Cell-mediated immunity and multi-locus heterozygosity in bluethroat nestlings
Frode Fossøy, Department of Biology, Norwegian University of Science and Technology, Høgskoleringen 5, NO-7431 Trondheim, Norway.
Tel.: +4773596284; fax: +4773596100; e-mail: firstname.lastname@example.org
Recent evidence suggests that marker-based heterozygosity-fitness correlations may be driven by only one or a few markers, indicating local heterozygosity effects caused by linkage disequilibrium with functional genes. In this study, we investigated the relationship between microsatellite heterozygosity and a measure of cell-mediated immunity (phytohaemagglutinin; PHA) in bluethroat (Luscinia s. svecica) nestlings using a full-sibling design. We found significant positive associations between PHA response and two different indices of microsatellite heterozygosity, i.e. multi-locus heterozygosity and mean d2. However, model comparisons disclosed that both associations were more likely caused by local effects rather than general effects and that the two local effects appeared to be realized through two different genetic mechanisms. Our results indicate that both the random assortment of parental chromosomes during meiosis as well as inbreeding can drive heterozygosity-fitness correlations.
Individual heterozygosity may be an important determinant of fitness, either because genetically depauperate individuals are more vulnerable to the negative effects of deleterious recessive alleles or because heterozygous individuals are more fit than homozygous ones because of genetic overdominance (e.g. Charlesworth & Charlesworth, 1987, 1999). There is currently extensive empirical evidence for a relationship between individual heterozygosity and fitness in both plants and animals (e.g. Coltman et al., 1998; Coulson et al., 1998; Amos et al., 2001; Höglund et al., 2002; Charlesworth, 2003; Coltman & Slate, 2003; Marshall et al., 2003; Hansson et al., 2004; Fossøy et al., 2008), although the genetic mechanisms behind such patterns remains obscure (David, 1998; Tsitrone et al., 2001; Hansson & Westerberg, 2002; Keller & Waller, 2002; Balloux et al., 2004; Slate et al., 2004). Commonly, multiple microsatellite markers are used to estimate heterozygosity with the assumption that these neutral markers reflect the genome-wide level of inbreeding, i.e. a general heterozygosity effect (David, 1998; Hansson & Westerberg, 2002, 2008). However, correlations between such marker-based heterozygosity and fitness traits may sometimes be driven by a single marker only (Bierne et al., 1998; Hansson et al., 2001, 2004; Coltman & Slate, 2003; Tiira et al., 2006). Presumably, linkage disequilibrium (LD) (the nonrandom association of alleles at different loci in gametes) causes a genetic correlation between the microsatellite locus and a functional locus important for fitness, i.e. a local heterozygosity effect (David, 1998; Hansson & Westerberg, 2002, 2008).
Weak LD can occur in small finite populations as a result of genetic drift, but is normally caused by genetic linkage, i.e. two or more loci located closely together on a chromosome with limited recombination between them (David, 1998; Hansson & Westerberg, 2002). The effect of LD is therefore normally restricted to a narrow chromosomal segment around the target locus and is hence referred to as a local effect. A general effect however, is dependent on identity disequilibrium (the nonrandom association of diploid genotypes in zygotes), which is mainly generated by partial inbreeding (David, 1998; Hansson & Westerberg, 2002). Identity disequilibrium is relatively insensitive to genetic linkage, and creates a correlation among all loci of the genome within an individual (David, 1998).
Host individual heterozygosity has been found to mediate resistance towards parasites and pathogens (Sorci et al., 1997a; Coltman et al., 1999; Arkush et al., 2002; Penn et al., 2002; MacDougall-Shackleton et al., 2005), and immune function may thus contribute to the maintenance of a link between heterozygosity and fitness. In birds, the swelling response to subcutaneously injected phytohaemagglutinin (PHA) is a regularly used in vivo assay to measure immune function. PHA is a mitogen that induces proliferation and differentiation of a variety of different immune cells (lymphocytes, thrombocytes, basophils, heterophils and macrophages), and causes a local swelling at the injection site (Stadecker et al., 1977; Lochmiller et al., 1993; Martin et al., 2006). Measuring the magnitude of this swelling gives a reliable estimate of cell-mediated immunity (Stadecker et al., 1977). However, a recent histological study of the PHA response in house sparrows (Passer domesticus), revealed that both innate and adaptive components of the immune system are involved (Martin et al., 2006).
