Altering plant carbon allocation to stems has distinct effects on rhizosphere soil microbiome assembly, interactions, and potential functions in sorghum

Altering plant carbon allocation from leaves to stems is key to improve biomass for forage, fuel, and renewable chemicals. The sorghum dry stalk (D) locus controls a quantitative trait for sugar accumulation, with enhanced carbon allocation in the stems of juicy green (dd) sorghum but reduced carbon allocation in that of dry white (DD) sorghum. However, it remains unclear whether altering sorghum sugar accumulation in stem affects below‐ground microbiome. Here we investigated sorghum rhizosphere soil microbiome in near isogenic lines with different magnitude of carbon allocations and accumulation in the stems. Results showed that enhanced carbon accumulation in stems of juicy green sorghum results in stronger selection in rhizosphere microbiome assembly. The rhizosphere soil microbial communities selected in juicy green sorghum tended to be fast‐growing microbial taxa which possessed potential functions that would promote higher potential capacity to use chemically labile carbon sources and potentially result in higher potential decomposition rates. We found the rhizosphere microbes selected by juicy green sorghum form weaker interactions than dry white sorghum. This is the first comprehensive study revealing how the different magnitude of carbon allocations to stems regulates microbial community assembly, microbial interaction, and microbial functions. This study indicates that future plant modification for bioenergy crops should also consider the impacts on belowground microbial community without compromising the sustainability.


| INTRODUCTION
Plant modification by increasing sink strength and carbon partitioning is critical for improving productivity and recovery of bioproducts (Reynolds et al., 2021).Efforts are thus underway to optimize carbon allocation to improve biomass of bioenergy plants (Li et al., 2019).Plants harbor complex microbial communities in rhizosphere soil that impact the health and development of host plants (Trivedi et al., 2020).By affecting the quantity and composition of root exudates such as primary and secondary metabolites, altering plant carbon allocation could directly select surrounding rhizosphere microbiomes (de Vries et al., 2020;Haichar et al., 2014;Korenblum et al., 2022;Zhalnina et al., 2018).In addition, rhizosphere microbiome can be indirectly influenced by modification of plant carbon allocation via altered soil physical and chemical properties such as soil pH, soil aggregation, and water holding capacity (Banerjee & van der Heijden, 2022;Haichar et al., 2014).Understanding how plant modification, especially carbon allocation from leaves to stems, influences the rhizosphere soil microbiome is critical for improving plant fitness and sustainability.
To understand how plants recruit associated microbiomes, current studies have mostly focused on comparing rhizosphere microbial community among different plant treatments (Garcia et al., 2022).However, these studies assume only host selection played a role in the assembly of microbial community, and largely ignored the fact that stochastic processes such as passive dispersal and ecological drift also contribute to microbial community establishment (Ning et al., 2020).As a result, studies employing techniques such as principal component/coordinate analysis to investigate plant microbiomes often end up with a large degree of overall microbial community variation not able to be explained by host effects.Therefore, to advance our understanding on how plants assemble their microbiomes, it is critical to separate the impact of stochastic processes from host selection when partitioning the variation of rhizosphere microbial community.Previous studies showed that the relative importance of stochastic processes decreased while that of species sorting (e.g.host selection) increased with increasing environmental heterogeneity in intestinal, aquatic, and terrestrial microbial communities (Adair et al., 2018;Martinson et al., 2017;Sprockett et al., 2020;Weinstein et al., 2021).However, in agriculture, it is yet unknown if plant modification, especially altering carbon allocation to different plant tissues, would also shift the relative importance of host selection versus stochastic processes in shaping microbiomes.
