Genetic architecture of heritable leaf microbes

25

26 Host-associated microbiomes are shaped by both their environment and host genetics, and often 27 impact host performance. The scale of host genetic variation important to microbes is largely 28 unknown, yet fundamental to the community assembly of host-associated microbiomes, and with 29 implications for the eco-evolutionary dynamics of microbes and hosts. Using Ipomoea 30 hederacea, Ivy-leaved morning glory, we generated matrilines differing in quantitative genetic 31 variation and leaf shape, which is controlled by a single Mendelian locus. We then investigated 32 the relative roles of Mendelian and quantitative genetic variation in structuring the leaf 33 microbiome, and how these two sources of genetic variation contributed to microbe heritability. 34

35 We found that despite large effects of the environment, both Mendelian and quantitative genetic 36 host variation were important in contributing to microbe heritability, and that the cumulative 37 small effect genomic differences due to matriline explained as much or more microbial variation 38 than a single large effect locus. Furthermore, our results are the first to suggest that leaf shape 39 itself contributes to variation in the abundances of some microbes in the leaf microbiome. 40

41 Our results demonstrate the roles of different scales of host genetic variation in the assembly of a 42 natural microbiome. The genetic architecture of plant-associated microbiomes depends on both 43 quantitative genetic variation and Mendelian traits, with similar contributions to microbe 44 heritability. The presence of Mendelian and quantitative genetic variation for host associated 45 microbes suggests that evolution in plant hosts-- at the leaf shape locus, and other quantitative 46 genetic loci-- has the potential to affect microbiome distribution, abundance, and composition. 47 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 3

leaf shape, phyllosphere, Ipomoea hederacea, microclimate, host age, leaf 48 microbiome, heritability, matriline, Mendelian, quantitative genetics 49 50

51 The community assembly of microbes on a host depends on the environment, the host’s 52 traits, and the microbes [1]. Microbes that colonize leaves and other aboveground plant parts, 53 known as the phyllosphere, can disperse to a leaf via air [2], rain [3], or soil [4], after which they 54 experience selection due to conditions on the leaf surface and microbe-microbe interactions [1, 55 5]. Host genetics play a role in the leaf microbiome; the similarity of leaf microbiomes across 56 species depends on host phylogenetic relatedness [6, 7], and the similarity of microbiomes 57 between individuals of the same species can depend on host intraspecific genetic variation [8–58 11]. Here we test how a single leaf shape gene affects leaf bacterial communities in Ipomoea 59 hederacea, and compare the magnitude of effect from that single locus to that of many small 60 effect loci. 61 The genetically-based physiological, morphological, or immune traits of hosts can lead to 62 consistent associations between the abundance or community composition of microbes and host 63 genotypes [12–15]. When consistent host-microbe associations are present, the host’s microbial 64 phenotype can be considered a heritable host trait, even when the microbes are environmentally-65 acquired [16–18]. In other words, horizontally-transmitted microbes can be heritable in host 66 populations when host traits have predictable effects on microbial recruitment and alleles for 67 these host traits are transmitted from host parents to host offspring. The heritability of a host’s 68 microbial phenotype is important to understanding host evolution and ecology due to microbial 69 effects on host fitness [19–21], and potential reciprocal selection between hosts and microbiomes 70 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 4 [17, 22, 23]. The variation in microbe heritability may have interesting evolutionary 71 repercussions on hosts, given that the phenotypic variance of many host traits can be microbially 72 mediated [17, 18]. Furthermore, if a host’s microbial phenotype is due to heritable host genetic 73 variation, then the evolutionary forces affecting those host traits-- selection, drift, mutation, 74 migration, non-random mating-- can in turn affect microbial abundances, distributions, and 75 microbiome composition [24]. Like all heritable traits, host microbial phenotypes can be 76 influenced by few loci of large effect or the cumulative effects of many small effect loci, and 77 understanding microbe heritability necessitates we determine the scale of host variation that 78 matters to microbes. 79 Mechanisms producing genotype-specific phyllosphere microbiomes are complex and 80 varied. In Arabidopsis thaliana, a genome-wide association study of the phyllosphere microbial 81 community found that loci related to defense and plant cell wall integrity affect microbial 82 community variation, while species richness was affected by loci involved with viral 83 reproduction, trichomes, and morphogenesis [9]. Host genotypes may differ in immune genes 84 and disease resistance, which have been linked to differences in switchgrass leaf fungal 85 community structure [25] and inconsistent effects on bacterial and fungal maize phyllospheres 86 [11]. Plant genotypes can also vary in leaf morphology and other leaf attributes; for example, 87 mutations in cuticle formation and ethylene production also affected the microbiome in a 88 synthetic A. thaliana phyllosphere [26]. The microclimate of a leaf is determined by many leaf 89 properties including surface temperature, surface area, thickness of the boundary layer of air, 90 chemical composition, trichomes, gas exchange, and nutrients [27]. The boundary layer of air is 91 closest to the leaf surface and has warmer, humid air; non-segmented leaves with larger surface 92 area often have a thicker boundary layer, and thicker boundary layers can impede gas exchange 93 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 5 from the leaves [28]. Compounds and water released by leaves during gas exchange can be an 94 important source of nutrients and resources for phyllosphere microbes [5], and they can impose 95 selection on which microbial species can establish. These findings suggest that leaf morphology, 96 and its effects on microclimates and resources, can potentially affect phyllosphere microbes. 97 Factors driving phyllosphere microbiome composition may depend strongly on the 98 environment and the interaction between environment and genotype, since the leaf microbiome 99 is primarily environmentally acquired [29]. Environmental factors like weather and seasonality 100 influence which microbes are available to disperse to leaves and with what frequency [29]. 