The swelling response to PHA correlates with a number of life history variables in birds (Tella et al., 2002), including survival in both nestlings (Sorci et al., 1997b; Christe et al., 1998) and adults (Soler et al., 1999). In song sparrows (Melospiza melodia), PHA response was related to the level of inbreeding (Reid et al., 2003), and in house finches (Carpodacus mexicanus) both PHA response and pathogen resistance correlated with microsatellite heterozygosity (Hawley et al., 2005), indicating that heterozygosity is important for PHA response in birds. Cross-fostering studies have revealed a strong environmental effect on nestling PHA response, whereas a genetic effect of nest of origin has been found in some studies (Saino et al., 1997; Brinkhof et al., 1999; Soler et al., 2003; Ardia, 2005), but not in others (Christe et al., 2000; Tella et al., 2000; Kilpimaa et al., 2005). In a recent ‘animal model’ study, Pitala et al. (2007) found little evidence of a true heritable effect and stronger support for a nonheritable nest-of-origin effect on nestling PHA response in collared flycatchers (Ficedula albicollis). This suggests that the PHA response is not under direct selection but rather depends on parental genetic compatibility, as for example reflected in offspring heterozygosity (Pitala et al., 2007).
In this study, we investigate the influence of microsatellite heterozygosity on PHA response in a 4-year dataset on the bluethroat (Luscinia s. svecica). Bluethroat females frequently participate in extrapair copulations and thereby enhance the PHA response of their offspring by being fertilized by a compatible extrapair male (Johnsen et al., 2000). In a recent study, we also found that extrapair offspring had a higher level of multi-locus heterozygosity than their withinpair half-siblings (Fossøy et al., 2008). However, the enhanced PHA response and increased heterozygosity among extrapair offspring appeared to be independent of each other, indicating little association between heterozygosity and PHA response in this species (Fossøy et al., 2008). In this study, we examine this relationship more closely in a full-sibling approach (i.e. removing the effect of different sires) to facilitate the detection of possible local effects of heterozygosity.
Material and methods
General field methods
Field work was carried out in Heimdalen, Øystre Slidre municipality, southern Norway (61°25′N, 8°52′E), during May and June in 1998, 1999, 2002 and 2003. Males were caught and colour-ringed for individual identification shortly after territory establishment and females shortly after clutch completion. Blood samples (5–25 μL) were obtained by brachial venipuncture and diluted in lysis buffer for genetic analyses.
All nests were visited frequently around the expected time of hatching. The hatchlings were individually marked with a permanent marker pen and/or nail clipping, and weighed. Blood samples (5–25 μL) were normally obtained on day 2 by puncturing the femoral vein. The nestlings were also weighed on day 5 through 8. The body mass at the time of the PHA treatment injection (day 7, see below) was used in the statistical models.
The protocol for the PHA assay is described in Johnsen et al. (2000). Briefly, 0.1 mg of PHA (product number L8754; Sigma-Aldrich, St Louis, MO, USA) was injected subcutaneously in the outer section (metacarpus) of the right wing on day 5 (sensitizing injection) and in the inner section (ulna) on day 7 (treatment injection). The PHA was dissolved in 40 μL phosphate buffered saline (PBS) in 1998 and 1999, and in 20 μL PBS in 2002 and 2003. The reduction of volume in 2002 and 2003 was made to ease the injection of the total solution. An identical volume of PBS only was injected at the respective positions in the left wing as a control. Two measurements of the thickness of each wing were taken with a micrometer (Mitutoyo Digimatic Model 543–681, Mitutoyo, Aurora, IL, USA) at the injection site immediately before injection and 24 ± 1 h after injection at the same position. The difference in swelling between the right and left wing from day 7 to day 8 was used as an estimate of cell-mediated immunity (Stadecker et al., 1977; Fairbrother et al., 2004). All measurements of a particular brood were performed by one person only. The repeatability of successive measurements at the same spot (i.e. precision) was high in the bluethroat (Johnsen et al., 2000). In European starlings (Sturnus vulgaris), the repeatability between the left and right wing injected at the same time (i.e. spatial repeatability) was highly significant (Granbom et al., 2005), suggesting that the swelling response represents a repeatable measure within an individual. Moreover, several cross-fostering studies have revealed an effect of nest of origin, indicating a genetic component of PHA response (Saino et al., 1997; Brinkhof et al., 1999; Soler et al., 2003; Ardia, 2005).
