Plants host diverse and abundant microbial communities that can be considered the “second genomes” of plants. Microbial communities that exist in close association with plants can be categorized into three groups: endophytic, epiphytic, and closely associated (Tringe et al., 2005). Plants and their associated microbes interact with each other and form assemblages of genotypes that are often referred to as “holobionts” (Vandenkoornhuyse et al., 2015; Hassani et al., 2018). Plants attract and select for beneficial microbiomes by releasing signal molecules and providing carbon metabolites as root exudates to endosphere and rhizosphere bacteria (Ryu et al., 2004; Guo et al., 2016; Lopes et al., 2016). Plants can influence net ecosystem changes through deposition of secondary metabolites into the rhizosphere that attract or inhibit the growth of specific microorganisms. This rhizodeposition was made up of small-molecular weight metabolites, amino acids, secreted enzymes, mucilage, and cell lysates (Grayston et al., 1998; Paterson and Sim, 2000). Soil microbes utilize this abundant carbon source, thereby implying that selective secretion of specific compounds may encourage beneficial symbiotic and protective relationships whereas secretion of other compounds inhibit pathogenic associations (Hoffland et al., 1992; Holden et al., 1999). A concrete example is the secretion of isoflavones by soybean roots, which attract a mutualist (Bradyrhizobium japonicum) and a pathogen (Phytopthora sojae) (Morris et al., 1998). In turn, plants benefit from these relationships, as the microbes change key nutrients into more usable forms (Long, 1989; Bolan, 1991; Zhang et al., 2009). The symbioses between plants and the associated microbes play important roles in the development, health and environmental adaptability of the plant hosts (Spor et al., 2011; Berendsen et al., 2012; Yuan et al., 2018). Previous studies on Arabidopsis thaliana (Durán et al., 2018), grapevine (Rolli et al., 2015), and citrus (Zhang et al., 2017) have demonstrated that the bacterial community plays an essential role in plant growth through a variety of mechanisms, including increasing nutrient acquisition, promoting plant hormone production, and protecting plants against pathogen attacks (Ritpitakphong et al., 2016; Álvarez-Pérez et al., 2017; Hassani et al., 2018). Many plant-associated microbes can induce systemic resistance in plants (Liu et al., 2019). For example, stem inoculation with the bacterial strains Bacillus amyloliquefaciens (GB03) and Microbacterium imperiale (MAIIF2a) mitigates Fusarium root rot in cassava (Freitas et al., 2019). Bacillus cereus AR156 is a plant growth-promoting rhizobacterium (PGPR) that induces resistance against a broad spectrum of pathogens in A. thaliana (Niu et al., 2011). Moreover, harnessing the plant microbiome to maximize crop production is increasingly considered a viable and sustainable approach for the future of agriculture (Geddes et al., 2015; Qiu et al., 2019).

