Mean values and results of the analyses of variance are listed in Table 1. As expected from the past selection, lines had different (P ≤ 0.001) mean values for the selected traits, but there were also significant line effects (P ≤ 0.05 to P ≤ 0.001) on four of the other measures of immunity.

Mean values and results of the analyses of variance are listed in Table 4. As in Group A, lines had different (P ≤ 0.01 or P ≤ 0.001) mean values for the traits which were under artificial selection, but a line effect (P ≤ 0.05 to P ≤ 0.001) was also present for all other traits except LPS before immunization.

Results obtained by comparing the three selected lines and the control line, showed that after 12 generations of selection, no difference was present between lines for specific antibody titers to the new T-cell dependent antigen tested (KLH) as shown from the principal component analysis, where individuals were not forming well defined clusters in relationship to their line of origin (Figure 1). These findings are in contradiction with previous studies, where higher titers were observed in the line selected for high response to SRBC 6 compared to the low responder line. An explanation for the difference could be that the selection criterion for the ND3 response (Th-1 like induced antibody response to an attenuated vaccine) has no effect on antibody production to a strong Th-2 like inducer antigen as KLH. This is in line with the results obtained in SRBC selected lines, because the experiment was done with KLH and SRBC which both induce the Th-2 like pathway. As a consequence, it might be interesting to add the KLH response in a fourth and separated selected line in the present selection experiment to add an antigen that induces a Th-2 like response in the panel of traits.

The present results, however, are contrary to the observation made on the second generation crosses of the same lines 16 that the selected lines differed significantly from the control line for specific antibodies to KLH. Yet, KLH is a very strong Th-2 like response inducer, and conditions of immunization might have been different in the two studies. Also, favourable recombinations might have occurred in the F2 generation between genes from different selected lines. In the present study, level of Nabs binding two T-cell independent antigens (LPS and LTA) have been tested, before and after immunization with KLH. While no differences were present between lines for KLH and LPS Nabs, lines ND3-L and PHA-L had higher titers to LTA and anti-Gal titers measured before the immunization protocol. Positive and very significative correlations were found between LPS titers measured post KLH immunization and LPS responsiveness with ND3 response. In addition, we found a positive phenotypic correlation between the level of specific antibodies against KLH and NAbs to LPS in the second generation crosses of the pure lines under study 16. Earlier, levels of antibodies to LPS were shown to be positively influenced by KLH pre-treatment 6. These findings suggest possible co-selection of innate immunity in the line selected for specific immune response to Newcastle Disease Virus vaccine.

Results obtained from principal components analysis summarise the differences observed between the selected lines for the responses to the new antigen (M. butyricum), the humoral innate components, and confirm the underlying correlations (Figure 2). The lines ND3-L and CC-L are located on the positive and the negative ends of the PC1 axis. The location of line ND3-L is confirmed by the ANOVA results (Table 4) since this line has higher titers for almost all immune measures which contribute most to the first axis of the PCA, except for anti-Gal Nabs and responsiveness to LTA that are both negatively correlated to this axis. The CC-L line has higher titers compared to the other lines for anti-Gal NAbs and responsiveness to LTA, and is therefore located on the left side of the figure, since the two traits are negatively correlated to the first principal component (PC1). The PHA-L line and the Control line show wider distribution of the individuals due to their intermediate values.

