Plant species identity and plant-induced changes in soil physicochemistry—but not plant phylogeny or functional traits - shape the assembly of the root-associated soil microbiome

Abstract The root-associated soil microbiome contributes immensely to support plant health and performance against abiotic and biotic stressors. Understanding the processes that shape microbial assembly in root-associated soils is of interest in microbial ecology and plant health research. In this study, 37 plant species were grown in the same soil mixture for 10 months, whereupon the root-associated soil microbiome was assessed using amplicon sequencing. From this, the contribution of direct and indirect plant effects on microbial assembly was assessed. Plant species and plant-induced changes in soil physicochemistry were the most significant factors that accounted for bacterial and fungal community variation. Considering that all plants were grown in the same starting soil mixture, our results suggest that plants, in part, shape the assembly of their root-associated soil microbiome via their effects on soil physicochemistry. With the increase in phylogenetic ranking from plant species to class, we observed declines in the degree of community variation attributed to phylogenetic origin. That is, plant-microbe associations were unique to each plant species, but the phylogenetic associations between plant species were not important. We observed a large degree of residual variation (> 65%) not accounted for by any plant-related factors, which may be attributed to random community assembly.


Introduction
The functional activities of the root-associated soil microbiome are of fundamental importance for plant health.The rootassociated soil microbiome influences plant nutrient acquisition, pathogen defense, induced systemic r esistance, gr owth, dr ought tolerance, and other traits (Berendsen et al. 2012, Philippot et al. 2013, Pieterse et al. 2014 ).Given the importance of the microbiome to plant fitness and health, soil fertility, and ecosystem producti vity, defining the k e y pr ocesses that sha pe micr obial assembl y has been an ongoing pursuit within microbial ecology (Marschner et al. 2001, Lundberg et al. 2012, Pér ez-Jar amillo et al. 2016, Fitzpatrick et al. 2018 ).Whilst many studies have performed observ ational r esearc h to describe patterns of microbial assembly and comm unity ecology (Pr osser 2020 ), fe wer studies hav e performed direct manipulations in controlled experiments to examine the factors sha ping r oot-associated soil micr obial assembl y during a plant's early phases of growth and establishment.
Se v er al studies hav e pr oposed that plant species (PS) themselv es ar e the primary determinants of their associated micr obiome (Becklin et al. 2012, Mendes et al. 2013, Bouffaud et al. 2014, Cha parr o et al. 2014 ).Theor eticall y, as the phylogenetic relatedness of PS influences their degree of shared developmental and functional traits , it ma y also influence the phylogenetic similarity of the micr oor ganisms that they recruit.T hus , with increasing phylogenetic similarity among PS, one may observe an increased relatedness of their microbiome.Findings supporting this hypothesis have been observed in several studies (Bouffaud et al. 2014, Lambais et al. 2014, Lei et al. 2019, Hartman et al. 2023 ).In contr ast, other studies hav e observ ed that the abiotic conditions of the soil en vironment (i.e .soil type , pH, nutrient a vailability, and C:N r atio) ar e of gr eater influence on micr obial comm unity assembly in the rhizosphere and root-associated soil environment (Girvan et al. 2003, Ulrich and Becker 2006, Lauber et al. 2009, Xiao et al. 2017, Yeoh et al. 2017, Veach et al. 2019, Ren et al. 2020 ).
This ecological conundrum dr aws par allels to the naturev ersus-nurtur e debate that has shaped research around human de v elopment for decades.Se v er al studies hav e pr oposed an assembl y model mor e akin to natur e-via-nurtur e, wher eby both the PS and soil shape the microbiome, and the relative strength of these different drivers will vary depending on the specific ecological context (Garbe v a et al. 2008, Berg and Smalla 2009, Tkacz et al. 2015, Müller et al. 2016, Lee and Hawkes 2020 ).In the naturevia-nurture model, soil provides the primary source of microbial inoculum available to plants and sets the boundaries from which plants may select their microbiome.The dominant influence of soil type and edaphic properties on determining the broad patterns of microbial biogeography was recognized by Fierer and Jackson ( 2006 ) and Lauber et al. ( 2009 ).Howe v er, thr oughout their de v elopment and life span, plants and their root systems exert species-specific influences on the rhizosphere and root-associated soil envir onment, whic h driv es envir onmental filtering of their micr obiome (Ber g and Smalla 2009, Cha parr o et al. 2014, Reinhold-Hurek et al. 2015, Hu et al. 2018 ).Additionally, the symbiotic associations of plant hosts (e.g.N 2 -fixing rhizobia, arbuscular mycorrhizal fungi) have been identified to shape the assembly of the r oot micr obiome (Hartman et al. 2023 ).
T here ha ve been two primary processes proposed that shape microbiomes: deterministic and stochastic (Goss-Souza et al. 2017 ).Nic he-based, deterministic models pr opose that the biotic and abiotic conditions of the local envir onment driv e micr obial selection (Carroll et al. 2011, Goss-Souza et al. 2017 ).Deterministic models can be further split into primary and secondary processes.Primary deterministic processes constitute a more direct mec hanism, wher eby the r elease of plant-specific rhizo-deposits selects or favours microbial taxa from the wider soil microbial community (Hu et al. 2018, Sasse et al. 2018, Zhalnina et al. 2018 ).Secondary deterministic processes function indir ectl y wher eby plant roots modify the general rhizosphere and soil conditions (pH, available P, nitrogen, etc.), and these changes, in turn, encourage the growth of microorganisms best adapted to that modified habitat space (Hinsinger 2001, Liang et al. 2002, Bell et al. 2015, van Veelen et al. 2020, Hernández-Cáceres et al. 2022 ).In contrast to deterministic models, stochastic models propose an element of randomness to community assembly (Dini-Andreote et al. 2015, Goss-Souza et al. 2017 ).These deterministic and stochastic processes do not occur independently of each other, and the challenge is determining the relative contribution of these under different experimental and ecological contexts.
We performed a plant experiment whereby a broad phylogenetic range of 37 different PS were grown in an identical blended soil medium.Following 10 months of growth, the root-associated soil microbiome was characterized using the 16S rRNA gene and ITS region sequencing.The differences in the structural variance of the root-associated soil microbiome between PS were related to their phylogenetic and functional traits, as well as to any plant-induced changes in soil physicochemistry (SC) that occurr ed thr oughout the experiment.By performing this experiment, our r esearc h aimed to partition the influences of primary and secondary deterministic pr ocesses (i.e.dir ect and indirect plant effects) and stochastic processes on microbial assembly in the root-associated soil en vironment.Furthermore , we hypothesize that the phylogenetic relatedness of the PS will be positively correlated to the phylogenetic similarity of their root-associated soil microbiome.

