The bacterial microbiota of a parasitic plant and its host

How plant-associated microbiota are shaped by, and potentially contribute to the unique ecology and heterotrophic life history of parasitic plants is relatively unknown. Here, we investigate the leaf and root bacterial communities associated with the root holoparasite Orobanche hederae and its host plant Hedera spp. We sequenced the V4 region of the 16S rRNA gene from DNA extracted from leaf and root samples of naturally growing populations of Orobanche and infected and uninfected Hedera. Root bacteria inhabiting Orobanche were less diverse, had fewer co-associations, and displayed increased compositional similarity to leaf bacteria relative to Hedera. Overall, Orobanche bacteria exhibited significant congruency with Hedera root bacteria across sites, but not the surrounding soil. Infection had localized and systemic effects on Hedera bacteria, which included effects on the abundance of individual taxa and root network properties. Collectively, our results indicate that the parasitic plant microbiome is derived but distinct from host plant microbiota, exhibits increased homogenization between shoot and root tissues, and displays far fewer co-associations among individual bacterial members. Host plant infection is accompanied by modest changes of associated microbiota at both local and systemic scales compared with uninfected individuals. Our results provide insight into the assembly and function of plant microbiota.


20
Plants harbour rich assemblages of microorganisms, which vary in diversity and 21 composition across host plant tissues, individuals, and species (Bulgarelli et al., 2013). 22 This variation is driven by innate plant immunity (Lebeis et al., 2015;Stringlis et al., 23 2018), the quality and quantity of plant-derived resources (Zhalnina et al., 2018), and 24 microbe-microbe interactions (Agler et al., 2016;Durán et al., 2018). Controlled 25 experiments indicate that microbiota may play a role in plant nutrient acquisition 26 (Castrillo et al., 2017) and tolerance to abiotic and biotic stress (Fitzpatrick et al., 2018), 27 including defense from pathogens (Vogel et al., 2016). However, host plant species vary 28 in the composition of their associated microbiota and the causes and ecological 29 consequences of this variation are poorly understood (Fitzpatrick et al., 2018). Increased 30 insight into the assembly and function of plant microbiota requires investigation of plant 31 species that occupy diverse ecological niches (e.g. Angel et al., 2016;Coleman-Derr et 32 al., 2016;Finkel et al., 2016). Such investigation provides an opportunity to address how 33 associated microbiota are shaped by, and potentially contribute to the functional 34 diversity found across plant species. 35 36 One of the most dramatic niche shifts undergone by plants is the transition from 37 autotrophy to heterotrophy. In heterotrophic plants, resource acquisition from 38 photosynthetic plants occurs indirectly through co-associated mycorrhizal fungi 39 (mycoheterotrophy), or directly through specialized parasitic organs called haustoria. 40 The haustorium attaches to the root or stem vasculature of host plants and acts as a 41 conduit through which material is exchanged primarily from host to parasite 42 uninfected roots (IUR; i.e. roots from the same ivy individual and within 50 cm of the 134 parasite, but not directly being parasitized). From uninfected patches, we harvested ivy 135 leaves and roots (UL and UR, respectively). All samples were standardized by fresh 136 weight and carefully selected to represent homologous organs between the parasite and 137 host plant (full details in Supplementary Data: Methods). Voucher and host information 138 is provided in Table S1. We processed sequences using the R package 'DADA2' v. 1.8.0 (Callahan et al., 2016) [Bates et al., 2015]). We calculated α -diversity as ASV richness (R), inverse Simpson's 167 diversity (D -1 ), and evenness (D -1 /R). We also calculated phylogenetic diversity (Faith, 168 1992), the sum of the total phylogenetic branch lengths in an assemblage, using the R 169 package 'picante' (Kembel et al., 2010). Data for α -diversity indices were ln-transformed 170 to meet assumptions of normality and homogeneity of variance. We performed principal 171 coordinates analysis (PCoA) using a weighted UniFrac distance matrix of the ASV 172 dataset (Supplementary Data: Methods). 173 We calculated β -diversity as the individual sample scores along the first three 174 PCoA axes. Plant species (Hedera spp. or O. hederae), organ type (leaf or root), the 175 interaction between species and organ type, and usable reads were treated as fixed 176 effects, and site and the interaction between species and site were treated as random 177 effects. Using the model object from 'lmer', we tested the significance of fixed effects 178 with type III ANOVA from the R package 'car' v. 3.0-0 using the Kenward-Roger degrees 179 of freedom approximation (Fox and Weisburg, 2011). To test the significance of random 180 effects we used 'ranova' from the R package 'LmerTest' v. 3.0-1 (Kuznetsova et al., 181 2015) to perform likelihood ratio tests comparing full and reduced models. Next (Love et al., 2014) and 'ALDEx2' v. 192 1.12.0 (Fernandes et al., 2013). With each method we tested whether bacterial taxa 193 exhibited differential abundance using selected contrasts: and Hedera infection status influenced bacterial community structure we inferred 201 bacterial co-association networks (Layeghifard et al., 2017) for each of the leaf and root 202 community types occurring in Orobanche and Hedera (i.e. PL, PR, UL, UR, IL, IR, IIR [n 203 = 12 for each type]). We used two methods, which utilize the raw read counts of 204 individual ASVs to infer co-association networks and are designed to be robust to the 205 compositional and sparse nature of microbiome datasets (SparCC [Friedman and Alm, 206 2012]; SPIEC-EASI [Kurtz et al., 2015]). To reduce the bias of anomalous ASVs found 207 at particular sampling sites we included only ASVs found with at least 10 reads in 50% 208 of samples (Berry and Widder, 2014). We applied this threshold for each community 209 type separately because some ASVs were unique to single community types (e.g.  Table S4) and thus would have been filtered out had the threshold been applied to 211 the entire dataset. This resulted in 7 ASV subsets representing a fraction of the total 212 sequenced reads for each community type (number of ASVs/fraction of total: IIR, 213 229/63%; IUR, 285/61%; UR, 289/60%; PR, 113/51%; IL, 11/50%; UL, 10/54%; PL, 214 23/56%). 215

