Network structure of resource use and niche overlap within the endophytic microbiome

Endophytes often have dramatic effects on their host plants. Characterizing the relationships among members of these communities has focused on identifying the effects of single microbes on their host, but has generally overlooked interactions among the myriad microbes in natural communities as well as potential higher-order interactions. Network analyses offer a powerful means for characterizing patterns of interaction among microbial members of the phytobiome that may be crucial to mediating its assembly and function. We sampled twelve endophytic communities, comparing patterns of niche overlap between coexisting bacteria and fungi to evaluate the effect of nutrient supplementation on local and global competitive network structure. We found that, despite differences in the degree distribution, there were few significant differences in the global network structure of niche-overlap networks following persistent nutrient amendment. Likewise, we found idiosyncratic and weak evidence for higher-order interactions regardless of nutrient treatment. This work provides a first-time characterization of niche-overlap network structure in endophytic communities and serves as a framework for higher-resolution analyses of microbial interaction networks as a consequence and a cause of ecological variation in microbiome function.


Hypervolume overlap
In this section, we explore an alternative measure of niche overlap, in which we treat each isolate's growth on the 95 nutrients as a (maximally) 95-dimensional resource niche, where ⃗ g is a vector of optical densities,m signifies a resource upon which isolate i had greater than 0 growth, and M signifies the total number of such resources. We then measure interactions by looking at the overlap of the partner isolate's hypervolume by one bounded in each direction by the lesser optical density between isolate i and the partner isolate j, Ω hv i→j = η (min 1≤n≤N (g i,n , g j,n )) η( ⃗ g j ) . Outdegree Indegree Figure S1: As Figure 3, but utilizing a hypervolume measure of niche overlap. Links are assigned when an isolate has greater than 82 % niche overlap with another isolate. Points do not fall exactly on integers due to slight jittering to improve readability. Colored points indicate the kingdom of the partner isolate: orange indicates interactions between kingdoms, while within kingdom links are colored according to the kingdom, using the same colors as in Figure 2 (pink for fungi, blue for bacteria). Black points indicate values when partner's kingdom is ignored. Thus, there are three points for each isolate in each panel: one for their interactions with other members of their same kingdom (blue or pink for bacteria or fungi, respectively), one for interactions with the alternate kingdom (orange), and one for their total number of interactions (black). Lines indicate best fitting linear models for each subset of the data. Slopes are significantly different from 0 in all cases except for the Bacteria → Bacteria interaction in the control treatment (blue points in top-left panel; Table S3).  Figure S1). This analysis utilizes a hypervolume measure of niche overlap and binary interaction strengths. An analysis of variance in these slopes is presented in Table S4.

Vector projection
In this section, we explore an alternative measure of niche overlap, in which we treat each isolate's growth on the 95 nutrients as a vector in 95-dimensional space, whose magnitude can be calculated as We then measure interactions by looking at how much an isolate's resource use vector is overlapped by a projection of another isolate's resource use vector into the same direction, Links are assigned when an isolate has 100 % niche overlap with another isolate. Points do not fall exactly on integers due to slight jittering to improve readability. Colored points indicate the kingdom of the partner isolate: orange indicates interactions between kingdoms, while within kingdom links are colored according to the kingdom, using the same colors as in Figure 2 (pink for fungi, blue for bacteria). Black points indicate values when partner's kingdom is ignored. Thus, there are three points for each isolate in each panel: one for their interactions with other members of their same kingdom (blue or pink for bacteria or fungi, respectively), one for interactions with the alternate kingdom (orange), and one for their total number of interactions (black). Lines indicate best fitting linear models for each subset of the data. All slopes are significantly different from 0 (Table S5).  Figure S2). This analysis utilizes a vector projection measure of niche overlap and binary interaction strengths. An analysis of variance in these slopes is presented in Table S6.  Outdegree Indegree Figure S3: As Figure 3, but incorporating weighted interactions. Colored points indicate the kingdom of the partner isolate: orange indicates interactions between kingdoms, while within kingdom links are colored according to the kingdom, using the same colors as in Figure 2 (pink for fungi, blue for bacteria). Black points indicate values when partner's kingdom is ignored. Thus, there are three points for each isolate in each panel: one for their interactions with other members of their same kingdom (blue or pink for bacteria or fungi, respectively), one for interactions with the alternate kingdom (orange), and one for their total number of interactions (black). Lines indicate best fitting linear models for each subset of the data. Slopes are significantly different from 0 in all cases except for the Bacteria → Bacteria interaction in the control treatment (blue points in top-left panel; Table S7). Removal of the outlying control network (see main text) results in all slopes being significant and a sign reversal in the case of Fungi → Bacteria interactions (orange points in top-right panel) Table S7: Linear model results for indegree by outdegree in each sympatric network, differentiated according to focal and partner isolates' Kingdom (each unique combination of color and panel in Figure S3). This analysis utilizes an average pairwise measure of niche overlap and weighted interaction strengths. An analysis of variance in these slopes is presented in Table S8. Table S8: Summary of an analysis of variance for linear models relating isolate outdegree to indegree using an average pairwise measure of niche overlap and weighted interaction strengths (reported in Table S7).  Colored points indicate the kingdom of the focal isolate: orange indicates interactions between kingdoms, while within kingdom links are colored according to the kingdom, using the same colors as in Figure 2 (pink for fungi, blue for bacteria). Black points indicate values when the partner's kingdom is ignored. Thus, there are three points for each isolate in each panel: one for their interactions with other members of their same kingdom (blue or pink for bacteria or fungi, respectively), one for interactions with the alternate kingdom (orange), and one for their total number of interactions (black). Lines indicate best fitting linear models for each subset of the data. All slopes are significantly different from 0 except for Fungi → Bacteria interactions in the control treatment (orange points in top-right panel; Table S9). Removal of the outlying control network (see main text) results in all slopes being significant.  Figure S4). This analysis utilizes a vector projection measure of niche overlap and weighted interaction strengths. An analysis of variance in these slopes is presented in Table S10.  Table S10: Summary of an analysis of variance for linear models relating isolate outdegree to indegree using a vector projection measure of niche overlap and weighted interaction strengths (reported in Table S9).  Colored points indicate the kingdom of the partner isolate: orange indicates interactions between kingdoms, while within kingdom links are colored according to the kingdom, using the same colors as in Figure 2 (pink for fungi, blue for bacteria). Black points indicate values when the partner's kingdom is ignored. Thus, there are three points for each isolate in each panel: one for their interactions with other members of their same kingdom (blue or pink for bacteria or fungi, respectively), one for interactions with the alternate kingdom (orange), and one for their total number of interactions (black). Lines indicate best fitting linear models for each subset of the data. Slopes are significantly different from 0 in all cases except for the Fungi → Bacteria interaction in the control treatment (orange points in top-right panel; Table S11). Removal of the outlying control network (see main text) results in all slopes being significant and a sign reversal in the case of Bacteria → Bacteria interactions (blue points in top-left panel).