DNA was extracted using QIAamp DNA Blood Kit (Qiagen, Venlo, the Netherlands). Microsatellite markers were amplified by polymerase chain reaction (PCR) on an ABI Prism GeneAmp PCR System 9700 (Applied Biosystems, Foster City, CA, USA), and run on an ABI Prism 3100 Genetic Analyser using fluorescently labelled primers. Allele sizes were determined using ABI Prism Genemapper™ Software version 3.0 (Applied Biosystems, Foster City, CA, USA). Eight microsatellite markers (Ase19, Cuμ4, FhU2, HrU7, Mcyμ4, PAT MP 2–43, Phtr2 and Ppi2) were analysed for all four years (mean for all years: 7.96 loci, range: 6–8 loci, for details on genotyping and locus characteristics see Johnsen et al. (1998) and Fossøy et al. (2006)). Offspring were considered as withinpair when they showed none or only one allelic mismatch (assumed mutation, N = 1) with their putative fathers. In this study, the putative fathers sired 491 of 631 (78%) nestlings.
We calculated two different standardized measures of individual heterozygosity using the microsatellite data. Multi-locus heterozygosity is the proportion of heterozygous loci across all loci genotyped for an individual, whereas mean d2 is the mean squared difference in the number of repeat units between the alleles across all loci. Whereas multi-locus heterozygosity is assumed to reflect recent inbreeding, mean d2 is thought to be a measurement of long-term relatedness, or outbreeding (Neff, 2004). Although the use of mean d2 is debated (e.g. Tsitrone et al., 2001), several empirical studies show a positive correlation between this index of individual heterozygosity and fitness-related variables in birds and mammals (Coltman et al., 1998; Coulson et al., 1998; Hansson et al., 2001, 2004; Höglund et al., 2002). We calculated standardized multi-locus heterozygosity (SH) by dividing the proportion of heterozygous loci for an individual by the mean observed heterozygosity for all loci typed for that individual (Coltman et al., 1999), and standardized mean d2 (Sd2) by dividing the d2 value for a given marker on maximum d2 for that marker before averaging across markers (Amos et al., 2001). All d2 values were log10 transformed before standardizing to improve normality (e.g. Hansson et al., 2004). The marker HrU7 has a very complex repeat unit (Primmer et al., 1995) and was therefore excluded from all d2 calculations. By standardizing, we avoid any bias towards individuals that were successfully genotyped on all markers compared to those that did not amplify on one or more markers (Coltman et al., 1999). Moreover, we also make sure that each marker will have a similar contribution to the calculation of overall heterozygosity (Amos et al., 2001). Thus, any local effect of a particular marker will be explained by the location of that microsatellite in the genome rather than its relative degree of heterozygosity. We also reran the analyses on unstandardized estimates of individual heterozygosity and mean d2, and both measures gave qualitatively similar results (data not shown).
We used mixed model analyses (maximum likelihood) to investigate the effect of SH and Sd2 on PHA response in the full-sibling comparisons. To investigate the possible occurrence of local effects, we substituted SH and Sd2 with the individual heterozygosity of each single marker for each estimate respectively and reran the models. Only broods containing at least two withinpair offspring were included in the analyses, i.e. 438 offspring in 100 broods for SH and 431 offspring in 98 broods for Sd2. The somewhat different sample sizes result from one marker (HrU7) not being included in Sd2 because of a complex repeat unit, and the occurrence of some null alleles which does not allow for the estimate of d2. Body mass is related to PHA response in this species (Johnsen et al., 2000) and was therefore included as a fixed factor in all analyses. Year was included as a random factor to control for annual differences in PHA response, and brood identity was included as a random factor, nested under year, to control for nonindependence among full-siblings within broods. No re-nesting broods were included in the analyses, and broods where one of the parents bred in a previous year were excluded, to avoid pseudo-replication.
In this study, we only utilize full-siblings to accommodate a comparison of individuals possessing the same level of inbreeding, but varying in heterozygosity (e.g. Hansson & Westerberg, 2008). However, this depends on the statistical characteristic of the mixed model to analyse solely the variation within the random factor (i.e. within-nest), which is not necessarily true (van de Pol & Wright, 2009). To determine if the relationship between PHA response and heterozygosity was caused by a within-nest or between-nest effect, we therefore used within-group centring, where the centred value and the mean of each group for the predictor in question is included in the same model (Model 2, van de Pol & Wright, 2009). A significant effect of the centred value then supports a within-group effect, whereas a significant effect of the mean value suggests a between-group effect.
All statistical analyses were carried out in R (R Development Core Team, 2008). The mixed model analyses were performed using the package lm4 (Bates, 2008), and P-values for fixed factors were calculated using the package languageR (Baayen, 2008). We tested for genotypic disequilibrium between each pair of microsatellite loci using F-Stat version 126.96.36.199.