Endophytic microbes and rhizosphere exophytic microbes are affected both by their host plants and by environmental stimuli. In some plants, the diversity and composition of endophytic communities are highly variable between cultivars (Liotti et al., 2018; López et al., 2018). For example, two Rosa cultivars with different powdery mildew susceptibilities were determined to share only 34.2% of operational taxonomic units (OTUs), and the resistant cultivar had significantly lower fungal diversity than the susceptible cultivar in the early stage of development (Zhao et al., 2018b). Some studies have shown that the species has a stronger influence on bacterial community composition by growing Populus, Quercus, and Pinus in three soils originating from different field sites (Bonito et al., 2014). The host plant species is the most important factor that determine the leaf endophytic bacterial communities collected from 5 species of plants (Asclepias viridis, Ambrosia psilostachya, Sorghastrum nutans, Panicum virgatum, and Ruellia humilis) (Ding et al., 2013). Moreover, the taxonomic composition of the extraordinarily diverse communities of microorganisms associated with plants is determined partly by the plant genotype (Korkama et al., 2006; Peiffer et al., 2013; van der Heijden and Schlaeppi, 2015). Genotype effects on the fungal and bacterial microbiomes have been detected in Triticum aestivum (Simonin et al., 2020), rice (Oryza sativa) (Edwards et al., 2015), maize (Zea mays) (Walters et al., 2018), and potato (Solanum tuberosum) (van Overbeek and van Elsas, 2008; Inceoğlu et al., 2010). The influence of plant genotypes on belowground microbiota can be attributed to differences in plant growth performance, as well as in the varying amounts of nutrients provided to soil through litter and root exudates (Korkama et al., 2006; van der Heijden and Schlaeppi, 2015; Hugoni et al., 2018) and through their symbionts (Smith and Read, 2008; Gorka et al., 2019). Rhizosphere-associated microbes obtain essential nutrients from plants through host root exudates, and crucial nutrients can be converted to more usable forms, including sugars, organic acids, amino acids, and peptides, by microbes before being assimilated by plants (Ryu et al., 2004). Therefore, through the release of a broad variety of secondary metabolites and root exudates, plants have the capacity to drive and shape plant-associated microbial communities (Raaijmakers et al., 2009). The plant genotype, by determining the community structure of its microbial partners, can be expected to exert cascading effects on ecosystem functions related to nutrient cycling. Overall, plant genetic control of the microbial community is of considerable interest for crop plant breeding and for exploring the possibility of designing a “healthy” microbiome (Morella et al., 2020).

Cassava (Manihot esculenta Crantz) is a member of the Euphorbiaceae family, has strong environmental adaptability and is tolerant of barrenness and drought (Luo, 2005). Owing to its starch-enriched tuberous root, cassava is an important cash crop in tropical and subtropical areas, and it can also be converted into a large number of products; for example, it is a major resource used in the production of starch, biofuel, and animal feed (Utsumi et al., 2012; Okogbenin et al., 2013). Previous studies have shown that the agronomical characteristics of cassava are significantly influenced by different PGPR strains (Suja et al., 2015). Microbial inoculation significantly improved the mineral nutrient uptake, yield, harvest index, and repression of root rot infection in cassava compared with those in uninoculated controls (Hridya et al., 2013). The application of beneficial bacteria plays an important role in increasing plant growth and protecting against pathogen infection in cassava (Freitas et al., 2019). With the development of next-generation sequencing technologies, culture-independent methods have been employed to determine the profiles of the cassava-associated microbial communities by 16S rDNA sequencing. Li et al. (2020) showed that cassava cultivars recruited various endophytic microbial taxa from tuberous roots to affect the ability of root rot resistance. Similarly, structure of microbiomes of cassava genotypes were analyzed, and revealed their potential roles in cassava bacterial blight resistance (Zhang et al., 2021). Such previous studies have been made on cassava-associated bacteria overall, and few have focused on the comprehensive investigation of endophytic and rhizospheric bacteria of cassava genotypes. Besides, we currently have a poor understanding of how tuber metabolites influence microbial community structure.

In this study, fourteen cassava genotypes with significant differences in genetic background were used to investigate the relationship between cassava-associated bacteria and genetic differences among cassava genotypes by 16S rDNA gene tag sequencing analysis. Moreover, we examined the effects of cassava tuber metabolites collected from three typical cassava genotypes on the different bacterial communities from the tuberous roots and rhizosphere soil. Our results will provide new insight into the linkages between cassava-associated bacteria and cassava genotypes.

Plants and microbes interact in order to obtain nutrients to improve their growth and stress resistance, especially through root-microbe interactions (Edwards et al., 2015; Wu et al., 2020). Microbial communities have been shown to be impacted by the microhabitat (Jin et al., 2017), soil type (Bonito et al., 2014), and host genotype (Ofek-Lalzar et al., 2014; Zhang et al., 2019). At present, the microbial diversity of soils, roots, leaves, and aerial organs has been extensively investigated by high-throughput sequencing (Buée et al., 2009; Abdelfattah et al., 2016a,b, 2017; Liu et al., 2018; Yuan et al., 2018). To extend our knowledge of bacterial diversity as it relates to cassava genotypes, we used culture-independent high-throughput sequencing technology to investigate the diversity and community structure of bacteria present in the tuber endospheres and rhizospheres of fourteen cassava genotypes (SC5, SC8, SC9, SC205, KU50, R72, XL1, FX01, SC16, 4612, 587, 045, S0061, and 1110). Our results showed that the diversity of bacterial communities in the tuber endosphere and rhizosphere depends on the plant genotype and the tuber metabolites.

