The mean prevalence of the three most common malaria lineages in breeding great reed warblers at our study site were 23% (GRW1), 8% (GRW2) and 20% (GRW4), and the prevalence varied considerably between cohorts ().

Individuals infected or non-infected with malaria parasite GRW2 tended to be genetically differentiated in terms of MHC alleles (23 MHC alleles, AMOVA_{1,339}, *F*_{st}=0.013, *p*=0.040), but this was not at all the case for infection with malaria parasite GRW1 (AMOVA_{1,339}, *F*_{st}=0.0019, *p*=0.23) and GRW4 (AMOVA_{2,339}, *F*_{st}=0.001, *p*=0.66).

The average number of MHC alleles was significantly higher in individuals infected with GRW2 (7.1±1.4 (s.d.), *n*=26) compared with uninfected individuals (6.3±1.8 (s.d.), *n*=314; *U*=2963, *p*=0.018; ). The prevalence of malaria infection GRW2 was significantly positively associated with the number of MHC alleles, an estimate of MHC heterozygosity (logistic regression, *n*=340, χ^{2}=7.17, *p*=0.0074; ). Furthermore, the squared number of MHC alleles contributed significantly to the model (logistic regression, *n*=340, χ^{2}=5.82, *p*=0.016; ), demonstrating that the there is a nonlinear relationship between the number of MHC alleles and the prevalence of GRW2. This nonlinear relationship is visualized in a cubic spline plot (). We cannot tell from our results whether the prevalence of GRW2 is highest among birds having an intermediate number of MHC alleles, because of the very large 95% confidence interval of the cubic spline function at greater than eight MHC alleles ( and ). The most parsimonious conclusion is therefore that the prevalence of GRW2 reached a plateau at eight MHC alleles where after a further increase in alleles has no effect on the prevalence (). In contrast, there was no correlation between either infection with GRW1 (logistic regression, *n*=340, χ_{1}^{2}=0.38, *p*=0.54), or infection with GRW4 (logistic regression, *n*=340, χ_{1}^{2}=0.033, *p*=0.85) and the number of MHC alleles.

| **Table 1**Results from logistic regression analyses of the variation in prevalence of the malaria parasite GRW2 in relation to number of MHC alleles (MHC_{alleles}) and MHC allele B4b in great reed warblers. Parameters entered the models at *p*<0.1. |

The prevalence of GRW2 was not correlated with genome-wide heterozygosity, measured at 18 microsatellite loci (logistic regression, *n*=340, χ_{1}^{2}=0.40, *p*=0.53). There was no correlation between infections with either GRW1 (logistic regression, *n*=340, χ_{1}^{2}<0.01, *p*=1.00) or GRW4 (logistic regression, *n*=340, χ_{1}^{2}=0.65, *p*=0.42) and microsatellite heterozygosity. Finally, there was no correlation between number of MHC alleles and microsatellite heterozygosity (*r*_{s}=0.039, *n*=340, *p*=0.38).

There was a significant positive association between the MHC class I allele B4b (this allele has been shown to vary significantly between years among cohorts of great reed warblers;

Westerdahl *et al*. 2004*a*) and infection with GRW2 (). There were no significant associations between the other two parasites and allele B4b (). Afterwards, we tested whether there were associations between any of the remaining 22 MHC alleles and infection with GRW1, GRW2 and GRW4, but found no such correlations (

*p*>0.03 in all cases, hence, they were far from significant after Bonferroni correction).

| **Table 2**Malaria parasite prevalence (parasite GRW1, GRW2 and GRW4, respectively) in great reed warblers with and without MHC allele B4b. |

There was a positive correlation between allele B4b and the number of MHC alleles (logistic regression, *n*=340, χ_{1}^{2}=11.59, *p*=0.0007). Thus, individuals with many MHC alleles were more likely to carry the B4b allele. However, the correlation between MHC alleles and prevalence of GRW2 remained significant also after excluding all individuals carrying allele B4b (). Finally, we included both the B4b allele and number of MHC alleles in a multiple logistic regression analysis, and found that both these factors appear to contribute to the model ().