-RFI chickens, suggesting RFI-related differences in host metabolism and paracellular nutrient absorption, as the effects were similar for the two feeding-level groups. Restrictive feeding, however, did not impair the mucosal barrier function as previously reported (11), as shown by equal jejunal permeability marker flux rates and R T values and tight-junction protein expression levels.
It is likely that, as a consequence of enhanced nutrient uptake in the upper small intestine, fewer nutrients reached the ileum and ceca of restrictively fed high-RFI chickens, thereby leading to modulation of the bacterial composition and activity in those segments. It can thus be speculated that the decrease in ileal acetate levels and hence in total SCFA concentrations may have been a consequence of the drastic decline in Turicibacteraceae and Enterobacteriaceae levels in restrictively fed chickens. However, since the proportional contributions of the single SCFA hardly changed, the lower substrate availability in restrictively fed birds than in ad libitum-fed birds may have been the main factor responsible for the lower SCFA concentration in ileal digesta. Concurrently, the FL ϫ RFI interactions for SCFA may have supported the lower ileal substrate flow in the restrictively fed high-RFI chickens than in the chickens in the other 3 groups, which apparently promoted the metabolic activity of propionate-, valerate-, caproate-, and isovalerate-producing species. Overall, the ileal isobutyrate levels contributed considerably to the ileal SCFA concentrations across all chicken groups and may have been related to the high protein content of the starter, grower, and finisher diets, as branched-chain fatty acids mainly result from fermentation of branched-chain amino acids (15). Although SCFA levels were not measured in jejunal digesta, the FL ϫ RFI interactions for the jejunal and cecal SMCT expression levels may indicate that changes in SCFA profiles underlie these findings. For instance, the higher proportion of cecal butyrate in both the ad libitum-fed high-RFI and restrictively fed low-RFI chicken groups showed the same pattern as the trend toward SMCT expression levels in these 2 chicken groups compared to the other 2 groups. Since butyrate is one of the primary energy sources for the intestinal mucosa (16), an enhanced rate of mucosal butyrate uptake may be behind the trend toward an enlargement of the ceca among the restrictively fed low-RFI chickens compared to the other 3 groups.
With respect to the bacterial taxonomy, however, FL ϫ RFI interactions were absent, indicating that the intestinal substrate availability may have more extensively affected the metabolic capacities of bacteria than the actual abundances. In support of this idea, both restrictively fed chicken groups had higher species richness and evenness in their ileal digesta than the ad libitum-fed chickens, suggesting a robust effect of restrictive feeding on the taxonomic richness of the bacterial microbiome. By contrast, RFIassociated bacterial differences were more or less absent in the ileum as indicated by diversity data and taxonomic and networking analyses. This may have been related to the fact that, although the ad libitum-fed high-RFI chickens ate more than the low-RFI chickens, the nutrient profiles in digesta of ad libitum-fed low-and high-RFI chickens are similar, as can be assumed from the equal levels of nutrient retention in these birds (4). Following this concept, the restrictively fed low-RFI chickens should have had a bacterial profile similar to that shown by the ad libitum-fed birds, which was not true in the present study. Therefore, aside from the total amount of nutrients available, other physiological adaptations caused by restrictive feeding may explain the present restrictive-feeding-associated taxonomic differences. Possible changes in the small intestinal passage (7), together with the enlarged crop, may have given the microbiota more time to utilize nondigested dietary material. Accordingly, Lactobacillaceae, Ruminococcaceae, and Lachnospiraceae, which have functional abilities to degrade nondigestible carbohydrates (e.g., resistant starch, hemicellulose, and cellulose) (17,18), were enriched with restrictive feeding in the ileum. Lactobacillaceae, Ruminococcaceae, and Lachnospiraceae are able to bind to and degrade host glycoproteins (19)(20)(21). Therefore, a greater mucus secretion to facilitate the digesta flow in restrictively fed chickens might have been another underlying mechanism. However, jejunal and ileal goblet cell numbers as well as the jejunal expression levels of MUC1 and MUC2 did not indicate changes in levels of mucus production between the chicken groups. In line with this, the decrease in the level of mucin-degrading Enterobacteriaceae (22) seen with restrictive feeding also indicates that dietary substrate-related changes are likely the cause for alterations in the ileal community.