101 Phyllosphere microbes may differ over time due to changing moisture and temperature 102 conditions abiotically filtering for microbes in the air and soil [30], and biotic interactions that 103 emerge at different times, e.g. changing microbe-microbe interactions and new microbes from 104 insects [31] or other plants [32]. Determining the relative roles of environment, phenology, and 105 plant genotype on the microbial community composition and abundance in the phyllosphere is an 106 unresolved empirical challenge. 107 One unstudied potential source of host genotype influence on the leaf microbiome is leaf 108 shape. Leaf shape is a key morphological difference between plant species [33] and could 109 contribute to interspecific variation in leaf microbiomes. The shape of a leaf determines the 110 temperature and humidity conditions it experiences and even its photosynthetic and gas exchange 111 abilities [34, 35], potentially affecting microbial communities. Comparing the effects of leaf 112 shape on the microbiome between plant species is confounded by other genetic and ecological 113 differences between taxa, and for this reason, it is preferable to use a single species that exhibits 114 intraspecific variation in leaf shape. Ipomoea hederacea, Ivy-leaved morning glory, is a 115 flowering annual plant whose leaf shape is determined by a single Mendelian locus [36–38]; 116 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 6 homozygous genotypes are either fully lobed or whole, with heterozygotes being partially lobed 117 (Figure 1). In I. hederacea, homozygous lobed genotypes have been shown to more protected 118 against extreme temperature changes compared to whole leaves; homozygous lobed leaves 119 remained warmer at night by a mean difference of 0.16-0.22°C, likely due to their altered 120 boundary layer resulting in leaves being more coupled with ambient air temperature [39, 40]. 121 The differences in Ipomoea hederacea leaf morphology, and its effects on temperature and 122 boundary layers that affect leaf microclimates, suggest that different leaf shape genotypes could 123 directly determine which colonizing microbes establish on the leaf, indirectly affecting microbe-124 microbe interactions, and structuring leaf microbial communities as a whole. 125 To assess the relative contributions of quantitative and Mendelian genetic variation we 126 first examined if a single large effect Mendelian locus underlying leaf shape generates leaf 127 microbiome differences within a host species. Then, we estimated the heritability of microbes 128 due to many small effect loci that differ between plant lines– in other words, the amount of 129 microbiome variation attributable to quantitative genetic causes. Finally, we compared the 130 relative importance of host genotype to other factors such as germination time, and therefore host 131 age, to assess the role of the environment in structuring leaf microbiomes. 132 133

134 Study system and crossing design 135 Ipomoea hederacea, Ivy-leaved morning glory, is a flowering annual plant, commonly 136 found in eastern North America in roadside ditches and agricultural fields. It is predominantly 137 selfing [41], and dies with the first hard frost in autumn. We used seeds derived from a cross by 138 Campitelli and Stinchcombe [39], where individuals from the two alternate homozygous leaf 139 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 7 shape phenotypes (i.e., fully lobed or whole) were collected from North Carolina, USA, selfed 140 for seven generations to generate homozygous parents (P1), and then crossed with each other 141 (Figure 1). A single F1 was allowed to self-fertilize, producing F2 progeny. We scored F2 plants 142 for leaf shape, and allowed them to self-fertilize; we refer to all the selfed progeny of an 143 individual F2 plant as a “matriline.” We used F3 seeds, set by F2 plants we had identified as 144 heterozygous for leaf shape, as our experimental seeds. In total, we used seeds from 82 145 matrilines. As expected, genotype frequencies did not significantly differ from a 1:2:1 ratio for 146 leaf shape (/i1 2 (2, 218)= 3, p = 0.22), with 55 plants with whole leaves, 112 plants with partially 147 lobed leaves, and 51 plants with fully lobed leaves. 148 Our breeding design has three consequences for leaf microbiomes, which we return to in 149 the Discussion. First, variation between matrilines is due to the combined effects of loci that 150 differed between the original parents of the cross and the effects of recombination. Second, 151 because of the effects of recombination, significant differences between leaf shape genotypes are 152 due to either the effects of the leaf shape locus itself or linked loci not broken up by 153 recombination. Third, within a matriline, we expect 25% of the loci that differed between the 154 parents to still be heterozygous, which is considerably more within-matriline genetic variation 155 than would be found in inbred lines or recombinant inbred lines (RILs), potentially reducing our 156 power and making our estimates of microbe heritability more conservative. 157 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 8 158 Figure 1. Ipomoea hederacea leaf polymorphism and crossing design. Leaf shape is determined 159 by a single Mendelian locus. First, both homozygous genotypes were collected from North 160 Carolina, USA and selfed for seven generations to create the parents (P1). To generate the plants 161 used in the field experiment, the two homozygous P1 individuals were crossed, then 162 heterozygote offspring were selfed for two more generations. Our experiment used F3 plants 163 from 82 matrilines. Photos taken by Julia Boyle, with brightness and colors edited for style and 164 clarity. 165 166 Field site and experimental design 167 In 2021, we planted a total of 240 I. hederacea seeds from 82 matrilines (2-3 168 seeds/matriline) in a common garden at Koffler Scientific Reserve (www.ksr.utoronto.ca) in 169 Ontario, Canada. We plowed and disked an old field for the common garden site, and we 170 assigned morning glories in 3 blocks of 80 plants each, planted 1 m apart with a 6 ft tall bamboo 171 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 9 pole to climb. We scarified seeds then planted them in a greenhouse in peat pots containing soil 172 from the field. We planted the first cohort on June 4th and a second cohort on July 2nd, because 173 of poor germination in the first cohort. After the majority of plants in a given cohort germinated 174 (approximately one week), we transplanted the pots into the field. The 218 surviving plants 175 consisted of 83 individuals from the first cohort, and 135 individuals from the second cohort. 