Heritability estimates of excretion traits
Genetic parameters of digestibility, feed efficiency and anatomy of the digestive tract have been discussed previously by de Verdal et al.  on the same data set and are not detailed further here. However, it should be noted that the D+ birds had 33.5% higher AMEn, 14.5% higher BW23 and 36.8% lower FCR than D- birds. Furthermore, AMEn and FCR heritabilities were estimated at 0.30 and 0.21, respectively .
While chicken manure can be used as fertiliser, at high levels it is considered a pollutant, increasing water eutrophication, excessive algae development and ammonia volatilisation in the air. Thus, in view of the problems related to the management and the environmental impact of chicken manure, the selection of birds producing reduced quantities of excreta is important.
To our knowledge, the present study is the first to present estimated genetic parameters of broiler excretion traits and their correlations with performance characteristics and GIT morphology. However, probably due to the low number of birds used, the standard errors were sometimes relatively high, and consequently, some results should be taken with caution. In some other cases, several parameters in the same analysis were close to the limit of parameter space (genetic correlation close to unity and h² of some traits close to 0), which makes convergence more difficult. It was for example the case when FEW/BW with low h² was included simultaneously with DEW/BW, with which it was very highly correlated. Similarly, even if FCR and AMEn had already been shown to be strongly correlated (-0.70 ), the genetic correlation of 0.98 between CDUDM and FCR was probably overestimated due to the presence of 3 highly correlated traits in the analysis.
Excretion traits were moderately heritable, showing that it should be possible to include such traits in poultry selection. The estimated heritability of PE was much higher than that reported by Zhang et al.  and Ankra-Badu et al.  who reported a value of 0.09 for phytic phosphorus bioavailability (PBA). However, even with this rather low heritability, Zhang et al.  obtained a divergence of 9.7% on P bioavailability after 3 generations. We can hypothesize that this wide difference between different studies is related to the diet used, as Zhang et al.  and Ankra-Badu et al.  used a corn-based diet that is easy to digest, while the wheat diet used in the present study made it easier to distinguish between animals with poor or high capacity of retention. Furthermore, these experiments differed from ours by genetic lines used, which showed a much slower growth than ours. Mignon-Grasteau et al.  showed that heritability estimates of metabolisable energy and coefficients of digestive use of proteins and lipids were much higher when animals were fed with poor wheat than with corn.
Phenotype differences between D+ and D- lines and genetic correlations for excretion traits
In the present study, we found that D+ birds had a 28.2% greater CDUDM than D- birds, showing that digestive utilization was improved in D+ compared to D- birds. This could be explained by the genetic correlations between CDUDM and GIT morphology. Indeed, it seems that selection on high CDUDM would increase the relative weight of the upper part of the GIT (proventriculus and gizzard) and conversely decrease the relative weight and the density of the small intestine, consistent with previous results . A larger gizzard and proventriculus would lead to greater nutrient accessibility in the small intestine and thus to better digestive efficiency. At the intestinal level, the genetic correlations were higher between CDUDM and the relative weights of the jejunum and the ileum than between the CDUDM and the relative weight of the duodenum, which could potentially be explained by the fact that absorption processes mainly take place in the jejunum and ileum .
This higher digestive utilisation in D+ birds leads to a 41.3 and 54.1% reduction in FEW and DEW, respectively, compared to D-. These differences are also present at later ages. Furthermore, the commercial line used at the beginning of the selection experiment excreted 31.5% more DEW/FI between 21 and 53 d of age than D+ birds . Selection for a better AMEn thus led to a reduced environmental impact of chicken production. Although WE was greater in D+ birds, the total quantity of water excreted and the FEW/FI ratio were 51.2% and 26.7% lower than in D- birds. As Williams et al.  explained that water consumption closely follows food consumption, and since FI was 27.4% higher in D- than in D+ birds, it could be hypothesized that D- birds consume almost 30% more water than D+ birds. This is important since this can have consequences in terms of health and welfare. An increase in the quantity of water excreted would lead to a more humid litter and consequently to an increase in the incidence of associated poultry diseases, such as breast blisters, skin burns, scabby areas, bruising, rejection or downgrades . Moreover, the litter moisture content is known to have a high impact on the ammonia losses by volatilization, which may cause respiratory disorders in birds and farmers and increase the imbalance between N and P in manure .
However, most of the studies related to environmental problems due to the spreading of manure focus on N and P content  and their deleterious environmental impact. The capacity of D+ birds to retain N and P was 34.9 and 19.0% higher, respectively, compared to the D-, as shown by NE/NI and PE/PI ratios. Thus, for each 100 g of BW, D+ and D- excreted 0.73 and 1.65 g of N and 0.17 and 0.32 g of P, respectively. It has already been shown that the lower NE/NI ratio in D+ can be linked to the 8.7 to 13.1% better ability of these birds to utilise proteins [10, 13, 25]. Moreover, the NE/NI ratio was more highly correlated genetically with the lower rather than the upper part of the GIT. This suggests a major contribution of the lower part of the intestine compared to the upper part in N utilization. Péron et al.  showed that the pancreas was heavier in D- than in D+ birds, and found negative phenotype correlations between pancreas weight in relation to BW and AMEn and lipid, protein and starch digestibility. These authors explained that the enlargement of the pancreas could be an adaptation to decreased digestion in D- birds.
