Phylogenetic Diversity of Nodule- and Soil-Dwelling Frankia
Our phylogenetic results broadly match previous studies ofFrankia using a variety of single and multiple loci (e.g., Ghodhbane-Gtari et al ., 2010; Normand et al ., 1996; Nouioui et al ., 2011; Pozzi et al . 2018a; Swensen & Benson, 2008). The four major host infection ‘Clusters’ described in prior studies are evident in our tree, though their relative positions differ slightly from the most comprehensive phylogenetic characterizations (Pozzi et al . 2018), a common result in single locus studies. The main ambiguity it adds to our results is whether our ‘Clade A’ is more closely related to Alnus -infective orElaeagnus -infective Frankia . The rest of our clone-derived sequences fell well within previously-described Alnus -infectiveFrankia , and formed distinct, well-supported clades based on whether or not they had been previously observed in nodules and, if so, which host species. Clades A and B contained sequences we have not found in nodules of either alder species native to our study area. Because of the rarity of alternative actinorhizal plants known to be compatible with Alnus -infective Frankia in our study sites (Hollingsworth 2022), we think these sequences are likely to represent free-living, non-symbiotic Frankia . In all of our unmanipulated plots, sequences from these clades were found in a clear majority of clones, ranging from 65.4% (early succession alder soil) to 87.7% (late succession alder soil) of all clones screened. This result is similar to the only two prior surveys of which we are aware that used DNA cloning to examine both nodule and soil-dwelling Frankia from the same field-collected soils (Mirza et al . 2009c; Pokharelet al . 2010). Mirza et al (2009c), in a sample of 247 clones derived from soils sampled on five continents, found that most soil-derived sequences were distinct from sequences found in nodules ofMyrica pennsylvanica trap plants grown in the same soil. Pokharelet al . (2010), found that soils from the rhizosphere of oneA. glutinosa individual yielded sequences that did not match any of those derived from nodules of 12 Alnus species and sub-species growing in the same site. Collectively, these results suggest a high abundance of non-symbiotic Frankia in soils.
Our results further suggest that non-symbiotic Frankia may be much more diverse than symbiotic genotypes in our sites. Our cloning method – optimized to detect sequences matching those we have found in host nodules – captured a diversity of sequences that we have never observed in nodules but clearly clustered with Frankia in our phylogenetic analysis. While PCR and DNA cloning are subject to well-known biases and artifacts (e.g., Acinas et al ., 2005; Chandler et al ., 1997; Sipos et al ., 2010; van Elsas & Boersma, 2011), we think this result is unlikely to be primarily artifactual, for three reasons. Firstly, the size of the difference is unlikely to be due to diversity-inflating sequence artifacts alone. Estimates of spurious sequence production rates in multi-template PCR due to copying errors, heteroduplexes and chimeras tend to be ≤20% of the total estimated sequence diversity (Acinas et al ., 2005; Qiuet al ., 2001; Schloss et al ., 2011; Speksnijder et al ., 2001). Depending on how we defined OTU clusters, richness of non-symbiotic OTUs was between ~3x and ~10x greater than for symbiotic OTUs in our data (Figure S6), much greater than known rates of spurious sequence production. Secondly, our soil clones yielded representative sequences closely related to all of the major phylogenetic clusters we have observed in nodules of both alder species present in our sites. This suggests that sources of bias in our methods – at least those that diminish our ability to detect specific genotypes – were collectively weak. Finally, we implemented several recommended precautions against known biases, including bead beating soils prior to DNA extraction (de Lipthayet al ., 2004), use of a high-performing DNA extraction kit (İnceoǧlu et al ., 2010) supplemented with lysozyme to enhance lysis of gram-positive bacteria (Robe et al ., 2003), a balance between low enough [template DNA] to dilute PCR-inhibiting soil compounds and high enough to minimize bias due to random priming (Chandler et al . 1997), use of degenerate primers to target a broader spectrum of template sequences, a high annealing temperature during PCR to enhance primer specificity (Sipos et al ., 2010), testing of our selected cloning kit for phylogenetic bias (Tayloret al ., 2007), detection and removal of sequence chimeras (which were rare), and analysis of sequence diversity by clustering into OTUs at multiple levels of similarity (Acinas et al ., 2005).