216
The network inference yields a set of bacterial ASVs (nodes) connected by edges, which 217 represent significant co-associations (either positive or negative co-association 218 occurring across samples). To compare networks across sample types we used the R 219 package 'igraph' (Csardi and Nepousz, 2006) to calculate whole network and ASV-level 220 properties thought to be related to individual and community function (Röttjers snd 221 Faust, 2018). For each network, we calculated two measures of individual ASV 222 centrality: degree, the number of edges connected to an ASV; and betweenness 223 centrality, the proportion of the shortest edge paths connecting other members in the 224 network occupied by an ASV (Freeman, 1978). ASVs with high degree number have 225 numerous co-associations with other network members and may be indicative of hub 226 species (Agler et al., 2015), while ASVs with high betweenness centrality may be 227 mediating interactions between other network members (Röttjers snd Faust, 2018). We 228 used a resampling approach to test whether networks varied in their measures of mean 229 ASV centrality (Supplementary Data: Methods). At the whole network level we 230 calculated edge density and network betweenness centrality. Edge density measures 231 the observed proportion of all possible co-associations among ASVs and network 232 betweenness centrality measures the evenness of betweenness centrality among 233 network members (Freeman, 1978 between two bacterial communities is given by the Procrustes correlation-like statistic 245 (t 0 ), which ranges from 0 (complete discordance) to 1 (perfect congruence). High 246 congruence between two community types (e.g. IIR and PR) indicates that 247 compositional shifts in one community are closely matched by parallel compositional 248 shifts in the other community (Fig. S7). To understand how individual bacterial taxa may 249 be contributing to the congruence or discordance between parasite and host microbiota 250 we used a leave-one-out approach (Wang et al., 2012). We removed all bacterial ASVs 251 from the parasite dataset classified to a given bacterial phylum, re-calculated weighted 252 UniFrac distances among all samples, obtained sample scores from a new PCoA, and 253 re-calculated t 0 . The effect of excluding a particular bacterial clade on the fit between 254 host and parasite microbiota is given by Δ t = (t excluded -t 0 ). After repeating this for each 255 bacterial phylum we iterated the leave-one-out approach across bacterial orders within 256 phyla whose exclusion lead to large Δ t. We performed the entire leave-one-out 257 approach on root and leaf bacterial communities separately (PR vs. IIR; PL vs. IIR). We 258 used the function 'protest' with 999 permutations from the R package 'vegan' v2.5-2 259 (Oksanen et al., 2018), which calculates t 0 , and performs the permutation test.

