Engineering multifunctional rhizosphere probiotics using consortia of Bacillus amyloliquefaciens transposon insertion mutants

  1. Jingxuan Li
  2. Chunlan Yang
  3. Alexandre Jousset
  4. Keming Yang
  5. Xiaofang Wang
  6. Zhihui Xu
  7. Tianjie Yang
  8. Xinlan Mei
  9. Zengtao Zhong
  10. Yangchun Xu
  11. Qirong Shen
  12. Ville-Petri Friman  Is a corresponding author
  13. Zhong Wei  Is a corresponding author
  1. Key Lab of Organic-based Fertilizers of China and Jiangsu Provincial Key Lab for Solid Organic Waste Utilization, Nanjing Agricultural University, China
  2. College of Life Science, Nanjing Agricultural University, China
  3. Department of Microbiology, University of Helsinki, Finland

Abstract

While bacterial diversity is beneficial for the functioning of rhizosphere microbiomes, multi-species bioinoculants often fail to promote plant growth. One potential reason for this is that competition between different species of inoculated consortia members creates conflicts for their survival and functioning. To circumvent this, we used transposon insertion mutagenesis to increase the functional diversity within Bacillus amyloliquefaciens bacterial species and tested if we could improve plant growth promotion by assembling consortia of highly clonal but phenotypically dissimilar mutants. While most insertion mutations were harmful, some significantly improved B. amyloliquefaciens plant growth promotion traits relative to the wild-type strain. Eight phenotypically distinct mutants were selected to test if their functioning could be improved by applying them as multifunctional consortia. We found that B. amyloliquefaciens consortium richness correlated positively with plant root colonization and protection from Ralstonia solanacearum phytopathogenic bacterium. Crucially, 8-mutant consortium consisting of phenotypically dissimilar mutants performed better than randomly assembled 8-mutant consortia, suggesting that improvements were likely driven by consortia multifunctionality instead of consortia richness. Together, our results suggest that increasing intra-species phenotypic diversity could be an effective way to improve probiotic consortium functioning and plant growth promotion in agricultural systems.

Editor's evaluation

The work is significant because it reports that intra-species phenotypic diversity in bacteria could be an effective way to improve probiotic consortia in agriculture. Two solid conclusions emerge from this important work: (1) Communities of near-clonal bacteria can be optimized based on functional traits to promote functional diversity. (2) Consortium multifunctionality – but not richness – is key to promoting bacterial colonization in roots and to protection against a plant pathogen. The work has broad practical implications for designing probiotic consortia that promote host health beyond the plant-microbe interaction field.

https://doi.org/10.7554/eLife.90726.sa0

Introduction

Bacterial rhizosphere microbiome diversity has been strongly associated with beneficial effects on plant growth in both natural and agricultural environments (Mendes et al., 2013; Yu et al., 2019). Such beneficial diversity effects are often associated with increased functional diversity and community multifunctionality (Saleem et al., 2019; Raza et al., 2020; Hu et al., 2021; Raza et al., 2021; Lindsay et al., 2021), where different taxa play complementary roles in plant growth promotion, specializing in nutrient solubilization, plant immune priming or for example by interacting with beneficial or pathogenic rhizosphere microbes (Hayat et al., 2010; Lugtenberg and Kamilova, 2009). Moreover, high microbial diversity can have other benefits such as providing functional redundancy in the case of species extinctions (Allison and Martiny, 2008), community stability (Hu et al., 2016), or provide emergent consortia-level properties that cannot be predicted based on the sum of individual species (Netzker et al., 2018; Abrudan et al., 2015; Tilman, 1982; Tilman, 2004). While several studies have tried to harness positive effects of bacterial diversity for crop production (Weidner et al., 2015; Hu et al., 2017; Wu et al., 2018), diverse inoculants often fail to produce the desired benefits in field conditions. This discrepancy could rise due to constraints set by external factors, such as resource availability (De-la-Peña et al., 2010; Gransee and Wittenmayer, 2000; Chaparro et al., 2013), that constrains the expression of plant growth promotion traits, or the presence of competitors, parasites, and predators that reduce the survival of the inoculated bacteria in the rhizosphere (Hibbing et al., 2010; Gao et al., 2019). Also intrinsic factors, such as interactions between inoculant consortia members, could constrain their survival in the rhizosphere (Wang et al., 2021; Jousset et al., 2011; Gu et al., 2020). It has thus been suggested that microbial inoculant design should aim to minimize negative interactions within the inoculated consortia without compromising consortia competitive ability, which is important for consortia establishment in the diverse rhizosphere microbiome, especially when being initially rare (Gu et al., 2020; Li et al., 2019; Wei et al., 2015).

