In this issue of New Phytologist, Croll et al. (pp. 672–687) describe a study in which they sampled over 40 individuals from a field population and scored each individual for its genotype at each of 13 polymorphic molecular loci. Such studies have been the routine fodder of population genetics since the 1970s, so why is this paper noteworthy? It is because the subjects of this study are arbuscular mycorrhizal fungi (AMF). These are strange organisms, and it has become obligatory to open every paper on AMF with two true statements: AMF are vital for the normal growth of most plants, and AMF are fiendishly difficult to study.
‘AMF are exceptional in that there is no stage in the life history where an individual is reduced to a single nucleus, because the large spores contain hundreds of nuclei.’
The first challenge in AMF population genetics is to sample a set of individuals from the tangled web of root infections in the field, which Croll et al. did by establishing multiple in vitro cultures. Nobody has yet succeeded in growing AMF in pure culture, as the fungus does not grow unless attached to plant roots, but it is possible to propagate AMF on root organ cultures in vitro, although only a few AMF species have so far been grown successfully in this way. This indirect approach is bound to introduce some bias reflecting culturability in the chosen hosts and conditions, but has major advantages. A simpler approach would be to collect spores from the field (Stukenbrock & Rosendahl, 2005), but this is the equivalent of studying a plant community by digging up the seeds in the soil; spore numbers are a poor reflection of fungal biomass and activity. The other problem with a spore is that it only has enough DNA for a small number of analyses (Stukenbrock & Rosendahl, 2005), whereas an in vitro culture can be multiplied indefinitely. Croll et al. were therefore able to type each isolate reliably at 13 genetic loci. Furthermore, each locus was targeted by AMF-specific PCR primers, so we can be confident that all the products are from the fungus itself rather than from the bacteria or nonAMF fungi that are frequently closely associated with AMF.
The first surprise is that, at all the loci described by Croll et al., they report just one allele in each isolate. This might seem a normal expectation for a fungus, given that most fungi are haploid most of the time, but in fact the authors did find some loci that had more than one allele in some individuals, although they did not use them in the study because this would have complicated the analysis (I. Sanders, pers. comm.). The history of multiple sequences within AMF began with the discovery that a single isolate may have two or more distinct sequences for the ribosomal RNA gene region (Sanders et al., 1995; Lloyd-Macgilp et al., 1996). Usually these differences are fairly small, but in the Glomus mosseae group of species, two internal transcribed spacer (ITS) sequences are found, and these are so different that one might expect them to indicate different genera, except that they have been reported to co-occur in the same isolate (Clapp et al., 2002). Nor is within-isolate diversity confined to the ribosomal RNA genes: it has also been seen for other genes (Helgason et al., 2003; Corradi et al., 2004). Whether this diversity reflects the cohabitation of genetically distinct nuclei within the same cytoplasm (a heterokaryon) or identical nuclei, each of which contains all the genetic variants (a homokaryon), is the subject of ongoing debate (Kuhn et al., 2001; Hijri & Sanders, 2004; Pawlowska & Taylor, 2004; Hijri & Sanders, 2005; Pawlowska, 2005; Corradi et al., 2007; Rosendahl, 2008). AMF have few cross-walls within their hyphal networks, so nuclei inhabit a common cytoplasm (a syncytium). This is not so unusual in fungi, but AMF are exceptional in that there is no stage in the life history where an individual is reduced to a single nucleus, because the large spores contain hundreds of nuclei. The nuclei in a spore, and in the mycelium derived from a spore, are therefore better considered as a population rather than as an individual. If they are collectively a homokaryon, we have to explain how the variation generated by mutation is purged, while if they are a heterokaryon, we must explain how diversity is maintained in the face of genetic drift.
There is not space here for a full discussion of all the evidence, so I shall merely point out that the most-cited study supporting the homokaryon hypothesis is far from conclusive. Pawlowska & Taylor (2004) studied two genes in Glomus etunicatum. The first, PLS1, was unusual in that 13 allelic variants coexisted in a single isolate. All the variants were maintained through successive generations, which would be unlikely if each variant were in a separate, independently segregating nucleus. However, under some models for heterokaryosis, such as the selfish nucleus hypothesis (Fig. 1), independent segregation of nuclei is not expected. Secondly, Pawlowska & Taylor (2004) amplified the ribosomal ITS from a number of individual nuclei, and showed that, whenever they succeeded in getting amplification, a nucleus carried all three of the ITS variants seen in that isolate. Ribosomal genes are unusual in that they are in multiple copies in the genomes of all organisms, and sometimes in several arrays that may diverge in sequence. Admittedly, the sequences in different nuclear lineages within a permanent heterokaryon would be expected to diverge, but Pawlowska & Taylor (2004) only succeeded in obtaining amplification for a handful of nuclei, which might have represented a particularly favourable genotype (perhaps with a large number of rDNA copies), rather than a random sample of all the nuclei in the spore. Indeed, in a syncytium, it is theoretically possible for some nuclei to survive even if they have no rDNA at all.
Multiple genetic markers allow a good number of genotypes to be distinguished, and Croll et al. discuss their spatial distribution and association with host species. Furthermore, with multiple loci in multiple individuals, patterns of association between loci can be explored. Here things start to get really interesting. AMF have been described as anciently asexual, because no convincing structures associated with sexual reproduction have been seen, and no unambiguous evidence for genetic recombination (Rosendahl & Taylor, 1997). Ancient asexuals are of interest because most asexual lineages are thought to be short-lived (Gandolfi et al., 2003). This issue, and others raised in this commentary, have been discussed in a recent Tansley review (Rosendahl, 2008). A completely asexual organism evolves as a branching tree of clonal lineages. A genetic variant that arises in one lineage is forever associated with the set of genotypes characteristic of that lineage. Without recombination, it cannot get together with an allele at another locus that arose in a different lineage (Fig. 2). On the face of it, the table of genotypes presented by Croll et al. seems to contradict this expectation. There are several instances where the same allele crops up in genotypes that, on the evidence of the other loci, clearly belong to different lineages. Is this evidence for genetic recombination in the supposedly asexual AMF?
The evidence is not conclusive, and the potential problem is homoplasy; that is, the possibility that the same allele arose independently more than once. This is a particularly prevalent problem with microsatellites (simple sequence repeats, SSR), and most of the loci in the study were of this kind. An SSR locus consists of a number of tandem repeats of a very short DNA motif, typically two or three bases. Mutations that increase or decrease the number of repeats occur frequently by strand-slippage during replication, so such loci are typically highly polymorphic in populations. This makes them handy as genetic markers, but homoplasy is rife at SSR loci because a variant with a particular number of repeats can be generated in many ways. Hence, SSRs do not seem a good choice for a critical study of interlocus associations. However, an examination of the allele sequences that Croll et al. provide in their supplementary material reveals that, although they chose loci because they included SSR motifs, most of the allelic differences that they observed are not the result of typical SSR length variation. In fact, the alleles differ in multiple ways, including short insertions or deletions and single nucleotide substitutions. When an allele has multiple unique features, it is very unlikely to have arisen more than once, so we can discount homoplasy. Some ambiguity remains, though, because Croll et al. do not provide a complete sequence for every gene in every individual. Most individuals are only characterized by the overall length of the PCR product from each locus. It is easy to see that quite different sequences can happen to be of the same overall length and, in the case of the nuclear intron locus, the authors demonstrate exactly this. Nevertheless, Croll et al. appear to have the tools for a rigorous assessment of recombination in AMF. Will we soon have to reassess the assumption that AMF are ancient asexuals?