265
Orobanche root microbiota are less diverse, compositionally dissimilar, and 266 display far fewer co-associations than Hedera 267 Orobanche roots, but not leaves, had reduced α -diversity relative to Hedera (Figs. 2A, 268 S1; Table S2). Both leaf and root bacterial communities of Orobanche were 269 compositionally distinct from those of Hedera (Figs. 2B,C, S2; We also found a large number of ASVs unique to Orobanche leaves and roots (Table  274 S4). Orobanche leaf and root communities exhibited greater compositional similarity 275 than those of Hedera ( Table S6). Compared to Hedera, the root bacterial network of Orobanche 293 displayed a near absence of co-associations among ASVs (Fig. 4A-D). This absence 294 was reflected in the measures of mean ASV centrality, which was approximately ten-fold 295 lower in Orobanche versus Hedera root bacterial networks (Figs. 4F, Smirnov test, D = 1, P < 0.001; results are qualitatively similar for degree number). The 297 lack of co-associations in the Orobanche root bacterial network is also evident from the 298 diminished number of associations per bacterial ASV (i.e. degree distribution) in the 299 Orobanche network (Fig. 4E). Additionally, the edge density and network betweenness 300 centrality of the Orobanche root bacterial network was substantially reduced relative to 301 Hedera ( Fig. 4A-D), indicating that the reduced number of co-associations in the 302 Orobanche root bacterial network is an attribute shared among all constituent ASVs. In 303 contrast to plant roots, we found few significant associations between members of leaf 304 bacterial communities ( Fig. S6; Table S6).  Table S7). Leaf and root communities within a given species were not 311 congruent (Orobanche: t 0 = 0.50, P = 0.11; Hedera: t 0 = 0.33, P = 0.53). Hedera root 312 communities were congruent with soil bacterial communities in infected but not 313 uninfected patches (infected patches: t 0 = 0.91, P = 0.04; uninfected: t 0 = 0.83, P = 0.33). 314 Orobanche roots displayed high but non-significant congruence to soil communities in 315 infected patches (t 0 = 0.95, P = 0.08). 316 317 Our leave-one-out approach revealed that a subset of bacterial taxa contribute strongly 318 to either the congruence or discordance between Orobanche and Hedera leaf and root 319 communities (Figs. S7-S10; Tables S8, S9). In root communities, excluding the 320 Burkholderiales (Proteobacteria) led to a large decrease in the Procrustes goodness-of-321 fit (Δ t 0 = -0.133). By contrast, excluding the Actinomycetales (Actinobacteria) and 322 Flavobacteriales (Bacteroidetes)  In contrast to our predictions, infection status had no effect on the overall diversity or 335 composition of Hedera leaf and root bacterial communities (Figs. 2A-C; Table S2, S3),  336 although, several ASVs were unique to infected leaves and roots (Table S4). Few 337 bacterial taxa were affected by infection status but these findings appeared to be 338 sensitive to analysis method (Fig. 3, S4; Table S5). Considering only DESeq2 results, a 339 number of taxa did exhibit differential abundance in infected Hedera roots and leaves 340 consistent with either localized or systemic effects of parasitic plant infection on host 341 plant microbiota (Fig. 3: highlighted taxa). Moreover, our network analyses revealed that 342 infected Hedera root bacterial communities had higher mean ASV and network 343 betweenness centrality than uninfected roots from both infected and uninfected 344 individuals ( Fig. 4F; Kolmogorov-Smirnov test: D = 0.45, P < 0.001). Thus, infection 345 leads to an increase in the mean but also unevenness in centrality among ASVs 346 associated with Hedera roots. 347 348