Diversity-ecosystem functioning experiments are widely used to study if communities perform better due to inherent benefits of diversity or because certain communities contain specific species that are important for the community functioning (i.e., species identity effects due to presence of certain microbial keystone species) (Banerjee et al., 2018; Jousset et al., 2014). In microbial context, such experiments have been mostly conducted by using taxonomically diverse communities, where associations between community diversity and functioning have been found to range from positive to neutral and negative (Jousset et al., 2011; Wei et al., 2015; Maynard et al., 2017). One reason for such variety of outcomes might be that increasing community diversity often also introduces competition between the community members, which might overshadow and constrain the benefits of inoculant diversity. For example, antimicrobial activity of pathogen-suppressing volatile organic compounds has been shown to peak at intermediate levels of bacterial community diversity (Raza et al., 2020), and a similar hump-shaped pattern has also been found between toxin production and bacterial community richness (Jousset et al., 2011). Likely explanation for these findings is that the same compounds that are responsible for pathogens suppression also affect the competition between other consortia members (Gu et al., 2020). One way to reduce this negative effect would be to quantify species interactions in advance to assemble multifunctional bioinoculants with weak within-consortia competitive interactions (Li et al., 2019; Wei et al., 2015). Such optimization could be achieved by using highly related bioinoculant strains because genetic relatedness is predicted to favor cooperation instead of conflict due to kin selection (Simonet and McNally, 2021; Crespi, 2001; Kolter and Greenberg, 2006; Raymond et al., 2012). For example, if bioinoculant species produce antimicrobials, it is likely that related strains share the same antibiotic biosynthesis and resistance genes, and hence, unlikely antagonize each other (Xia et al., 2022). Increasing intra-species diversity could further lead to increased bioinoculant consortia multifunctionality, which has been shown to play an important role in the functioning of microbial ecosystems (Van Rossum et al., 2020; Fields et al., 2021; Raffard et al., 2019; Nicastro et al., 2020; Blake et al., 2021; Dragoš et al., 2018). Multifunctionality could be optimized based on ecological complementarity by assembling consortia from strains that use different niches or specialize in different functions within the same niche (Dragoš et al., 2018; Martin et al., 2016). This might allow more efficient expression or production of different compounds at the consortia level and help to overcome any antagonistic pleiotropic effects experienced at the individual strain level (i.e., trade-offs between plant growth promotion traits). Moreover, interactions between near clonal specialist genotypes could allow division of labor, where shared workload by two specialists leads to higher productivity relative to one generalist as has been demonstrated with Bacillus subtilis bacterium during biofilm matrix production (Dragoš et al., 2018). Such overperformance (or overyielding) could also be driven by other mechanisms such as resource partitioning, abiotic facilitation, or biotic feedbacks (Barry et al., 2019), leading to higher consortia functioning than predicted based on the sum of its individual members.

Here, we used a combination of genetics, molecular biology, and biodiversity-ecosystem functioning theory (Saleem et al., 2019; Hu et al., 2016) to test if increasing phenotypic diversity of a single plant growth-promoting Bacillus amyloliquefaciens T-5 bacterium offers a viable strategy to improve bioinoculant consortia multifunctionality. We chose B. amyloliquefaciens T-5 strain as our model species because it originates from the tomato rhizosphere (Tan et al., 2013) and has previously been shown to protect plants from various diseases, including bacterial wilt caused by phytopathogenic Ralstonia solanacearum bacterium (Jiang et al., 2017). To increase functional diversity within a single species, we first created a B. amyloliquefaciens mutant library using TnYLB-1 transposon mutagenesis (Le Breton et al., 2006), resulting in 1999 unique insertion mutants. A representative subset of 479 mutants were chosen for high-throughput phenotyping in vitro in the lab regarding four important plant growth promotion traits: biomass production, biofilm formation, swarming motility, and pathogen suppression (Lugtenberg and Kamilova, 2009; Turnbull et al., 2001; Raaijmakers et al., 2002). After phenotyping, a subset of 47 mutants that performed better or equally well as the wild-type strain were taken forward to plant experiments to determine if in vitro phenotyping could predict mutant success in terms of tomato root colonization and plant protection from R. solanacearum in vivo. Finally, we tested if plant growth promotion could be improved by combining phenotypically distinct mutants into multi-strain consortia. We predicted that increasing mutant consortia richness would lead to improved root colonization and plant protection if mutants are phenotypically dissimilar, which could result in positive diversity-ecosystem relationships or trait multifunctionality. These two hypotheses were tested by comparing the performance between phenotypically dissimilar and randomly assembled mutant consortia. Our results demonstrate that transposon insertion mutagenesis is an effective way to improve and identify genes underlying plant growth promotion traits in B. amyloliquefaciens. Importantly, we show that in vitro phenotyping can be used to optimize inoculant consortia functioning in vivo and that diverse mutant consortia are better at colonizing and protecting tomato plants when they are assembled based on phenotypic dissimilarity.

Results

Effects of transposon insertions on B. amyloliquefaciens T-5 mutant traits measured in vitro and in vivo

