Resistance

Initial experiments regarding resistance to disease were undertaken as early as in the 1930's, (Beaumont et al., Reference BEAUMONT, DAMBRINE, CHAUSSE and FLOCK2003a). The goal was to reduce the frequency of the disease and resulting mortality, at a time when many animals died from acute salmonellosis. No experimental inoculation was needed to observe diseased animals and experiments could be achieved directly on the field. The higher resistance of White Leghorns was then described by Robert and Card, quoted by Hutt and Scholes (1941). But De Volt et al. (Reference DE VOLT, QUIGLEY and BYERLY1941) showed that the former were less resistant than selected Rhode Island Red hens; this result and others contributed to the development of selection for a higher resistance.

The first selection experiment for resistance to salmonellosis after deliberate inoculation was undertaken in 1932; and proved to be efficient. Indeed, heritability of reduced mortality was estimated at 0.15 by Beaumont et al. (Reference BEAUMONT, PROTAIS, GUILLOT, COLIN, PROUX, MILLET and PARDON1999) and at 0.12 by Janss and Bolder (Reference JANSS and BOLDER2000), respectively, at day one and 14 of age, respectively. However, the latter authors also noticed that estimated genetic parameters strongly suggested that more resistant animals would survive longer and could thus contribute to a higher risk for consumers if they were still carriers of the bacteria.

Increased food safety can only result from a decreased risk for apparently healthy animals to carry bacteria (i.e. from an increase in genetic resistance to carrier-state). Since heritability estimates showed to be of moderate or even high value (Berthelot et al., Reference BERTHELOT, BEAUMONT, MOMPART, GIRARD-SANTOSUOSSO, PARDON and DUCHET-SUCHAUX1998; Beaumont et al., Reference BEAUMONT, PROTAIS, GUILLOT, COLIN, PROUX, MILLET and PARDON1999), an experiment of divergent selection was initiated by Beaumont et al. (Reference BEAUMONT, CHAPUIS, PROTAIS, SELLIER, MENANTEAU, FRAVALO and VELGE2009). The base population was issued from a layer-type line and two series of divergent lines were selected, for increased or decreased resistance after inoculation at one week of age or at the peak of lay. These will are referred to as Saly+ and Saly- for the lines selected at a young age to increase (Saly+) or decrease (Saly-) the level of Salmonella contamination while Sala+ and Sala- are the names of lines selected at the adult age. Adult hens were orally contaminated and bacterial contamination of caeca, spleen, liver and ovary assessed four weeks later, as described in Protais et al. (Reference PROTAIS, COLIN, BEAUMONT, GUILLOT, LANTIER, PARDON and BENNEJEAN1996). The selection criterion was an all-or-none trait called global contamination and coded ‘1’ if at least one organ was found positive and ‘0’ in the other cases. In chicks, resistance was assessed from the logarithm of the number of cfu. per gram of caeca (i.e. contamination level) five weeks after oral inoculation, as described in Duchet-Suchaux et al. (Reference DUCHET-SUCHAUX, LECHOPIER, MARLY, BERNARDET, DELAUNAY and PARDON P.1995). A total of 3817 animals were thus measured; 1408 at the adult age and 2409 at the younger age (Beaumont et al., Reference BEAUMONT, CHAPUIS, PROTAIS, SELLIER, MENANTEAU, FRAVALO and VELGE2009).

Most probably because the Saly+ and Saly- lines were only separated from the Sala+ and Sala- lines at the third generation, the difference between the Sala+ and Sala- lines remained small, i.e. from 2.19 in the Saly- to 2.56 log(cfu) in the Saly+ line. Larger differences were observed between the Sala- and Sala+ lines. At the 5th generation, average contamination levels were equal to 30 and 44% in the caeca, 5 and 6% in liver and 15 and 30% in spleen respectively, so that the mean percentages of global contamination (adult_g) were equal to 41 and 60% respectively. Selection may, therefore, be efficient in reducing the level of Salmonella carrier-state (Beaumont et al., Reference BEAUMONT, CHAPUIS, PROTAIS, SELLIER, MENANTEAU, FRAVALO and VELGE2009)