Plant and soil sample collection
Plants from 37 different species were grown in a blended soil media or obtained from a commercial nursery (Southern Woods Plant Nursery, New Zealand).The selected PS cov er ed a broad range of phylogenetic groups and included r epr esentativ es fr om three plant classes, 12 orders, 14 families, and 31 genera.The PS cov er ed a range of different life spans (annual, perennial, or long-lived), functional groups (e .g. grass , shrub, or tree), and pr ov enances (exotic or native to Aotearoa New Zealand).They also included species with different mycorrhizal associations (arbuscular, AMF, ectomycorrhizal, EMF, and no association) and N 2 fixation (presence or absence).Plant metadata was primarily obtained from PS profiles on the New Zealand Plant Conservation Network ( https:// www.nzpcn.org.nz/ ) and from literature searc hes wher e additional information was needed.The full list of PS used in this study and their associated metadata are provided in Supplementary Table S1 .
The seeds or cuttings of each PS were planted in individual 10-L pots containing a blend of field-collected live soils mixed with a pasteurized soil: sand carrier for bulk.Between 12 and 20 r eplicate pots wer e established for eac h PS.The collection of the live soils was conducted across 12 sub-alpine , grass , and shrubdominated sites that included the ranges of the plants being experimentall y e v aluated.This sampling design was performed to allow for the microbiomes associated with these species to be available for plant 'recruitment'.Ho w ever, it is important to state that we did not examine the microbial background of these 'live' soils before setting up the experiment.Details regarding the collection, handling, treatment, and mixing of these soils were first provided by Wakelin et al. ( 2021 ).The plants were randomized within a glasshouse and grown with regular watering and supplemental lighting when r equir ed.No fertilizers or other chemicals were added to the pots, and weeds were removed when apparent.
After 10 months of plant gr owth, r oot-associated soil samples were collected from each plant pot.Samples were collected from between four and se v en r eplicate pots of eac h plant depending on plant availability [i.e.plants that had grown to full health and were not required for other research (Wakelin et al. 2021 )].All root-associated soil samples were collected aseptically by pressing an open 50-mL conical centrifuge tube into the soil adjacent to the stem(s) of the plant in each pot directly into the root zone.Thr ee samples fr om ar ound the circumfer ence of eac h pot wer e collected and pooled to provide a single sample for each replicate of each PS.Pooled soil samples wer e sie v ed to 2 mm and stored at either 4 ˚C until physicoc hemical anal ysis or -80 ˚C until DNA extraction.

