MATERIALS AND METHODS

The genetics of vegetative incompatibility in C. parasitica were described recently for all vc types found in Italy (Cortesi and Milgroom 1998) and the majority of vc types from three populations in the eastern United States (Milgroom and Cortesi 1999). Six polymorphic vic loci, each with two alleles, have been identified in Europe (Cortesi and Milgroom 1998), although additional polymorphic vic loci are likely to exist in Europe (Robinet al. 2000) and in Asia (Y.-C. Liu and M. G. Milgroom, unpublished data). We use the nomenclature for vic genotypes described by Cortesi and Milgroom (1998), where the full genotype is abbreviated by the allele numbers only, 1 and 2 (Table 1). For example, the genotype vic1-2, vic2-2, vic3-1, vic4-2, vic6-1, vic7-2 for six vic loci is denoted simply as 2212-12. No allele is designated for locus vic5 (shown as -in the genotype abbreviation), because polymorphism at this locus cannot be detected using our assay, and vic5 has no detectable effect on transmission (Huber 1996; Baidyaroyet al. 2000). Heteroallelism at three vic loci has been shown to prevent heterokaryon formation in C. parasitica (Huber 1996).

Virus transmission assays: Virus transmission was assayed by growing donor and recipient isolates together on solid medium as described previously (Liu and Milgroom 1996). Assays were performed in petri plates containing a modified potato dextrose agar [1 liter contains 24 g potato dextrose broth (Difco, Detroit), 20 g agar no. 1 (Oxoid, Basingstoke, Hampshire, UK), 2 g yeast extract, 7 g malt extract (Difco), and 0.8 g tannic acid (Fluka Chemie, Buchs, Switzerland)]. A mycelial plug from a virus-infected donor was placed on the medium adjacent to a mycelial plug of a virus-free donor and incubated for 7 days at 24° in the dark. Host isolates infected with Cryphonectria hypovirus 1 (CHV-1) typically produce mycelium with markedly less pigmentation when grown under low light (Hillmanet al. 1990). Virus transmission from donor to recipient isolates is detected by the growth of mycelium with less pigmentation in the recipient, while lack of transmission is evident when the recipient isolate maintains an orange-pigmented colony typical of virus-free isolates (Figure 1).

Initially, two CHV-1-infected isolates were used as sources of viruses in this study. E13 was isolated from Valesone (Domodossola, Italy) in 1976 (Bisiachet al. 1988); TE9 was isolated from Teano, Italy (Cortesiet al. 1996). On the basis of previous studies comparing different hypovirus species (Huber 1996; Liu and Milgroom 1996) and preliminary observations (results not shown), we assumed that virus transmission was not affected by virus strains. Viruses were transmitted into virus-free laboratory isolates to create a set of virus-infected isolates to be used as donors. Transmission was not always possible directly from the original source isolates to every laboratory isolate. In some cases we transmitted viruses into other laboratory isolates first and then from these isolates to other donor isolates (Anagnostakis 1983; Peeveret al. 2000). Most isolates were single ascospore cultures (Cortesi and Milgroom 1998), which we assumed were virus free because of pigmented phenotypes and because CHV-1 is not transmitted into ascospores (Elliston 1985; Anagnostakis 1988). In a few cases we used field isolates (Cortesiet al. 1996).

Effect of heteroallelism at each vic locus: For each vic locus, we assayed virus transmission between at least three pairs of vic genotypes, with each pair heteroallelic only at the vic locus being tested. With few exceptions, at least 15 independent trials (1 per petri dish) were performed for each pair of isolates, although many tests had 20 or more (Table 1). To provide independent replication of the same vic genotypes, assays were repeated using isolates derived from different crosses or field isolates from different populations.

Asymmetric virus transmission: For most pairs of isolates used for estimating the effects of vic alleles on transmission, each isolate was used as both a donor and recipient (reciprocal transmission) to determine whether transmission was the same in both directions.

Interactions between vic loci: To test for independent effects at different vic loci, we assayed virus transmission between isolates heteroallelic at two loci; higher-order interactions were not investigated. For each two-locus combination, we paired vic genotypes to account for all possible two-gene differences. Donors with two-locus genotypes 11, 12, 21, and 22 (at the two vic loci being studied) were paired with recipients with two-locus genotypes 22, 21, 12, and 11, respectively (all other vic alleles were held constant between donors and recipients in any given pairing). For each two-locus combination (with the exception of those involving vic4) we replicated at least three sets of vic genotypes for each two-allele pairing (Table 1). Fewer combinations were tested for vic4 because this locus had little effect on virus transmission (Huber 1996; Huber and Fulbright 1994).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Virus transmission assays in Cryphonectria parasitica. Virus-infected donors, D(V), and virus-free recipients, R, are co-cultured on solid medium in petri plates. Lack of virus transmission is evident in the plate on the left from the uniformly pigmented mycelium in the recipient colony. In contrast, successful virus transmission is evident in the plate on the right because the virus-infected recipient, R(V), produces mycelium with less pigmentation after transmission.