We first quantified the phenotypic effects of transposon insertions on B. amyloliquefaciens T-5 mutant traits across the whole mutant library (1999 mutants in total, Figure 1—figure supplement 1, Supplementary file 1a). To make the number of mutants more manageable for in vivo experiment, we randomly selected a subset of 479 mutants for further analyses (Supplementary file 1b). While we likely lost certain unique mutants in the process, the sampled subset was phenotypically representative of the original collection based on four measured traits (Mantel test; r=0.7591, p=0.04167). Within this subset, most insertions had negative effects on the four measured phenotypic traits, with more than half of the mutants showing reduced swarming (58.7%), biomass production (67.2%), and biofilm formation (60.8%) compared to the wild-type strain (Figure 1A). In contrast, the median effect of insertions affecting the pathogen suppression was neutral, and 51.1% of the mutants showed only a moderate increase in their suppressiveness (Figure 1A, Supplementary file 1b). In line with this finding, the distribution of effects of insertions on each trait was skewed, where beneficial mutations resulted mainly in a moderate improvement, while harmful mutations often led to severe reductions in measured traits (Figure 1A). Moreover, several insertions caused trade-offs, where improvement in one trait led to a reduction in another trait (Figure 1B). For example, swarming motility correlated negatively with biomass production, while biomass production led to a trade-off with both biofilm production and pathogen suppression (Figure 1B). These results thus suggest that transposon insertions constrained the simultaneous improvement of multiple traits, leading to specialized B. amyloliquefaciens T-5 mutants, which could be clustered in three phenotypic groups based on K-means clustering (Adonis test: R2=0.5283, p<0.001, Figure 1C, Figure 1—figure supplement 2). Compared to the other two clusters, mutants belonging to the cluster 1 showed significant increases in biofilm formation and pathogen suppression but reduced biomass production (Figure 1D, Supplementary file 2a). Mutants in the cluster 2 showed improved swarming motility and reduced pathogen suppression, while mutants in the cluster 3 had poor performance overall, showing highly reduced swarming motility and pathogen suppression (Figure 1D, Supplementary file 2a).

Figure 1 with 4 supplements see all
The effects of transposon insertions on the traits of 479 B. amyloliquefaciens T-5 mutants measured in vitro and in vivo.

(A) shows distribution of fold changes regarding mutants’ swarming motility, biomass production, biofilm formation, and pathogen suppression relative to the wild-type strain (black dashed line equaling 1). (B) displays pairwise covariance matrix between individual traits, where red cells indicate negative trait correlations (trade-offs) and blue cells positive traits correlations; white cells indicate no correlation between the given traits. (C) shows principal coordinates analysis showing the clustering of all mutants and the wild-type (black point). Mutants were assigned to different clusters based on K-means algorithm of four measured traits. (D) shows mean trait differences between clusters based on unpaired two-samples Wilcoxon test. The wild-type was assigned in the cluster 1 based on K-means clustering and its trait values are shown as dashed black lines. (E) displays root colonization of representative 47 B. amyloliquefaciens mutants from clusters 1–3 relative to the wild-type strain (black dashed line) based on cell densities in the root system 30 days post pathogen inoculation (dpi). (F) shows plant protection of representative 47 B. amyloliquefaciens mutants from clusters 1–3 relative to the wild-type strain (black dashed line), and negative ‘pathogen-only’ control (red dashed line), quantified as bacterial wilt disease incidence 30 dpi. Shaded areas in (E and F) represent the mean ± SEM. Since (D) displays the normalized trait values, variation for the wild-type strain is not shown. Pairwise differences in (D–F) were analyzed using unpaired two-samples Wilcoxon test: *** denotes for statistical significance at p<0.001; ** denotes for statistical significance at p<0.01; * denotes for statistical significance at p<0.05; ns denote for no significance.

To test if the mutants clustered in different phenotypic groups also differed in their tomato root colonization and ability to protect plants from R. solanacearum infections, 47 mutants representing a subset of three clusters were randomly selected for a greenhouse experiment (the specific effects of insertions on biological processes, cellular components, and molecular function for all mutants are shown in Figure 1—figure supplement 3B and Supplementary file 1c). Compared to the wild-type, 57% of these mutants (27/47) reached lower population densities in the rhizosphere (30 days post-pathogen inoculation [dpi]), and this was especially clear for mutants belonging to clusters 2 and 3. In contrast, mutants belonging to the cluster 1 retained more efficient root colonization compared to mutants from the other clusters, and more than half (16 of 27) of the cluster 1 mutants showed improved root colonization relative to the wild-type (30 dpi, Figure 1E, Supplementary file 2b). Around 93% of all mutants (44/47) exhibited reduced plant protection relative to the wild-type strain. However, mutants from the cluster 1 showed higher plant protection compared to the other two clusters, and specifically, two of the cluster 1 mutants showed improved plant protection relative to the wild-type strain (30 dpi, Figure 1F, Supplementary file 2b). Together, these results show that while most transposon mutants had reduced performance relative to the wild-type strain, some of the mutants showed improvement in at least in one plant growth promotion trait, which often resulted in trade-offs with some other traits.

Phenotypic trait variation explains mutant success in rhizosphere colonization and plant protection in vivo with tomato