Plant DNA extraction and matK gene sequencing
The DNA of each PS was extracted using the DNeasy Plant Mini Kit (QIAGEN), utilizing cryogenic tissue grinding of plant leaves with a sterilized mortar and pestle in liquid nitrogen.For phylogenetic inference, the Maturase K gene (matK) was amplified using the primers MatK472F (5 -CCRTCA TCTGGAAA TCTTGGTT-3 ) and MatK1248R (5 -GCTRTRA T AA TGA GAAA GATT TCTGC-3 ) (F atima et al. 2019 ).PCR conditions consisted of an initial denaturation step of 94 ˚C for 5 min follo w ed b y 35 cycles of 94 ˚C for 30 sec, 56 ˚C for 30 sec, and 72 ˚C for 42 sec, follo w ed b y a final extension at 72 ˚C for 10 min.The PCR r eaction mixtur e consisted of 1 × PCR buffer, 0.5 mmol L −1 dNTPs, 0.25 μmol L −1 of each primer, 1 U Taq pol ymer ase, and 5-50 ng of template DNA.The PCR products were purified using the QIAquick PCR Purification Kit, and the purified DN A w as sequenced using Sanger sequencing at Macrogen (Seoul, K orea).T he quality of the sequencing data was c hec ked and edited using Sequencer software version 5.4.6 (Genecodes Corp, Ann Arbor, MI, USA).MEGA X (Kumar et al. 2018 ) was then used for sequence alignment and phylogenetic analysis.Briefly, matK genebased sequences were aligned using MUSCLE, and overhanging nucleotides were removed and then re-aligned.Distance matrices and phylogenetic tr ees wer e constructed using the maximum likelihood method and the Tam ur a-Nei model (Tam ur a and Nei 1993 ).

Soil DNA extraction and 16S rRNA gene/ITS region sequencing
Soil DN A w as extr acted fr om 0.25 g of soil using a DNeasy Po w erSoil Kit (QIAGEN) accor ding to the manufacturer's protocol and quantified using a Nanodrop spectrophotometer.Subsequent Illumina amplicon sequencing follo w ed the Earth Microbiome Pr oject's (EMP) pr otocol (Ca por aso et al. 2012 ).In short, the bacterial 16S rRNA gene was amplified using the primers 515F (5 -GTGYCAGCMGCCGCGGTAA -3 ) and 806R (5 -GGA CTA C-NVGGGTWTCT AA T -3 ) targeting the V4-V5 regions as described pr e viousl y (Apprill et al. 2015(Apprill et al. , P ar ada et al. 2016 ) ).The fungal ITS region was amplified using the primers ITS1f (5 -CTTGGTCATT-TA GA GGAA GTAA -3 ) and ITS2 (5 -GCTGCGTTCTTCA TCGA TGC -3 ) as described pr e viousl y (Bokulic h andMills 2013 , Hoggard et al. 2018 ).After PCR amplification, samples were purified using a Magnetic Bead PCR Cleanup Kit (GeneaidTM) and pooled in equimolar concentrations .T he purified PCR products were used to prepare DN A libraries follo wing the Illumina TruSeq DNA library preparation protocol using the Illumina MiSeq Reagent Kit v2.Illumina sequencing was performed at the Australian Genome Research Facility (Melbourne, Australia) using 2 × 150 bp pair-end chemistry on a MiSeq platform following the manufacturer's guidelines.