349
The bacterial microbiota of a root holoparasite 350 The lower bacterial diversity we found in Orobanche roots parallels their reduced 351 morphological and anatomical structure ( Fig. 2A) (Tate, 1925), perhaps due to the 352 availability of fewer niches for microbes. In addition to overall diversity, Orobanche and 353 Hedera significantly differed in the composition of both leaf and root bacteria (Fig. 2B, 354 C). Variation in functional traits (e.g. leaf and root mass per area, leaf nitrogen content) 355 and ecological strategies among plants are thought to contribute to differences in leaf 356 and root microbiota among plant species (Kembel et al., 2014;Laforest-Lapointe et al., 357 2016;Fitzpatrick et al., 2018). Our results support this paradigm for bacteria composition 358 and root diversity, but foliar bacteria appear less sensitive to a shift to heterotrophy in 359 their host that root-associated bacteria (Fig 2B, Fig. S4, Table S5). 360 361 Like other ecological communities, interactions among microbial species are thought to 362 be an important determinant of the overall composition and function of plant microbiota 363 (Agler et al., 2016;Durán et al., 2018). Remarkably, we found a near absence of 364 microbial co-associations in the roots of Orobanche, while in Hedera roots we were able 365 to robustly identify numerous co-associations among bacterial taxa (Fig. 4). Shi et al. 366 (2016) proposed that the increased complexity of microbial networks found in 367 rhizosphere versus bulk soil could be due to increased interactions among rhizosphere 368 taxa including microbial cross-feeding, competition and other forms of antagonism. 369 However, niche differences can also result in significant negative and positive co-370 associations among microbial taxa as environmental variation across habitats, including 371 hosts or host organs, drives species co-occurrence (Zhou et al., 2011;Shi et al., 2016;372 Freilich et al., 2018). In the context of plant roots, simpler bacterial networks could be 373 the result of fewer persistent microbial interactions and/or available niches across 374 sampled plants. This is not to say that microbial interactions or environmental filtering 375 are absent but rather that they are inconsistent across Orobanche roots, suggesting a 376 greater role for stochastic processes in bacterial community assembly, though this 377 remains to be tested experimentally. parasites. Nonetheless, based on the reduced diversity and simpler network structure in 386 this host-parasite system, we propose that the concept of "parasitic reduction syndrome" 387 (Colwell, 1994) may be expanded to include microbiome reduction as well. 388 389

Assembly of the parasitic plant microbiota 390
Holoparasitic plants obtain their requisite energy, nutrients, and water from their hosts 391 by way of haustoria, which also allow symplastic and apoplastic transfer of nucleic acids, 392 proteins, and microorganisms (LeBlanc et al., 2012;Spallek et al., 2017). Consequently, 393 the assembly of parasitic plant microbiota is likely shaped by factors associated with 394 host plants (Sheng-Liang et al., 2014;Kruh et al., 2017;Cui et al., 2018). In this study, 395 Orobanche leaf and root bacterial communities displayed compositional shifts congruent 396 with root bacterial communities from infected Hedera (Fig. 5; Table S6). Importantly, we 397 found strong congruency between surrounding soil bacterial composition and the root 398 communities of Hedera, but not Orobanche, indicating that soil environmental features 399 are not solely driving corresponding shifts in parasite and host root microbiota ( Fig. 5; 400 Table S6). Instead, the haustorial transfer of microorganisms or plant-derived molecules, 401 which may act to structure microbiota occurring in both Orobanche roots and leaves, 402 could drive this congruence (Kruh et al., 2017). Though we cannot exclude the 403 possibility that Orobanche may be shaping the microbiota of their hosts, we think it 404 unlikely due to the predominantly (albeit not exclusively) one-way flow of haustorial 405 transfer from host to parasite (Serghini et al., 2001). For example, O. hederae can 406 sequester antimicrobial polyacetylenes from ivy hosts (Avato et al., 1996;407 Sareendenchai and Zidorn, 2008). Orobanche sequestration coupled with host plant 408 variation in the identity or abundance such molecules would lead to congruence in the 409 composition of associated microbiota between parasite and host. Further study using a 410 holoparasite with a wider host breadth such as Aphyllon purpureum, which parasitizes 411 various members of the Apiaceae, Asteraceae, and Saxifragales (Schneider et al., 412 2016), could test this hypothesis. 413 414 Particular bacterial taxa contributed most strongly to our observed congruence or 415 discordance between Orobanche and Hedera microbiota (Figs. S7, S8; Tables S8, S9). 416