The NMDS-based ␤-diversity analysis supported the idea of the importance of the intestinal nutrient flow for the bacterial communities in the ceca. In fact, the cecal microbiota of the ad libitum-fed high-RFI chicken group, which was the chicken group with the highest nutrient flow into the ceca, clustered distinctly from the cecal community of restrictively fed high-RFI chickens, with presumably the lowest cecal flow among the four chicken groups. Moreover, a depletion of the cecal digesta in their preferred substrate may have caused the restrictive-feeding-associated reduced abundances of Ruminococcaceae and Lachnospiraceae, which comprise important acetateand butyrate-producing species (17,23,24), thereby providing a growth advantage for other Clostridiales bacteria. According to the changes in SCFA profiles, the latter may have comprised several propionate producers and protein utilizers. These assumptions are supported by the integration analysis data, showing positive and negative correlations for isobutyrate and acetate levels, respectively, with several unclassified Clostridiales species (e.g., OTU4, OTU74, OTU92, and OTU137) for component 1. Despite the limitations of predictions of metabolic capabilities from unclassified OTUs, circos plots indicated that these OTUs may have modified the cecal mucosal response related to TLR2 and cytokine signaling (e.g., TNFA) as well as SCFA uptake.
Reduced energy and nutrient availability in restrictively fed chickens may have impaired the adaptive mucosal immune defense in the ileum and ceca as indicated by the trend toward lower levels of lymphocytes in the lamina propria in those chickens than in ad libitum-fed chickens. Notably, cecal NFKB expression levels were lower in high-RFI chickens than in low-RFI chickens. According to the circos plot for component 2, this may have been related to the inhibitory effect of cecal SCFA (i.e., acetate, isovalerate, and caproate) on proinflammatory signaling pathways (16), which, in turn, seemed to be related to the increased Anaerotruncus abundance in cecal digesta of high-RFI chickens compared to low-RFI chickens.
In conclusion, the present results provided novel information with respect to the importance of the FL for RFI-associated intestinal bacterial and host physiological variation in chickens. By using quantitative restrictive feeding to impose the same feed intake on all chickens, we were able to show that restrictive feeding was the main driver for the ileal and cecal abundances of the predominant bacterial families Turicibacteraceae, Ruminococcaceae, and Enterobacteriaceae. In contrast, the cecal abundance of Anaerotruncus was mainly associated with high RFI, a finding supported by relevance network analysis. Results also demonstrated host-related physiological adaptations in response to the RFI-associated nutritional needs of high-RFI chickens to compensate for the restricted feeding and reduced intestinal nutrient availability. This may have contributed to the decreased RFI level in restrictively fed high-RFI chickens. In addition, low-RFI chickens developed energy-saving mechanisms (i.e., shallower crypts and fewer goblet cells in ceca and a trend toward a lower-weight liver) and a stronger jejunal barrier function. Data corresponding to dependencies among each bird's cecal microbiota and SCFA levels and expression levels of innate immune and SCFA transporter genes were further supported by multigroup supervised integration analysis. Animals, housing, and experimental design. One-day-old Cobb 500 female (n ϭ 57) and male (n ϭ 55) broiler chicks were used in 2 consecutive replicate batches (56 chickens per batch), with 1 more female and 1 fewer male in batch 1 than in batch 2. Housing and environmental conditions have been previously described (12). All chickens had free access to demineralized water from manual drinkers and were fed the same starter (1 to 8 dph), grower (9 to 20 dph), and finisher (21 to 37 dph) corn-soybean meal-based diets (see Table S7 in the supplemental material). All diets were free of antibiotics and coccidiostatics. From 1 to 8 dph, chicks (n ϭ 5 to 6) of the same sex were housed in groups in stainless steel metabolic cages and had ad libitum access to feed to ensure sufficient feed intake in the first days of life. From 9 dph until the end of the experiment (33 to 37 dph), chickens were individually housed to determine their individual levels of feed intake and were randomly assigned to 2 different treatments. Half of the chickens had ad libitum access to feed (for both replicate batches together, n ϭ 29 females and n ϭ 28 males), whereas the other half were restrictively fed (for both replicate batches together, n ϭ 28 females and n ϭ 27 males). The daily feed allocation of restrictively fed chickens was aimed to correspond to 90% to 95% of the average daily ad libitum feed intake observed in a former chicken trial in which we assessed the FE of the same chicken line (12). Optimally, feeders of the restrictively fed chickens were empty the next morning. A further aim was that all chickens in the restrictive feeding group would eat the same amount of feed. Therefore, the feed amount for the restrictively fed chickens was additionally adjusted daily based on observations of the female and male chickens with the lowest feed intake the day before. For both treatment groups, fresh feed was provided at 9:00 h, and feeders were refilled at 15:00 h.