Comparison of our diversity results with prior molecular studies of soilFrankia is difficult, due to differences in study systems, size and geographic scale of samples, and methodology among the few studies available. Nevertheless, our results seem to contrast with prior observations. Pokharel et al . (2010) found restricted diversity of Frankia in soil compared to nodules from the same site, but only examined a single soil sample from beneath one of the twelve sympatric Alnus taxa included in their nodule sample. Mirzaet al . (2009c) found similar diversity of soil Frankia andFrankia forming root nodules on trap plants of the ‘promiscuous’ host Morella pensylvanica in six soils collected from five continents. Rodriguez et al . (2016), examining soils from three widely different ecosystems, only observed 17 unique sequences across more than 86,000 individual Illumina sequencing reads. Clustering based on 97% similarity yielded eight OTUs found in their soil samples, four of which were only found in soil, and four of which matched OTUs from reference cultures produced from host nodules.
Methodological factors could underlie the differences between these studies and ours. All three prior studies utilized nifH, a non-neutral locus responsible for encoding one of the subunits of nitrogenase, the N2-fixing enzyme complex. The primers they used were developed using Frankia isolates (Normand et al . 1988), and have been used very commonly on nodules (e.g., Higgins & Kennedy, 2012; Kennedy et al ., 2010a,b; Lipus & Kennedy, 2011; Mirzaet al . 2009b,c; Pokharel et al ., 2010; Polme et al ., 2014; Welsh et al ., 2009a,b), but were not redesigned for the possibility of wider diversity in soils. By contrast, the non-coding intergenic spacer we targeted is likely to vary more widely amongFrankia due to its selective neutrality (Rocha 2018), and the degeneracy we incorporated into our primers is likely to have targeted a broader assemblage of Frankia . Additionally, Mirza et al . (2009c) and Rodriguez et al . (2016) utilized a nested PCR protocol, reducing the amount of information available for sequence analysis from 606 bp to approximately 260 bp per sequence.
While methodological factors probably help explain the difference in observed diversity between ours and prior studies, there are both empirical and theoretical reasons to expect soil-dwelling assemblages to have higher diversity than nodule-dwelling assemblages. Among symbiotic genotypes, diversity in soil has been observed in a large number of bioassays to be higher than in nodules of any single host species, due to specificity of host and symbiont associations (e.g., Mirza et al ., 2009b). For this reason, it is standard practice for such studies to utilize multiple host species in order to capture as wide a diversity as possible from soil (Chaia et al ., 2010; McInnes et al ., 2004). Diversity of free-living genotypes should also be higher than that of endosymbiotic types, based on several evolutionary considerations. Firstly, based on the relative evolutionary ages of plant hosts and symbiotic bacterial taxa, the free-living lifestyle is likely to be ancestral to symbiosis (e.g., Normand et al ., 1996; Sachs et al ., 2014), so symbiotic lineages should, all else being equal, tend to be nested within deeper non-symbiotic lineages. If the rate of transition to symbiosis is less than or equal to the diversification rate of free-living types, and rates of extinction are similar, then the phylogenetic process will also produce a broader range of free-living lineages. In our data, the nesting of soil-derived and nodule-derived sequences relative to each other is unclear, since the branches involved had weak statistical support, and/or occurred in the portion of our tree in which placement of large clusters did not match prior studies. However, our putatively non-symbiotic clades were clearly deeper than symbiotic ones (Figure 1, Figure S2). Secondly, soil habitats almost certainly contain a wider range of environmental conditions than the environment within host nodules, providing more opportunities for niche differentiation among free-living than symbiotic lineages. Finally, selection imposed by hosts should exert strong purifying selection on symbionts, restricting phenotypic and genetic diversity among symbionts (Denison & Kiers, 2004), although horizontal transfer of symbiotic genes complicates this expectation (e.g., Epstein & Tiffin, 2021).