Removing the Burkholderiales reduced the congruence between Orobanche and 417
Hedera root microbiota, indicating that their abundance is tightly linked across hosts and 418 parasites (Fig. S8, S9A). In contrast, removing the Actinomycetales increased 419 Orobanche and Hedera root microbiota congruence, indicating that their abundance is 420 decoupled across parasitic plants and hosts during infection (Fig. S9B). The fact that 421 different bacterial taxa contribute to the congruence of either Orobanche leaf or root, 422 and Hedera root microbiota, respectively, (Figs. S7, S8; Tables S8, S9) lends support to 423 our previous finding that the mechanisms structuring parasite leaf and root communities 424 are not entirely overlapping. Though their phylum-level makeup was distinct (Fig. 2C), 425 the leaf and root communities of Orobanche were compositionally similar (Fig. 2B inset). 426 Two non-exclusive explanations for this include increased overlap in the microbial 427 habitats of Orobanche leaves and roots versus Hedera, or perhaps a larger role for 428 stochastic processes governing the assembly of Orobanche microbiota. The reduced 429 complexity found in the Orobanche root bacterial network further supports a diminished 430 role of microbe-microbe interactions and niche-based processes in microbiome 431 assembly. The patterns of Orobanche bacterial diversity and community assembly 432 exhibit intriguing parallels with the microbiota of eukaryotic parasites of animal hosts 433 (e.g. Husnik, 2018) and suggest that general microbial dynamics may exist in host-434 parasite systems across plant and animal kingdoms (Dheilly, 2014;Dheilly et al., 2015). 435 436

The effect of infection status on host plant microbiota 437
Infection by a parasitic plant can induce host plant morphological, physiological, 438 molecular, and transcriptional responses (Westwood et al., 1998;Castillejo et al., 2009;439 Hiraoka et al., 2009;Hegenauer et al., 2016). In response to infection by the root 440 holoparasite O. cernua, sunflowers synthesize coumarins (Serghini et al., 2001), 441 compounds that are known to inhibit fungal pathogens and also reshape the root 442 microbiome due to antimicrobial effects (Stringlis et al., 2018). In our study, several 443 bacterial taxa exhibited differential abundance consistent with either localized or 444 systemic effects of parasitic plant infection on host microbiota (Fig. 3)

. The genus 445
Phytohabitans was reduced in only directly infected roots, while Flavobacterium was 446 reduced in both infected and uninfected roots from infected individuals, indicative of 447 systemic effects of infection on Hedera root microbiota. Kruh et al., (2017)

found that 448
Flavobacterium was enriched in the post-attachment but pre-inflorescence stage of the 449 parasitic Phelipanche aegyptiaca during infection of tomato plants. Interestingly,450 Pseudonocardia, was enriched only in infected Hedera leaves, which suggests that 451 microbial perturbations as a result of infection can be localized to tissues not directly 452 infected, potentially linking above and belowground host ecology (Press and Phoenix, 453 2005). Members of the Pseudonocardia are plant endophytes and also act as antifungal 454 mutualists for various species of attine ants (Sen et al., 2009). In spite of these shifts in 455 taxon abundance, we found that infection status had no effect on the overall diversity or 456 composition of Hedera microbiota. In contrast, Kruh et al., (2017) found that the 457 microbiota of tomato roots parasitized by P. aegyptiaca in a greenhouse setting 458 exhibited compositional similarity to that of their parasite when infected, but also were 459 distinct from the roots of uninfected hosts. However, we found that infected Hedera root 460 bacterial networks exhibited elevated network and ASV-level betweenness centrality 461 relative to uninfected Hedera roots (Fig. 4), suggesting that particular ASVs become 462 increasingly co-associated to others in infected roots. Alternatively, under infection, host 463 roots may promote the development of novel microbial niches, which could perturb the 464 co-associations among bacterial taxa resulting in an increase in the co-occurence of a 465 small number of taxa (Shi et al., 2016). Recent work demonstrates that root associated 466 microbiota of host plants may play an important role in mitigating the negative effects of 467 parasitic plant infection (Sui et al., 2018). To mechanistically link infection status and 468 microbiota, future work should characterize the effect of microbial inoculations across 469 resistant and susceptible genotypes during experimentally controlled parasitic plant 470 infections (e.g. Castillejo et al., 2009;Castrillo et al., 2017).   Zhou J, Deng Y, Luo F, He Z, Yang Y. 2011. Phylogenetic molecular ecological network of soil microbial communities in response to elevated CO2. mBio 2, e00122-11 Figure 1 An overview of the study system and sampling design. (A) We sampled leaves and roots from Orobanche (inflorescence in foreground) and from infected and uninfected Hedera (green plant in background). Note that the abbreviations used here are used throughout the text. (B) We carefully excavated individual Orobanche and its Hedera host to sample leaves and roots from both organisms. For infected Hedera, we distinguished between infected roots (IIR), which were physically attached to the parasite, and uninfected roots (IUR), which did not exhibit direct physical attachment to any parasite. For uninfected Hedera, we sampled leaves and roots in a similar fashion. (C) We sampled sampled leaves (PL) and roots (PR) for Orobanche, which are homolgous but lack the functionality of leaves and roots found in non-parasitic plants.
(D) We sampled plants and soil from paired infected and uninfected ivy patches at four sites on the University of California, Berkeley campus. From infected patches we collected Orobanche and Hedera samples and from uninfected patches we collected only Hedera samples.