MATERIALS AND METHODS
Determination of FE. For both treatment groups, the feed intake of each chicken was recorded on 9, 14, 21, 28, and 30 dph. In order to determine the feed intake, feed refusals were collected daily before morning feeding, and feed spills were collected weekly. Chickens were weighed on dph 1, 7, 9, 14, 21, 28, and 30. The RFI used as a metric for FE was determined for the period from 9 to 30 dph. For this, TFI, metabolic mid-test BW (MMW), and TBWG were used to estimate each chicken's RFI as the residuals over the test period (9 and 30 dph) using a nonlinear mixed model (SAS Stat Inc., version 9.4; Cary, NC, USA) (12). In each replicate batch (separately for females and males and balanced for batch), the chickens with the lowest RFI (good FE) and highest RFI (poor FE) values in each FL group were selected to be sampled. This resulted in 14 low-RFI (n ϭ 7 per sex) and 15 high-RFI (n ϭ 8 females; and n ϭ 7 males) ad libitum-fed chickens and 14 low-RFI (n ϭ 7 per sex) and 14 high-RFI (n ϭ 7 per sex) restrictively fed chickens in both batches together. Intestinal parameters were analyzed for only those selected chickens.
Sample collection. Chickens were weighed before being euthanized with an overdose of thiopental (medicamentum pharma GmbH; Allerheiligen im Mürztal, Austria) (50 to 100 mg/kg of body weight) by intravenous (i.v.) injection into the caudal tibial vein between 33 and 37 dph. After the abdominal cavity was opened, the visceral organs were removed and their weight was recorded using a method similar to that described previously by Metzler-Zebeli et al. (5). Visceral organ weight and total and segmental intestinal length were expressed per kilogram of body weight to account for differences in body weight among individual chickens. Intestinal digesta was collected and thoroughly homogenized using a spatula before snap-freezing in liquid nitrogen and long-term storage at Ϫ80°C for microbiota analysis or short-term storage on ice and storage at Ϫ20°C for SCFA analysis. Pieces of the intestinal tube (1-cm diameter) for morphometric measurements were collected from the Meckel's diverticulum, the first centimeter of the proximal ileum, and proximal to the blind end of the ceca. Those pieces were thoroughly washed in phosphate-buffered saline and fixed in neutrally buffered (pH 7.0) formalin (4% [vol/vol]). A 20-cm-long tissue tube piece for the Ussing chamber experiment was collected distal to the Meckel's diverticulum, immediately transferred into ice-cold and pregassed (carbogen gas [95% O 2 -5% CO 2 ]) transport buffer (5), and transported to the laboratory within 10 min of the death of the animal. Thereafter, the remaining intestinal segments were opened at the mesenterium, washed in neutrally buffered saline, and blotted dry with paper tissue. A glass slide was used to scrape the mucosa from the jejunum between the Meckel's diverticulum and 35 cm toward the duodenum as well as from both ceca. Mucosa samples were immediately snap-frozen in liquid nitrogen, and aliquots were stored at Ϫ80°C for RNA isolation.
Histomorphology. Histomorphological measurements were performed as described previously by Metzler-Zebeli et al. (5). After fixation of the intestinal tube pieces, the tube pieces were dehydrated in ethanol, cleared in xylene, and embedded in paraffin. Three discontinuous 3-to-4-m-thick sections per intestinal site were routinely stained with hematoxylin and eosin and examined on a Leica DM2000 light microscope (Leica Microsystems, Wetzlar, Germany) fitted with a digital camera (Leica DFC425C) and using Leica Application Suite V3.7 software. The images were analyzed with ImageJ software (Version 1.47; National Institutes of Health, MD, USA). In total, 15 intact well-oriented, crypt-villus units were selected, with the criteria for villus selection based on the presence of intact lamina propria. Villus height and width were measured at ϫ4 magnification and crypt depth at ϫ10 magnification. The circular and longitudinal muscular layers were measured. Numbers of goblet cells per 250 m of villus or crypt epithelium were determined using 15 replicates per intestinal section at ϫ10 magnification. Intraepithelial lymphocytes were counted per 400 m of villus epithelium using 12 replicates per intestinal section at ϫ20 magnification.
