Genetic Basis of Natural Variation in D. melanogaster Antibacterial Immunity

See allHide authors and affiliations

Science  19 Mar 2004:
Vol. 303, Issue 5665, pp. 1873-1876
DOI: 10.1126/science.1092447


Many genes involved in Drosophila melanogaster innate immune processes have been identified, but whether naturally occurring polymorphism in these genes leads to variation in immune competence among wild flies has not been tested. We report here substantial variability among wild-derived D. melanogaster in the ability to suppress infection by a Gram-negative entomopathogen, Serratia marcescens. Variability in immune competence was significantly associated with nucleotide polymorphism in 16 innate immunity genes, corresponding primarily to pathogen recognition and intracellular signaling loci, and substantial epistasis was detected between intracellular signaling and antimicrobial peptide genes. Variation in these genes, therefore, seems to drive variability in immunocompetence among wild Drosophila.

Efficacy of immune response is a critical determinant of fitness, and higher eukaryotes have accordingly evolved sophisticated mechanisms for suppressing bacterial infection. For invertebrates, this is mediated by generalized, or innate, immune pathways, which include phagocytosis by scavenging macrophages and the extracellular circulation of short antibiotic peptides (1, 2). Artificially generated mutations that abolish the function of key genes have resulted in severe immune deficiencies [(1) and citations therein]. In addition, studies of population-level variation have suggested that Drosophila immunity genes evolve under positive natural selection (35). However, phenotypic effects of naturally occurring polymorphism in innate immune genes have not previously been studied in invertebrates. We sought to measure variability in immunocompetence among Drosophila melanogaster in wild populations with the aim of attributing that variation to gene candidates involved in immune function. We initially focused on loci encoded on the D. melanogaster second chromosome, examining 21 protein-coding genes of diverse function (Table 1). Seven candidate genes are hypothesized to be involved in microbial recognition: four class C scavenger receptors [SR-CI, SR-CII, SR-CIII, and SR-CIV (6)] and a three-gene cluster of putative peptidoglycan recognition proteins [PGRP-SC1A, PGRP-SC1B, and PGRP-SC2 (7)]. Candidates involved in signal transduction are three Toll-like receptors (Tehao, 18-Wheeler, and Toll-4), the rel transcription factor Dif, rel inhibitor cactus, and intracellular signaling genes imd and IK2 (2). The remaining seven genes encode the secreted antibacterial peptides Defensin, Metchnikowin, Attacins A, B and C, and Diptericins A and B (812).

Table 1.

Distribution of molecular markers among candidate loci. “Cyt. pos.” is the cytological position of each locus. Length of the survey region is measured in kilobases. “Sites” is the number of polymorphic markers genotyped within each locus, without regard to linkage relationships among markers. Genic locations and site genotypes in each line are available in fig. S1. Disequilibrium relationships among markers are illustrated in fig. S2.

Locus Cyt. pos. Length (kb) Sites
Pathogen recognition
    PGRP-SC1A, -SCIB, -SC2 44E 10.0 10
    SR-CI, SR-CIII 24D 4.3 13
    SR-CII 48E 4.4 7
    SR-CIV 23F 4.6 7
Toll-like receptors
    Toll-4 30A 6.8 8
    Tehao 34C 7.8 16
    18-Wheeler 56F 7.6 6
Signal transduction
    DIF 36C 20.7 8
    cactus 35F 14.8 7
    IK2 38D 5.1 5
    Immune deficiency 55C 4.0 7
Antibacterial peptides
    Attacin AB 51C 4.9 8
    Attacin C 50A 3.1 7
    Defensin 46D 1.4 5
    Diptericin AB 55F 4.7 7
    Metchnikowin 52A 1.8 6

We generated 101 D. melanogaster lines, each homozygous for an independent chromosome 2 isolated directly from a wild North American population and crossed into a common genetic background (13). Consequently, all genetic variation among test lines could be attributed to naturally occurring polymorphism in chromosome 2 loci. Adult flies from each line were manually infected with S. marcescens at 4 to 8 days after eclosion, and viable pathogen load was measured by quantitative plating at prescribed times after infection (13). In total, more than 16,000 flies were infected, and colony counts were obtained from 6924 plates. Homogenates from all sham-infected flies failed to yield colonies.