Sta tistical anal ysis
Following sequencing, paired-end fastQ files were processed into amplicon sequence variants (ASVs) using the D AD A2 version 1.18 w orkflo w (Callahan et al. 2016 ).Briefly, the forw ar d and r e v erse r eads wer e quality-filter ed, trimmed, and denoised befor e being merged into ASVs.Chimeric ASVs were removed, and taxonomies were assigned to each ASV using the Ribosomal Database Project (RDP) Classifier (Wang et al. 2007 ) and the UNITE (Abar enk ov et al. 2021 ) databases.Following D AD A2 processing, ASV count tables wer e filter ed to r emov e unidentified and unwanted phyla (i.e.Cyanobacteria/Chloroplasts) and singletons .T he ASV count tables wer e r ar efied to adjust for differ ences in libr ary size between samples.Befor e r ar efaction, samples with low r ead counts wer e r emov ed to avoid excessive data loss.Rarefaction curves displaying the number of ASVs in each sample have been provided in Supplementary Figs S1 and S2 .The number of replicates per PS that were included in the rarefied 16S (henceforth reported as 'bacterial') and ITS (henceforth reported as 'fungal') ASV datasets is displayed in Supplementary Table S1 .In total, all the PS had at least three replicates in the rarefied fungal ASV dataset.In the r ar efied bacterial ASV dataset, 35 out of the 37 PS had at least thr ee r eplicates; ho w e v er, onl y two r eplicates r emained for the PS Chionochloa conspicua and Trifolium repens following rarefaction.
The r ar efied bacterial and fungal ASV datasets wer e anal ysed separ atel y using the m ultiv ariate statistical anal yses outlined below.Maximum likelihood phylogenetic trees were built using Fast-Tree2 (Price et al. 2010 ).To provide estimates of alpha diversity, Faith's phylogenetic diversity (PD) and species richness (SR) were calculated for each sample in Picante R (Kembel et al. 2010 ).The PD index assesses the PD of a community and is defined as the sum of the total phylogenetic br anc h length separating taxa in a community (Faith 1992, Kembel et al. 2010 ).In contrast, the SR index calculates the total number of taxa in a community based on their identity alone-no phylogenetic information is factored into the calculation.The differences in the PD and SR index between plant host-related factors were tested for significance using Kruskal-Wallis tests and pairwise Wilcoxon tests with Bonferroni correction.
To estimate the phylogenetic distances in microbial community composition between samples, weighted UniFrac distances were calculated on rarefied bacterial and fungal ASV count tables (Lozupone et al. 2011 ).Differences in bacterial and fungal community composition between the plant-related factors were tested for significance using permutational multiple analysis of variance (PERMANOVA) on distance metrics using the adonis2 (by = 'terms') function in vegan R (Oksanen et al. 2019 ) and pair-wiseAdonis (Martinez Arbizu 2020 ).The differences in community composition were visualized using non-metric multidimensional scaling (NMDS) ordination plots.To estimate the withingr oup v ariance amongst samples , the a v er a ge distance of individual samples to the group centroid (beta dispersion) was calculated using the betadisper function in phyloseq R (McMurdie and Holmes 2013 ).Permutation tests were used to determine significant differences in the within-group variance between plantrelated factors.
The w eighted UniF rac distances w er e corr elated to differ ences in soil physicochemical properties using Mantel tests.In addition, w eighted UniF rac distances w er e corr elated to matrices of matK sequence similarity using Mantel tests.MatK similarity matrices were constructed to represent the phylogenetic relatedness between the different PS under investigation.Observations of the phylogenetic tr ee gener ated fr om the matK phylogen y sho w ed sensible grouping of PS to their taxonomic positioning.Hierarchical clustering analysis was performed on the w eighted UniF rac distance matrices and matK distance matrices using the complete linkage method in Stats R. Following this, dendrograms were constructed to visually compare differences in the clustering patterns of PS based on the weighted UniFrac distances of their fungal and bacterial communities versus their matK distances in ape R (Paradis and Schliep 2019 ).
Variance partitioning (VP) analysis was performed in vegan R to partition the variance observed in bacterial and fungal community composition (as r epr esented by weighted UniFrac distances) to the plant-related factors (Oksanen 2019 ).Four explanatory matrices were constructed to represent the different influencing factors .T hese were: PS; plant life history (PLH) (i.e .pro venance + life span + functional group); plant rhizosphere traits (PRT) (i.e.mycorrhizal association + N 2 fixation); and SC.All unexplained (residual) v ariation fr om VP anal ysis was tentativ el y assigned to r epresent the influence of stochastic processes.Following VP analysis, distance-based redundancy analysis (db-RDA) was performed to test the significance of each explanatory matrix whilst conditioning for the other three matrices.In addition, forw ar d stepwise selection was performed to identify the soil physicochemical properties that best accounted for the community variance that was partitioned to the influence of SC.Pairwise differences in the soil physicochemical properties selected by the forward selection model between the 37 different PS were identified using pairwise t -tests with Holm correction.
The r ar efied bacterial and fungal ASV tables wer e used as input for differential abundance analysis.First, the R package pime was used to select bacterial and fungal ASVs that best defined the microbiome of each PS (Luiz Fernando 2020 ).Prevalence intervals with an out-of-bag (OOB) error rate of 0% were selected as cut-offs.For fungal ASVs, this was a pr e v alence of 75%, whic h r etained 217 ASVs and 1 168 389 sequences.For bacterial ASVs, this was a pr e v alence of 80%, whic h r etained 771 ASVs and 433 795 sequences.PIME-filtered ASV count tables were used as input for differ ential abundance anal ysis using meta genomeSeq R (P aulson et al. 2013 ), where the log change estimate of each ASV between different PS was calculated using the fitLogNormal function.Significant differences in the log change estimates of ASVs between PS were determined using permutation tests ( n = 999) with correction for multiple comparisons using the Holm-Bonferroni method (holm).Heatma ps wer e pr oduced using pheatma p R (Kolde and Kolde 2018 ) to display (a) the bacterial and fungal ASVs with significant log change estimates across PS and (b) the correlation shar ed between differ ent PS (Pearson's) based on the log change estimates of their bacterial and fungal ASVs.