The mean swelling response to PHA for the 480 withinpair offspring was 0.82 mm (SD = 0.40). Both SH and Sd2 showed significant positive associations with PHA response in the full-sibling general effect models (Table 1). Whereas the effect of SH could be attributed to within-nest variation, the effect of Sd2 were mainly caused by between-nest variation in PHA response (Table 2). An analysis where the heterozygosity of each individual locus was included separately, revealed that only one or two loci for each estimate appeared to be responsible for the significant associations in the general models, i.e. Ase19 and PAT MP 2–43 for SH, and Ase19 and Phtr2 for Sd2, although Ase19 was not entirely significant for Sd2 (Table 3). We therefore ran two new models where we only included the individual heterozygosity of these two loci for each estimate and compared these models with the respective general effect models using likelihood-ratio tests (Table 4). For neither SH nor Sd2 did the general model explain more of the variance in PHA response than did the local model (Table 4). On the contrary, the local effect models showed higher likelihoods for both measures of heterozygosity, although not significantly so. We found no evidence of LD between any pair of our microsatellites (data not shown).
Table 1. Mixed model analyses showing the associations between nestling PHA response and (a) standardized multi-locus heterozygosity (SH) and (b) standardized mean d2 (Sd2).
|(a) PHA response (Intercept)||−0.16 ± 0.21||−0.76||0.45|
| Body mass||0.05 ± 0.01||4.01||< 0.001|
| SH||0.27 ± 0.11||2.45||0.015|
|(b) PHA response (Intercept)||−0.06 ± 0.19||−0.30||0.76|
| Body mass||0.05 ± 0.01||3.94||< 0.001|
| Mean d2||0.50 ± 0.20||2.49||0.013|
Table 2. Mixed model analyses based on within-group centring to reveal the within-nest and between-nest relationships between PHA response and (a) standardized multi-locus heterozygosity (SH) and (b) standardized mean d2 (Sd2).
|(a) PHA response (Intercept)||−0.10 ± 0.26||−0.40||0.69|
| Body mass||0.05 ± 0.01||4.01||< 0.001|
| Within-nest||0.30 ± 0.14||2.19||0.029|
| Between-nest||0.22 ± 0.19||1.16||0.25|
|(b) PHA response (Intercept)||−0.12 ± 0.21||−0.56||0.58|
| Body mass||0.05 ± 0.01||3.88||< 0.001|
| Within-nest||0.40 ± 0.25||1.65||0.10|
| Between-nest||0.71 ± 0.36||2.00||0.046|
Table 3. Mixed model analysis showing the independent effects of (a) heterozygosity (H) and (b) d2 (log10 transformed) of each marker on nestling PHA response. Influencial loci are shown in bold.
|(a) PHA response (Intercept)||−0.10 ± 0.27||−0.42||0.67|
| Body mass||0.05 ± 0.01||4.00||< 0.001|
| Ase19 (H)||0.11 ± 0.06||2.22||0.027|
| Cuμ4 (H)||0.07 ± 0.06||1.30||0.20|
| FhU2 (H)||0.01 ± 0.05||0.26||0.80|
| Mcyμ4 (H)||−0.04 ± 0.07||−0.66||0.51|
| Pat MP 2–43 (H)||0.12 ± 0.07||1.99||0.047|
| Phtr2 (H)||0.03 ± 0.06||0.58||0.56|
| Ppi2 (H)||−0.08 ± 0.11||−0.83||0.41|
| HrU7 (H)||0.06 ± 0.04||1.47||0.14|
|(b) PHA response (Intercept)||−0.02 ± 0.19||−0.11||0.91|
| Body mass||0.05 ± 0.01||3.70||< 0.001|
| Ase19 (log10d2)||0.06 ± 0.03||1.80||0.073|
| Cuμ4 (log10 d2)||0.02 ± 0.03||0.73||0.47|
| FhU2 (log10 d2)||0.01 ± 0.05||−0.08||0.94|
| Mcyμ4 (log10 d2)||−0.01 ± 0.03||−0.44||0.66|
| Pat MP 2–43 (log10 d2)||0.03 ± 0.03||0.84||0.40|
| Phtr2 (log10d2)||0.07 ± 0.03||2.24||0.026|
| Ppi2 (log10 d2)||0.04 ± 0.03||1.09||0.28|
Table 4. Likelihood-ratio model comparisons between the general models (using SH and Sd2) and local models (only including the individual heterozygosity of the two influential loci in Table 3) for (a) multi-locus heterozygosity and (b) mean d2 (a: 100 broods, 438 offspring; b: 98 broods, 431 offspring).