Bacteria
Table S11: Linear model results for indegree by outdegree in each sympatric network, differentiated according to focal and partner isolates' Kingdom (each unique combination of color and panel in Figure S5). This analysis utilizes a hypervolume measure of niche overlap and weighted interaction strengths. An analysis of variance in these slopes is presented in Table S12.    Table S7). Removal of the outlying control network (see main text) results in the two former cases becoming significant as well.

Alternative cutoff for binary link definition
Table S13: Linear model results for indegree by outdegree in each sympatric network, differentiated according to focal and partner isolates' Kingdom (each unique combination of color and panel in Figure S6). This analysis utilizes an average pairwise measure of niche overlap and binary interaction strengths. An analysis of variance in these slopes is presented in Table S14.   Table S15: As Table 5, but incorporating using an alternative cutoff of 0.5 to define link presence/absence. Summary test for higher-order interactions in endophytic microbial networks. Empirical p values (left) and z scores (right) for network structure metric comparisons between empirical and randomly-rewired networks. Each column represents an individual leaf, ordered as in Figure 2. Values less than 0.05 (−2) signify that the empirical value is significantly smaller than expected, and are represented by  ( for values less than 0.01 (−4)). Likewise, values greater than 0.95 (2) signify that the empirical value is larger than expected and are represented by  ( for values greater than 0.99 (4)). Dashes (−) signify values between 0.05 and 0.95 (−2 and 2 ; i.e. non-significant differences) and  indicates cases where all randomizations resulted in the same value for this metric/community combination. Note that, for this cutoff, all networks had sufficient unique configurations to be included.  Table S7). Note also that there were no Bacteria → Fungi interactions in the control treatment.
Table S16: Linear model results for indegree by outdegree in each sympatric network, differentiated according to focal and partner isolates' Kingdom (each unique combination of color and panel in Figure S7). This analysis utilizes an average pairwise measure of niche overlap and binary interaction strengths. An analysis of variance in these slopes is presented in Table S17.  Table S17: Summary of an analysis of variance for linear models relating isolate outdegree to indegree using an average pairwise measure of niche overlap and binary interaction strengths (reported in Table S16).  Table S18: As Table 5, but incorporating using an alternative cutoff of 0.875 to define link presence/absence. Summary test for higher-order interactions in endophytic microbial networks. Empirical p values (left) and z scores (right) for network structure metric comparisons between empirical and randomly-rewired networks. Each column represents an individual leaf, ordered as in Figure 2. Values less than 0.05 (−2) signify that the empirical value is significantly smaller than expected, and are represented by  ( for values less than 0.01 (−4)). Likewise, values greater than 0.95 (2) signify that the empirical value is larger than expected and are represented by  ( for values greater than 0.99 (4)). Dashes (−) signify values between 0.05 and 0.95 (−2 and 2 ; i.e. non-significant differences) and  indicates cases where all randomizations resulted in the same value for this metric/community combination. Note that, for this cutoff, all of the control leaves and two of the NPK leaves (N1 and N6) were omitted from this analysis due to a lack of unique network configurations.  Table S20: χ 2 -test results comparing the distribution of isolate kingdom across node groupings found using a spinglass algorithm (57). A filled box signifies a significant p value, i.e. a case where the distribution of kingdoms across groups is significantly non-uniform, with the intensity of the color indicating the α level of the significance □ and □ corresponding to < 0.001 and < 0.01, respectively). The fifth binary control network did not divide into two groups with more than one isolate in each, and thus was ineligible for a χ 2 analysis.

Control NPK Supplemented
Binary   Figure S10: Indegree by outdegree for each isolate (similar to Figure 3), colored according to the class of the focal isolate. Lines indicate linear model fits to isolates within each class. Note that three classes (Alphaproteobacteria, Betaproteobacteria, and Agaricomycetes) were represented by a single isolate each, and thus do not have associated linear models (but the points are still presented).