Previous reports have suggested that microbial density is generally higher in the rhizosphere than in the root and that bacterial diversity and richness gradually decrease from soils to epiphytes to endophytes (Bulgarelli et al., 2012, 2015; Lundberg et al., 2012; Edwards et al., 2015; Hacquard et al., 2015). In this study, the diversity and richness of bacteria in the rhizospheres of all cassava genotypes were higher than those in the tuber endospheres. Moreover, the cassava genotype did not significantly influence the endophytic bacterial community structure (Figure 4). Similar results have also been reported in previous studies: there were no significant differences in richness or diversity among the endophytes of different cassava genotypes (Li et al., 2020). Studies on A. thaliana have consistently suggested that root endosphere bacterial communities are strongly influenced by soil type and soil properties but that host genotype has a limited effect on the root microbiome (Bulgarelli et al., 2012; Lundberg et al., 2012; Thiergart et al., 2020). However, we detected a significant effect of cassava genotypes on the rhizosphere bacterial community structure (Figure 5), which is consistent with other previous findings (Miethling et al., 2000; Marschner et al., 2004; Garbeva et al., 2008; Bonito et al., 2014; Burns et al., 2015; Leff et al., 2018). For example, a significant effect of plant genotype on rhizosphere microbial communities was observed by comparing the rhizospheres of different experimental crops grown in soils of the same type (Liu et al., 2020). Some previous studies of the tree phyllosphere and maize rhizosphere separately showed that host genetics played an important role in shaping the bacterial microbiome (Laforest-Lapointe et al., 2016; Walters et al., 2018). Our results provide comprehensive empirical evidence for the selection of the microbial community by cassava and a theoretical framework for the coevolution between cassava and microbes; in this framework, cassava plants use exudates to recruit, filter, and enrich certain microbial taxa that have specific functions (Müller et al., 2016; Martin et al., 2017; Sasse et al., 2018), and competition among microbes for these resources drives their rapid evolutionary radiation and consequent divergence to reduce competition (Foster et al., 2017).

Generally, genetic based-interactions among genotypes are complex and have been recently gaining attention (Rasche et al., 2006; Xu et al., 2009; Aira et al., 2010; Ýnceoğlu et al., 2012; Cheng et al., 2020), and even minor genotype differences as between genetically modified and parental lines are believed to affect the microbial colonization of plant, particularly in vegetatively propagated crop. The seed stem-associated bacterial communities, independently of the genotypes and the soil type, is also a possible factor determining the specificity of the bacterial community in the tuber root system compartments. Nevertheless, the influence of genotype in our study is very evident.

In all cassava tuber samples, the dominant bacterial phyla were Firmicutes, Proteobacteria, and Actinobacteria (>1% of high-quality sequences) (Figure 1A). It has been previously reported that Proteobacteria, Firmicutes, and Actinobacteria are the dominant bacterial phyla in cassava (Li et al., 2020). Similar results based on both culture-dependent and culture-independent approaches have previously been reported for endophytes of other plants (Khan Chowdhury et al., 2017; Yang et al., 2017). In ginseng, Proteobacteria was found to have the highest abundance, followed by Firmicutes and Actinobacteria (Khan Chowdhury et al., 2017). Similarly, in peony, Proteobacteria, Firmicutes, and Actinobacteria have been reported to be the dominant bacterial phyla (Yang et al., 2017). In all cassava rhizosphere samples, the dominant bacterial phyla were Actinobacteria, Proteobacteria, Acidobacteria, Firmicutes, Chloroflexi, Planctomycetes, and Verrucomicrobia (>1% of high-quality sequences) (Figure 1A); these findings are in accordance with Sarr et al. (2017), who reported that the same soil bacterial communities were associated with cassava cultivation in Cameroon. Actinobacteria have also been shown to be enriched in exophytes of other plants, such as Pinus pinaster and maize-wheat (Triticum aestivum)/barley (Hordeum vulgare) rotation systems (Pérez-Izquierdo et al., 2019; Xiong et al., 2021); these bacteria are used as biocontrol agents to control soil- and seed-borne plant diseases (Priyadharsini and Dhanasekaran, 2015). These results suggest that rhizosphere microorganisms could play an important role in cassava cropping in tropical regions that experience various recurrent plant diseases.