Microbial taxa do not act alone.Instead, microbes interact with each other to play important roles in plant performance and fitness (Hassani et al., 2018).Previous studies have reported that plant hosts increased the complexity of microbial interaction networks in rhizosphere soil (Shi et al., 2016).However, most of the networks were constructed using the overall microbial communities, which often include a substantial number of spurious correlations but fail to focus on the microbial interactions actively regulated by hosts.Plants could mediate cooperative and competitive interactions among microbiota to confer benefits for plant growth and fitness (Sasse et al., 2018;Trivedi et al., 2020).For example, tomato variety Hawaii 7996 can suppress the wilt pathogen through altering competitive interactions between pathogen and bacteria in rhizosphere soil (Kwak et al., 2018).In addition, maize has been shown to benefit from phosphorous uptake by enhancing the cooperation between arbuscular mycorrhizal fungi and bacteria (Jiang et al., 2021).Therefore, by targeting the interactions of microbes directly selected by plants, we can better understand how microbes coordinate to confer plant fitness.
Sorghum (Sorghum bicolor) is a C4 plant widely cultivated for food, fodder, fiber, and fuel (Doggett, 1988).Sorghum is also a promising candidate biofuel crop because of its high biomass yield, abiotic stress tolerance, high water and nitrogen use efficiency, and rich genomic resources (Calviño & Messing, 2012;McCormick et al., 2018).Sorghum accumulates soluble sugars in the stem with a high biomass yield, making it a model for carbon accumulation for other bioenergy crops (Wang et al., 2013).Studies on sorghum genetics have revealed that a single locus, D, located on chromosome 6, encodes a NAC transcription factor that controls the expression of genes involved in plant pith parenchyma cells, which are responsible for carbon allocation in sorghum (Fujimoto et al., 2018).While changing patterns of carbon allocation by manipulating gene expression can be enticing for sorghum production, it remains unclear how altering carbon accumulation in the stem affects belowground microbiome community assembly, microbial interactions, and microbial potential functions.
To better understand how modification of plant carbon allocation impacts belowground microbiomes, we investigated the microbial community and potential function from near-isogenic lines of dry white (DD) and juicy green (dd) sorghums.By genetically engineering D gene expression, juicy green (dd) sorghum has higher sugar accumulation in stems, higher biomass, higher photosynthesis in leaves, and higher carbon allocation from leaves to stems than dry white (DD) sorghum (Zhang et al., 2018).Furthermore, the near-isogenic lines (NILs), juicy green (dd) and dry white (DD), mainly differ in sugar content with minimal background variability (Xia et al., 2018).First, we asked whether higher carbon allocation from leaves to stems in juicy green (dd) sorghum rhizosphere soil will result in the plants harboring different microbial taxa in the rhizosphere, impose different strength of selections on microbial community, and support different microbial potential functions compared to dry white (DD) sorghum.Second, we asked whether higher carbon allocation from leaves to stems in juicy green (dd) sorghum rhizosphere soil could alter microbe-microbe interactions compared to dry white (DD) sorghums.