Microbial species richness and phylogenetic di v ersity
Ther e wer e no significant differences in the SR of bacterial comm unities acr oss an y of the plant-r elated factors (Table 1 ).Howe v er, the Faith's div ersity of bacterial comm unities was significantly higher in native versus exotic plants and in non-N 2 fixing versus N 2 fixing plants.
For fungal communities, both Faith's diversity and SR were significantly higher in annual versus long-lived plants (Table 1 ).The PD of fungal communities was significantly lo w er in tr ees v ersus shrubs and grasses and significantly higher in non-mycorrhizal versus ectomycorrhizal plants .T he mean ( ± SD) values for Faith's diversity and SR of fungal and bacterial across the different plant metadata factors can be seen through Supplementary Tables S2 -S11 .

Microbial beta-di v ersity and comm unity composition
Bacterial and fungal micr obial comm unity composition was significantl y differ ent between all plant-r elated factors (Table 2 ).PS and genus were the factors that reported the highest R 2 values, thus accounting for most of the explained variation in bacterial and fungal community composition (Table 2 ; see also Supplementary Figs S3 and S4 in Supplementary Data ).Although significant, the R 2 values for many of the plant-related factors r epr esenting functional plant traits (i.e .pro venance , functional group, primary mycorrhizal association, life span, and N 2 fixation) were all low ( R 2 < 0.07).Bacterial and fungal communities both exhibited a heterogeneous dispersion and a high degr ee of within-gr oup v ariability.The degr ee of beta-dispersion (' F v alue') observ ed in bacterial comm unities was significantl y differ ent acr oss the following factors: PS, plant genus, plant family, plant order, and mycorrhizal association (Table 2 ).For fungal comm unities, significant beta-dispersion v alues wer e observ ed by plant family, plant order, plant class, pro venance , and mycorrhizal association (Table 2 ).

MatK gene sequence similarity
The distances in matK gene sequence similarity between the different PS did not significantly correlate to the corresponding w eighted UniF rac distances for their bacterial (Mantel r = 0.134, P
When the other explanatory matrices were conditioned out of the model, plant identity alone accounted for 9.52% of bacterial comm unity v ariance and 3.65% of fungal comm unity v ariance.In contrast, SC accounted for 5.67% of bacterial community variance and 9.40% of fungal community variance.PLH (Bacteria: F value = 0.00, P value > 0.05, Fungi: F value = 0.00, P value > 0.05) and PRT (Bacteria: F value = 0.00, P value > 0.05, Fungi: F value = 0.00, P value > 0.05) did not significantly account for any bacterial or fungal community variation.
The composition of the root-associated soil microbiome may be indir ectl y influenced by plant-induced modification of the physicoc hemical envir onment.When looking at soil physicoc hemical properties that best accounted for bacterial community variation, forw ar d selection models identified Olsen P ( F value = 10.76,P value < 0.05), sulphate sulphur ( F value = 2.54, P value < 0.05), and pH ( F value = 4.54, P value < 0.05) to be significant.For performed on weighted UniFrac distance matrices, which were calculated using bacterial and fungal ASV tables .T he results of permutationbased tests of beta-dispersion are display ed ( F values), w ere performed to identify significant differences in the within-group variance of bacterial and fungal communities for each plant-related factor.fungal communities, forw ar d selection models identified Olsen P ( F value = 9.95, P value < 0.05), AMN:TN ( F value = 3.69, P value < 0.01), and volume weight ( F value = 2.85, P value < 0.05) to be significant.The values for these soil properties were variable between the 37 different PS (Fig. 2 ).Pairwise t -tests identified that Olsen P was significantly higher ( P adjusted < 0.05) in Acaena caesiiglauca (vs.Achillea millefolium , Dactylis glomerata , and Poa colensoi ), Alnus glutinosa (vs.Ach .millefolium , D. glomerata , Holcus lanatus , Ozothamnus leptoph yllus , and P. colensoi ), and Pinus radiata (vs.Ach .millefolium , D. glomerata , and Po .colensoi ).Volume weight was significantly higher in Ho. lanatus (vs.Hebe odora , O .leptophyllus , Olearia virgata , and Sophora microphylla ) and Muehlenbeckia complexa (vs.He .odora and S. microphylla ).Soil pH was significantly higher in D. glomerata and Ho .lanatus (vs.O. leptophyllus , Pi .contorta , Pi. r adiata , Br ach yglottis gre yi , Coprosma robusta , and Ulex europaeus ), Ach .millefolium (vs.Pi .radiata ), and Po .colensoi and P .cita (vs.Pi. radiata, O. leptophyllus , and B. gre yi ).Although forw ar d selection models identified AMN:TN and sulphate sulphur to significantly influence fungal and bacterial community composition, no significant pairwise differences w ere determined betw een the 37 PS for these properties .T he mean ± SD values for all soil physicochemical properties associated with eac h PS ar e pr esented in Supplementary Table S12 .