Logistic regression model: A common model for describing the presence or absence of a trait is the logistic regression model (Hosmer and Lemeshow 1989). Logistic regression posits a nonlinear model for the probability of the trait and can flexibly incorporate categorical or continuous predictors. We thus model the probability of virus transmission between a pair of isolates as a function of the heteroallelic vic genes, i.e., as fixed effects. In addition, different isolates of the fungus were used, which may differ with regard to genes other than at the six vic loci; this is defined as the genetic background effect and was modeled as being selected from a distribution, i.e., as random effects.

Let p_ij be the probability of transmission from donor isolate i to recipient isolate j. Our logistic regression model is log[pij∕(1−pij)]=μ+∑kβkHTAijk+∑kγkASYijk+∑k∑lθklEPIijkl+donori+recipj, (1)where HTA_ijk = 1 if alleles in isolates i and j are heteroallelic at vic locus k and 0 otherwise; ASY_ijk = −½ if locus k is heteroallelic for isolates i and j with allele 1 in the recipient, ½ if locus k is heteroallelic with allele 2 in the recipient, and 0 if alleles at locus k are the same; and EPI_ijkl is an epistasis (or interaction) indicator variable that is equal to 1 if the pairing of isolates i and j is heteroallelic at both loci k and l and is 0 otherwise. The interpretation of the parameters is as follows: μ is the intercept, β_k is the effect of heteroallelism for locus k, θ_kl is the effect of epistasis (interaction) between heteroallelic loci k and l, and γ_k is the effect of asymmetry for locus k (i.e., the difference between heteroallelism with a recipient with allele 1 and a recipient with allele 2). The rationale for using ½ and −½ for asymmetry indicator variables is to estimate the average effect of heteroallelism at a vic locus, with the effects of different alleles in the recipients canceling out, and, at the same time, the differences between alleles can be estimated. In other words, the average effect (on the logit scale) for the kth locus is averaged over the two possible alleles in the recipients as [(μ+βk+γk∕2)+(μ+βk−γk∕2)]∕2=μ+βk, while the difference between recipients with allele 1 and allele 2 is (μ+βk+γk∕2)−(μ+βk−γk∕2)=γk. The donor and recipient effects are random factors and we model them as selected from normal distributions: donori~Normal(0,σD2)recipj~Normal(0,σR2). (2)This parameterization allows for estimation of a correlation among responses measured on the same isolate used as a donor or recipient.

Estimation was performed by maximum likelihood using a simulation-based maximization technique (McCulloch 1997). Tests were performed by simulation-based likelihood-ratio tests (Geyer and Thompson 1992). All computations were programmed and performed in MATLAB (Version 5, 1999; The MathWorks, Natick, MA).

Tests of the individual coefficients and groups of coefficients were performed to assess the influence of each heteroallelic vic locus, asymmetry, and epistasis effects, using a likelihood-ratio test. For example, to test for epistasis, the likelihoods of the fitted model above and the model with all the θ_kl set to zero were compared. Because the likelihood cannot be calculated explicitly, a simulation was performed to approximate the value of the likelihood-ratio test (McCulloch 1997).

To test if there are other genes affecting transmission we tested whether there is any residual variation in the probabilities associated with the donors and recipients not captured by the heteroallelism, asymmetry, and epistasis variables. That is, if there are no other genes affecting the transmission of the virus, then all isolates with a given set of vic genes (fixed effects) will behave the same. This is implemented by testing H0:σD2=0 , σR2=0 , again with a likelihood-ratio test.

Independent test of the regression model: To evaluate model predictions, virus transmission was tested empirically between field isolates sampled from populations from Italy and the United States. We used isolates from previous collections for which vic genotype data were available (Cortesiet al. 1996; Liuet al. 1996; Milgroom and Cortesi 1999). We sampled with replacement to form 20 pairs of isolates from each of three populations in northern Italy (Bergamo, Pigna, and Crevoladossola) and 40 pairs of isolates from Maryland. The vic alleles are in gametic equilibrium in the Maryland population (Milgroom and Cortesi 1999); therefore, we assume that vic genes are randomly associated with any genetic background effects in this population. The same assumption is only partially satisfied for the Italian populations, but they are relatively diverse with respect to vic genotypes and show some evidence for recombination (Milgroom and Cortesi 1999).

All field isolates were screened for CHV-1 by colony morphology and by the presence of double-stranded RNA (dsRNA) to ensure they were virus free. Virus screening was done by immunoblotting (Peeveret al. 1997) or by miniprep for dsRNA (Morriset al. 1983). Virus-free isolates were obtained from virus-infected isolates by sampling single-conidial isolates when necessary.

One member of each isolate pair was randomly designated as the donor and was infected with CHV-1 as described above; the other isolate was designated as the recipient. Virus transmission tests were conducted between these isolates as described above to estimate the proportion of trials with successful virus transmission; 10 or more trials were conducted for each pair of field isolates. The observed proportion of trials with successful virus transmission was compared to model predictions on the basis of the vic genotypes of donors and recipients. The predicted probability of transmission between individuals i and j was calculated from the logistic regression model by back transformation of the predicted log[p_ij/(1 − p_ij)] from Equation 1 for each pair of vic genotypes. We also compared virus transmission results from a published report by Bissegger et al. (1997), for which vic genotype data are available (Cortesiet al. 1998), to predictions from logistic regression.