To test if phenotypic trait variation measured in vitro correlates with beneficial effects on plants in vivo, the rhizosphere colonization and plant protection of 47 phenotypically distinct mutants was quantified after 5, 15, and 30 dpi in a greenhouse experiment. Results showed that B. amyloliquefaciens inoculations led to approximately 20.6% mean reduction in bacterial wilt disease incidence at the final time point of the experiment (30 dpi, Figure 2A–B). To establish a link between phenotypic variation measured in vitro and in vivo, we correlated mutant trait variation with root colonization and plant protection during different phases of the experiment. Trait correlations with the root colonization and plant protection became more significant in time and the most significant correlations were observed at the final time point (30 dpi), followed by middle (15 dpi) and early (5 dpi) sampling time points (Table 1). Specifically, swarming motility predicted the root colonization during the seedling stage (5 dpi, Table 1; Figure 2C, F1,46=7.65, R2=0.1239, p=0.0082), while swarming motility was positively associated with plant protection at the flowering stage (30 dpi, Table 1; Figure 2H, F1,46=15.08, R2=0.2306, p<0.001). Similarly, biofilm formation had positive associations with root colonization (Table 1; Figure 2E, 15 dpi: F1,46 = 4.40, R2=0.0675, p=0.0416, 30 dpi: F1,46 = 7.62, R2=0.1209, p=0.0089) and plant protection during vegetative and flowering stages (at 15 and 30 dpi), respectively (Table 1; Figure 2J, 15 dpi: F1,46 = 6.44, R2=0.1038, p=0.0146, 30 dpi: F1,46 = 8.69, R2=0.1406, p=0.0050). Pathogen suppression was positively associated with root colonization and plant protection at 30 dpi (Table 1; Figure 2F, F1,46=7.65, R2=0.1239, p=0.0082; Figure 2K, F1,46=15.65, R2=0.2538, p<0.001), while biomass production was not significantly associated with either root colonization or plant protection at any time points (Table 1; Figure 2D and I). As a result, the mean performance of mutants (‘Monoculture average performance’ index based on mean of all traits; see Materials and methods) was significantly positively correlated with both root colonization (Figure 2G, F1,46=7.77, R2=0.1259, p=0.0077) and plant protection (Figure 2L, F1,46=28.26, R2=0.3671, p<0.001) at the flowering stage (30 dpi). Together, these results suggest that while mutants with high trait values in biofilm formation, swarming motility, and pathogen suppression had positive effects on root colonization and plant protection, their relative importance varied depending on the growth stage of the plant.

Regression analysis explaining root colonization and plant protection of representative 47 B. amyloliquefaciens mutants based on their trait values measured in vitro at different sampling time points.

(A and B) show the dynamics of root colonization and plant protection, respectively. The red and black lines in (B) show the disease incidence of pathogen-only control and B. amyloliquefaciens mutant treatments, respectively. (C–G) and (H–L) show root colonization and plant protection, respectively, correlated with different traits at 5 days post pathogen inoculation (dpi) (gray), 15 dpi (light blue), and 30 dpi (dark blue) time points. Significant relationships and R-squared values are shown in panels with colors corresponding to the sampling time points (‘ns’ denotes for non-significant relationship).

Table 1
Analysis of variance (ANOVA) table summarizing the effects of mutant traits measured in vitro on the root colonization and plant protection.

Separate models were run for each dependent variable at different time points (5, 15, and 30 days post pathogen inoculation [dpi]) and all response variables were treated as continuous variables (bacterial abundances were log-transformed before the analysis). Table data represent only the most parsimonious models based on the Akaike’s information criterion (AIC) where ‘NA’ denotes variables that were not retained in the final models, ‘df’ denotes degrees of freedom, and ‘R2’ denotes total variance explained by regression coefficient of determination. The arrows represent the direction of coefficient values: ↑: coefficient >0; ↓: coefficient <0. Significant effects (p<0.05) are highlighted in bold.

Day post pathogen inoculation (dpi)
5 dpi15 dpi30 dpi
Mutant traitdfFpdfFpdfFp
Root colonization (Bacillus abundance – log CFU g–1 rhizosphere soil)
Swarming motility15.320.0259↑10.0060.9404↑10.0080.9302↑
Biomass production10.460.5032↓10.580.4512↑10.070.7886↓
Biofilm formation10.690.4103↑14.340.0433↑15.640.0220↑
Pathogen suppression12.040.1609↓10.0060.9410↓18.100.0068↑
No. of residuals43R2=0.087543R2=0.019343R2=0.1730
AIC: 16.18AIC: 3.82AIC: 25.06
Plant protection (disease incidence [%])
Swarming motilityNANA10.340.5620↑16.890.0119↓
Biomass productionNANA10.940.3369↓10.230.6339↑
Biofilm formationNANA15.710.0213↓112.340.0016↓
Pathogen suppressionNANA10.140.7090↓120.90<0.001↓
No. of residualsNA43R2=0.062543R2=0.4294
NAAIC: –101.24AIC: –112.38

Designing and testing the performance of mutant consortia in vitro and in vivo

As transposon insertions mainly improved the performance of mutants regarding only one phenotypic trait, we tested if B. amyloliquefaciens T-5 performance could be improved by using mutants as phenotypically diverse consortia. To this end, we chose eight mutants that showed improved performance relative to the wild-type strain regarding one of the plant-beneficial traits measured in vitro (Figure 3—figure supplement 1, Supplementary file 2d; two representative mutants per each measured trait were included). We first tested if these mutants showed negative effects on each other growth in vitro. Based on agar overlay assays, none of the strains clearly inhibited each other in direct contact in co-cultures. Similarly, only slightly negative (up to 12.7%) or positive (up to 7.1%) effects on strains’ growth were observed in supernatant exposure experiments (Figure 3—figure supplement 2, Supplementary file 2d). While direct co-culture experiments are needed to quantify the level of competitiveness between the mutants in the future, this data suggests that transposon insertions made mutants only slightly more inhibitory to each other. Mutants were then used to assemble a total of 37 consortia that varied in their richness level (1, 2, 4, or 8 mutants) and community composition, following a substitutive design where each mutant was equally often present at each richness level (see left panel key of Figure 3—figure supplement 3 for detailed composition of consortia). We hypothesized that consortia could show improved performance due to phenotypic complementarity or multifunctionality, where different mutants would ‘specialize’ respective to one of the four phenotypic traits, overcoming trade-offs and potential antagonistic pleiotropy experienced at the individual strain level (Figure 1B). We first tested the consortia performance regarding the four traits measured in vitro. We found that relative to wild-type strain, only a few consortia showed improved performance regarding swarming motility (8 of 37), biomass production (2 of 37), biofilm formation (4 of 37), or pathogen suppression (15 of 37) (Figure 3—figure supplement 3A–D). Moreover, consortia performance did not show clear relationship with increasing richness regarding to any of the measured traits (Figure 3—figure supplement 4). We further tested if the consortia performance could be predicted based on the trait averages of individually grown mutants, assuming that mutant performance is not affected by interactions between the consortia members. Only one significant positive relationship was found between the predicted pathogen suppressiveness, and the size of the inhibition halo observed in vitro lab measurements (Figure 3A–D). This suggest that individually measured mutant traits poorly predicted observed consortia performance in vitro except for the pathogen suppression.