This experiment allowed estimation using a large data set and thus with a good precision of parameters of genetic resistance and its links with production traits. Heritability of resistance was estimated at 0.16 in chicks and at 0.18 for global contamination of hens. In adult fowls, the genetic control appeared to be partly dependant on the organ. Heritability estimates varied from 0.14 to 0.23 and genetic correlations between contamination of the different organs ranged from 0.46 to 0.67. This result is in agreement with current biological knowledge on the differences between infection of systemic organs (spleen and liver) and intestinal tracts. Indeed, caecal contamination, which occurs earlier and for longer periods than systemic contamination (Duchet-Suchaux et al., Reference DUCHET-SUCHAUX, LECHOPIER, MARLY, BERNARDET, DELAUNAY and PARDON P.1995), was the most heritable. On the whole, the positive values of estimated correlations were favourable for selection. Increasing resistance to global contamination should also decrease ovarian contamination, which is the major feature for the risk of human transmission but whose occurrence is too low for efficient selection. Global contamination of each animal appears to be the most efficient selection criteria. It allows the assessment of the carrier state more precisely through the results of three tests, thus reducing the risk of false negative results. Moreover, it combines several traits, which are positively correlated between each other. But, these results also underline that assessment of resistance depends on the organ in which Salmonella spp are searched for, which reinforces the importance of a very accurate definition of the resistance trait.

An important feature was the negative (-0.50) correlation between adult global contamination and chick caecal contamination. This negative relationship showed that increasing genetic resistance of hens will reduce resistance in chicks. Partially different genetic controls of adult and chicken resistance have been suspected, but not to such an extent. This result implies that results observed at one age may not be extrapolated to the other without experimental validation.

This study allowed estimating of genetic correlations between resistance and production traits. Most estimated genetic correlations were of moderate values, showing loose biological relations between resistance and production traits, suggesting that selection for an increased resistance should not have much effect on other traits, which is favourable. In particular, no clear putative effect of selection for increased laying rate on resistance could be observed. Negative correlations between egg colour and adult contamination were observed but must be confirmed.

Identification of marker genes for resistance

Although these results may seem promising, selection for an increased resistance would be very difficult to implement in breeder flocks, as they require experimental infections in protected areas, which are very expensive. Nevertheless, it should now be possible to alleviate the need of such measures by using molecular markers related to resistance. Genomic selection based on a large number of anonymous markers should also be very efficient on such traits (Goddard and Hayes, Reference GODDARD and HAYES2007)

Pioneer studies of Bumstead and Barrow (Reference BUMSTEAD and BARROW1988; Reference BUMSTEAD and BARROW1993) examined the impact on inbred lines differing in resistance to mortality after experimental inoculation. Their higher genetic homogeneity increased the statistical power (i.e. the probability of detecting QTL or of evidencing differences of expression) of genomic studies, leading to the identification of QTLs for resistance to disease (Mariani et al., Reference MARIANI, BARROW, CHENG, GROENEN, NEGRINI and BUMSTEAD2001) and to their finer mapping (Fife et al., Reference FIFE, SALMON, HOCKING and KAISER2009) as well as allowing testing candidate genes (Hu et al., Reference HU, BUMSTEAD, BURKE, FA, SKAMENE, GROS and MALO1995), even if other crosses have been used since (Kramer et al., Reference KRAMER, MALEK and LAMONT2003 or Lamont et al., Reference LAMONT, KAISER and LIU2002).

Initial investigations have been made using candidate genes, i.e. genes chosen according to an a priori knowledge of their effect in Salmonella resistance. In particular, many studies focused on SLC11A1 and TLR4, two genes known to be involved in resistance to Salmonella in mice. SLC11A1 was first called Nramp1 (natural resistance-associated macrophage protein) and shown to be responsible for the difference between mice strains in resistance to inoculation with Salmonella typhimurium, (Roy and Malo, Reference ROY and MALO2002), Mycobacterium bovis and Leishmania donovani, because of its role in the control of the intracellular replication of parasites in phagosomes. After its identification, it was recognised as a member of the solute carrier family and renamed Slc11a1 (Vidal et al., Reference VIDAL, TREMBLAY, GOVONI, GAUTHIER, SEBASTIANI, MALO, SKAMENE, OLIVIER, JOTHY and GROS P.1993) The chicken homologue of Nramp1 was mapped on the chicken chromosome 7 (Hu et al., Reference HU, BUMSTEAD, BURKE, FA, SKAMENE, GROS and MALO1995) and subsequently cloned (Hu et al., Reference HU, BUMSTEAD, SKAMENE, GROS and MALO1996). The second candidate gene, TLR4 (Toll-like receptor 4) was previously named Lps, because its mutation results in a lack of response to LPS (lipo-polysaccharides specific of the Gram negative bacteria, including Salmonella) and a higher susceptibility to those bacteria. It was shown to belong to a family of innate immune system receptors (Toll-like receptors) and was mapped to the chicken micro chromosome 17 and cloned (Leveque et al., Reference LEVEQUE, FORGETTA, MORROLL, SMITH, BUMSTEAD, BARROW, LOREDO-OSTI, MORGAN and MALO2003).

