Abstract
In the late 1980s, Smith, Bruhn, and Anderson1 discovered a genetic individual of the fungus Armillaria gallica that extended over 37 hectares of forest floor and encompassed hundreds of tree root systems in Northern Michigan. Based on observed growth rates, the individual was estimated to be at least 1500 years old with a mass of 100,000 kg. The conclusion was that the Michigan individual of A. gallica was among the largest and oldest organisms on earth, a remarkable claim given that Armillaria is essentially a microorganism existing largely as microscopic hyphae embedded in their substrate. Nearly three decades on, we returned to the site of the large Michigan individual with the tools of whole-genome sequencing. Here, we show (a) that the large individual is still alive on its original site and (b) that mutation has occurred within the somatic cells of the large individual, reflecting its historical pattern of growth from a single point. The overall rate of mutation, however, was extremely low. On the spectrum of mutability in somatic cells, Armillaria occupies the extreme of stability, opposite the extreme of instability as typified by cancer2.
Main
Since the discovery of the Michigan individual1,3, much has been learned about genomes, gene content, and gene expression in Armillaria. A general picture of how individuals of this fungus become established and persist in woodlands has emerged. Because of its broad host range, capacity for enzyme production, and rhizomorphs, its unique organs of local dispersal, Armillaria appears to be adapted for persistent growth in local habitats4. Since its discovery, it is clear that the Michigan individual of A. gallica is not unique in its size and age. Any stable, relatively undisturbed forest could support large, old Armillaria individuals. Indeed, at least two others have been reported to exceed the area covered by the Michigan individual5–7.
Armillaria lives both as a saprophyte on dead wood and as a necrotrophic parasite, killing host plant tissues ahead of the advance of growing mycelia by secreting hydrolytic enzymes and possibly virulence factors8. The range of potential host plants is broad encompassing both angiosperm and conifer trees9. The large genome size of Armillaria relative to its sister taxa is due to increased gene content, particularly in plant cell wall degrading enzymes, and not to the proliferation of transposable elements8. Individuals of Armillaria are established when mating occurs between haploid spores or their germlings10,11. After mating, a diploid mycelium is established rather than the dikaryotic mycelium typical of other basidiomycetes. From mycelium, rhizomorphs develop and push through the soil and decaying woody substrates in much same way as a plant root. Rhizomorphs function as foraging organs in locating new sources of food in dead wood or weakened trees and they can also persist in soil in stasis until new substrates become available. Gene expression in rhizomorph development shares much the same signature of multicellularity found in fruit body development8. Rhizomorphs also share a pattern of gene expression underlying enzyme production with vegetative mycelium8. Against this background, the focus of this study was on the mutability of Armillaria - detecting genomic change in the spatial record left from the proliferation of somatic cells in space from a single zygote ancestor.
The large individual of A. gallica in Michigan was first identified in the late 1980s by making spatially mapped collections of the fungus and then genotyping them over multiple loci1,3. We refer to this individual here as C1, corresponding to Clone 1 in the original publication1 and to “The Humungous Fungus” as named by the news media at the time. All collections of C1 had the same multilocus genotype and shared an identical mitochondrial type. Other nearby individuals had different multilocus genotypes and mitochondrial types3,12. Another individual of lesser spatial extent, designated C2 here, was described in addition to C1.
In 2015-2017, we revisited the site of the large individual and made 245 collections linked to GPS coordinates (Supplementary Table 1). Figure 1 shows distribution of isolates representing C1, C2, and all other genotypes, which were excluded here from further analysis.
Fifteen collections of C1 were Illumina sequenced (Supplementary Table 2) to approximately 100X average coverage (NCBI Accession: PRJNA393342). The sequence reads were initially aligned with a 98 kb mtDNA reference (JGI project number), which was derived from another individual of A. gallica from Ontario, Canada13; JGI Armillaria gallica 21-2 v1). Each of C1, C2, and three additional individuals from the Michigan site had a unique mtDNA genotype based on well-defined SNPs (Supplementary Table 2). The recent Illumina sequenced collections of C1 have the same mtDNA genotype with a living strain of C1 that was collected in the late 1980s. In addition, the recent collections of C2 are linked to a C2 strain from the late 1980s by having the identical mtDNA genotype (Supplementary Table 2).
To test for the signature of mutation, we searched for variation in the nuclear genome of the Illumina sequenced strains. After all filters were applied, 163 variants were found (Supplementary Table 3). In this search, the key requirements for eliminating false calls were to require (a) that candidate sites had at most two alleles among the sequence reads, one reference and one alternate, (b) that one or more strains have 0% or 100% of the reads representing the alternate allele (a “purity” criterion), and (c) that in the heterozygous sites the allele frequencies hovered around 50% (30-70%).