Figure 3 with 5 supplements see all
The relationships between predicted and observed consortia performance measured in vitro and correlations between plant performance, consortia richness, and consortia average performance measured in vivo.

(A–F) show correlations between predicted and observed consortia performance regarding swarming motility, biomass production, biofilm formation, pathogen suppression, root colonization, and plant protection, respectively (blue dashed lines show 1:1 theoretical fit and solid black lines show the fitted regression between predicted and observed values). (G and H) show regression models where root colonization and plant protection were explained by B. amyloliquefaciens consortia richness, respectively. (I and J) show regression models where root colonization and plant protection were explained by B. amyloliquefaciens consortia average performance measured in vitro, respectively. In all panels, the black dashed lines show the performance of the wild-type strain, while red dashed lines in (F, H, and J) show the disease incidence of pathogen-only control treatment. In all panels, shaded areas show the confidence interval around the mean.

We next tested the consortia performance regarding root colonization and plant protection in vivo. While only 7 of 37 of consortia showed improved rhizosphere colonization, around half of them (18 of 37) exhibited a clear increase in plant protection compared to the wild-type strain (at 30 dpi, Figure 3—figure supplement 3E–F). While root colonization or plant protection could not be predicted based on the consortia performance measured in vitro (Figure 3E–F), increasing consortia richness improved both root colonization (Figure 3G, F1,35 = 6.52, R2=0.1330, p=0.0152) and plant protection (Figure 3H, F1,35 = 18.64, R2=0.3289, p<0.001), which were also positively correlated with the consortia average performance measured in vitro (root colonization: F1,35 = 6.47, R2=0.1319, p=0.0156; plant protection: F1,35 = 8.82, R2=0.1786, p=0.0053; Figure 3I–J). We also analyzed the significance of mutant identity effect on the consortia performance in vivo by excluding each strain from the dataset and comparing model fit and significance of explanatory variables. The presence of M54 mutant (efficient in biofilm formation) significantly increased consortia root colonization, while the presence of mutants M59 (efficient in biomass production) and M143 (efficient in biofilm formation) significantly improved plant protection (Figure 3—figure supplement 5, Supplementary file 2e). Crucially, the effect of consortia richness remained significant after sequential removal of each mutant and refitting of the model, which demonstrates that the diversity effect was robust and relatively more important in explaining root colonization and plant protection compared to mutant identity effects (Supplementary file 2f). Together, these data suggest that mutant consortia diversity was positively linked with consortia performance in vivo, indicative of positive diversity-ecosystem functioning relationship.

To test if the positive diversity effect was potentially driven by consortium multifunctionality, we compared the performance of the ‘optimized’ 8-mutant consortium (no. 37; Figure 3—figure supplement 3) used in the diversity-ecosystem functioning experiment with eight randomly assembled 8-mutant consortia, which could be considered as phenotypically ‘unoptimized’. We found that the ‘optimized’ consortium was more effective at both plant root colonization and plant protection compared to ‘unoptimized’ 8-mutant consortia (Figure 4A and B). We also correlated the relative performance of all 8-mutant consortia with both in vivo traits and found that both root colonization (Figure 4C, F1,7 = 26.56, R2=0.7616, p=0.0013) and plant protection (Figure 4D, F1,7 = 6.39, R2=0.4026, p=0.0394) improved along with the increase in the relative performance of consortia. Together, these results suggest that in vitro and in vivo phenotyping could reliably predict the root colonization and plant protection of 8-mutant B. amyloliquefaciens consortia.

Optimized and randomly assembled 8-mutant consortia show contrasting effects on root colonization and plant protection.

(A–B) compare differences between optimized (white bar) and randomly assembled (gray bars) B. amyloliquefaciens 8-mutant consortia on root colonization and plant performance based on Student’s t-test (n=3): *** denotes for statistical significance at p<0.001; ** denotes for statistical significance at p<0.01; * denotes for statistical significance at p<0.05. In both (A and B), Y-axes show the consortia performance as a fold change relative to wild-type strain and shaded areas represent the mean ± SEM. (C–D) show positive and negative correlations between consortia relative performance (calculated based on average performance and trait deviance, see Materials and methods) with root colonization (C) and disease incidence (D); optimized and randomly assembled 8-mutant consortia are shown as white and black circles, respectively (shaded area shows the confidence interval around the mean).

