Host Selection
Host choice theory predicts that N-fixing plants should choose the best-performing symbiont types from those available (Denison, 2000; Kiers & Denison, 2008; Simms & Taylor, 2002), and evolutionary constraints suggest that symbiont quality should trade off with the ability to survive and/or reproduce in the soil environment if the cost of maintaining both abilities is high (Roff & Fairbairn, 2007). Based on these considerations, we expected: 1) relative abundance ofFrankia genotypes in soil to be independent from that in nodules, both within and among successional stages, 2) nodulating genotypes to be more abundant in soils beneath their host species than in areas lacking hosts, and 3) up-regulation of nodulation via P-fertilization to increase abundance of symbiotic types relative to non-symbiotic, and down-regulation via N-fertilization to do the opposite. Our study yielded mixed results with respect to these expectations.
Our expectation that assemblages of symbiotic Frankia in soil and nodules would vary independently across succession was not met. In fact, the two assemblages were strikingly similar in all three successional stages we examined. Early succession soils were strongly dominated by sequences closely related to the dominant type in early succession nodules (‘RF7’), with a low proportion of sequences related to the dominant types in late succession nodules. In both mid- and late succession soils, the only symbiotic types we found were related to sequences we have found in mid- and late-succession nodules; the early-succession type – which we have never found in mid- or late-succession nodules – was also not detected in mid- or late-succession soil. Symbiont choice exerted by alder plants could produce this pattern via two opposite scenarios: 1) hosts do not exert choice, simply associating with symbiotic genotypes in proportion to their occurrence in soil, or 2) hosts do exert choice, which selectively amplifies the chosen genotypes’ representation in soils. While our study cannot distinguish between these two scenarios, we think that the bulk of evidence – from both our study and prior studies of theAlnus-Frankia system – favors the latter.
A great deal of circumstantial evidence supports the possibility of non-random selection of Frankia genotypes by alder species. A wide range of field studies has consistently reported genetic differences in Frankia occupying nodules collected from different alder species (Higgins & Kennedy, 2010; Lipus & Kennedy, 2011; Pozziet al ., 2018b), even when they occur in the same field site (Anderson et al ., 2009; Pokharel et al ., 2010; Simonetet al ., 1989). Bioassay studies of field soils have reported differences in nodulating Frankia between different Alnusspecies grown in the same soil (Ben Tekaya et al ., 2018; Lipus & Kennedy, 2011), and cross-inoculation studies using crushed nodules or isolates have indicated differences in compatibility between host species and specific Frankia strains (Du & Baker, 1992; Markham, 2008; Prat 1989). Recent microcosm studies have provided more direct evidence. Ben Tekaya et al . (2018) used qPCR and Illumina sequencing to compare the relative proportions of Frankiagenotypes in soils with those in associated nodules of three European and North American Alnus species. Nodule assemblages differed strongly among the two alder species that formed nodules in their soils, and also varied independently from soil assemblages. Importantly, soil assemblages were characterized after a short enough time (7 months) that nodule senescence driven feedback was unlikely to have contributed strongly to their result. Vemulapally et al . (2022a) conducted a very similar study, but included one additional North American alder species and a treatment in which equal mixtures of genetically distinctFrankia isolates were also inoculated into field soils. Both inoculated and indigenous genotypes grew to different densities in soils over a 3-month timeframe, but similar relative proportions of genotypes occurred across different host species. By contrast, nodules contained different mixtures of bacterial genotypes among alder species, and again the relative proportions in nodules were independent from those in associated soils. The selective ability of alder indicated by these lab results is supported by a cloning study of field-collected soils by the same group of researchers, in which alder nodules were found to contain high frequencies of some genotypes that were undetectable in soil clones (Mirza et al ., 2009c).
Selection of specific genotypes by host plants has strong potential to feed back to soil assemblages of N-fixing bacteria (West et al ., 2002). Density of bacterial cells in nodules is many orders of magnitude greater than in surrounding soils (Denison 2000; Denison & Kiers, 2004), and nodule-dwelling bacteria maintain connections with soil assemblages via extra-nodular extensions of hyphae (Diem et al ., 1982) or infection threads (Denison 2000), or release of viable cells (Denison 2000) or spores (Pozzi et al ., 2015) during nodule senescence. This host-enhanced density of symbiotic genotypes is thought to be an important factor that maintains mutualistic traits in soil assemblages against the selective advantages of cheating (Denison, 2000; Denison & Kiers, 2004; West et al ., 2002). In Frankia , such enhancement could also help overcome any tradeoffs that may occur between symbiotic and free-living traits. Previous authors have suggested that such tradeoffs are not likely to be important in rhizobial symbioses, because of the facultative nature of N-fixation in these bacteria (Denison & Kiers, 2004). Unlike rhizobia, however,Frankia appears to fix N in the free-living state (Pawlowski & Sprent, 2008), so that traits that enhance N-fixation in symbiosis may trade off with those that optimize non-symbiotic reproduction and N-fixation. Additionally, tradeoffs may occur between soil and nodule-dwelling traits more generally if, for example, survival in soil requires a broader suite of metabolic capabilities than the more specialized nodule-dwelling lifestyle. Rhizobial densities have been observed to increase during nodule senescence (Brockwell et al ., 1987 in Denison & Kiers, 2011), and nodulating capacities of soil-dwelling Frankia often increase near host plants (Chaiaet al ., 2010), providing circumstantial evidence of the importance of nodule feedback in maintaining symbiotic populations.