Two general kinds of genetic changes were seen. First, for 151 cases, only single sites experienced the change and adjacent sites in the genome were not affected. We interpret these changes as point mutations resulting in a gain of heterozygosity. The heterozygous strains after mutation had a nearly equal mix of reference and alternate alleles among the Illumina reads, while the homozygous strains had purely one allele. Second, there were regions in strains that were homozygous at a number of adjacent sites that were otherwise heterozygous other strains; we interpret these events as loss of heterozygosity (LOH) in which a string of adjacent SNP sites that were formerly heterozygous all become homozygous simultaneously, as can happen with mitotic gene conversion. That only six LOH tracts were observed (Supplementary Table 3), and that the preponderance of the original heterozygosity in the individual has been maintained, is remarkable given the potential for mitotic gene conversion and crossing over, which would lead to homozygosity over substantial portions of chromosome arms.
The predominant pattern of both point mutation and LOH was that genetic changes were observed as singletons, present in no more than one Illumina sequenced strain (130 of 163 sites, Supplementary Table 3). As expected for recent mutations mostly unaffected by selection14, the majority (73%) of point mutations were due to C to T transitions (insert in Supplementary Table 3).
On average, each strain had about 10 singleton changes that were not shared among the other strains. A minority of changes (33 sites), however, were shared among two or more of the strains. These shared changes were particularly informative because they reflect the historical growth pattern of the individual, while the singletons reflect a history of units of the individual persisting in place over time, presumably after the period of general expansion had occurred. This interpretation arises from the fact that individuals of Armillaria cannot exist in stasis over the long term. Existing food sources become exhausted and new sources of food become available, meaning that growth of the fungus is necessary even to remain in the same place.
To examine the spatial relationships of the mutations, we constructed a phylogeny of 14 of the 15 Illumina sequenced isolates of C1 (one strain, no. Ar188, had lower-than-expected coverage and was excluded from this analysis) using the variants in Supplementary Table 3 as characters and maximum parsimony as the optimality criterion (Fig. 2). The phylogeny shows a high degree of internal consistency and the nesting of the clades in space can be interpreted as reflecting past growth and colonization patterns. Three sites of the 33 sites, however, showed homoplasy in the phylogeny; at these sites, changes occurred two or three times in different branches of the tree. These examples are unlikely due to independent mutation. With a low mutation rate and approximately 100 mb in the genome, the candidate sites for mutation are essentially infinite. The alternative explanation is that the parallel changes could represent single somatic recombination events during the early expansion phase after the individual was established. The potential of fungi for somatic recombination has ample precedent15,16.
Next, for an independent test of the phylogenetic pattern among Illumina sequenced strains, we tested nine sites in Supplementary Table 3 among all isolates of C1. Once again, the changes showed a nesting pattern in space, with the spatially discrete sectors reflecting their mutational history (Fig. 3). A plausible branching pattern for the mycelium would reflect the phylogeny of nuclei over what is likely several thousands of years of growth by C1.
Despite our ability to identify and map mutant alleles in the Armillaria mycelium, estimating the rate of mutation is problematic because we do not know the number of cell divisions intervening between any two isolates. We can estimate the number of genetic differences among the Illumina sequenced isolates experimentally, but the number of intervening cell divisions is inseparable from the mutation rate (no. differences = no. cell divisions times the mutation rate). Mutation rates, however, have been estimated in other organisms17 and we can therefore assume an average mutation rate on the low end of the spectrum, 10-10 per base, per cell replication. (Note that mutation rate varies across the genome and among individuals18.) With a genome of 100 mb, a mutation is expected every 100 cell divisions in a haploid genome and every 50 cell divisions in a diploid genome, such as that of A. gallica. The pairs of strains were separated by an average of ca. 20 mutations, even those spatially separated by 1 km, the minimum path of growth separating them. This translates to only 1000 cell divisions over the 1 km, or only one division every 1 m of growth. If we assume a higher mutation rate, then the estimate of one cell division per meter of growth increases even further. Such a dearth of cell division is hard to reconcile with the microscopic size of fungal hyphae.
How might individuals of Armillaria protect themselves from mutation from cell division during growth? We see three possibilities, which are not mutually exclusive. First, the tips of rhizomorphs, which represent the inoculum potential for colonization, may remain relatively quiescent with respect to cell division, much like the apical meristem of plant roots13,19 and germline cells in mammals20. The rhizomorph tip, however, may be propelled forward by cell division and elongation behind the tip. In this way, the rhizomorph tip may minimize cell division even as it moves through its substrate. There is also precedent for avoidance of cell division in the shoots of plants in which axial meristems are derived from apical meristems with remarkably few intervening cell divisions21. A second possibility is that repair processes may have been driven to higher efficiency by natural selection up to the point where genetic drift negates any additional diminishing fitness benefit22,23. Also, Armillaria exists in environments that may be of low mutagenic potential. UV radiation, for example, is low in such environments and damage such as pyrimidine dimer formation may be lower than in other environments. The third possibility is that the distribution of DNA strands after replication is asymmetric; cells perpetuating the lineage tend to receive old DNA while cells committed solely to local development and not to perpetuating the lineage receive new DNA24. In Armillaria, this would mean that cells in the rhizomorph tip would retain the old DNA, with the subtending cells committed to local, dead-end development would receive the new DNA. The rhizomorphs tips perpetuating the lineage would retain fewer mutations than cells committed to local differentiation and not to perpetuating the lineage. The extent to which each of these mechanisms may contribute to stability remains to be determined.