176 There were no significant differences in the frequencies of leaf shape genotypes between the two 177 cohorts (/i1 2 (2, 218)= 0.31, p = 0.86; Table S1). The common garden was moderately weeded until 178 the I. hederacea plants had established; neighboring plants in the common garden were 179 predominantly Cirsium arvense, Canada thistle. On September 1st, we collected similarly-sized 180 leaves within one foot from the ground by cutting them with sterilized scissors into a sterile 181 plastic bag, after which they were stored in a -80°C freezer until DNA extraction. We avoided 182 leaves that looked very young or very old, as well as leaves with herbivory damage. At the time 183 of leaf collection, plants in the first germination cohort were 89 days old with 96% of plants 184 flowering, and plants in the second germination cohort were 61 days old with 49% of plants 185 flowering. We photographed collected leaves against a 1cm2 grid background, and measured leaf 186 surface area and shape using ImageJ [42]. From these photographs, we calculated leaf circularity 187 as 4π (Area/Perimeter2), which has a value between 0 and 1, with 1 indicating a perfect circle. 188 189 Sequencing and QIIME2 analysis 190 We performed extractions using the whole leaves with QIAGEN DNeasy® PowerSoil® 191 Pro Kits; both epiphytic and endophytic microbes were extracted. We sent samples to Génome 192 Québec (Montréal, Canada) for Illumina MiSeq PE 250bp 16S rRNA gene amplicon sequencing 193 on the conserved hypervariable V4 region (primer pair 515F-806R). We used Quantitative 194 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 10 Insights Into Microbial Ecology 2 (QIIME2) v.2022.2 [43] to trim the sequences for quality and 195 we denoised the sequences with DADA2 [44]. Samples had a median read frequency of 35,436 196 reads. Using QIIME2, we removed amplicon sequence variants (ASVs) that had fewer than 10 197 reads across all samples, and assigned taxonomy using the ‘sklearn’ feature classifier and 198 Greengenes 16S rRNA gene V4 region reference [45], then filtered out reads assigned as 199 cyanobacteria and mitochondria to remove plant DNA. After these steps, we had 4,934 ASVs 200 across the 218 leaf samples, with samples having a median read frequency of 4,523 reads. After 201 visualizing a rarefaction curve, we created a rarefied dataset with 4,695 ASVs total across 183 202 samples, with 1809 reads/sample; the first and second cohorts retained 77 and 106 samples 203 respectively. Finally, we constructed a phylogeny using QIIME2’s MAFFT [46] and FastTree 2 204 [47] to obtain a rooted tree. 205 206 Statistical analysis 207 For statistical analysis, we used R v4.2.0 [48], with the tidyverse [49], lmerTest [50], 208 phyloseq [51], vegan [52], and microbiome [53] packages. The general model structure we used 209 included leaf shape genotype and germination cohort as fixed effects, and random effects of 210 block and matriline, unless otherwise specified. All linear models were adjusted with type III 211 ANOVAs, calculated in the car package [54]. 212 First, we compared leaf shape surface area and circularity between genotypes using linear 213 mixed effect models. Next, we identified the core microbiota by filtering the rarefied dataset to 214 genera found in at least 80% of samples. To assess the effect of leaf shape and germination 215 cohort on microbial composition, we calculated a weighted UniFrac distance matrix using 216 rarefied data, then we used an adonis2 PERMANOVA (999 permutations) followed by a 217 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 11 permutational test of dispersion on significant groups using betadisper. The vegan package does 218 not allow for random effects, so we did not include block and matriline. To confirm that 219 germination cohort sample sizes did not affect our results, we also downsampled to equal cohort 220 sizes 100 times, and used the same permutation tests for composition and dispersion (Figure S3). 221 To visualize the community, we used a weighted UniFrac Principal Coordinate Analysis (PCoA) 222 using rarefied relative abundance data. 223 We estimated broad-sense heritability (H2) of microbial community phenotypes and 224 taxonomic groups using linear mixed effects models, taking a similar approach as Wagner et al. 225 [10] and Grieneisen et al. [14]. Here, the random effects of block and matriline explain variance 226 in the microbe abundance phenotype after accounting for the mean effects of the germination 227 cohort and leaf shape, meaning our estimates of H2 are conditioned on the fixed effects of the 228 germination cohort and leaf shape (as is commonly the case [55]). To estimate H2 from the 229 random effects, we calculated the ratio of matriline genetic variance (VG) to the sum of matriline, 230 spatial block, and residual variance (VP = VG + VBlock + VE). We generated the community 231 phenotypes using rarefied data and included observed species richness (base 10 log-232 transformed), evenness, Shannon diversity, and the first three axes of the weighted UniFrac 233 PCoA as the response variables in individual linear mixed effects models. To estimate the 234 heritability of taxonomic groups in a more compositionally aware way [56], we aggregated the 235 non-rarefied dataset to genera that were in at least 30% of samples, added 1 to each abundance, 236 and center log ratio transformed the abundance matrix. Next, we fit a linear mixed effects model 237 to the abundance of each genus that was in at least 30% of samples. To assess the statistical 238 significance of the matriline effect, we compared the log-likelihood of our models with and 239 without line as a random effect; the difference in log-likelihoods is X2 distributed, with one 240 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 12 degree of freedom [57]. Finally, solely for the purposes of a qualitative comparison of the 241 magnitude of variance explained by the leaf shape locus, matriline effects, spatial blocks, and 242 residual variation, we used the same mixed effects model except with leaf shape as a random 243 effect. 244 245

246 Genotype morphology 247 Leaf shape genotypes of I. hederacea differed quantitatively. Leaf shape genotypes 248 differed significantly in measures of circularity (Wald X2 (2,218)=1520, p<0.001; Table S2), with 249 the whole genotype having the highest circularity followed by heterozygotes (Figure S1). While 250 there was a trend for leaf surface area to be the largest in whole leaves and lowest in the lobed 251 leaves, the difference was not significantly predicted by leaf shape genotype (Wald X2 (2,218)=4.15, 252 p>0.05; Table S2) or germination cohort (Wald X2 (1,218)=2.22, p>0.05; Table S2) (Figure S1). As 253 we specifically chose to sample similarly sized leaves, with the intent of minimizing differences 254 in total leaf area, these results suggest our sampling strategy worked and that we can attribute 255 differences in microbes due to leaf shape as not simply being the result of differences in leaf 256 area. 257 258 Community composition and diversity 259 The core I. hederacea leaf microbes were the genera Methylobacterium, Sphingomonas, 260 Deinococcus, Pseudomonas, Arthrobacter, and Hymenobacter. Communities were structured by 261 the high relative abundances of Methylobacterium and Sphingomonas genera (Figure 3 and 4), 262 and their relative abundances were negatively correlated (Figure S2). The first PCoA axis was 263 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 13 strongly positively correlated with Methylobacterium relative abundance while the second axis 264 was strongly negatively correlated with Sphingomonas relative abundance (Figure S2). 