|(a) General model||−191.87|| || || |
| Local model||−190.75||2.23||1||0.14|
|(b) General model||−185.65|| || || |
| Local model||−184.41||2.48||1||0.12|
We found that both SH (standardized multi-locus heterozygosity) and Sd2 (standardized mean d2) were significantly and positively associated with nestling PHA response among full-siblings in the bluethroat (Table 1). Whereas the association between SH and PHA response appeared to be caused by within-nest variation, the relationship between Sd2 and PHA seemed to be caused by between-nest variation (Table 2). The within-nest association between SH and PHA response corroborates previous findings of a relationship between fitness and variation in heterozygosity that is unrelated to the level of inbreeding, but rather stems from the random assortment of parental chromosomes (Hansson et al., 2001, 2004; Hansson & Westerberg, 2008). As we only utilized full-siblings in this study, all offspring within each nest have identical levels of inbreeding, and any variation in heterozygosity between them must therefore stem from random recombination events in the meiotic process. However, the between-nest relationship between Sd2 and PHA suggests that there is also an additional component of inbreeding in the association between heterozygosity and PHA immunity. An effect of heterozygosity per se may either be explained by homozygotes displaying recessive deleterious alleles, or by heterozygotes having an advantage due to genetic overdominance (Charlesworth & Charlesworth, 1987). However, an effect of d2 suggests that the difference in number of repeats between the alleles is of significance, indicating an advantage of allelic dissimilarity among heterozygotes. Hence, outbred offspring should benefit from having parents with highly dissimilar genotypes.
A model comparison revealed that one or two loci for each estimate appeared to be responsible for the significant associations in the general models (Tables 3 and 4), suggesting a local effect of heterozygosity on PHA response in this species. Local effects of heterozygosity are thought to arise from LD between the neutral microsatellites and functional genes important for the trait in question (David, 1998; Hansson & Westerberg, 2002). High levels of LD can be found in finite populations, caused by genetic drift, as for example in newly founded or recently bottlenecked populations, but is normally caused by genetic linkage, i.e. two or more loci located closely together on a chromosome with limited recombination between them. However, recent evidence also suggests that selection may be powerful in generating LD, and that LD may be much more common in natural populations than previously assumed (Hansson et al., 2004). Our results indicate that some of our markers may be in LD with functional genes important for PHA immunity. A histological study of the PHA response in house sparrows (P. domesticus), revealed both innate and adaptive components of the immune system, displaying intensive infiltration of many immune cell types, including basophils, eosinophils, heterophils, lymphocytes, macrophages and thrombocytes (Martin et al., 2006). Hence, the immune response to an injection of PHA appears to be very complex and thus likely affected by multiple genes, which further suggests that several different local effects could be expected.
In a recent model simulation study, Hansson & Westerberg (2008) discovered that whereas general heterozygosity effects are unlikely to cause within-family heterozygosity-fitness correlations, such correlations can arise through local heterozygosity effects. Our results are in line with this finding, as the association between SH and PHA response appeared to be caused by a local heterozygosity effect and arose from variations within nests containing full-siblings only. Furthermore, this also supports Hansson & Westerberg’s (2008) conclusion that heterozygosity-fitness correlations should not always be interpreted as evidence of inbreeding. However, we also found evidence of an additional between-family inbreeding effect in this study, suggesting that heterozygosity-fitness correlations can be very complex.
Other studies have also found an association between PHA response and genetic variation in passerine birds. For example, Hawley et al. (2005) found a correlation between multi-locus heterozygosity and both PHA response and disease severity after an experimental inoculation of a bacterial pathogen in adult house finches. An exclusion analysis on the correlation between multi-locus heterozygosity and disease severity revealed no evidence of a local effect (Hawley et al., 2005). However, they did not test for local effects on the correlation between multi-locus heterozygosity and PHA response. Reid et al. (2003) found an association between the level of inbreeding based on pedigree data and PHA response in fledged juvenile and adult song sparrows. Inbreeding is thought to generate identity disequilibrium and a general effect can therefore be expected.
We conclude that the significant correlations between two estimates of microsatellite-based heterozygosity and PHA response in bluethroat nestlings were both more likely caused by local than general effects, but probably realized through two different genetic mechanisms. Our results provide empirical evidence that both random assortment of parental chromosomes during meiosis as well as inbreeding can drive heterozygosity-fitness correlations.
This work was supported by several grants from the Norwegian Research Council. The immune experiment was approved by the Norwegian Animal Research Authority. We would like to thank Vegard Andersen, Roger Dahl, Kåre Fossøy, Vidar Fossøy, Vegard A. Larsen, Olivier Putot, Henrik Pärn, Magnus Snøtun and Christine Sunding for field assistance, Gro Bjørnstad for laboratory assistance and Brian D. Neff for advices on the calculation of mean d2.