Discussion

In this work, we tested if increasing intra-species diversity of B. amyloliquefaciens T-5 bacterium via mutagenesis could offer a viable strategy for improving mutant consortia multifunctionality and plant health (Figure 3G–H). Our results show that mutations that improved bacterial performance regarding one trait often led to specialism and reduced performance regarding other traits. Such trait trade-offs experienced at the individual genotype level could be overcome by assembling consortia of phenotypically distinct mutants, that showed increase in average trait performance. Crucially, the consortia richness and average trait performance correlated positively with increased root colonization and plant protection, indicative of increased consortia multifunctionality and improved plant health. Especially, diverse 8-mutant consortium that consisted of phenotypically distinct mutants performed better compared to randomly assembled 8-mutant consortia consisting of phenotypically similar mutants. Together, these findings suggest that increasing intra-species functional diversity could offer an easy solution for improving the performance of bacterial consortia.

We specifically focused on four B. amyloliquefaciens traits that have previously been linked to plant growth promotion: swarming motility, biomass production, biofilm formation, and direct pathogen suppression via antibiosis (Chen et al., 2013; Huang et al., 2020; Fira et al., 2018; Moreno-Velandia et al., 2019). While most transposon insertions reduced the strains’ performance, some of them improved at least one measured trait. However, all trait improvements were costly and reduced mutants’ performance regarding other traits, indicative of trade-offs and antagonistic pleiotropy (Eyre-Walker and Keightley, 2007). Such costs of adaptation are common with microbes and have previously been linked to a wide range of functions, including metabolism, antibiotic production, motility, and stress resistance (Yang et al., 2019; Ferenci, 2016). Overall, transposon insertions were identified in several genes associated with broad range of functions (Figure 1—figure supplement 3, Supplementary file 1c). With eight phenotypically distinct mutants that were used for consortia assembly experiment, increased swarming motility was associated with insertions in parE (DNA topoisomerase IV subunit B) and DeoR (DNA-binding transcriptional repressor) genes, which has previously been linked to antibiotic resistance and bacterial deoxyribonucleoside and deoxyribose utilization (Saxild et al., 1996), respectively. While it remains unclear how these genes were linked with swarming motility, their disruption also affected other traits as evidenced by reduced biofilm formation. Increased biomass production was linked to disruption of comQ (competence protein) and hutI (imidazolonepropionase) genes and trade-offs with the other three measured traits. ComQ gene controls the production of ComX pheromone (Bacon Schneider et al., 2002) and has recently been linked to antimicrobial activity (Esmaeilishirazifard et al., 2018), which could explain reduction in the pathogen suppression by this mutant. Insertion in histidine utilization (hut) system gene, hutI, could have potentially impaired catabolite and amino acid repression, resulting in improved biomass production with one of the mutants (Eda et al., 1999; Bender, 2012). Interestingly, increased biofilm formation was also linked to insertion in hut system (hutU) with one mutant, while the other mutant had insertion in YsnB gene, which encodes for a putative metallophosphoesterase. While both insertions were linked to trade-offs with swarming motility and biomass production, they are not commonly associated with biofilm formation in Bacillus (Dragoš et al., 2018; Prágai et al., 2001). Finally, moderate improvement in pathogen suppression was observed with two mutants that had insertions in nhaC (sodium-proton antiporter) and dfnG (difficidin polyketide synthesis) genes. The nhaC gene is known to act as repressor for Pho regulon in Bacillus (Prágai et al., 2001), while mutations in Pho regulon have been linked to increased antibiotics production with several Streptomyces species (Santos-Beneit, 2015). Moreover, the dfnG gene controls the production of difficidin antibiotic, which has previously been linked to biocontrol activity against fire blight and Xanthomonas oryzae rice pathogen (Wu et al., 2015) and could have also suppressed the growth of R. solanacearum. Similar to the other phenotypically distinct mutants, insertions associated with increased pathogen suppression led to trade-offs with other measured traits. While more work is required to unequivocally link these mutations with associated traits at the molecular level, our findings show that all above trait improvements achieved via mutagenesis resulted in trade-offs with other traits and that this technique could be used to identify genes underlying plant growth promotion and pathogen suppression.

To overcome trait trade-offs experienced at the individual mutant level, we tested if we could improve the plant growth promotion by combining phenotypically distinct mutants into multifunctional consortia. We found that mutant performance measured in monocultures was a poor predictor of B. amyloliquefaciens performance in consortia in vitro, except for the pathogen suppression. This suggest that while the selected mutants did not show direct antagonism toward each other, they likely interacted in other ways leading to unpredictable trait expression when growing together (e.g. via certain emergent effects; Goldford et al., 2018). Despite this, we found clear diversity effects, where consortia richness and average performance were positively associated with both plant root colonization and plant protection from R. solanacearum pathogen infection. While important mutant identity effects were also observed, omission of these strains did not change the significance of underlying diversity effects, highlighting the importance of interactions between the consortia members in determining the positive effects on the plant health. Together, these results suggest that consortia that performed well regarding all measured traits on average had improved ability to colonize rhizosphere and suppress the pathogen. To test if these effects were driven by diversity per se or underlying trait variation between consortia members, we compared the optimized 8-mutant consortium with eight randomly assembled 8-mutant consortia that were phenotypically more similar. We found that randomly assembled consortia performed less well on average, while consortia functioning improved along with consortia relative performance, suggesting that diverse consortia performed better only when they had been assembled from phenotypically dissimilar mutants. While similar positive diversity-ecosystem functioning relationships have previously been found in more complex Pseudomonas (Hu et al., 2021; Hu et al., 2016; Hu et al., 2017), leaf bacterial (Laforest-Lapointe et al., 2017) and grass-land soil microbiomes (Wagg et al., 2019), we here show that this pattern also holds along with intra-species diversity gradient. While the measured phenotypic traits are considered to be robust to inoculum densities, it will be important to evaluate in the future if the absolute abundances of each mutant play a role in the consortia functioning.