265 Germination cohort was the main driver of overall microbial community composition. 266 While microbiomes from both cohorts overlapped in composition, they were significantly 267 different from each other (F(1,182)=10.4, p=0.001; Table S3); older plants from the first cohort had 268 more Methylobacterium, and younger plants from the second cohort had more Sphingomonas 269 (Figure 4ACD). The younger cohort’s microbiomes were more distinct from one another, 270 creating significantly higher dispersion (F(1,182)=10.4, p=0.001; Table S3). Germination cohorts 271 always remained significantly different in composition and dispersion when cohort sample sizes 272 were equalized (Figure S3), suggesting this is a true biological phenomenon. Germination cohort 273 significantly predicted observed richness (Wald X2 (1,182)=10.1, p=0.001; Table S4), with older 274 plants having higher ASV richness than younger plants, but it did not predict Shannon diversity 275 (Wald X2 (1,182)=3.33, p>0.05; Table S4) or evenness (Wald X2 (1,182)=2.96, p>0.05; Table S4). 276 Despite differences in morphology and microclimate that we a priori predicted to affect 277 the leaf microbiome, we did not detect strong leaf shape effects on community composition as a 278 whole. Leaf shape did not significantly structure composition (F(2,182)=1.83, p>0.05; Table S3), 279 or affect Shannon diversity, evenness, or observed richness (Wald X2 (2,182)0.05 for all; 280 Table S4). While whole community composition was not affected, we still identified microbial 281 genera significantly affected by leaf shape, as discussed below. 282 283 Heritable microbes and community phenotypes 284 Microbial community phenotypes did not have a strong genetic basis, yet we still 285 discovered significant heritable variation in some microbial traits. Community phenotypes had 286 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 14 low broad-sense heritability (0≤ H2≤ ~0.05) with the third axis of the PCoA having the highest H2, 287 followed by Shannon diversity (Figure 2A). All community phenotypes with non-zero H2 were 288 significantly heritable, as the models’ log-likelihoods were improved by including matriline; 289 community phenotypes with no H2 (evenness and PCoA axis 2) showed no difference in fit when 290 matriline was included (Figure S4). The main effect of the germination cohort significantly 291 affected observed richness, as previously described, as well as the first two axes of the PCoA 292 (Wald X2 (1,182)>6.71, p<0.01 for both axes; Table S4), whereas leaf shape did not significantly 293 affect community phenotypes (Wald X2 (2,182)0.05 for all; Table S4). In our qualitative 294 comparison of the magnitudes of variances explained, plant matriline had a larger effect on 295 community phenotypes’ H2 than leaf shape, with the exception of observed richness (Figure 2C, 296 Table S5). 297 Host genetics had a stronger influence on the abundance of microbial genera than 298 community phenotypes, with both quantitative genetic and Mendelian traits affecting microbe 299 abundance. Considering the genera in at least 30% of samples, we identified eight genera with 300 non-zero and significant broad-sense heritability. The most heritable genera (0.05≤H 2≤ 0.07) 301 were Nocardioides, Kineococcus, Pseudomonas, and Agrobacterium (Figure 2B). Our likelihood 302 ratio test results showed that including plant matriline improved model fit for 10 out of 15 genera 303 (which included all genera with non-zero H2), there was no difference in fit for 3 genera 304 (Deinococcus, Hymenobacter, Sphingomonas), and slightly reduced model fit for 2 genera 305 (Actinotelluria, Rubellimicrobium) (Figure S4). The germination cohort significantly affected the 306 abundance of more than half of the most common genera (Wald X2 (1,218)>3.83, p<0.05 for all; 307 Table S6, Figure 2B). Four genera, Sphingomonas, Nocardioides, Methylobacterium and 308 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 15 Agrobacterium, had significant main effects of leaf shape on their phenotype (Wald 309 X2 (2,218)>7.03, p<0.05 for all; Table S6, Figure 2B). When we examined whether the magnitude 310 of plant matriline variance was qualitatively comparable to leaf shape locus, we found that these 311 four genera had equal to larger magnitudes of variance attributable to leaf shape than plant 312 matriline (Figure 2D, Table S7). In heritable genera with a non-significant main effect of leaf 313 shape, plant matriline explained more variance than leaf shape (Figure 2D, Table S7), as 314 expected. 315 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 16 316 Figure 2. Broad-sense heritability (H2) of community phenotypes and common microbial 317 genera. Genera are ordered by percent variance explained by plant matriline in the H2 model. A 318 and B: H2 of community phenotypes and genera in at least 30% of samples. H2 was calculated as 319 the percent variation explained by genetic line compared to the total of matriline, spatial block, 320 and residual variances. Significant fixed effects of germination cohort and leaf shape are 321 indicated by diamonds and triangles respectively. C and D: H2 of community phenotypes and 322 16 as .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 17 genera in at least 30% of samples, but with leaf shape as a random effect to directly compare the 323 magnitude of variance explained to that of matriline. 324 325 326 Figure 3. Relative abundance of bacterial families in the Ipomoea hederacea leaf microbiome, 327 indicated by color. Samples are ordered by Sphingomonadaceae abundance and grouped by 328 germination cohort. The first germination cohort (A) was 89 days old (n=83) and the second 329 germination cohort (B) was 61 days old (n=135) at the time of sampling. For this visualization 330 we included only families in at least 30% of samples and at least 1% relative abundance. 331 17 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 18 332 Figure 4. Weighted UniFrac PCoA of I. hederacea leaf microbiomes. Germination cohort 333 significantly predicted community composition (A), while leaf shape was non-significant (B). 334 Core microbiota structured communities (C-F). For C-F, note that the relative abundance legend 335 scales differ. 336 337