| Study design and sample collection
This study was conducted in Urbana, IL (40°05′12.8″N, 88°13′31.7″W) in summer 2020.Paired rows of nearisogenic lines (NILs) sorghum (juicy green (dd) and dry white (DD)) were planted in 20 replicates in rows 10′ long with 30″ row spacing in the field.The juicy green (dd) and dry white (DD) sorghum have the same phenological period.Carbon allocation of dd and DD sorghum is most active at ~30 days after flowering, which results in substantial differences in sugar yield between the two sorghum lines (Fujimoto et al., 2018;Xia et al., 2018).Ten individual plants of each genotype (juicy green (dd) and dry white (DD)) were harvested ~30 days after flowering.For the above-ground plant, we measured stem juice content, sugar concentration (Brix), stem sugar content, and biomass according to methods as previously described (Xia et al., 2018) and the data is presented in Table S1.All soil parameter data is presented in Table S1.Below-ground, the rhizosphere soil (soil tightly adhering to the sorghum root surface) was collected and bulk soil was collected approximately 12 inches away from the sample plants using a soil corer.A total of 30 soil samples were collected from bulk soil and dry white (DD) and juicy green (dd) sorghum rhizospheres.All soil samples were transported to the laboratory on ice and preserved at −20°C until further processing.

| Bioinformatic analysis
DNA sequences were obtained as fastq files.Pairedend 16S rRNA and ITS sequences were merged using Fast Length Adjustment of SHort reads (FLASH) software (Magoč & Salzberg, 2011).Quality filtering of fastq files was performed using the FASTX-Toolkit software (Gordon & Hannon, 2010); primer sequences and sequence reads with a quality score of less than 30 and with fewer than 90% of bases were removed.Sequences were binned into discrete operational taxonomic units (OTUs) based on 97% similarity and generated an OTU table using USEARCH (Edgar, 2010).OTUs with representative sequences were assigned to taxonomic lineages using the RDP classifier within the SILVA database (release 128) for bacteria and archaea and using BLAST within the UNITE database (Nilsson et al., 2019) for fungi.Sequences identified as plants, protists, chloroplasts, and mitochondria were removed prior to statistical analysis.A total of 2,010,551 raw reads were obtained from the V4 region and 3,510,494 raw reads were obtained from the ITS2 region.Library size ranged from 49,688 to 83,361 sequences per sample from the prokaryoticV4 region with a mean of 67,018 sequences per sample, and 74,108 to 171,572 sequences per sample for the fungal ITS region with a mean of 117,016 sequences per sample.Read counts were rarefied to 49,688 reads for 16S rRNA and ITS rRNA soil samples.After rarefying, we were left with: 16S rRNA data for 30 soil samples with 7488 OTUs; fungal ITS data for 30 soil samples with 1214 OTUs.

| Statistical analysis
2.4.1 | Microbial richness and community composition analysis related to different patterns of carbon allocation Most statistical analyses were conducted in R v.4.3.36.Rarefaction curves were computed for all prokaryotic and fungal sequences from soil samples collected from bulk, dry white (DD), and juicy green (dd) sorghum rhizosphere to evaluate the comprehensiveness of the sampling strategy using the VEGAN R package (Oksanen et al., 2013).The Kruskal-Wallis test (McKight & Najab, 2010) and Dunn's post hoc multiple comparisons test (Trawiński et al., 2012) was performed to assess the differences in the microbial diversity among bulk soil, dry white (DD) and juicy green (dd) sorghum rhizosphere soil.Distance matrices for the soil prokaryotic and fungal community from bulk soil, dry white (DD), and juicy green (dd) sorghum rhizosphere soil were constructed by calculating dissimilarity with the Bray-Curtis method on Hellinger-transformed OTU read data.To investigate patterns of soil prokaryotic and fungal community structures in bulk soil, dry white (DD) and juicy green (dd) sorghum rhizosphere, unconstrained PCoA (for principal coordinates PCoA1 and PCoA2) ordination of analysis was performed based on the Bray-Curtis dissimilarity matrices.Then, differences among treatments in prokaryotic and fungal community compositions were tested by conducting a permutational multivariate analysis of variance (PERMANOVA).PERMANOVA using the adonis command in the VEGAN R package was also performed to disentangle the relative effect of stem juice content, sugar concentration (Brix), stem sugar content, and biomass on soil prokaryotic and fungal community compositions.