Positive diversity effects have previously been explained by facilitation, ecological complementarity, and division of labor, which can reduce competition between the consortia members within or between niches (Dragoš et al., 2018; Martin et al., 2016; Bulleri et al., 2016). Moreover, high diversity could provide stability for consortia functioning via insurance effects by increasing the likelihood of certain consortia members surviving in the soil after the inoculation (Yachi and Loreau, 1999; Boles et al., 2004). While our experiments were not designed to disentangle the relative importance of these potential mechanisms, we analyzed which mutant traits could significantly explain the dynamics of root colonization and plant protection at seedling, vegetative, and flowering stages of the tomato growth by focusing on 47 mutants. While the effect of Bacillus biomass production was consistently non-significant, both motility and biofilm formation were positively associated with the root colonization. However, motility was significant only at the seedling stage, while biofilm became significant during vegetative and flowering stages. In line with ecological succession often taking place in the rhizosphere (Chen et al., 2013), high motility might have allowed faster colonization of relatively ‘sterile’ young roots by Bacillus, while biofilm formation could have promoted stress tolerance and resource competition in more diverse and mature microbial communities during the later stages of tomato growth (Bais et al., 2004; Xu et al., 2013). Interestingly, increase in pathogen suppression, biofilm formation, and motility were positively associated with improved plant protection at the flowering stage, which suggests that all these traits were positively associated with the ecosystem functioning in terms of plant health. It is thus possible that consortia were together able to overcome the trait trade-offs experienced at the individual mutant level, leading to improved and more stable ecosystem functioning for the whole duration of tomato growth cycle. In addition, it is possible that some of the measured plant growth promotion traits might act as public goods (Lee et al., 2010; Driscoll et al., 2011), which could have been shared between different mutants, leading to overall improvement in the mutant population fitness. Such micro-scale mechanisms could be potentially validated in the future using transcriptomics and barcoded Tn-seq mutants, which would allow estimating activity and changes in mutant frequencies during bioinoculation.

In conclusion, we here demonstrate that the beneficial effects provided by a single B. amyloliquefaciens bacterium can be improved by increasing consortia functional diversity using transposon mutagenesis. Our approach highlights the importance of intra-species genetic diversity for the ecosystem functioning and provides a trait-based approach for designing microbial communities for biotechnological applications. Our approach does not require a priori knowledge on specific genes or molecular mechanisms, but instead relies on generation of trait variation which is screened and selected by the experimenter. Our method can also help to identify novel functional roles of previously characterized and uncharacterized genes. While the benefit of this method was here demonstrated in the context of agriculture, it could be applied in other biotechnological contexts, such as biofermentation, waste degradation, and food manufacturing. Future work focusing on the population dynamics, metabolism, and gene expression of different mutants will help to understand the relative importance of ecological complementarity, division of labor, insurance effects, population asynchrony (Blüthgen et al., 2016), and facilitation for the consortia ecosystem functioning.

Materials and methods

Bacterial strains and culture conditions

Request a detailed protocol

We used phytopathogenic R. solanacearum QL-Rs1115 (Wei et al., 2011) and B. amyloliquefaciens T-5 biocontrol strains (Tan et al., 2013; Wang et al., 2017) as our model bacterial species. The B. amyloliquefaciens T-5 can suppress the growth of R. solanacearum QL-Rs1115 by competing for space and nutrients in the rhizosphere (Tan et al., 2016) and by producing various antibacterial secondary metabolites (Yang et al., 2019). Both bacterial stocks were cryopreserved at –80°C in 30% glycerol stocks. Prior starting the experiments, active cultures were prepared as follows: B. amyloliquefaciens T-5 was grown at 37°C in Lysogeny Broth (LB-Lennox, 10.0 g L–1 Tryptone, 5.0 g L–1 yeast extract, 5.0 g L–1 NaCl, pH = 7.0) and R. solanacearum QL-Rs1115 was grown at 30°C in Nutrient Broth (NB, 10.0 g L–1 glucose, 5.0 g L–1 peptone, 0.5 g L–1 yeast extract, 3.0 g L–1 beef extract, pH = 7.0) for 24 hr.