Many studies have dealt with these two genes. Both are involved in the early response of chicks to an acute infection with S. typhimurium (Hu et al., Reference HU, BUMSTEAD, BARROW, SEBASTIANI, OLIEN, MORGAN and MALO1997). NRAMP1 was also shown to be involved in early stages of systemic Salmonella infection in meat-type chicks (Kramer et al., Reference KRAMER, MALEK and LAMONT2003), in a cross between layer-type strains or breeds (Lamont et al., Reference LAMONT, KAISER and LIU2002; Liu et al., Reference LIU, KAISER and LAMONT2003) or in pullets (Girard-Santosuosso et al., Reference GIRARD-SANTOSUOSSO, LANTIER, LANTIER, BUMSTEAD, ELSEN and BEAUMONT2002). While a potential role of Nramp1 in later stages of infection was also shown in mice inoculated with S. enteritidis (Caron et al., Reference CARON, LOREDO-OSTI, LAROCHE, SKAMENE, MORGAN and MALO2002), it is of note that the NRAMP1 allele coding for a better resistance to an early and acute infection was responsible for a higher excretion rate in later stages. While not yet investigated in fowls, this result implies, at least in mice, that animals selected for the NRAMP1 allele coding for resistance to disease could be more susceptible to carrier-state, showing the importance of the choice of the selection criteria and the use of a relevant protocol of experimental inoculation. In hens, NRAMP1 was shown to be associated with later spleen contamination of hens (Beaumont et al., Reference BEAUMONT, PROTAIS, PITEL, LEVEQUE, MALO, LANTIER, PLISSON-PETIT, COLIN, PROTAIS, LE ROY, ELSEN, MILAN, LANTIER, NEAU, SALVAT and VIGNAL2003b), while the role of TLR4 could not be confirmed in the same study.

In addition to these two genes, several QTL detection programs were carried out in crosses between inbred lines differing for resistance to disease for a large range of Salmonella serotypes (Bumstead et al., 1988; 1993) or to Salmonella carrier-state (Tilquin et al., Reference TILQUIN, BARROW, MARLY, PITEL, PLISSON-PETIT, VELGE, VIGNAL, BARET P.V., BUMSTEAD and BEAUMONT C.2005). Several QTLs were identified (Mariani et al., Reference MARIANI, BARROW, CHENG, GROENEN, NEGRINI and BUMSTEAD2001, Tilquin et al., Reference TILQUIN, BARROW, MARLY, PITEL, PLISSON-PETIT, VELGE, VIGNAL, BARET P.V., BUMSTEAD and BEAUMONT C.2005) in these genotypes. However, the latter study used selective genotypings (i.e. only those of the most extreme animals) which could have resulted in biases. Calenge et al. (Reference CALENGE, LECERF, DEMARS, FEVE, VIGNOLES, PITEL, VIGNAL, VELGE, SELLIER and BEAUMONT2009) genotyped all animals in the formerly identified QTL regions, confirming two out of the four QTLs. The same authors tested the effect of the genome regions identified by Tilquin et al. (Reference TILQUIN, BARROW, MARLY, PITEL, PLISSON-PETIT, VELGE, VIGNAL, BARET P.V., BUMSTEAD and BEAUMONT C.2005) in animals close to commercial ones. They used data and samples issued from the selection experiment and could thus confirm the role of the QTL on chromosome 1 and of SLC11A1 in both chicken and laying hens. Development of a large scale arrays thanks to the identification of a large number of SNPs should soon result in the identification of new genome regions and in a shorter term, SNP-assisted or even genomic selection.

In parallel, differences in gene expression after a S. enteritidis infection were identified between poultry lines, either for candidate immune genes (Sadeyen et al., Reference SADEYEN, TROTEREAU, VELGE, MARLY, BEAUMONT, BARROW, BUMSTEAD and LALMANACH2004; Reference SADEYEN, TROTEREAU, PROTAIS, BEAUMONT, SELLIER, SALVAT, VELGE and LALMANACH2006; Kaiser et al., Reference KAISER, CHEESEMAN, KAISER and LAMONT2006), or using large scale arrays as in Van Hemert et al. (Reference VAN HEMERT, HOEKMAN, SMITS and REBEL2006; Reference VAN HEMERT, HOEKMAN, SMITS and REBEL2007). These results contributed to the identification of the genes involved in resistance, provided that the mutations responsible for resistance result in differences in their expression levels.