| Microbial assembly rules in different magnitude of carbon allocations
To explore how different patterns of carbon allocation contribute to microbial assembly rules, a neutral community model was used to determine the contribution of selective process and stochastic process to microbial community assembly by predicting the relationship between the frequency with which taxa occur in a set of local communities and their abundance across the wider metacommunity (Sloan et al., 2006).This model predicts that less abundant taxa will be lost from individual samples due to ecological drift (i.e., the stochastic loss and replacement of individuals), while more abundant taxa are more likely to be dispersed by chance and therefore present in more samples.In the model, the estimated migration rate is a parameter for evaluating the probability that a random loss of an individual in a local community would be replaced by dispersal from the metacommunity, and, therefore, is a measure of dispersal limitation (Burns et al., 2016;Sloan et al., 2006).The parameter R 2 represents the overall fit to the neutral model.Calculation of 95% confidence intervals around all fitting statistics was done by bootstrapping with 1000 bootstrap replicates.
To further confirm the how the different magnitude of carbon allocation affects microbial taxa actively selected by plant, we analyzed deviations from neutral model predictions.The model is validated because this model adopted a conceptual framework in which we consider the microorganisms associated with individual sorghum plant to be local communities that are a part of a broader metacommunity consisting of the microorganisms associated with all the sorghum plants in each treatment (dd, DD or bulk soil) (Burns et al., 2016;Costello et al., 2012;Leibold et al., 2004).Samples belonging to the same treatment were first pooled and OTUs from this pool were subsequently sorted into three partitions depending on whether they occurred more frequently than ("above prediction" partition, considered as host positive selection), less frequently than ("below prediction" partition, considered as host negative selection) or within ("neutral" partition, considered as neutral dispersal) the 95% confidence interval of the neutral model predictions (Burns et al., 2016).Each partition was then treated as a distinct community sample for further analysis.To identify microbial taxonomic groups that distinguish host positive selection partition from neutral and host negative selection partitions of the metacommunity, unconstrained PCoA (for principal coordinates PCo1 and PCo2) ordination of analysis was performed based on the Bray-Curtis dissimilarity matrices.To confirm the stronger relationship between plant carbon allocation traits and community composition variation of taxa that were host positive selection (compared to neutrally and host negative selection microbial OTUs), the dissimilarity of the prokaryotic and fungal community from host positive selection, neutrally, and host negative selection partitions and the Euclidean distance of the plant carbon allocation traits (stem sugar yield) were estimated based on Spearman correlations using the Mantel test in R (VEGAN package).

| Microbial functional traits in different magnitude of carbon allocations
To confirm the host plant carbon allocation traits affected microbial ecological fitness traits, the dormancy potential and heterotrophic strategy were investigated in host negative selection prokaryotic OTUs.The dormancy refers to an organism's ability to enter a reversible state of low metabolic activity when faced with unfavorable environmental conditions (Lennon & Jones, 2011).Moreover, changes in the environmental cues that control dormancy may result in the rapid loss of dormancy as an adaptive trait (Lennon & Jones, 2011).The dormancy potential was predicted by PICRUSt2 (Douglas et al., 2020).One of the most representative characteristics of some dormant microorganisms is sporulation, which has been relatively well documented and explored.Hence, from the PICRUSt2-predicted metagenomes, genes related to sporulation were searched in the KEGG database to represent the dormancy potential.The KO identifiers of the genes were provided in Table S2.
To investigate the heterotrophic life history strategy of prokaryotic OTUs in the host positive selection partition, we used the operon copy number method, which has been demonstrated to be positively correlated with the capacity to respond to resource availability (Roller et al., 2016).Microbes with multiple operons (>2) tend to be r-strategists (Copiotroph), which rapidly respond to the nutrient availability and have high reproductive rates; those with few rrn operons (≤2) tend to be K-strategists (Oligotroph), which are more competitive when resources become limited (Wu et al., 2017).The ribosomal operon copy number of each species was obtained from the rrndb database (version 5.6, released on 25 October 2019) (Stoddard et al., 2015).Then, the weighted mean copy number was calculated as described previously (Nemergut et al., 2016) by summing the relative abundance of each taxon multiplied by its copy number to represent the heterotrophic strategy.For host positive selection fungal OTUs, copiotroph strategy was identified for OTUs within phylum Ascomycota, Mortierellomycota, and Mucoromycota (Ho et al., 2017).
In order to identify whether the different magnitude of carbon allocations affect potential functions of host positive selection OTUs, prokaryotic functional profiles were predicted using functional annotation of host positive selection prokaryotic taxa (FAPROTAX) (Louca et al., 2016).Fungal functional guilds were inferred (guild assignments with confidence rankings "Highly probable" and "Probable" were retained) using the program FUNGuild (Nguyen et al., 2016) and the identification method is based on this study (Ji et al., 2019).

