Mycorrhizal dominance reduces forest-tree species diversity

Ectomycorrhizas and arbuscular mycorrhizas, the two most widespread plant-fungal symbioses, are thought to differentially influence tree species diversity, with positive plant-soil feedbacks favoring locally abundant ectomycorrhizal tree species and negative feedbacks promoting species coexistence and diversity in arbuscular mycorrhizal forests. While seedling recruitment studies and cross-biome patterns of plant diversity and mycorrhizal dominance support this hypothesis, it remains to be tested at the forest stand level over continental scales. Here, we analyze ∼85,000 forest plots across the United States to show that both ectomycorrhizal-dominated and arbuscular mycorrhizal-dominated forests show relatively low tree diversity, while forests with a mixture of mycorrhizal strategies support a higher number of tree species. Our findings suggest that mycorrhizal dominance, rather than mycorrhizal type, shapes tree diversity in forests.


INTRODUCTION 28
Mycorrhizas -the most widespread terrestrial symbiosis on Earth -have long been 29 known for their nutritional benefits to plants 1  conspecific seedlings from soil-borne pathogens and improved nutrient acquisition, 40 relative to AM forests 6,9 . However, we do not know whether these short-term effects 41 on recruitment dynamics translate into persistent effects on canopy tree species 42 composition and diversity. Indeed, neither the historical biome-level observations nor 43 the individual-level studies of seedling recruitment directly test the hypothesis that 44 EcM-dominated forests sustain lower tree species diversity than AM-dominated 45 forests; broad-scale analyses at the forest-tree community level are needed to resolve 46 this. 47 4 and tree species diversity declined under increasing dominance by AM trees (Fig. S4; 74 See "the mycorrhizal mixture hypothesis" in Fig. 2A). As such, tree species diversity 75 was lowest in forests dominated by either the EcM or AM strategies, and highest 76 when there was an approximate mixture of both strategies (Fig. S3, S4). 77

78
Mycorrhizal distributions are known to be correlated with environmental factors that 79 also influence plant diversity 12 , but the pattern in the bivariate relationship remained 80 strong after controlling for environmental variables (Fig. 2B). In models including 81 effects of local abiotic factors (climatic, topographic and physiographic properties), 82 we found that tree diversity was influenced by these factors, especially temperature, 83 topography and water availability, but the negative effects of mycorrhizal dominance 84 on tree diversity were strongest (Fig. 2B, Fig. 3). 85

86
Our results are also robust to the potential statistical artefact that forest plots with only 87 one mycorrhizal type have lower potential diversity (i.e., only tree species of that type 88 can be present). Local plant diversity is determined by environmental filtering from 89 the regional flora 13 , and species diversity depends on the size and composition of the 90 regional species pool 14 . As such, we applied a null model to determine if mycorrhizal 91 dominance had detectable effects on tree species diversity beyond what could be 92 explained by the tree species composition of the regional flora. We assessed the 93 expected relationship between tree diversity and the proportion of EcM tree basal area 94 based on random sampling from the regional tree species pool, from which we 95 calculated the deviation between observed and expected values for each plot (which 96 we call "alpha-deviation"). The null model re-assigned a species identity to each 97 individual tree in a given plot based on a random draw of species of the same 98 mycorrhizal type (simplified as either EcM or "other" strategies) from the regional 99 pool, with probability weights proportional to species regional abundances. Our null 100 model thus preserved the value of EcM proportion in each plot. Regional pools were 101 defined within each of 25 ecoregions 15 , which represent geographical areas with 102 relatively similar ecological and environmental conditions (Fig. S5). Results of our 103 null model analysis showed that the hump-shaped relationship with tree species 104 diversity being lowest at low or high EcM basal area persisted (Fig. 2C, Fig. S6). This 105 means that the lower local tree species diversity observed in plots dominated by either 106 the EcM or AM strategy was not the result of the regional species pool containing a 107 smaller number of tree species from either one of these strategies, but rather an 108 outcome of local processes reducing species diversity in EcM-or AM-dominated 109 plots. 110

DISCUSSION 112
Our results strongly suggest that dominance by either EcM or AM strategy, and not 113 only the EcM strategy type, reduces local tree diversity across the forested United 114 States. Several mechanisms involving mycorrhizal type may combine locally to 115 influence plant diversity 2 . First, positive plant-soil feedbacks commonly reported for 116 EcM species at the seedling recruitment stages 7,8 , could also apply to AM species, but 117 at later life stages (e.g. saplings or small sub-canopy trees), eventually leading to 118 canopy dominance by certain AM tree species (Fig. S7). While it is widely accepted 119 that EcM plants may benefit more from EcM fungi in terms of mineral nutrition and 120 protection against root pathogens 9 , the higher maintenance costs of EcM fungi 121 compared to AM fungi could mean that the net benefits of the two mycorrhizal types 122 are similar 16 , thereby equalizing fitness differences among strategies. Furthermore, 6 fine-scale niche partitioning could promote coexistence of different mycorrhizal 124 types 17 and ecosystems with a mixture of mycorrhizal strategies may also create 125 environments that are more diverse and spatially heterogeneous 18 . Together, these 126 processes could locally promote diversity where multiple plant nutrient-acquisition 127 strategies such as mycorrhizal types co-occur 8,19 . Following the "National hierarchical framework of ecological units" 15 we defined 25 217 ecoregions and assigned one for each plot depending on its location (Fig. S5). 218 Ecological units are defined as areas of similar surficial geology, lithology, 219 geomorphic processes, soil groups and subregional climate. 220 221

Null model 222
We used a null model that re-assigned the species identity of the individuals in each 223 plot based on random draws from the regional pool of tree species within ecoregions 224 (Fig. S5), while keeping the total number of individuals per plot and the proportion 225 EcM constant. Each species' abundance (i.e., its probability of being chosen by the 226 null model) was calculated as the number of tree stems in the ecoregion, divided by 227 the total number of stems across species. We ran 100 randomizations from which we 228 calculated the diversity "deviation" (or "corrected" diversity) as the observed diversity