The line selected for ND3 showed significantly higher specific antibody response to M. butyricum, and this agrees well with the Th-1 pathway expected to be selected for in line ND3-L and the Th-1 like response induced by M. butyricum 4. Previous studies on commercial chickens differing in residual feed intake (RFI) demonstrated no difference between the efficient and inefficient animals in specific antibody production to M. Butyricum, suggesting that this response does not influence feed efficiency 17 and confirming that antibody production is a not an energy demanding process 1819. This experiment was focused on layers, however, and it would be interesting to investigate the effect of selection for immune traits on other production traits and for other types of poultry. The positive correlations between ND3 and natural antibodies to LPS and LTA in the M. butyricum treatment group suggest a possible co-selection of the innate immune responses. They confirm also the results obtained in the second generation crosses of the same pure lines 16. In lines divergently selected for SRBC, correlations between primary antibody response to SRBC and natural antibodies to LPS were low but significant 6. Recently, levels of natural antibodies and specific antibodies to Newcastle disease virus (NDV) have been measured in 12 purebred lines of laying hens in relation to their survival rate. Results showed no correlation between innate and acquired immune response although the lines that had a higher response of antibodies to NDV showed also higher levels of natural antibodies 20.

Comparing responses in KLH- versus M. butyricum challenged birds, it is noteworthy that line differences in the KLH treated birds were found in the day 0 samples. This suggests that the selection for the three immune traits changed the immune status in the 'non-challenged' birds. All lines responded equally to KLH on the level of Nabs and specific antibodies, which might be due to the strong Th2 nature of the antigen (KLH) that overrules the selection criteria.

Otherwise, the current data suggest that, regardless of selection, the lines can mount compensatory responses as illustrated by the anti-Gal responses. It is clear also that the M. butyricum challenge revealed more line differences than KLH for the three measures of immune response selected in the pure lines and for correlated traits, possibly because of the nature of M. butyricum itself and of the strong Th-1 like response it induces.

The selected lines were developed by 12 generations of selection from a cross between an experimental White Leghorn line segregating for the sex-linked dwarf (dw) gene and a commercial Babcock^® White Leghorn line. Three immunity-related traits were measured on all birds, at each generation, but each line was under selection for one of them only: high antibody response to Newcastle disease virus (HB1 vaccine) 3 weeks after vaccination (ND3) in L1 (ND3-L), high cell-mediated immune response, using the wing web response to PHA at 9 weeks of age in L2 (PHA-L), high phagocytic activity, measured as clearance of carbon at 12 weeks of age in L3 (CC-L), and random selection in L4 (Control). Every year, all birds were hatched in a single batch, and 15 males and 30 females were chosen as breeders out of 100 candidates per sex and line, by within-family mass selection based on individual phenotypic performance. Mating was at random, but full and half sib matings were prohibited.

After hatching, chicks were housed and lines intermingled, in a three-tier battery of group cages, in which they remained until the end of the experiment. Each cage housed about 13 animals. Both sexes were kept in separate cages but housed in the same room. Artificial light was 16 h/d. The birds were fed a layer diet (2,685 kcal ME/kg and 175 g/kg crude protein) ad libitum, with free access to water. Birds were vaccinated against Marek's disease at hatch by intramuscular injection; spray vaccinated for infectious bronchitis at d 1 and 63; against Gumboro disease via drinking water at d 19, 37 and 84; against Newcastle disease by intraocular way (eye drop) at d 26 and in drinking water at d 62; against Avian encephalomyelitis via drinking water at d 98, and spray-vaccinated foravian infectious Rhinotracheitis or Swollen Head Syndrome at d 77. All birds were weighed at 8 wk of age.

The treatment consisted in an immunization protocol with KLH Keyhole Limpet Hemocyanin (group A) or with M. butyricum (group B) which started when the animals were 14 weeks of age. Each treatment group was composed of 200 birds, 50 per line (3 selected lines and 1 control line) belonging to 12^th generation of selection. The two groups A and B were designed to be as similar genetically as possible, by choosing for each line and in each full sib family two animals of the same sex to be assigned to each group. Blood samples were taken on day 0 (at 14 wk of age, before primary immunization) and 7 days post immunization with KLH (group A) and 11 days post immunization with M. butyricum (group B). Immunization doses and days of sampling were based on previous studies on selected chicken lines 7. Responsiveness was measured for the four non-selected immune traits as the difference between antibody titer at day 7 (after immunization with KLH) or day 11 (after immunization with M. butyricum) and antibody titer at day 0.