| Microbial interaction in different magnitude of carbon allocations
To understand how the microbial interactions changed between plants with contrasting carbon allocation traits, prokaryotic and fungal networks were constructed using the "WGCNA" R package (Langfelder & Horvath, 2008) based on the Spearman correlation index.The nodes and the edges in the network represent prokaryotic and fungal OTUs and the significant interactions between pairs of OTUs, respectively.OTUs with relative abundance less than 0.01% and frequency less 50% of samples in each treatment were excluded because they were poorly represented.The p-values for multiple testing were calculated using the Benjamini and Hochberg false discovery rate (FDR) test controlling procedure.Only the rank correlation coefficient with values above 0.6 or below −0.6 and a statistically significant adjusted-p value lower than 0.001 were considered as a valid correlation in the network.The networks of the prokaryote and fungi were graphically displayed in Gephi (http:// gephi.github.io/ ).To confirm how the microbial network complexity changed with increasing carbon allocation from leaves to stems, we generated sub-networks for each sample from the network meta-matrix by preserving OTUs presented at each sample.For each network, we calculated linkage density, connectance, and modularity in the R package igraph (Csardi & Nepusz, 2006).To explore how the interactions changed between host positive selection taxa and other taxa with increasing carbon allocation from leaves to stems, we calculated changes in the proportion of host positive selection taxa links in bulk soil, dry white (DD), and juicy green (dd) sorghum rhizosphere soil.Erdös-Réyni model random networks with the same number of nodes and edges as the observed networks were also constructed for each treatment.

| Characterization of sequencing data
We achieved richness asymptotes for both datasets, suggesting that sequencing efforts were sufficient to capture comparative dynamics and diversity (Figure S1a,b).The total richness observed at this rarefaction depth was 1214 fungal and 7488 prokaryotic OTUs.The prokaryotes were dominated by Proteobacteria, Actinobacteria, and Acidobacteria in bulk soil, dry white (DD), and juicy green (dd) sorghum rhizosphere soil.The relative abundance of Proteobacteria was higher in juicy green (dd) sorghum rhizosphere soil, whereas Actinobacteria and Chloroflexi showed higher relative abundance in bulk and dry white (DD) sorghum rhizosphere soil than in juicy green (dd) sorghum rhizosphere soil (Figure S2a).The fungi were dominated by Ascomycota and Basidiomycota, with varying relative abundances among different treatments (Figure S2b).