Intestinal electrophysiology. Differences in intestinal electrophysiological parameters and permeability marker flux were evaluated for four chickens per sampling day (5,25). This resulted in six observations per RFI and FL group and sex. The jejunal tube pieces were opened at the mesenterium and rinsed with transport buffer. Clean tissue pieces were stripped of the outer serosal layers (5). Three consecutive pieces were cut from the proximal 10 cm of the jejunal tube, mounted in Ussing chambers (exposed area of 0.91 cm 2 ), and incubated in a total volume of 10 ml of serosal and mucosal buffer solution (pH 7.4, 38°C) (5). Continuous gassing with carbogen using a gas lift was provided on both the mucosal and the serosal sides to ensure oxygenation and circulation of the buffer. In using two pairs of dual-channel current and voltage Ag-AgCl electrodes, which were connected via 3% agar bridges filled with 3 M potassium chloride, the potential difference (mV), I sc (measured in microamperes per square centimeter), and R T (⍀ ϫ square centimeter) were continuously recorded using a microprocessor-based voltage-clamp device and software (version 9.10) (Microclamp; Mussler, Aachen, Germany). The tissue was alternatively pulsed with a positive or negative pulse of 20 A and 100-ms duration. After an equilibration period of 20 min under open-circuit conditions, the tissue was short-circuited by clamping the voltage to zero. After electrophysiological measurements were recorded for 5 min, fluorescein 5(6)-isothiocyanate (FITC) (Sigma-Aldrich, Schnelldorf, Austria) (389.38 g/mol) and horseradish peroxidase (HRP) (Carl Roth GmbH & Co. KG, Karlsruhe, Germany) (44,000 g/mol) were added to reach final concentrations of 0.1 mM and 1.8 M, respectively, to assess the mucosa-to-serosa flux (5). The glucoseabsorptive activity of the tissue samples was assessed by adding glucose to reach a final concentration of 10 mmol/liter to the buffer at the mucosal side at 45 min after short-circuiting the tissue (25). The chemical effect on glucose transporter function was measured by comparing the I sc and R T values for 1 min before glucose was added to the peak current, and the resistance response of the exposed tissue (ΔI sc and ΔR T ) was obtained within 2 min after the addition of glucose.
Mucosal gene expression. Total RNA was isolated from jejunal mucosal scrapings of low-and high-RFI chickens (26) using mechanical homogenization (FastPrep-24 instrument; MP Biomedicals, Santa Ana, CA) and the RNeasy Mini kit (Qiagen, Hilden, Germany). The RNA isolates were treated with DNase I (RNA Clean & Concentrator-5 kit, Zymo Research, Irvine, USA) before transcribing 2 g of total RNA into single stranded cDNA using the High Capacity Reverse Transcription kit (Life Technologies Foster City, USA). The quality (q) of the isolated RNA was verified using the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, USA), showing RNA integrity numbers between 8 and 10.
Primers for target genes and potential housekeeping genes (HKG) used were published previously by our group (5) (Table S8). Amplifications were performed on the Stratagene Mx3000P QPCR System (Agilent Technologies, Santa Clara, CA) using the following conditions: 95°C for 5 min, followed by 95°C for 10 s, 60°C for 30 s, and 72°C for 30 s for 40 cycles, followed by the generation of dissociation curves. Each 20 l reaction consisted of 50 ng cDNA, 10 l Fast Plus Eva Green master mix with low ROX (Biotium, Hayward, CA, USA), 100 nM (each) forward and reverse primers, and DEPC-treated water in a 96-well plate (VWR, Vienna, Austria). All reactions were run in duplicate. Negative controls and reverse transcription controls (R T minus) were included in order to control for residual DNA contamination. Of the 6 tested HKG, ACTB and B2M were the most stably expressed, and their data were analyzed using NormFinder (27) and BestKeeper (28). The geometric mean of the expression levels of ACTB and B2M was used for normalization of target gene expression levels. For this, the mean raw gene expression data (obtained as quantification cycle [Cq] values) from the identified HKG were subtracted from the Cq values determined for the target genes to determine ΔCq values. Gene expression levels were calculated relative to those determined for the chicken with the lowest expression of the corresponding genes using the 2 ϪΔΔCq method. Data representing the amplification efficiencies [E ϭ 10 (Ϫ1/slope)Ϫ1 ] of all primer sets are provided in Table S2 and were prepared by using a 5-fold serial dilution of samples.