Total immunoglobulin titers to KLH (Sigma-Aldrich GmbH, Schnelldorf, Germany), and levels of NAB binding to Salmonella Enteriditis Lipopolysaccharide LPS(Sigma Chemical Co., St Louis, MO) and to Lipoteichoic acid LTA and KLH in plasma of 200 birds previously immunized at 14 wk of age with 1 mg/ml of KLH were measured by an indirect two step ELISA procedure at day 0 and d 7 post immunization. Total antibody titers binding to M. butyricum and levels of NAB binding Salmonella Enteriditis Lipopolysaccharide LPS(Sigma Chemical Co., St Louis, MO) and Lipoteichoic acid (LTA) were measured in plasma samples of 200 birds previously immunized with 1 mg/ml of M. butyricum protein (Difco, Laboratories, Detroit, USA) at day 0 and day 11 days post immunization. Briefly, 96 well plates were coated either with 1 μg/mL of KLH, 4 μg/mL of M. butyricum, 10 μg/mL of LTA or 4 μg/mL of LPS.

After washing with tap water and 0.05% Tween, plates were incubated with serial dilutions of plasma. Binding of total antibodies to KLH, M. butyricum and of natural antibodies to LPS, LTA, KLH and M. butyricum was detected using 1:20,000 diluted rabbit anti chicken (IgG_H+L)(Nordic, Tilburg, The Netherlands) coupled to peroxidase (PO). Tetramethylbenzidine and 0.05% of H₂0₂ were added and incubated for 10 min at room temperature. The reaction was stopped with 1.25 M of H₂SO₄ and absorbencies were measured with a Multiscan (Labsystems, Helsinki, Finland) at a wavelength of 450 nm. Titers were expressed as the log₂ values of the highest dilution giving a positive reaction. Titers were derived from the absorbance values of a positive control serum present on every microtiter plate.

Anti-Gal antibody titer was determined by an agglutination test to rabbit red blood cells (RRBC) because these cells express high levels of the α-Gal epitope 12. The test was performed in 96-well U-bottom plates. Plasma samples were added to the first and second wells (25 μl), and 25 μl of PBS were then added to columns 2 to 11. Next, serial dilutions (1:2) were made from column 2 to 11. Column 12 contained only 25 μl of PBS, as a negative control. Then, 25 μl of a 1% solution of rabbit red blood cells were added to each well. Plates were vortexed gently for a few seconds and incubated overnight at room temperature. Titers were recorded as the column number of the last plasma dilution showing clear evidence of agglutination.

Plasma Ab titers to KLH, ND3, M. butyricum, skin swelling to PHA, and levels of NAB binding LPS, LTA and Anti-Gal were analyzed by a mixed 3-way ANOVA with the sex and the line as fixed effects, the sire as a random effect nested within the line, and with all possible interactions. All analyses were done with the GLM procedure of SAS 21. The sire mean square error was used to test the overall significance of the effect of the line. The mean square error of the interaction between sex and sire within line was used to test the significance of the effect of sex and of the interaction between the sex and the line. When a main effect was found to be significant, the corresponding line means were compared by using Duncan's multiple range test. Phenotypic correlations were computed within line, and linear regression coefficients have been estimated across all lines. In addition, residuals obtained from adjusting a linear model including the effects of line, sex and interaction were obtained for each measure of immunity, and their correlations (residual correlations) were estimated.

Principal component analysis (PCA) was applied to the entire dataset and was performed with the SPAD 6.0 software 22. The normality of each measure was checked using UNIVARIATE procedure of SAS 21 because PCA is only used for measures normally distributed, and for individuals measured for several traits. The objective of this analysis is to find a small number of factors (the "principal components") which are linear combinations of the original variables, and which best explain the total variation between animals. Each variable is associated to a principal component by an eigenvector and an eigenvalue, which indicate respectively the contribution and the correlation of the trait to the specific principal component axis.