Figure 2
Diversity and composition of the bacterial communities found in Orobanche and Hedera. (A) Inverse Simpson's diversity across leaves and roots of both plant species (n = 12 for each community type). Roots exhibited nearly 4-fold higher diversity than leaves (Table S2; F 1,21 = 17.17, P < 0.001) and parasites had reduced root, but not leaf diversity relative to hosts (Table S2; F 1,5 = 5.45, P = 0.06). Hedera infection status had no effect on leaf or root diversity (Table S2). (B) Principle Coordinates Analysis of the weighted UniFrac dissimilarity among bacterial communities. Community composition strongly varied between leaves and roots (Table S3; F 1,22 = 64.04, P < 0.001) and plant species (Table S3; F 1,13 = 51.32, P < 0.001). We also found a significant interaction between plant species and organ type (Table S3; F 1,5 = 65.50, P < 0.001), reflecting the larger compositional differences found between leaf and root microbiota for Hedera versus Orobanche (H versus O, shown in 2b inset; paired t-test: t = -2.52, P = 0.01). Hedera infection status had no main effect on the bacterial community composition, though we found a significant interaction between root infection status and sampling site (Table S3; χ 2 = 9.12, P = 0.003). (C) The relative abundance of the major bacterial phyla found across leaves and roots of both plant species (n = 12 for each community type).

Figure 3
Differential abundance of bacterial genera across plant species and infection status. We tested whether bacterial genera (labelled according to phylum classification), exhibited differential abundance across five specific contrasts (n = 12 for each contrast level). For example, "PR vs. IIR" tested whether bacterial families in Orobanche roots (PR) exhibited enriched or reduced abundance relative to infected Hedera roots (IIR). We used two analytical methods to test for differential abundance, DESeq2 and ALDEx2, and display the overlapping results in darker shades and the results unique to DESeq2 in lighter shades (e.g. both methods found that the Amycolatopsis [Actinobacteria] were reduced in Orobanche roots relative to Hedera, but only DESeq2 found that they were enriched in uninfected versus infected leaves). Contrasts highlighted with arrows represent localized or systemic effects of infection status on Hedera microbiota, and differences between Orobanche and Hedera. We repeated the analysis at all bacterial taxonomic ranks (see Fig. S4 and Table S5 for full results).

Figure 4
Orobanche and Hedera root bacterial networks inferred by SPIEC-EASI. We inferred the bacterial network of (A) infected roots of infected Hedera, (B) uninfected roots of infected Hedera, (C) roots of uninfected Hedera, and (D) Orobanche roots (n = 12 for each community type). Node colour and size represent bacterial phylum classification and abundance (centered log-ratio transformed), respectively. Edge colour and width represent sign (green = positive association, red = negative association), and strength of co-association, respectively. At the whole network-level, we found large differences in the edge density and betweenness centrality between Hedera and Orobanche, but not across infected and uninfected Hedera roots (see Table S6), as reflected in the (E) degree distribution (number of associations per node) among community types. (F) We also found large significant differences in mean betweeness centrality of individual taxa among the root bacterial networks of Hedera and Orobanche, as well as infected and uninfected Hedera roots. We tested significance using a series of Kolmogorov-Smirnov tests on the distributions of mean node-level betweenness centrality estimated from 50 nodes sampled with replacement 10,000 times (see Materials and Methods).

Figure 5
Congruence in compositional change across host and parasite leaves and roots. Circle size is proportional to the Procrustes correlation between PCoA ordinations of weighted UniFrac dissimilarity. For example, compositional change among Orobanche leaf bacterial communities (PL) was not correlated (t 0 = 0.5, P > 0.05) with compositional change in Orobanche roots (PR). Instead turnover among PL communities was positively correlated with compositional change among both infected (IIR: t 0 = 0.62, P < 0.05) and uninfected (IUR: t 0 = 0.61, P < 0.05) roots of infected hosts. See Table S7 for additional comparisons.