Taxonomic differentiation across plant species
Out of the 771 bacterial ASVs that wer e r etained following PIME filtering and used as input for differential abundance analysis, only 10.12% (78 ASVs) were identified to be differ entiall y abundant amongst PS ( P adjusted < 0.05).Furthermore, out of the 217 fungal ASVs retained following PIME filtering, only 16.59% (36 ASVs) wer e differ entiall y abundant amongst PS ( P adjusted < 0.05).Figures 3 and 4 display the bacterial and fungal ASVs that had significantl y differ ent log c hange estimates acr oss the PS under inv estigation.These r esults highlight that ther e wer e no lar ge patterns of taxonomic differentiation amongst PS, that is, PS did not hav e markedl y distinct taxonomic compositions.One exception was Agrostis capillaris (common bent or brown top grass), whose bacterial and fungal taxa were more evidently differentiated compared to the other PS.All the PS shared a significant ( P adjusted < 0.05) positiv e corr elation based on the log change estimates of Figur e 2. T he mean ± SE values of the soil properties that were identified by forward selection models to significantly account for the variation in root-associated soil microbial communities .T he soil properties Olsen P (mg/L), sulphate sulphur (mg/kg), and pH significantly accounted for bacterial comm unity v ariation, whilst Olsen P (mg/L), AMN:TN ratio, and v olume w eight (g/mL) significantl y accounted for fungal comm unity v ariation.their bacterial (Pearson's r correlation; 0.70 ± 0.08 SD) and fungal ASVs (Pearson's r correlation; 0.69 ± 0.08 SD).These results indicate a low div er gence of PS based on their root-associated soil microbiome ( Supplementary Figs S7 and S8 ).

Discussion
The root-associated soil microbiome provides fundamental roles in supporting plant health, productivity, and resilience against abiotic and biotic stressors (Mendes et al. 2011, Berendsen et al. 2012, Penton et al. 2014 ).T hus , pinpointing how different components of the plant-root-soil interface drive microbial selection and establishment is k e y for us to manage plant and soil health into the future.Ho w ever, identifying the primary processes that drive micr obial assembl y is complex and is suggested to be by the interacting influences of plant genotype, developmental stage, root exudates, r oot mor phology, PLH, soil type, and pr e vious soil history (Cha parr o et al. 2014, Zhao et al. 2019, Zhou et al. 2020, Cordovez et al. 2021 ).By controlling for the starting soil mixture and surrounding en vironmental conditions , our r esearc h aimed to identify how the different phylogenetic, functional, and ecological traits of PS wer e r elated to the assembly of their root-associated soil microbiome.

The phylogenetic relatedness of plant hosts shared no relationship to the similarity in their root-associated soil microbiome
Our r esearc h aimed to test whether the phylogenetic relatedness of PS was correlated with the phylogenetic similarity of their root-associated soil microbiome-a hypothesis that has been supported by pr e vious r esearc h (Bouffaud et al. 2014, Lambais et al. 2014, Yeoh et al. 2017, Lei et al. 2019, Kaplan et al. 2020, Hartman et al. 2023 ).Our results did not support this hypothesis, as the phylogenetic similarity in root-associated soil microbiomes did not correlate with the phylogenetic similarity between different PS.With the increase in phylogenetic ranking from PS to class level, we Figur e 3. T he log change estimates of the 78 bacterial ASVs that were identified to have significant differential abundance values ( P adjusted < 0.05) across the PS under in vestigation.T he taxonomic identity of each bacterial ASV is presented, with each bacterial ASV represented by the most refined taxonomic rank that could be accurately assigned.
Figur e 4. T he log change estimates of the 36 fungal ASVs that were identified to have significant differential abundance values ( P adjusted < 0.05) across the PS under in vestigation.T he taxonomic identity of each fungal ASV is presented, with each fungal ASV represented by the most refined taxonomic rank that could be accurately assigned.observed a consistent decline in the degree of microbial community variation that could be accounted for by plant phylogenetic origin.T hat is , higher phylogenetic r ankings suc h as plant class and order only explained a small amount of compositional variation compared to PS-level identity.This suggests that, whilst PS may be used as a predictor of the root-associated soil microbiome, higher taxonomic rankings of PS cannot.Similar findings were observ ed by Fitzpatric k et al. ( 2018), who identified that although PS identity was a significant factor shaping rhizosphere assembl y, the emer gent structur e of the rhizospher e micr obiome shar ed no relationship with the phylogenetic relatedness between plant hosts.