Systemic mean loads of S. marcescens differed by only marginal significance at 7, 15, 26, and 39 hours after artificial infection [analysis of variance (ANOVA), P = 0.02], suggesting that bacterial density generally reached a stable plateau over that time. Because no statistically significant interaction occurred between the parameters of time after infection and genetic line (ANOVA, P = 0.67), data from different time points could be pooled. Genetic line was a significant determinant of bacterial load at all times after infection, and was highly significant when all times were pooled (ANOVA, P < 10–4) with extreme lines differing by 10 phenotypic standard errors (Fig. 1). For technical convenience, flies assayed at 7 and 26 hours after infection had been infected in the morning, and flies assayed at 15 and 39 hours after infection had been infected in the evening. A strong effect of circadian period at time of infection was observed in bacterial loads (ANOVA, P < 10–4), with higher bacterial loads occurring in PM-infected flies. However, the interaction of circadian time of infection with genetic line was not statistically significant (ANOVA, P = 0.14). This effect of circadian time is consistent with previous studies showing that expression of multiple D. melanogaster immune response genes cycle over the circadian day (14). In no case did we observe a significant effect of Drosophila sex on bacterial load, nor did we observe any significant interaction between sex and genetic line. After controlling for the effects of time after infection and circadian time of day, 47.2% of the variance in bacterial load could be attributed directly to genetic differences among lines. Due to the large statistical effect of the circadian time of day of infections, further association tests were done after attempting to control for this variable with analysis of covariance (ANCOVA) and after parsing the phenotype data into sex and circadian time categories.

Fig. 1.

Variability among D. melanogaster lines in the ability to suppress systemic growth of the Gram-negative entomopathogen S. marcescens. Bacterial load was estimated by counting viable S. marcescens colonies grown from flies homogenized at prescribed times after artificial infection (13). Genetic line was a highly significant determinant of bacterial load (ANOVA, P < 10–4), with extreme lines differing by 10 phenotypic standard errors.

Polymorphism was ascertained in each of the 21 candidate genes by sequencing 10 to 12 alleles chosen independently of prior phenotypic information (12, 13). More than 100 kb of unique sequence encompassing coding regions, introns, and 1 to 2 kb of upstream putative regulatory sequence were surveyed for polymorphism, with 127 of the uncovered sites typed across all 101 test lines (13) (fig. S1). Linkage disequilibrium among typed markers was generally nonsignificant, with the exception of a small number of sites showing intralocus correlations (fig. S2). The effects of the 127 genotyped markers in the 21 candidate genes on mean bacterial load were tested for each of the 9 phenotypic constructions: male, female, or pooled-sex flies in each or both circadian periods (13). Thirty polymorphic sites were detected in 16 genes at which allelic state was significantly associated with variation in systemic S. marcescens load (Fig. 2A) in at least one phenotypic construction. Most of these markers explain less than 10% of the phenotypic variance observed (Fig. 2B), although that variance may be underestimated if the typed markers are in linkage disequilibrium with but are not themselves the sites causal to the polymorphism. The five genes that had no polymorphic sites associating with variability in resistance to S. marcescens all encoded antibacterial peptides. Many of the polymorphic markers that associated with variability in bacterial load showed effects that were limited by sex or circadian time of day, although this may partially stem from differences in statistical power when the data are subdivided. To generally assess the broader relationship between candidate gene markers and the immunity phenotypes observed, a composite probability was generated for each site based on the strength of association in the nine subphenotypes. These composite probabilities were similar to those calculated from the data set pooled across sexes and circadian times but were less sensitive to statistical effects resulting from sex- or circadian-limited genetic effects. Fifteen sites among the 127 immune gene polymorphisms had a composite P value smaller than 0.01, the vast majority of which were located in genes involved in pathogen recognition and intracellular signaling (Fig. 2A).

Fig. 2.

(A) Sites in candidate genes that are significantly associated with variation in bacterial growth suppression in males, females, or pooled-sex analyses from flies infected in circadian AM and/or PM, representing a total of nine phenotype constructions. Several sites have effects that are limited by sex or environment (circadian day). The composite probability is derived from a comprehensive assessment of the multiple combinations of sex and circadian time (13). Associations that have a comparison-wise significance between 0.05 and 0.01 are shaded in blue. Associations with a null probability smaller than 0.01 are shaded in red. Numbers in parentheses are the unique nucleotide coordinates from Release 3.1 of the D. melanogaster genome sequence assembly (25). More detail on each polymorphism is presented in fig. S1; disequilibrium relationships among sites are illustrated in fig. S2. (B) Percentage of the phenotypic variance explained by each site. Percentages shown are of the observed phenotypic variance attributable to genetic line. Because the statistical significance of an association is partially dependent on allele frequency, the rank orders of strength of association and percent of variance explained are not equivalent.