In conclusion, many genome regions have already been identified regarding their effect on resistance to Salmonella, but most of them were only studied using one protocol of experimental inoculation (i.e. one serovar, one route of inoculation, one single animal age and genotype). These results cannot thus be extrapolated to other conditions without confirmation. Particular care must be taken regarding the negative genetic correlation between resistance of chicks and laying hens, which might lead to different strategies for layer- and meat-type lines.

Modelling

All the former results show the complexity of genetic control for resistance to disease carrier states. This first depends on the animal's age but also on the target organ, so that several selection criteria should be used to increase bacterial clearance. This choice is an important issue. The same holds true for the global strategy of genetic improvement. For instance, should the resistance of all animals be improved or would it be more efficient to increase it in smaller proportion of animals but to a larger extent? Answering such a question will clearly benefit from the comparison of the impact of these different options on the rate and levels of contamination of laying hens flocks. Modelling Salmonella propagation within a flock allows such comparisons. Prevost et al. (2006; Reference PREVOST, MAGAL, PROTAIS and BEAUMONT2008) derived mathematical models and, in addition to naïve (also called susceptible) birds, they distinguished three steps in the contamination:

• • digestive contamination when only the digestive tract harbours Salmonella spp.

• • systemic infection (when systemic organs such as liver or spleen are contaminated after translocation of bacteria through the digestive barrier, which may also result in egg contamination) and

• • bacterial clearance leading to recovery.

The development of the models and, in particular, the fitting of key parameters used in the results from the selection experiment, allowed identification of the most important factors for the risk of egg contamination and thus to human contamination risks. These included:

• • the recovery rate is dependent on the ability of hens to eliminate Salmonella. This influences both the maximal prevalence and the duration of the epizooty.

• • the rate of return to the susceptible state depends on the loss of protective immunity.

This modelling showed that the effect of selection of both parameters were not the same when selecting for a higher level of bacterial load than for a lower one: The recovery rate was altered when selecting for reduced carrier-state while the rate of return to the susceptible line was mainly responsible for the higher susceptibility of Sala+ lines. This result is in agreement with results on immune response of lines differing (among other traits) in resistance to carrier states, whether inbred (Sadeyen et al., Reference SADEYEN, TROTEREAU, PROTAIS, BEAUMONT, SELLIER, SALVAT, VELGE and LALMANACH2006), or outbred (Proux et al., Reference PROUX, JOUY, HOUDAYER, PROTAIS, DIBB-FULLER, BOSCHER, GILLARD, GRACIEUX, GILBERT, BEAUMONT and DUCHET-SUCHAUX2002; Protais et al., Reference PROTAIS, NAGARD, BOSCHER, QUEGUINER, BEAUMONT and SALVAT2003). Moreover, in the latter study, vaccine efficiency was found to be better in the more resistant lines. This difference in immune response also suggests the vaccines will be less efficient in lines more susceptible to carrier-state. This difference was found in simulations when comparing responses to selection in populations mimicking selected lines (Prevost et al., Reference PREVOST, MAGAL, PROTAIS and BEAUMONT2008). It is important to notice that the combination of vaccination and genetic selection results in a very low percentage of contamination, very similar to the level the European community is requesting.

This study also compared the kinetics of infection between a homogenous population corresponding to the base population from which selection was undertaken, or a mixture of the Sala+ and Sala- lines with the same geometrical mean, which is the same duration of the different steps of infection. Introducing Sala+ animals reduces the peak of infection (i.e. the maximal percentage of infected animals), since accelerated bacterial clearance of Sala- animals reduces the risk of cross contamination. But it also delays the extinction of the epizooty, because Sala- animals require more time to recover. Inversely, results of genotype comparisons will not be the same when animals with different degree of bacterial clearance are reared together or not. This point should be further studied in practice.

These models may be further improved toward a finer modelling of the animal innate immune response which may allow the animal to get rid of any digestive contamination. Individual variations need to be more thoroughly investigated through individual-based models. This will allow consideration at the individual level and not, as previously, for the whole population, summarising it through the proportions of the different categories of animals (naïve, infective at the digestive level, systematically infected or recovered). Such investigations are currently underway.

Acknowledgements

Grant funding from Aliment Demain SF20 from French Ministries of Research and of Agriculture, RESISAL from French National Agency for Research, as well the help of the Region Centre and of the EADGENE Network of Excellence are greatly acknowledged.