338 18 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 19 Microbes can be considered an extended host phenotype with potentially adaptive 339 functions for the host [58], and heritable microbes are more likely to consistently affect host 340 phenotype and fitness through time. Here we examined the relative contributions of Mendelian 341 and quantitative genetic variation to the heritability of host-associated microbes to determine 342 what types of host genetic variation mattered to microbial ecology. We found that both 343 Mendelian and quantitative genetic host variation contribute to microbe heritability, and that the 344 cumulative small effect genomic differences due to matriline explained as much or more 345 variation than a single large effect locus. We were able to identify heritable microbial variation 346 despite a large effect of germination timing, which incorporated host age and initial microbial 347 environment, and which was a major factor structuring community composition. We discuss the 348 implications of our results below. 349 350 Genetic influence on the microbiome 351 There was genetic variation at several scales mediating how well microbes can establish 352 and perform in the phyllosphere. It is important to note that, as always, our estimates of 353 heritability are specific to the population [59] and conditioned on fixed effects in the model [55], 354 in our case, the germination cohort and leaf shape. A single leaf shape gene significantly 355 explained the abundance of several common microbes, suggesting the genetic differences in leaf 356 microclimate (or other, unknown features of the leaf shape genotypes) impact microbial 357 establishment and persistence, and can contribute to intraspecific differences in the phyllosphere. 358 When comparing the magnitude of effect, both leaf shape and matriline were important factors 359 influencing microbe abundance, although plant matriline affected more of the microbes tested. 360 Community phenotypes like species diversity and weighted UniFrac distances showed lower 361 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 20 heritability than individual genera, but since the majority of microbes are not likely to be 362 heritable in an environmentally acquired microbiome, very low heritability at a community scale 363 was perhaps unsurprising. Our estimates of heritability are low to commensurate compared to 364 other heritable microbiome studies (summarized in Table S8; [10, 12–14, 16, 60–64]; while the 365 range of heritabilities for plant-associated microbes across other studies was as broad as H2=0-1, 366 the average significant heritability of microbes and community phenotypes was usually low 367 (H2≤ 0.10) (Table S8). Our results therefore follow the trend of low heritability for 368 environmentally acquired microbes, with a strong influence of the environment. 369 In our experiment, each plant line represented a different mosaic of the original parental 370 genotypes, similar in concept to recombinant inbred lines (RILs). However, because the seeds we 371 used in our experiment were F3 individuals, approximately 25% of the loci initially heterozygous 372 in the F1 (i.e., those that differed between the parents) remain heterozygous within matrilines. 373 One consequence of this is that there is genetic variation between matrilines due to the effects of 374 recombination, and genetic variation within matrilines (due to heterozygosity and segregation 375 within a matriline). These effects make it more difficult to detect quantitative genetic variation in 376 microbial phenotypes, especially compared to RIL populations that have been made homozygous 377 within lines. Our results may thus underestimate the prevalence and magnitude of H2 due to 378 reduced power. Furthermore, our crossing design included only two parental genotypes, 379 potentially limiting the amount of genetic variation segregating in the cross. For instance, had we 380 sampled more genotypes to create MAGIC or Nested Association Mapping lines [65, 66], more 381 genetic variation would be captured in our population. It is difficult to determine, a priori, 382 whether this would lead to increased or decreased estimates of genetic variation. On the one 383 hand, two parental genotypes is a limited sample of the genetic variation found in many 384 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 21 populations. Although putatively neutral genetic variation in I. hederacea populations is often 385 quite low [67], there is frequently quantitative genetic variation [68, 69]. On the other hand, the 386 selfing rate in I. hederacea is quite high (~93%) [67, 70]. Rare outcrossing events, followed by 387 selfing, may produce recombinant populations not that dissimilar from our experimental 388 population. Similar arguments have been made for A. thaliana [71, 72], which is also highly 389 selfing. As a qualitative investigation of the influence of including limited genetic variation in 390 our crossing design, we gathered estimates of H2 for life history and quantitative traits in 391 Arabidopsis thaliana, specifically comparing studies that used either multiple accessions or 392 RILs. These data (Table S9), while an imperfect comparison, suggest that if anything, our 393 estimates of heritability could be under-estimated from using two parental lines in the initial 394 cross. 395 396 Evolutionary implications 397 We found similar magnitudes of variance in microbial phenotypes attributable to plant 398 line and leaf shape genotype for several common bacterial genera, which is important because 399 there is the potential for different evolutionary forces to act on quantitative genetic loci versus 400 Mendelian loci. For example, Mendelian traits like leaf shape have high potential to show 401 sampling effects due to drift, especially in small populations [73]. In contrast, drift is much less 402 likely to produce a change in quantitative traits which are influenced by many genes [73]. Thus, 403 drift can potentially lead to bigger effects on leaf shape-associated microbes. Furthermore, high 404 selfing in I. hederacea will reduce the frequency of heterozygous leaf shape genotypes by as 405 much as 50% per generation. In this way, the mating system may have outsized effects on 406 microbes linked to Mendelian traits. If natural selection acts on host traits, as it appears to with I. 407 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 22 hederacea’s leaf shape [38], there may be indirect effects of selection on the microbial 408 phenotype in the host population. 409 The microbes we identified as having heritable variation in abundance are also linked to 410 host performance and fitness in other systems. These findings from other systems suggest 411 intriguing hypotheses for further study. For example, one of the most heritable genera that was 412 also significantly affected by leaf shape, Agrobacterium, has pathogens linked to causing tumor-413 like galls in plants [74], suggesting a potential link of leaf shape to disease risk. 414 Methylobacterium spp. are linked to improving plant growth [75], and increased growth rate 415 improves fitness in I. hederacea [39]. Pseudomonas spp. have a wide range of positive or 416 negative effects on plant performance [76]. The most heritable genus Nocardioides is commonly 417 endophytic on plant leaves and stems [77, 78] and has species capable of fixing nitrogen [79] and 418 reducing nitrate [80]. Evaluating whether any of these microbial genera which have effects on 419 host performance in other systems do so in I. hederacea will require further empirical work. If 420 heritable microbes affect plant fitness, then there is higher potential for reciprocal selection 421 between the microbes and plant hosts. 