Plant species and plant-induced changes in soil physicochemistry were the strongest predictors of microbial assembly
Although patterns in micr obial assembl y did not relate to the phylogenetic relationships among PS, species identity and differences in SC were the two most significant factors that accounted for bacterial and fungal comm unity v ariation-a finding also observed by Burns et al. ( 2015 ).Whilst both factions were significant, PS identity accounted for a greater proportion of bacterial community v ariance than SC.Se v er al studies hav e r eported PS identity to be a significant factor in shaping microbial assembly and community structure (Garbeva et al. 2008, Berg and Smalla 2009, Becklin et al. 2012, Burns et al. 2015 ).These plant-species-dependent effects on micr obial assembl y hav e been attributed to the r elease of carbon-rich root exudates, which selectively enrich and recruit specific root-associated soil microorganisms (Bais et al. 2006 ), with the quality and composition of root exudates varying according to PS and plant de v elopmental sta ge (Badri andViv anco 2009 , Zhalnina et al. 2018 ).
For fungal communities, SC was identified to account for a higher amount of community variance than PS identity.In our experiment, all PS were planted in the same starting soil mixtur e.As suc h, these effects ar e not associated with differ ences in soil type or edaphic properties per se but are changes that the plants themselv es hav e dir ectl y expr essed on the rhizospher e and soil envir onment.Furthermor e, plant-driv en c hanges in the composition of root-associated microorganisms throughout the early sta ges of micr obial assembl y ma y also ha v e indir ectl y driv en the shifts observed in SC.Plants can dir ectl y modify the conditions of their surrounding soil physicochemical environment via nutrient uptake/loss or by the chemical signatures of their leaf litter, r oots, and r oot exudates.Plants can also shape their soil physicoc hemical envir onment indir ectl y by driving c hanges in the activity and composition of their root-associated microorganisms (Rengel and Marschner 2005, Waring et al. 2015, Henneron et al. 2020 ).Root-associated micr oor ganisms hav e k e y r oles in the tr ansformation and mobilization of inorganic and organic substrates into more plant-accessible soil nutrients, meaning that they can have a tr ansformativ e impact on soil nutrient cycling (Finzi et al. 2015, Dlamini et al. 2022 ;Dotaniya and Meena 2015 ).Plant-induced c hanges in SC pr ovide an example of how secondary deterministic pr ocesses can indir ectl y sha pe micr obial assembl y.As plant r oots modify the conditions of the rhizosphere and root-associated soil envir onment, this encour a ges the gr owth of micr oor ganisms that can occupy the modified habitat space (Hinsinger 2001, Liang et al. 2002, Bell et al. 2015, van Veelen et al. 2020, Hernández-Cáceres et al. 2022 ).
Plant-av ailable P, suc h as that measur ed by Olsen P (bicarbonate extractable), is a k e y measure of soil fertility and ecosystem productivity (Vitousek et al. 2010 ).In our research, Olsen P had a particularl y str ong r elationship with c hanges in r oot-associated soil fungal and bacterial communities.In particular, the rootassociated soils from the PS Pi .radiata , Al .glutinosa, and Ac .caesiiglauca had high Olsen P values compared to the other PS.These observ ations may demonstr ate the pr ocess of plants mobilizing soil nutrients essential for their individual growth and fitness (Will 1978, Chen et al. 2002, Tallec et al. 2009, Varin et al. 2009 ) and how these are linked to changes in soil microbiology.When rootinduced changes in soil chemistry influence microbial assembl y, this ultimatel y impacts plant health and performance, and thereby success in the ecosystem.These form plant-soil feedback mechanisms that amplify over successive life cycles (van der Putten et al. 2013 , Bennett andKlironomos 2019 ) and are profoundly connected with ecosystem-le v el pr ocesses.

The functional traits of plant species did not influence microbial community assembly
In our study, plant functional tr aits suc h as life span, functional gr oup, pr ov enance, N 2 fixation, and mycorrhizal association were not identified as strong drivers in microbial community assembly.It is important to consider that we examined root-associated soil microbes, but not microbes that colonize and de v elop symbiotic relationships with plant roots such as endophytes or mycorrhizal fungi.Had we examined the assembly patterns of plant symbiotic microbes and not free-living soil microbes, we may have observed the functional traits of host plants to have had a more pronounced impact on patterns of microbial assembly.Our findings are complementary to Hartman et al. ( 2023 ), who identified that the symbiotic associations of plant hosts significantly impact the root microbiome.Unlike Hartman et al. ( 2023 ) and Bodenhausen et al. ( 2019 ), who sampled the root-associated microbiome, our study examined soil adjacent to plant roots.Thus, discrepancies between the findings of our r esearc h and Hartman et al. ( 2023 ) are due to the clearly different sampling methodologies, as we sampled soils at a greater physical distance from the r oot.Additionall y, our r esearc h inv estigated r oot-associated soil micr obial assembl y follo wing (a) a single life c ycle of the plant and (b) at a single time point during the plant's de v elopmental sta ge.T hus , the absence of clear div er gences in micr obial assembl y between plants with contrasting functional traits may be a consequence of our experiment's r elativ el y short dur ation or other factors .For example , although we studied plants with different life cycle strategies (i.e. annual vs. perennial), we did not study them ov er r epeated life cycles, wher e the outcomes of their contrasting life histories ma y ha ve modified their soil environment to a degree that influenced microbial assembly.Several studies have reported the soil microbiome to shift according to plant de v elopment, influenced by changes in plant root morphology and exudate release with each developmental stage (Micallef et al. 2009, Cha parr o et al. 2014 ).Furthermor e, the div er gence in micr obial assembly between PS may amplify over successive life cycles (Cordovez et al. 2021 ).This increasing divergence is driven by plantsoil feedback mechanisms, whereby successive modifications in soil biotic and abiotic conditions by plants exert greater selec-tion pr essur es on their r oot-associated soil micr obiota (Hu et al. 2018 ).