The β-diversity analyses showed that bacterial communities in the rhizosphere soil varied across the different plant genotypes. PCoA indicated that the bacterial communities in the tuberous roots were not clearly separated by cassava genotype but that those in the rhizosphere soil were clearly separated by cassava genotype. These results were also supported by heatmap analyses at the genus level. Moreover, HCA demonstrated that the rhizosphere bacterial communities of the fourteen cassava genotypes were clustered based on the cassava genotypes (Figure 5C); these findings support the view that the cassava genotype influences the bacterial rhizosphere community. Similarly, Schlemper et al. (2017) reported that the community structures of the rhizosphere microbiome were significantly different among seven different sorghum cultivars. However, further studies are needed to confirm this hypothesis and to confirm the effects of genetic diversity on the compositions of root-associated bacterial communities.

We noted a few genera that were consistently enriched in the cassava tuberous roots, including Lactococcus, Bacillus, Oceanobacillus, and Carnobacterium (Figure 2A). Lactococcus and Bacillus improve plant resistance to diseases such as bacterial blight and root rot in cassava and wilt disease in cucumber (Xu et al., 2014; Freitas et al., 2019; Zhang et al., 2021). In addition, Pseudomonas was also a dominant genus in the tuber endosphere of 045. Pseudomonas species have been used to alleviate heavy metal toxicity and the negative effects of saline sodic field growth on wheat (Hassan et al., 2017). Eight genera (Gp1, Gaiella, Bacillus, Lactococcus, Gp3, Ktedonobacter, Rhodoplanes, and Spirillospora) were highly abundant in all rhizosphere samples (Figure 2B). This result is in accordance with Bao et al. (2019), who reported that the dominant bacteria in a paddy soil included Bacillus, Acidobacteria/Gp1, Acidobacteria/Gp3, and Ktedonobacter. Wu et al. (2021) found that the enrichment of beneficial bacteria, mainly Gaiella, contributed to the ability of ramie to tolerate poor soils. Our results showed that the cultivation of different cassava genotypes recruited different unique and core microbes to the cassava tubers and rhizospheres and that the microbes were significantly different in identity and relative abundance (Figure 3). This result is mainly attributed to the process by which cassava plants recruit different microbes, i.e., the release of a wide variety of exudates from tubers (Garbeva et al., 2004; Raaijmakers et al., 2009). LEfSe analysis identified specialized bacterial groups exclusively in the rhizosphere soil (Supplementary Figure 2), which suggests that these groups may play critical roles in maintaining the structure and function of rhizospheric soil bacterial communities. For example, Burkholderia, found in the R72 rhizosphere, was more abundant in a pineapple-banana crop rotation soil than in a banana monoculture soil, has the capacity to suppress Fusarium wilt of banana (Wang et al., 2015). Other reports who have demonstrated that bacteria from the Burkholderia genera possess a potential biocontrol ability through the production of varying compounds that inhibit plant pathogens (Mendes et al., 2011; Raaijmakers and Mazzola, 2012; Tenorio-Salgado et al., 2013). Spartobacteria_genera, found in the S0061 rhizosphere, was stimulated potentially to suppress the Fusarium wilt disease by sustainable biofertilizer application (Shen et al., 2015). Furthermore, we also found that most core microbes in the tubers and rhizosphere had plant growth-promoting potential. These core genera are known to produce various antibiotics, including bacillibactin and lipopeptides (produced by Bacillus) (Li et al., 2014; Liu et al., 2017), 2, 4-diac-etylphlor-oglucinol and phenazines (produced by Pseudomonas) (Mazurier et al., 2009; Hu et al., 2017), fusaricidin (produced by Paenibacillus) (Finch et al., 2018; Li and Chen, 2019), and thiopeptide and ectoine (produced by Streptomyces) (Cha et al., 2016). Overall, identifying these core and unique microbiomes is important for understanding the responsive microbial components associated with different plant genotypes.