422 423 Environmental and phenological effects on the leaf microbiome 424 Community-wide differences between germination cohorts could be due to seasonal 425 differences in what microbes were present in the plant’s early life and the amount of time 426 selection had to act on the microbial communities. Input of new microbes to the phyllosphere can 427 be low relative to environments like soil rhizospheres, and these sources may change across the 428 growing season [29]. The cumulative input of microbes may matter as well, since we found that 429 the older plant phyllospheres had a higher species richness than younger plants; this is possibly 430 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 23 because the leaves themselves were older and had more time to collect rare microbes dispersing 431 from the environment. Additionally, as plants age and change phenologically, this may change 432 the leaf environment for microbes [81–83]. Whilst in a time of growth and development, young 433 plants have weaker immune responses [84], meaning they may impose weaker selection on their 434 phyllospheres. Thus, the disparate microbiomes on young plants could be due to weaker 435 selection from the environment and host over a shorter time. In contrast, older plants showed 436 more similar and clustered microbiomes suggesting that over time there is selection on microbes 437 by the I. hederacea leaf microclimate, and the microbial community converges. The merging 438 community compositions between young and old plants could suggest microbial succession 439 occurred in the phyllosphere [85, 86]. 440 The large effect of germination cohort is not unexpected, given that most studies on 441 microbiome heritability find a very large effect of environment and host age on the microbiome 442 [10, 12, 16, 87]. For example, Walters et al. [16] found that plant age was the largest driver of 443 the maize rhizosphere, but in one year they nonetheless found 143 heritable root microbes with 444 H2=0.15-0.25. While our results suggest genera like Methylobacterium and Sphingomonas were 445 significantly affected by host age and initial microbial environment, host genotype still 446 significantly mediated these environmental and phenological changes in microbe abundance. Our 447

add to recent evidence that the assembly process of host-associated microbiomes is 448 governed by both stochastic forces and host-based selective forces [87, 88]. 449 450

451 Our results show that while I. hederacea leaf microbe composition differences were 452 primarily shaped by host age and the environment, there exists a heritable subset of core 453 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 24 microbes in the I. hederacea microbiome. A Mendelian trait and quantitative genetic variation 454 across matrilines explained similar amounts of variation in microbial abundance, with 455 implications for plant-microbe eco-evolutionary dynamics across time. Furthermore, we show 456 for the first time that leaf shape itself may contribute to differences in phyllosphere microbial 457 abundances within a species. 458 459 Declarations 460 Ethics approval and consent to participate 461 Not applicable. 462 Consent for publication 463 Not applicable. 464 Competing interests 465 The authors declare that they have no competing interests. 466 Funding 467 Funding sources include NSERC graduate funding to JAB and NSERC discovery grants to MEF 468 and JRS. 469 Authors' contributions 470 JAB, MEF, and JRS collaborated on the ideas; JAB, MEF, and JRS designed the methodology, 471 JAB performed the experiments, and JAB, MEF, and JRS wrote and edited the manuscript. JAB, 472 MEF, and JRS analyzed the data. All authors contributed to this manuscript and approved its 473 publication. 474

475 We thank Brandon Campitelli for generating the morning glory lines and the staff of Koffler 476 Scientific Reserve in 2021 for helping to maintain the field site. We acknowledge Génome 477 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 25 Québec (Montréal, Canada) for sequencing samples. Finally, we gratefully acknowledge the 478 Swedish Collegium for Advanced Study, and Jennifer Mack, Natuschka Lee, Iva Luč ić , and 479 Arthur Asseraf, for support. 480 481

482 1. Leveau JH. A brief from the leaf: latest research to inform our understanding of the 483 phyllosphere microbiome. Curr Opin Microbiol. 2019;49:41–9. 484 2. Vokou D, Vareli K, Zarali E, Karamanoli K, Constantinidou HIA, Monokrousos N, et al. 485 Exploring Biodiversity in the Bacterial Community of the Mediterranean Phyllosphere and its 486 Relationship with Airborne Bacteria. Microb Ecol. 2012;64:714–24. 487 3. Lindow SE. Role of Immigration and Other Processes in Determining Epiphytic Bacterial 488 Populations. In: Morris CE, Nicot PC, Nguyen-The C, editors. Aerial Plant Surface 489 Microbiology. Boston, MA: Springer, Boston, MA; 1996. p. 155–68. 490 4. Massoni J, Bortfeld-Miller M, Widmer A, Vorholt JA. Capacity of soil bacteria to reach the 491 phyllosphere and convergence of floral communities despite soil microbiota variation. Proc Natl 492 Acad Sci U S A. 2021;118:e2100150118. 493 5. Dastogeer KMG, Tumpa FH, Sultana A, Akter MA, Chakraborty A. Plant microbiome–an 494 account of the factors that shape community composition and diversity. Curr Plant Biol. 495 2020;23:100161. 496 6. Redford AJ, Bowers RM, Knight R, Linhart Y, Fierer N. The ecology of the phyllosphere: 497 geographic and phylogenetic variability in the distribution of bacteria on tree leaves. Environ 498 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 26 Microbiol. 2010;12:2885–93. 499 7. Kembel SW, O’Connor TK, Arnold HK, Hubbell SP, Wright SJ, Green JL. Relationships 500 between phyllosphere bacterial communities and plant functional traits in a neotropical forest. 501 Proc Natl Acad Sci. 2014;111:13715–20. 502 8. Bodenhausen N, Horton MW, Bergelson J. Bacterial Communities Associated with the Leaves 503 and the Roots of Arabidopsis thaliana. PLoS One. 2013;8:e56329. 504 9. Horton MW, Bodenhausen N, Beilsmith K, Meng D, Muegge BD, Subramanian S, et al. 505 Genome-wide association study of Arabidopsis thaliana leaf microbial community. Nat 506 Commun. 2014;5:5320. 507 10. Wagner MR, Lundberg DS, Del Rio TG, Tringe SG, Dangl JL, Mitchell-Olds T. Host 508 genotype and age shape the leaf and root microbiomes of a wild perennial plant. Nat Commun. 509 2016;7:12151. 510 11. Wagner MR, Busby PE, Balint/i4 Kurti P. Analysis of leaf microbiome composition of 511 near/i4 isogenic maize lines differing in broad/i4 spectrum disease resistance. New Phytol. 512 2019;225:2152–65. 