Root-associated soil microbiomes exhibited a large degree of unexplained v aria tion
Niche-based theories of microbial community assembly assert that deterministic processes govern community structure, such as ada ptiv e species tr aits, biotic inter actions , and en vironmental filtering (Dini-Andreote et al. 2015 , Zhou andNing 2017 ).As discussed, aside from plant identity and plant-induced changes in SC, the functional plant tr aits measur ed in our study accounted for very little of the variation observed in root-associated soil micr obial comm unities.Our r esults identified that a lar ge amount of the compositional variation in root-associated soil communities remained unexplained, with over 73% of fungal community variation and 65% of bacterial comm unity v ariation unaccounted.Giv en the br eadth of v ariables we assessed, m uc h of this variation may represent elements of stochastic processes driving random community assemblage.More recently, there has been a growing body of liter atur e r ecognizing the degr ee to whic h stoc hastic pr ocesses ma y go v ern the r esulting structur e of micr obial comm unities (Caruso et al. 2011, Zhang et al. 2016, Zhou and Ning 2017, Chen et al. 2019, Hou et al. 2021, Huang et al. 2022 ).
T he PS under in v estigation in this study wer e at r elativ el y earl y sta ges of succession and growth (plants were grown for 10 months), which may explain the large amount of unexplained compositional variation we observed.Stochastic processes are reported to dominate micr obial assembl y during the earl y sta ges of community establishment, as the roots of plant seedlings release an abundant supply of exudates, whic h r educes competitiv e biotic interactions (Dini-Andreote et al. 2015 ).Ho w ever, throughout comm unity de v elopment, micr obiomes tr ansition fr om r andom comm unity assembl y to mor e highl y structur ed, nic hediffer entiated assembla ges because of functional ada ptations to the environmental selection pressures (Aguilar andSommaruga 2020 , Hu et al. 2020 ).As plants de v elop, they alter the bioav ailability of resour ces accor ding to their needs; thus, deterministic pr ocesses incr easingl y dominate micr obial comm unity assembl y as the surrounding environment is increasingly modified by plant gr owth (Dini-Andr eote et al. 2015 ).The modification of soil physicoc hemical pr operties b y PS w as observ ed for se v er al of the PS in our study, with Pi .radiata , Ac. caesiiglauca, and Al .glutinosa driving changes in Olsen P, for example.It is possible that if the microbial communities were measured over longer periods, community assembly would be more evidently niche-differentiated as each plant exerted unique selection pressures within the rootassociated soil environment.

Conclusion
Our r esearc h findings identified that during the earl y sta ges of plant growth and establishment, PS identity and plant-induced c hanges in SC wer e the most significant factors that shaped rootassociated soil microbial assembly.The functional traits of the PS under inv estigation, suc h as their life span, pr ov enance, gr owth form, and mycorrhizal associations, did not significantly account for any of the structural variation observed in bacterial or fungal communities between plants.Although PS identity was determined to be a significant factor driving micr obial assembl y, the phylogenetic r elationships shar ed between the 37 PS under inv estigation shar ed no r elationship to the similarity of their root-associated soil microbiomes .T hus , our findings reject the hypothesis that plant phylogenetic relatedness can be used to predict the emer gent structur e of the r oot-associated soil micr obiome.

Figur e 1 .
Figur e 1. T he proportion of explained variance attributed to each explanatory matrix, as identified using VP analysis.VP analysis was performed on weighted UniFrac distances of (A) bacterial ASVs and (B) fungal ASVs.Explanatory matrices used as input were PS, PLH, PRT, and SC.The significance of each explanatory matrix in accounting for comm unity v ariation is displayed, wher eby * * * denotes P value < 0.001.

Table 1 .
The results of Kruskal-Wallis tests that were performed to identify significant differences in the SR and Faith's diversity of root-associated bacterial and fungal ASVs across different plant-related factors.

Table 2 .
The results of PERMANOVA tests ( R 2 values) that were performed to identify significant differences in the composition of rootassociated microbial communities across different plant-related factors.