To detect potential epistatic interactions among sites and loci, associations between bacterial load and genotypes consisting of all observed pairs of sites were tested. Of 5815 pairwise tests conducted, 392 were significant at a nominal P ≤ 0.05, compared with 291 that would be expected by chance (Fig. 3A). This suggested a partial nonadditive nature of the observed allelic effects. The preponderance of significant epistatic interactions were detected among intracellular signaling loci and between intracellular signaling loci and antimicrobial peptide genes (Fig. 3B). This contrasted with the single-site analyses, where polymorphism in antibacterial peptides was generally nonsignificant (Fig. 2A) and highlights the importance of considering genetic background when measuring genetic effects.

Fig. 3.

(A) Distribution of tail P values for two-site interaction tests. The proportion of interaction tests in each probability class is plotted in the histogram. The horizontal line indicates the expected proportion of tests in each class under the null hypothesis of no biological interactions. (B) Distribution within and among the loci of two-site interaction tests significant at or below a nominal P = 0.05. Red boxes indicate that more than half of the two-site interaction tests performed across a pair of loci are significant at the 5% level. Locus pairs at which 30 to 50% of the tests were significant are shaded in yellow, and pairs at which 10 to 30% of the tests were significant are shaded in blue.

Several conclusions about the structure of naturally occurring variation in D. melanogaster innate immunity can be drawn from this work. First, the continuous and normally distributed phenotypic variation among lines (Fig. 1) and the generally small contributions of individual markers to the total genetic variance imply that most naturally occurring variation in immunocompetence is derived from mutations of minor effect. However, mutations of larger effect (explaining 10 to 15% of the measured phenotypic variance) can be found segregating in intracellular signaling molecules (Fig. 2B). Second, the polymorphic sites most significantly associated with variance in suppression of S. marcescens are found in pathogen recognition and intracellular signaling molecules, with antibacterial peptide genes harboring markedly few polymorphisms that produce individually significant effects. The apparent strength of influence of recognition and signaling molecules may result from the regulatory control these proteins exert over a variety of immune-related cellular processes (1, 2, 15). In particular, because activation of Drosophila immune signaling pathways induces the expression of components of those pathways (16, 17), small changes in expression or activity of signaling components may show an amplified effect on the phenotype. In contrast, because antimicrobial peptides are the most downstream targets of immune regulation and are partially redundant (18), these genes might be expected to make weaker contributions to phenotypic variability. Third, although polymorphic markers were typed in both transcribed and regulatory regions of candidate genes (fig. S1), many of the significantly associated sites are located upstream of protein-coding sequences. Because the markers may not themselves be the sites causing phenotypic variability, it is possible that upstream markers are in linkage disequilibrium with untyped amino acid variants in the coding region. However, it is also likely that variation in transcriptional regulation of immunity genes may influence a substantial proportion of observed phenotypic variation. This hypothesis is particularly attractive in light of the epistatic interactions observed between intracellular signaling molecules and antimicrobial peptide genes.

D. melanogaster immune competence and the phenotypic effects of genetic polymorphism were found to vary over the circadian day. Notably, a complex of sites upstream of the imd gene showed strong associations with bacterial load sustained by AM-infected flies, but not by PM-infected flies, with one site explaining 16.4% of the AM phenotypic variance. imd is an important regulator of the response to Gram-negative bacteria. Circadian cycling of the expression of immune-related genes, including imd, has been previously documented in the absence of immune challenge (14). The current results suggest that circadian cycling could have an important genetic influence on the immunocompetence of D. melanogaster in the wild. Interactions between genotype and sex or environment in D. melanogaster quantitative traits are not unprecedented, as studies of sensory bristle number, wing shape, and longevity have previously identified multiple quantitative trait loci that are either strongly influenced by or completely dependent on sex and/or environment (1922).

It is not apparent why phenotypically strong mutations in a vital trait such as immune competence are allowed to persist in the population. Potentially, these may exist in mutation-selection-drift balance, as the result of other physiological trade-offs (23, 24) or as polymorphisms balanced against multiple environmental pathogens. Further study of sequence level and phenotypic variability in innate immune systems is warranted to address this question.

Supporting Online Material

Materials and Methods

Figs. S1 and S2


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article