513 12. Peiffer JA, Spor A, Koren O, Jin Z, Tringe SG, Dangl JL, et al. Diversity and heritability of 514 the maize rhizosphere microbiome under field conditions. Proc Natl Acad Sci U S A. 515 2013;110:6548–53. 516 13. Deng S, Caddell DF, Xu G, Dahlen L, Washington L, Yang J, et al. Genome wide 517 association study reveals plant loci controlling heritability of the rhizosphere microbiome. ISME 518 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 27 J 2021 1511. 2021;15:3181–94. 519 14. Grieneisen L, Dasari M, Gould TJ, Björk JR, Grenier JC, Yotova V, et al. Gut microbiome 520 heritability is nearly universal but environmentally contingent. Science (80- ). 2021;373:181–6. 521 15. Wagner MR, Tang C, Salvato F, Clouse KM, Bartlett A, Vintila S, et al. Microbe-dependent 522 heterosis in maize. Proc Natl Acad Sci. 2021;118:e2021965118. 523 16. Walters WA, Jin Z, Youngblut N, Wallace JG, Sutter J, Zhang W, et al. Large-scale 524 replicated field study of maize rhizosphere identifies heritable microbes. Proc Natl Acad Sci U S 525 A. 2018;115:7368–73. 526 17. Henry LP, Bruijning M, Forsberg SKG, Ayroles JF. The microbiome extends host 527 evolutionary potential. Nat Commun. 2021;12:5141. 528 18. Mueller UG, Linksvayer TA. Microbiome breeding: conceptual and practical issues. Trends 529 Microbiol. 2022;30:997–1011. 530 19. Wagner MR, Lundberg DS, Coleman-Derr D, Tringe SG, Dangl JL, Mitchell-Olds T. Natural 531 soil microbes alter flowering phenology and the intensity of selection on flowering time in a wild 532 Arabidopsis relative. Ecol Lett. 2014;17:717–26. 533 20. Petersen C, Hamerich IK, Adair KL, Griem-Krey H, Torres Oliva M, Hoeppner MP, et al. 534 Host and microbiome jointly contribute to environmental adaptation. ISME J. 2023;17:1953–65. 535 21. Pogoda CS, Keepers KG, Reinert S, Talukder ZI, Smart BC, Attia Z, et al. Heritable 536 differences in abundance of bacterial rhizosphere taxa are correlated with fungal necrotrophic 537 pathogen resistance. Mol Ecol. 2023;00:1–14. 538 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 28 22. Baltrus DA. Adaptation, specialization, and coevolution within phytobiomes. Curr Opin 539 Plant Biol. 2017;38:109–16. 540 23. Koskella B, Bergelson J. The study of host–microbiome (co)evolution across levels of 541 selection. Philos Trans R Soc B. 2020;375:20190604. 542 24. Johnson MTJ, Stinchcombe JR. An emerging synthesis between community ecology and 543 evolutionary biology. Trends Ecol Evol. 2007;22:250–7. 544 25. VanWallendael A, Benucci GMN, da Costa PB, Fraser L, Sreedasyam A, Fritschi F, et al. 545 Host genotype controls ecological change in the leaf fungal microbiome. PLOS Biol. 546 2022;20:e3001681. 547 26. Bodenhausen N, Bortfeld-Miller M, Ackermann M, Vorholt JA. A Synthetic Community 548 Approach Reveals Plant Genotypes Affecting the Phyllosphere Microbiota. PLOS Genet. 549 2014;10:e1004283. 550 27. Pincebourde S, Woods HA. Climate uncertainty on leaf surfaces: the biophysics of leaf 551 microclimates and their consequences for leaf-dwelling organisms. Funct Ecol. 2012;26:844–53. 552 28. Runkle E. The Boundary Layer and Its Importance. Greenhouse Product News. 2016;:34. 553 29. Vacher C, Hampe A, Porté AJ, Sauer U, Compant S, Morris CE. The Phyllosphere: 554 Microbial Jungle at the Plant–Climate Interface. Annu Rev Ecol Evol Syst. 2016;47:1–24. 555 30. Núñez A, García AM, Moreno DA, Guantes R. Seasonal changes dominate long-term 556 variability of the urban air microbiome across space and time. Environ Int. 2021;150:106423. 557 31. Humphrey PT, Nguyen TT, Villalobos MM, Whiteman NK. Diversity and abundance of 558 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 29 phyllosphere bacteria are linked to insect herbivory. Mol Ecol. 2014;23:1497–515. 559 32. Meyer KM, Porch R, Muscettola IE, Luisa A, Vasconcelos S, Sherman JK, et al. Plant 560 neighborhood shapes diversity and reduces interspecific variation of the phyllosphere 561 microbiome. ISME J 2022. 2022;:1–12. 562 33. Nicotra AB, Leigh A, Boyce CK, Jones CS, Niklas KJ, Royer DL, et al. The evolution and 563 functional significance of leaf shape in the angiosperms. Functional Plant Biology. 2011;38:535–564 52. 565 34. Givnish TJ. Comparative Studies Of Leaf Form: Assessing The Relative Roles Of Selective 566 Pressures And Phylogenetic Constraints. New Phytol. 1987;106:131–60. 567 35. Yates MJ, Verboom GA, Rebelo AG, Cramer MD. Ecophysiological significance of leaf size 568 variation in Proteaceae from the Cape Floristic Region. Funct Ecol. 2010;24:485–92. 569 36. Elmore CD. Mode of Reproduction and Inheritance of Leaf Shape in Ipomoea hederacea . 570 Weed Sci. 1986;34:391–5. 571 37. Bright KL, Rausher MD. Natural selection on a leaf-shape polymorphism in the ivyleaf 572 morning glory (Ipomoea hederacea). Evolution (N Y). 2008;62:1978–90. 573 38. Campitelli BE, Stinchcombe JR. Natural selection maintains a single-locus leaf shape cline in 574 Ivyleaf morning glory, Ipomoea hederacea. Mol Ecol. 2013;22:552–64. 575 39. Campitelli BE, Stinchcombe JR. Testing potential selective agents acting on leaf shape in 576 Ipomoea hederacea: Predictions based on an adaptive leaf shape cline. Ecol Evol. 2013;3:2409–577 23. 578 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 30 40. Campitelli BE, Gorton AJ, Ostevik KL, Stinchcombe JR. The effect of leaf shape on the 579 thermoregulation and frost tolerance of an annual vine, Ipomoea hederacea (Convolvulaceae). 580 Am J Bot. 2013;100:2175–82. 581 41. Hull-Sanders HM, Eubanks MD, Carr DE. Inbreeding depression and selfing rate of Ipomoea 582 hederacea var. integriuscula (Convolvulaceae). Am J Bot. 2005;92:1871–7. 583 42. Schneider CA, Rasband WS, Eliceiri KW. NIH Image to ImageJ: 25 years of image analysis. 584 Nat Methods. 2012;9:671–5. 585 43. Bolyen E, Rideout JR, Dillon MR, Bokulich NA, Abnet CC, Al-Ghalith GA, et al. 586 Reproducible, interactive, scalable and extensible microbiome data science using QIIME 2. Nat 587 Biotechnol 2019 378. 2019;37:852–7. 588 44. Callahan BJ, McMurdie PJ, Rosen MJ, Han AW, Johnson AJA, Holmes SP. DADA2: High-589 resolution sample inference from Illumina amplicon data. Nat Methods. 2016;13:581–3. 590 45. McDonald D, Price MN, Goodrich J, Nawrocki EP, Desantis TZ, Probst A, et al. An 591 improved Greengenes taxonomy with explicit ranks for ecological and evolutionary analyses of 592 bacteria and archaea. ISME J . 2011;6:610–8. 593 46. Katoh K, Misawa K, Kuma KI, Miyata T. MAFFT: a novel method for rapid multiple 594 sequence alignment based on fast Fourier transform. Nucleic Acids Res. 2002;30:3059–66. 595 47. Price MN, Dehal PS, Arkin AP. FastTree 2 – Approximately Maximum-Likelihood Trees for 596 Large Alignments. PLoS One. 2010;5:e9490. 597 48. R Core Team. R: A language and environment for statistical computing. 2022. 598 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 31 49. Wickham H, Averick M, Bryan J, Chang W, McGowan L, François R, et al. Welcome to the 599 Tidyverse. J Open Source Softw. 2019;4:1686. 600 50. Kuznetsova A, Brockhoff PB, Christensen RHB. lmerTest Package: Tests in Linear Mixed 601 Effects Models . J Stat Softw. 