In our study, enhancement of symbiotic assemblages by hosts is suggested by two specific results. First, the proportion of total clones yielding symbiotic genotypes found in rhizosphere soils decreases monotonically from early (0.346) to mid (0.233) to late-succession (0.061). This pattern parallels both the density of host plants and the density of nodules within host rhizospheres across succession, indicating a correspondence between the availability of host interactions and the relative representation of symbiotic genotypes in soil assemblages. Second, this trend continues in the frequency of AT clade genotypes, particularly compared with non-symbiotic ones, in host rhizospheres (proportion of symbiotic genotypes = 0.061) versus non-alder soils (0.035) in late succession. Interestingly, the highest frequency of typical Alnus genotypes observed in our study also occurred in late succession non-alder soil. We have only rarely observed this group in A. tenuifolia nodules, which raises the intriguing possibility that non-host-specific symbiotic types may be inhibited in rhizospheres of this alder species, in addition to enhancement of specific nodule-forming genotypes. This speculation, of course, will require further evidence to confirm.
Our most surprising finding, compared with a priori expectations, occurred in response to long-term fertilization in the mid-succession site. Addition of N and P have well-known effects on nodule formation by alders that are opposite in direction (Wall & Berry, 2008). In our study, nodule formation differed among treatments in the expected direction: P significantly increased the number of nodules per unit area (4631 ± 1051 clusters m-2), and N significantly decreased it (346 ± 159 clusters m-2), compared with unfertilized control plots (911 ± 199 clusters m-2) (Ruess et al . 2013). Based on the relative differences in opportunities for host interactions and soil feedback provided by these two responses to fertilization, we expected P to enhance host-selected genotypes in host-associated soils, particularly compared with non-symbiotic types, and N to decrease this proportion. Consistent with the former expectation, the proportion of total clones yielding symbiotic genotypes was nearly double in soils under long-term P fertilization (0.512) compared to the value in control soils (0.233). Contrary to expectation, however, it reached its highest value (0.843) under N-fertilization.
In a system as complex as soil, the number of hypotheses we could propose to explain this unexpected result is nearly infinite. A useful distinction between kinds of explanation is the dichotomy between mechanisms mediated by hosts and those that are not. The size of the fertilization effect, together with the fact that symbiotic types were disproportionately enhanced, would seem to implicate host-mediated mechanisms. Two non-exclusive processes seem most likely to be involved. First, addition of N may have produced an influx of nodule-dwellingFrankia into surrounding soil by triggering large-scale nodule senescence. There is some evidence that this may have occurred: live nodule biomass was lower in N-fertilized (21.0 ± 3.4 g/m2) than control plots (33.5 ± 5.9 clusters/m2), and dead biomass was higher under N-fertilization (6.7 ± 1.8 g/m2) than in control (4.3 ± 1.2 g/m2) plots. Though neither difference was statistically significant, it is plausible that they could have amounted to a difference in input of nodule-dwelling genotypes. The latter difference could have amplified the difference in nodule biomass changes, particularly in response to our second hypothetical process: N-enhanced growth of host plants. Though we did not measure any growth-related variables in our host plants, mineral N is known to increase growth in seedlings of several alder species under laboratory conditions, even when the seedlings are symbiotically fixing N (Ingestad, 1980; Radwan, 1987; Stewart & Bond, 1961). Even as nodulation continued to decrease over subsequent years, any initial influx of symbiotic genotypes from senescent nodules could have been further enhanced in soil by, for example, plant growth-stimulated production of leaf litter (as observed by Mirza et al . 2007; 2009a) and/or unidentified host-associated compounds (as observed by Mirza et al ., 2007; Samant et al ., 2015; 2016), both of which have been observed to preferentially enhance soil-dwelling populations of specific symbiotic Frankia genotypes. Obviously, this scenario is speculative and will require further study to confirm. Overall, our results indicate that the transition between soil and nodule-dwelling states is likely to be a significant contributor to the eco-evolutionary complexity that increasingly appears to define these symbiotic interactions.