| Microbial richness and community composition in different magnitude of carbon allocations
Total prokaryotic and fungal richness was similar in the bulk soil, dry white (DD), and juicy green (dd) sorghum rhizosphere soil (Figure 1a,b).The PCoA ordination analysis showed that the prokaryotic and fungal community assemblages in the bulk soil, dry white (DD), and juicy green (dd) sorghum rhizosphere soil were clearly distinct (p < 0.001) (Figure 1c,d).Variations in prokaryotic community composition among samples from bulk soil, dry white (DD), and juicy green (dd) sorghum rhizosphere soil were primarily explained by stem sugar yield and stem wet weight (Table S3).Variations in fungal community composition among samples from bulk soil, dry white (DD), and juicy green (dd) sorghum rhizosphere soil were primarily explained by stem sugar yield, stem volume, and stem wet weight (Table S3).
F I G U R E 1 Microbial richness and community composition in bulk soil, dry white (DD) and juicy green(dd) rhizosphere soil.(a) Prokaryotic and (b) fungal OTU richness.Horizontal lines within boxes denote medians.Top and bottom segments of boxes denote the 75th and 25th percentiles, respectively.Upper and lower whiskers extend to data no more than 1.59 the interquartile range from the upper edge and lower edge of the box, respectively.Principal coordinate analysis (PcoA) using the Bary-Curtis disimalarity of prokaryote (c) and fungi (d) in bulk soil, dry white (DD) and juicy green (dd) rhizosphere soil.To explore how the different magnitude of carbon allocations contribute to microbial assembly rules, we found the neutral model was well fitted to the prokaryotic and fungal communities for all treatments (Figure 2a-f), with the lowest R 2 value in juicy green (dd) sorghum rhizosphere soil (Figure 2c,f), suggesting host plant selection increased and stochastic processes decreased with increasing carbon allocation from leaves to stems.In total, 85% of prokaryotic and fungal OTUs were neutrally distributed, while only 11% were consistently under host positive selection and 3.6% were under host negative selection (Figure 3a).These results suggest neutral processes play an important role in microbial community assembly in all treatments.
The migration rates (m) tended to be lower from bulk soil to juicy green (dd) sorghum rhizosphere soil (Figure 3b), suggesting that communities become increasingly dispersal limited with increasing carbon allocation from leaves to stems and roots.
To further confirm the how different magnitude of carbon allocations affect microbial taxa actively selected by plant, we found the richness of microbial taxa positively selected by the host significantly decreased with increasing carbon allocation from leaves to stems (Figure S3a-f), suggesting increasing carbon allocation to the stem might be a possible ecological filtering mechanism.We also found the microbial taxa positively selected by the host within Proteobacteria and Mortierellomycota were relatively higher in juicy green (dd) sorghum rhizosphere soil (Figure S4a,b).The relative abundance of Proteobacteria and Mortierellomycota in host positive selection partitions of the metacommunity among bulk soil, dry white (DD) and juicy green (dd) were significantly associated with sugar yield in the stem (Figure S5a,b).
To identify microbial taxonomic groups that distinguish host positive selection from neutral and host negative F I G U R E 2 Fit of the Sloan neutral model.The predicted occurrence of frequencies for prokaryotic (a-c) and fungal (d-f) communities in bulk, dry white (DD) and juicy green (dd) sorghum rhizosphere soil.OTUs that occur more frequently than the value predicted by the model are shown in reen while those that occur less frequently than predicted are shown in yellow.Black points represent the frequency of occurrence within the 95% confidence interval (grey dotted lines) ranging around the model prediction (light blue line).The values of m and Nm indicate the estimates of dispersal rate between communities and the metacommunity size times immigration, respectively; R-square indicates the fit to this model.selection partitions of the metacommunity, we found that partitions clustered strongly based on how they deviated from the neutral prediction across treatments (Figure 3c).The taxa comprising neutral and non-neutral partitions of the metacommunity also tended to be related to the different magnitude of sugar yield in the stalk (Figure S6a-f).Our analysis confirmed the relationship between plant carbon allocation traits and the microbial community members in each partition (host positive selection, host negative selection, and neutrally distributed OTUs): we found that the relationships between sugar yield and microbial OTUs in host positive selection partition are stronger than those between sugar yield and neutral distributed microbial OTUs (Figure S6a,b).This result further confirmed that the plant carbon allocation from leaves to stems mainly affected the microbial assembly via regulating host positive selection community variations in rhizosphere soil.

| Microbial potential functions changed in different magnitude of carbon allocations
Ecological fitness traits enriched in microbial taxa positively selected by host include prokaryotic sporulation gene abundance and fungal and bacterial traits related to a copiotrophic lifestyle: fungal copiotroph relative abundance and bacterial weighted ribosomal copy number (>2) significantly increased in the juicy green (dd) sorghum (Figure 3d).We also found most of microbial potential functions were similar between juicy green (dd) and dry white (DD) sorghum (Figure S7) except for aromatic carbon degradation gene abundance and fungal saprotrophic abundance.The aromatic carbon degradation gene abundance and fungal saprotrophic abundance were significantly increased in juicy green (dd) sorghum compared to bulk soil and dry white (DD) sorghum (Figure 4a,b).

| Microbial interactions changed in different magnitude of carbon allocations
To understand how the microbial interactions differed between hosts with contrasting magnitude of carbon allocations, we assessed the co-occurrence patterns of prokaryotic and fungal communities (Figure 5a-f).We found the microbial co-occurrence network characteristics such as linkage density and connectance increased but modularity decreased with increasing carbon allocation in stems, suggesting microbial co-occurrence network became more complex with increasing carbon allocation in stems (Figure S8a-f).The proportion of host positive selection taxa interactions with other taxa decreased in the rhizosphere communities of juicy green (dd) sorghum (Figure 5g).