2017;82:1–26. 602 51. McMurdie PJ, Holmes S. phyloseq: An R Package for Reproducible Interactive Analysis and 603 Graphics of Microbiome Census Data. PLoS One. 2013;8:e61217. 604 52. Oksanen J, Blanchet FG, Friendly M, Kindt R, Legendre P, Mcglinn D, et al. Package 605 “vegan” Version 2.5-7. 2020. 606 53. Lahti L, Shetty S. microbiome R package. 2019. 607 https://bioconductor.org/packages/release/bioc/html/microbiome.html. Accessed 3 Apr 2023. 608 54. Fox J, Weisberg Sanford. An R Companion to Applied Regression. 3rd edition. Sage 609 Publications; 2019. 610 55. Wilson AJ. Why h2 does not always equal Va / Vp/i4 ? J Evol Biol. 2008;21:647–50. 611 56. Bruijning M, Ayroles JF, Henry LP, Koskella B, Meyer KM, Metcalf CJE. Relative 612 abundance data can misrepresent heritability of the microbiome. Microbiome. 2023;11:222. 613 57. Buse A. The Likelihood Ratio, Wald, and Lagrange Multiplier Tests: An Expository Note. 614 Am Stat. 1982;36:153–7. 615 58. Friesen ML, Porter SS, Stark SC, Von Wettberg EJ, Sachs JL, Martinez-Romero E. 616 Microbially Mediated Plant Functional Traits. Annu Rev Ecol Evol Syst. 2011;42:23–46. 617 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 32 59. Falconer D, Mackay T. Introduction to Quantitative Genetics. 4th edition. England: Prentice 618 Hall, Harlow; 1996. 619 60. Schweitzer JA, Bailey JK, Fischer DG, LeRoy CJ, Lonsdorf E V., Whitham TG, et al. Plant–620 Soil–Microorganism Interactions: Heritable Relationship Between Plant Genotype And 621 Associated Soil Microorganisms. Ecology. 2008;89:773–81. 622 61. Goodrich JK, Waters JL, Poole AC, Sutter JL, Koren O, Blekhman R, et al. Human Genetics 623 Shape the Gut Microbiome. Cell. 2014;159:789–99. 624 62. Wallace JG, Kremling KA, Kovar LL, Buckler ES. Quantitative genetics of the maize leaf 625 microbiome. Phytobiomes J. 2018;2:208–24. 626 63. John Wallace R, Sasson G, Garnsworthy PC, Tapio I, Gregson E, Bani P, et al. A heritable 627 subset of the core rumen microbiome dictates dairy cow productivity and emissions. Sci Adv. 628 2019;5:8391–4. 629 64. Edwards JA, Saran UB, Bonnette J, MacQueen A, Yin J, Nguyen T uyen, et al. Genetic 630 determinants of switchgrass-root-associated microbiota in field sites spanning its natural range. 631 Curr Biol. 2023;33:1926-1938.e6. 632 65. Cavanagh C, Morell M, Mackay I, Powell W. From mutations to MAGIC: resources for gene 633 discovery, validation and delivery in crop plants. Curr Opin Plant Biol. 2008;11:215–21. 634 66. Yu J, Holland JB, McMullen MD, Buckler ES. Genetic Design and Statistical Power of 635 Nested Association Mapping in Maize. Genetics. 2008;178:539–51. 636 67. Campitelli BE, Stinchcombe JR. Population dynamics and evolutionary history of the weedy 637 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 33 vine Ipomoea hederacea in North America. G3 Genes, Genomes, Genet. 2014;4:1407–16. 638 68. Stock AJ, Campitelli BE, Stinchcombe JR. Quantitative genetic variance and multivariate 639 clines in the Ivyleaf morning glory, Ipomoea hederacea. Philos Trans R Soc B Biol Sci. 640 2014;369:20130259. 641 69. Henry GA, Stinchcombe JR. G-matrix stability in clinally diverging populations of an annual 642 weed. Evolution (N Y). 2023;77:49–62. 643 70. Ennos RA. Quantitative studies of the mating system in two sympatric species of Ipomoea 644 (Convolvulaceae). Genetica. 1981;57:93–8. 645 71. Weinig C, Dorn LA, Kane NC, German ZM, Halldorsdottir SS, Ungerer MC, et al. 646 Heterogeneous Selection at Specific Loci in Natural Environments in Arabidopsis thaliana. 647 Genetics. 2003;165:321–9. 648 72. Weinig C, Stinchcombe JR, Schmitt J. Evolutionary Genetics Of Resistance And Tolerance 649 To Natural Herbivory In Arabidopsis thaliana. Evolution (N Y). 2003;57:1270–80. 650 73 . Vasema/i10 gi A. The Adaptive Hypothesis of Clinal Variation Revisited: Single-Locus Clines 651 as a Result of Spatially Restricted Gene Flow. Genetics. 2006;173:2411–4. 652 74. Lippincott JA, Lippincott BB. Tumour-initiating Ability and Nutrition in the Genus 653 Agrobacterium. J Gen Microbiol. 1969;59:57–75. 654 75. Lee HS, Madhaiyan M, Kim CW, Choi SJ, Chung KY, Sa TM. Physiological enhancement 655 of early growth of rice seedlings (Oryza sativa L.) by production of phytohormone of N2-fixing 656 methylotrophic isolates. Biol Fertil Soils. 2006;42:402–8. 657 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 34 76. Wang NR, Wiesmann CL, Melnyk RA, Hossain SS, Chi M-H, Martens K, et al. Commensal 658 Pseudomonas fluorescens Strains Protect Arabidopsis from Closely Related Pseudomonas 659 Pathogens in a Colonization-Dependent Manner. MBio. 2022;13:e02892-21. 660 77. Kämpfer P, Glaeser SP, McInroy JA, Busse H-J. Nocardioides zeicaulis sp. nov., an 661 endophyte actinobacterium of maize. Int J Syst Evol Microbiol. 2016;66:1869–74. 662 78. Wang P, Kong X, Chen H, Xiao Y, Liu H, Li X, et al. Exploration of Intrinsic Microbial 663 Community Modulators in the Rice Endosphere Indicates a Key Role of Distinct Bacterial Taxa 664 Across Different Cultivars. Front Microbiol. 2021;12:629852. 665 79. Nafis A, Raklami A, Bechtaoui N, El Khalloufi F, El Alaoui A, Glick BR, et al. 666 Actinobacteria from Extreme Niches in Morocco and Their Plant Growth-Promoting Potentials. 667 Diversity. 2019;11:139. 668 80. Dastager SG, Lee J-C, Pandey A, Kim C-J. Nocardioides mesophilus sp. nov., isolated from 669 soil. Int J Syst Evol Microbiol. 2010;60:2288–92. 670 81. Escudero A, Mediavilla S. Decline in photosynthetic nitrogen use efficiency with leaf age 671 and nitrogen resorption as determinants of leaf life span. J Ecol. 2003;91:880–9. 672 82. England JR, Attiwill PM. Changes in leaf morphology and anatomy with tree age and height 673 in the broadleaved evergreen species, Eucalyptus regnans F. Muell. Trees. 2006;20:79. 674 83. Pantin F, Simonneau T, Muller B. Coming of leaf age: control of growth by hydraulics and 675 metabolics during leaf ontogeny. New Phytol. 2012;196:349–66. 676 84. van Butselaar T, Van den Ackerveken G. Salicylic Acid Steers the Growth–Immunity 677 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint 35 Tradeoff. Trends Plant Sci. 2020;25:566–76. 678 85. Redford AJ, Fierer N. Bacterial succession on the leaf surface: a novel system for studying 679 successional dynamics. Microb Ecol. 2009;58:189–98. 680 86. Copeland JK, Yuan L, Layeghifard M, Wang PW, Guttman DS. Seasonal Community 681 Succession of the Phyllosphere Microbiome. Mol Plant-Microbe Interact. 2015;28:274–85. 682 87. Dove NC, Veach AM, Muchero W, Wahl T, Stegen JC, Schadt CW, et al. Assembly of the 683 Populus Microbiome Is Temporally Dynamic and Determined by Selective and Stochastic 684 Factors . mSphere. 2021;6. 685 88. Xu CCY, Lemoine J, Albert A, Whirter É Mac, Barrett RDH. Community assembly of the 686 human piercing microbiome. Proc R Soc B Biol Sci. 2023;290. 687 688 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2024. ; https://doi.org/10.1101/2024.01.17.576069doi: bioRxiv preprint