Furthermore, NE/NI and PE/PI ratios were more genetically correlated with jejunum and ileum relative weights and densities than with those of the duodenum. This illustrates the major contribution of the lower part of the intestine in N and P absorption [32, 33]. Nevertheless, in contrast to the NE/NI ratio, the PE/PI ratio was positively genetically correlated with PRW, indicating a major contribution of this segment to P availability. These results are probably related to the morphological and functional differences in the upper GIT characterizing both lines [12, 34]. Indeed, the greater development of the upper part of the GIT in D+ birds may underlie an increase in the synthesis of hydrochloric acid. Moreover, the mean retention time in the upper part of the GIT is greater in D+ than in D- birds . All of these phenomena could lead to a lower pH of digesta that promote solubility of mineral phosphates  in D+ birds and the capacity of residual endogenous phytase of the feed . They can also favour the hydrolysis of phytic P by endogenous bacteria .
These high levels of differences in N and P excretion between D+ and D- birds could explain why the ratio of the NE to PE was 25.4% higher in D- than in D+ birds. French and European regulations limit the amounts of N and phosphates (P2O5) that can be spread on fields to 170 kg. ha-1 and 100 kg.ha-1, respectively, the ideal ratio of N to P2O5 on spread manure should thus be 1.7 . Considering the litter and the water part of the manure being spread, the ratio of N to P2O5 that would be found in the manure would be 1.95 and 2.33 for the D+ and the D- birds, respectively. However, since 50% of the N excreted by chickens is lost between excretion and spreading , these ratios would become 0.976 for the D+ and 1.166 for the D- birds for manure ready to be spread on fields, implying that the manure of both lines is too rich in P2O5 compared to N. This suggests first that N losses should be limited to increase the N/P2O5 ratio in manure and secondly that this limitation should take into account the genotype of birds. Indeed, N losses in manure should be limited to 15% in D+ birds and 37% in D-birds, whereas the usual value is closer to 50%.
A second way to improve manure quality would be to combine genetic and nutritional approaches, i.e. by reducing the P rate in the diet and adjusting the phytase quantity added to the diet to each genotype in order to reduce P excretion .
Direct selection on excretion traits vs indirect selection on efficiency
It is often assumed that excretion can be reduced by selection on feed efficiency. By providing a full set of genetic parameters of excretion traits and efficiency, our study allows comparison of the expected responses to direct selection on excretion traits and digestibility and to indirect selection on feed efficiency. Using equations  and , for most excretion traits (CDUDM, WE, NE/NI, PE/PI and NE/PE) the expected responses to selection on FCR would be reduced by 12.3 to 50.5% as compared to selection on AMEn or CDUDM, and selection on RFI would lead to expected responses reduced by 19.9 to 85.7% compared to AMEn or CDUDM selection. While the indirect expected response was higher for selection on AMEn or CDUDM than on FCR for DEW/BW and FEW/BW, a selection on RFI would be 16.4 to 20.1% more efficient than on AMEn or CDUDM. At the opposite, for the FEW/FI ratio, the expected responses were similar for selection on AMEn, CDUDM or FCR, but selection on RFI would be 26.4 to 28.9% less efficient.
It therefore appears that, in order to reduce environmental pollution, selecting chickens on AMEn or CDUDM would be more effective than selection on feed efficiency, all the more true that actual methods (as NIRS) allow measuring these traits at a very moderate cost.
Besides AMEn or CDUDM selection, direct selection on excretion traits could be considered. Using equation , it appears that for FEW/FI, DEW/FI, NE/PE and NE/NI ratios, indirect selection on AMEn or CDUDM would be more effective than direct selection, with improvements ranging from 1 to 78%. For the other excretion traits, indirect selection could be almost as effective as direct selection. Indeed, the responses of the FEW/FI ratio to indirect selection on AMEn or CDUDM were 88.2% of those of direct selection. Similarly, using AMEn or CDUDM as indirect selection criterion of WE would also be very effective (ranging from 62 to 69% of the direct response). Moreover, the PE/PI ratio would be considerably modified by selection on AMEn or CDUDM, with indirect responses ranging between 73 and 82% of the direct response. Consequently, introducing AMEn or CDUDM in selection schemes could be a good way to reduce excretion and hence the environmental impact of chicken production. Finally, if evolution of genetic values in D+ and D- are symmetric, it is not the case for phenotypic values, which is commonly observed in divergent selection experiments. To draw a definitive conclusion on practical interest of such a selection, it would be necessary to compare to a control line (CL) such as the line used at the beginning of selection experiment. First elements brought by such a comparison indicated that DEW/FI was 31.5% lower in D+ than in CL birds , between 21 and 53 d (age at which birds reached commercial market weight).