| DISCUSSION
In this study, we demonstrated that with increasing carbon allocation from leaves to stems, juicy green (dd) sorghum imposed stronger selection on rhizosphere microbiome assembly.We also found that the magnitude of plant carbon allocation from leaves to stems mainly affected the microbial assembly via regulating host positive selection community variations in rhizosphere soil.We found that the host positive selection microbial communities exhibited lower dormancy and higher fast-growing ecological traits since higher carbon allocation from leaves to stems and roots would supply heterotrophic microbial communities in the rhizosphere with more available carbon and energy, promoting microbial activity and the cycling of all elements.We also demonstrated that microbes selected in the juicy green (dd) sorghum rhizosphere soil have higher aromatic carbon degradation and higher decomposition metabolic functions than in dry white (DD) and bulk soil, where fewer carbon resources are predicted to be available.Finally, we found microbes selected in the higher carbon allocation sorghum (dd) rhizosphere soil form weaker interactions than that in dry white (DD) sorghum, which may indicate that the host is relieving resource limitations.

| Plant carbon allocation altered microbiome assembly
The community variations of prokaryote and fungi in bulk soil, dry white (DD), and juicy green (dd) sorghum rhizosphere soil were significantly related to sugar yield in stem (Figure 1c,d; Table S3), suggesting altering aboveground carbon allocation could affect belowground rhizosphere microbial community composition.This finding is supported by a previous study, where under conditions that decreased photosynthesis, the host plant transferred carbon to root and rhizosphere fungal communities and gram-negative bacteria and thus affected microbial composition (Bahn et al., 2013).This might indicate that altering plant aboveground carbon allocation could affect the amount of plant carbon allocation to root and rhizosphere and thus result in changing in quantity and quality of root exudates which affect microbial activity and composition (Bahn et al., 2013;de Vries et al., 2019;Hartmann et al., 2020).
The values of NCM parameter R 2 decreased from bulk soil, dry white (DD), to juicy green (dd) sorghum rhizosphere soil (Figure 2a-e), indicating that neutral processes on microbiome become relatively less important, but host selection on microbiome become more important with increasing carbon allocation from leaves to stems and roots.Further, regarding the prokaryotic and fungal community immigration rate, the m value was lower in juicy green (dd) than in dry white (DD) sorghum rhizosphere soil, indicating the dispersal ability of prokaryotic and fungal communities in juicy green (dd) was lower than in dry white (DD) sorghum rhizosphere soil.Consistently, previous studies have shown that increasing carbon exudates released into soil could enhance host plant selection in the rhizosphere soil microbiome (Weisskopf et al., 2008).propose this is the case in juicy green (dd) sorghum, where higher carbon allocation to stems resulted in higher amount of labile carbon released into the soil, supplying microbes with more available carbon and energy, and thus exerting stronger host plant selection.
Neutral models also allow recognition of microbial OTUs that are more widespread than predicted and these taxa generally are associated with host plant selection.Our study has confirmed that the microbial taxa by host positive selection were significantly different from neutrally distributed OTUs (Figure 3c).Proteobacteria and Mortierellomycota were actively selected in juicy green (dd) sorghum rhizosphere soil (Figure S4a,b), and selection for these taxa was significantly related to sugar yield (Figure S5a,b).With increasing carbon allocation into stems and roots, more labile carbon may be available in the rhizosphere, potentially stimulating the growth and reproduction of copiotroph taxa such as Proteobacteria and Mortierellomycota (Ali et al., 2018;Levy et al., 2018;Trivedi et al., 2013).

| Plant carbon allocation altered microbial potential functions
Juicy green (dd) sorghum selected fewer prokaryotic OTUs with sporulation genes and favored those with higher weighted rrn copy numbers (r strategists) in rhizosphere soil (Figure 3d), indicating plant with higher carbon allocation to stems could select microbes with traits that favor rapid response to resource availability.Previous studies confirmed that sporulation genes represent microbial dormancy potential (Douglas et al., 2020), and the microbial dormancy rate is reduced in environments with high resource availability (Jones & Lennon, 2010).Our results demonstrated that higher carbon allocation to stems might lead to more labile carbon available in rhizosphere soil, creating a nutrient rich environment that recruited fast-growing microbes.Furthermore, the enhanced microbial activity induced by C addition can affect several important soil processes, such as C and N cycling, thereby providing more nutrients for plant growth (Bornø et al., 2018).In addition, juicy green (dd) sorghum selected higher copiotroph fungal abundance in rhizosphere soil, suggesting copiotroph fungi could easily consume the labile carbon released into soil.In our study, we found that higher carbon allocation from leaves to stems in plant could enhance aromatic degradation metabolic functions in juicy green (dd) sorghum rhizosphere soil, indicating more carbon allocation to roots or soil could select prokaryotic microbes which have higher capacity of aromatic carbon degradation (Liu et al., 2022).Thus, increased available carbon derived from juicy green (dd) sorghum could trigger the decomposition of plant litter and result in upregulation of the expression of genes encoding enzymes involved aromatic carbon degradation (Rineau et al., 2013).Moreover, we found more carbon allocation from leaves to stems and roots could increase saprotrophic fungal abundance, suggesting that fungal decomposition organic matter function could also be enhanced by greater carbon allocation to roots and soil (Clemmensen et al., 2013).