In this study, the bacterial communities in the rhizosphere soils conditioned by the tuber metabolomes collected from three typical cassava genotypes, KU50, SC205, and SC9, were significantly different from each other taxonomically; in contrast, the bacterial communities in the tuberous roots were not significantly different. The correlation analysis between bacterial phyla and the produced exudates revealed that these differences could be linked to the exudation of certain tuber metabolites (Figure 6). These results thus support the notion that specific compounds within changing plant exudate profiles may drive soil microbial dynamics (Badri et al., 2013). Although the potential to mobilize soil nutrients is clearly already present in the soil microbiome, tuber exudates increase the functional potential of soil microbial communities. In addition, exudate concentrations can play a major role in shaping the abundances of microbial functional genes, which may be beneficial to plants (Badri et al., 2013). Moreover, the contents of measured metabolites in the three genotypes of cassava and their biological roles were significantly different; the metabolites included sugars, sugar acids, sugar alcohols, organic acids, amino acids, amides/amines, and others (Supplementary Figure 3). These differences could reflect a level of microbial community functional redundancy that was stimulated by the metabolomes. In a study of the rhizosphere microbiome of A. thaliana, Chaparro et al. (2013) suggested that plants exude sugars that are used by a wide variety of microorganisms as well as more specific exudates, such as phenolic compounds, that may be intended to attract more specific microbes. Amino acids, as specific chemoattractants for microorganisms, promote the chemotaxis of soil microbes to the rhizosphere (Barbour et al., 1991). Organic acids play a crucial role in nutrient acquisition (P, Fe, and Mn) by plants growing in nutrient-poor soils (Dakora and Phillips, 2002). Our result is in accordance with the finding that specific exudates of different Sorghum bicolor genotypes may influence the rhizosphere microbial community composition (Funnell-Harris et al., 2008). The available mineral nutrients in soils are not sufficient to meet the requirements of plants for optimal growth; thus, plants have evolved systems to recruit symbiotic microbial partners that increase the availability of nutrients (Tinker, 1984; Landeweert et al., 2001; Gyaneshwar et al., 2002; Adesemoye and Kloepper, 2009). Many studies of root exudate-mediated microbial defenses have developed robust models of coevolution between plants and soil microbes; for example, in the “cry for help” and “legacy effects” models (Weller et al., 2002; Yuan et al., 2018), plants select for microbial communities that help to suppress plant pathogens. The exudates released by different genotypes of the same plant species can vary, which affects the microbial community composition of the rhizosphere (Micallef et al., 2009; Inceoğlu et al., 2010).

The different cassava genotypes did not affect the richness or diversity of the endophytic bacterial community, but they affected the richness and diversity of the exophytic bacterial community. Furthermore, the cassava genotype shaped the endophytic and exophytic community structures and affected the relative abundances of core bacterial genera. The bacterial community structures varied between the tuber endosphere and the rhizosphere across cassava genotypes. The more dominant bacterial phyla associated with tubers and the rhizosphere were Firmicutes and Actinobacteria, respectively. Moreover, the majority of genera were associated with the tuberous roots and rhizosphere soils of specific cassava genotypes; this may have been due to the exudation of certain metabolites from the cassava tubers. These results suggest that plant genotypes affect the community composition of endophytic bacteria and may affect the community composition of exophytic bacteria through the exudation of metabolites.

The raw sequences were deposited in the NCBI Sequence Read Archive under BioProject accession number PRJNA750582.

YC and KL conceived and designed the experiments. JH, YG, RZ, YZ, XN, and XC performed the experiments and data analyses. JH drafted the manuscript. JH, YG, YC, and KL revised the manuscript. All authors have read and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.