Here, we followed clonal evolution within cell lineages of a single fungal individual of Armillaria in a spatial context. This follows an earlier analysis of a much smaller Armillaria individual in Ontario13. Our picture of clonal evolution in individuals of Armillaria closely parallels that in cancer progression within single individuals25–28. Cancer progression, however, is accompanied by extreme genomic instability2, not necessarily due to loss of function in DNA repair processes, but rather to loss of control of DNA replication; the rate of replication increases to the extent that fidelity suffers and DNA damage accumulates rapidly. Evolution in cancer occurs on a time span shorter than the life span of the individual affected. Evolution occurs similarly in Armillaria individuals, but over a span of centuries and millennia, and is characterized by extreme genomic stability. The genomic stability of Armillaria and the underlying mechanisms allowing such stability may provide a useful counterpoint to cancer.
Methods Summary
In 2015 – 2017, we made 245 collections of Armillaria linked to GPS coordinates on the site of the large individual of A. gallica. DNA was extracted from each of the collections, which were initially genotyped with polymorphic DNA markers and were tested for somatic compatibility with one another in order to identify collections that represent the large individual C1 (110 collections). Fifteen collections of C1 were Illumina sequenced in order to discover genomic variants. The variants were then spatially mapped over the Illumina sequened strains, and, in nine cases, over the 110 collections of the large individual.
Online methods
Sample
In 2015-2017, we revisited the site of the large individual of A. gallica and made 245 collections linked to GPS coordinates (Supplementary Table 1). Collections were mostly pure cultures from rhizomorphs, but in the fall of 2015 and of 2016, samples also included fruit-body tissues which were used directly in DNA analysis. Subsequent methods generally followed Anderson and Catona13.
Culturing and DNA extraction
Rhizomorphs were cut into 2 cm segments and placed in 2.5% hypochlorite bleach for ten minutes to surface disinfect. The rhizomorphs were then trimmed to less than 1 cm and placed on 2% malt-extract agar medium. Liquid cultures were in 2% malt extract without agar. Mycelium was harvested, flash frozen in liquid N2 and then lyophilized. DNA was extracted with a CTAB-low salt CTAB precipitation method as described earlier29.
Somatic compatibility testing
To determine whether or not a collection represented C1, we tested for somatic compatibility (Supplementary Table 1), which distinguishes the ability of growing mycelia of the same individual to merge seamlessly in culture and mycelia of different individuals to react with a zone of cell death and pigmentation30. We noted one large grouping of 110 isolates that were later confirmed to represent C1 and a smaller grouping of eight isolates that matched the C2 described earlier1,3. C1 and C2 are still centered on the same respective localities as reported in 19928.
Initial genotyping
Polymorphic molecular markers were also used to test whether a new collection represented the C1 identity or not (Supplementary Table 1). For example, in one segment of DNA, homozygosity for absence of a MboI site has a frequency of 0.64 in the general population12. In addition, the 3’ end of the 25S rRNA gene is heterozygous in C1 for a length polymorphism, a genotype that has a frequency of 0.21 in the population. The frequency of the combined genotype of the C1 individual over the two DNA regions is 0.13. The combination of the two DNA regions and somatic compatibility drives the probability of a spurious match in a non-C1 individual much lower.
Illumina sequencing
HiSeq Illumina sequencing was by paired-end with 155 bp reads at the Centre for Applied Genomics at the Hospital for Sick Children, Toronto. The Illumina sequences for 15 collections of C1 are deposited as accession PRJNA393342 in the SRA at NCBI.
Bioinformatics
The pipeline mapped the raw fasta files onto the reference genome, produced the pileup files from the resulting Bam files, and then filtered the pileup files to discover the variation among the Illumina sequenced strains. The pipeline is described in detail in Supplementary Bioinformatics 1.
Contributions
J.N.B., J.B.A., and M.L.S. planned the study and did the field collecting. D.K. did the laboratory culturing and somatic compatibility tests, DNA extractions, PCR, Sanger sequencing, and preparation of DNA for Illumina sequencing. N.R. and H.W. did the bioinformatic analysis of the Illumina sequences and subsequent nuclear SNP discovery. J.B.A. did the analysis of mtDNA variation. J.N.B., J.B.A., and M.L.S. wrote the paper with input from D.K., N.R., and H.W.
Competing interests
The authors declare no competing financial interests.
Acknowledgements
J.B.A, M.L.S. and N.R. were supported by Discovery Grants from the Natural Sciences and Engineering Research Council of Canada.