| Plant carbon allocations altered within and between host positive selection microbial interactions
Prokaryotic and fungal co-occurrence networks became more complex with increasing carbon allocation in stems, suggesting that enhancing labile carbon inputs into soil may improve overall microbial interactions (Ji et al., 2023).More species and root exudates in the rhizosphere soil of plants with higher carbon allocation likely stimulated the microbial activities and caused the increased metabolites, which complicated the nutrient interaction network among species (such as cross feeding) (Landi et al., 2006;Murillo-Roos et al., 2022).We also found that the proportion of within host positive selection taxa interactions and between host positive selection taxa and other taxa decreased from bulk soil, dry white (DD), to juicy green (dd) sorghum, suggesting that host-selected microbial taxa interactions became weaker with increasing carbon allocation from leaves to stems.The increased labile carbon secreted by plants may relieve competition or cooperation between plant-selected microorganisms for resources (Hu et al., 2018).While network analyses suggest potential for interactions among microbial taxa, co-occurrence networks cannot reflect the true microbial interactions and should be treated as a method of generating hypotheses that further experiments can focus on.

| CONCLUSION
Our results highlight the importance of investigating the rhizosphere soil microbial communities selected by host plants in response to different genetic modifications, such as altering the magnitude of carbon allocation to stems.In response to increased carbon accumulation in stems, the sorghum plant affected the rhizosphere soil microbial assembly via regulating the host positive selection community variation.The rhizosphere soil microbial communities selected in juicy green sorghum tended to be fast-growing microbial taxa which possessed potential functions that would promote higher potential capacity to use chemically labile carbon sources and potentially result in higher potential decomposition rates.Overall, our study demonstrates that future plant breeding for bioenergy crops (particularly the "stems as factories" strategy) should also consider the impacts on belowground microbial community without compromising sustainability.

F
Patterns of selection and neutral assembly of microbial community in bulk, dry white (DD) and juicy green (dd) sorghum rhizosphere soil.(a) The proportion of prokaryotic and fungal OTUs in each treatment that fit in the Sloan neutral model.(b) The estimated migration rate (m), or the probability that a random loss of an individual in a local community will be replaced by dispersal from the metacommunity for each treatment in prokaryote and fungi.(c) Principal coordinate analysis using the Bray-Curtis distance for each of the three data partitions of prokaryote and fungi from three treatments.(d) Microbial functional traits of host positive selection in each treatment.

F
Aromatic carbon degradation gene abundance (a) and saprotrophic fungal abundance (b) of host positive selection partition in the bulk soil, dry white (DD) and juicy green (dd) sorghum rhizosphere soil.Kruskal−Wallis, p = 0

F
Co-occurrence networks of the prokaryotic (a-c) and fungal (d-f) communities in bulk soil, dry white (DD), and juicy green (dd) sorghum rhizosphere soil.Node colors represent the OTU distributed types resulting from the Sloan neutral model.(G) Proportion of associations between host positive selection partition prokaryotic OTUs and between host positive selection partition fungal OTUs in bulk soil, dry white (DD) and juicy green (dd) sorghum rhizosphere soil.host posi ti ve sel ecti onOTU l i nks(%) Host positive selection OTU Neutral distributed OTU Host negative selection OTU P