Generation of B. amyloliquefaciens T-5 transposon mutant library

Request a detailed protocol

To increase the intra-species diversity of B. amyloliquefaciens, we generated a random transposon insertion mutant library by using a TnYLB-1 transposon derivative, carried in the thermosensitive shuttle plasmid pMarA (Supplementary file 2g), which was electrotransformed to bacteria as previously described by Zakataeva et al., 2010; Ito and Nagane, 2001. The cells with intact pMarA plasmid contained resistance cassettes to both erythromycin and kanamycin, while the cells with integrated transposons were resistant only to kanamycin. Transposon mutant library was created as follows. An overnight B. amyloliquefaciens T-5 cell culture grown in neutral complex medium (NCM, 17.4 g L–1 K2HPO4, 11.6 g L–1 NaCl, 5 g L–1 glucose, 5 g L–1 tryptone, 1 g L–1 yeast extract, 0.3 g L–1 trisodium citrate, 0.05 g L–1 MgSO4.7H2O, and 91.1 g L–1 sorbitol, pH = 7.2) was diluted 25-fold with fresh NCM supplemented with 5 mg mL–1 of glycine and grown at 30°C for 3 hr on a rotary shaker (170 rpm). After 1 hr incubation (at an optical density OD600∼ 0.8), cells were cooled on ice, harvested by centrifugation (8000×g for 6 min at 4°C) and washed four times with ice-cold electrotransformation buffer (ETM, 0.5 M sorbitol, 0.5 M mannitol, and 10% glycerol). Resulting pellets were resuspended in ETM buffer supplemented with 10% PEG 6000 and 1 mM MgCl2, yielding approximately 1010 cells mL–1. Cells were then mixed with 500 ng of plasmid DNA in an ice-cold electrotransformation cuvette (2 mm electrode gap), and after 1–3 min incubation at room temperature, exposed to a single electrical pulse using a MicroPulser Electroporator (Bio-Rad Laboratories) at field strength of 7.5 kV cm–1 for 4.5–6 ms. Immediately after the electrical discharge, cells were transferred into 1 mL of LB, incubated with gentle shaking at 30°C for 3–8 hr, and plated on LB agar containing 10 μg mL−1 erythromycin. Transformants were selected after 36–48 hr incubation at 30°C. To generate final transposon library, erythromycin-resistant colonies with plasmids were individually transferred to fresh LB and incubated overnight at 30°C, after cultures were diluted, spread on LB plates supplemented with 10 μg mL−1 kanamycin, and incubated for 24 hr at 46°C. As the plasmid cannot replicate at 46°C, only cells with an integrated transposons grew and could be separated. A total of 1999 transformed colonies were isolated and individually cryopreserved in 30% glycerol at –80°C.

Phenotypic characterization of B. amyloliquefaciens T-5 mutant library in vitro

Request a detailed protocol

The wild-type strain and 1999 mutants were phenotyped for following plant-growth promoting traits: swarming motility, biomass production, biofilm formation, and pathogen suppression via production of antimicrobials (see below). These traits were selected due to their known importance for B. amyloliquefaciens competitiveness in the rhizosphere and their involvement in pathogen suppression (Chen et al., 2013; Huang et al., 2020; Fira et al., 2018; Moreno-Velandia et al., 2019). To prepare bacterial inoculants, frozen colonies were picked and pre-grown overnight in LB at 37°C, washed three times in 0.85% NaCl and adjusted to initial OD600 of 0.5 (∼ 107 cells mL–1, based on OD vs colony forming unit [CFU] calibration curve, Figure 1—figure supplement 4). In addition to each individual trait, we also calculated the average of all measured traits and used the resulting ‘monoculture average performance’ (Wagg et al., 2014) index to compare mutants’ overall performance. Of the 1999 phenotyped mutants, a subset of 479 mutants were randomly selected for more detailed analysis and probiotic bioinoculant design. While we likely missed certain mutants with this method, the 479 mutants represented a similar phenotypic diversity as the 1999 mutant collection (Mantel test; r=0.7591, p=0.04167), indicating that our sampling captured a phenotypically representative subsample of mutants (Supplementary file 1a and b).

Swarming motility was measured using a previous method described by Kearns (Kearns et al., 2004). Briefly, 2 μL of each B. amyloliquefaciens mutant was inoculated into the center of 0.7% agar LB plates supplemented with 10 μg mL–1 of kanamycin. After 24 hr incubation at 30°C, swarming motility was evaluated as the radius of the colony. Three replicates were used for each mutant.

Biomass production was assessed on 96-wells microtiter plates (at 30°C with agitation) in 200 μL of ‘recomposed exudate’ medium (abbreviated as ‘RE’, which contained: 0.5 g L–1 MgSO4.7H2O, 1.0 g L–1 K2HPO4, 0.5 g L–1 KCl, 1.0 g L–1 yeast extract, 1.2 mg L–1 Fe2(SO4)3, 0.4 mg L–1 MnSO4, 1.6 mg L–1 CuSO4, 2 g L–1 (NH4)2SO4, 0.8 g L–1 glucose, 1.3 g L–1 fructose, 0.2 g L–1 maltose, 0.02 g L–1 ribose, 5.6 g L–1 citrate, 1.4 g L–1 succinate, 0.2 g L–1 malate, 0.8 g L–1 casamino acids Nihorimbere et al., 2012). After 24 hr of growth, OD600 was measured as a proxy for biomass production (Figure 1—figure supplement 4). Three replicates were used for each mutant.

Biofilm formation was assessed as described previously (Hamon and Lazazzera, 2001) using 24-well polyvinyl chloride microtiter plates instead of 96-well plates. Briefly, 10 μL B. amyloliquefaciens cells were added into 1 mL of biofilm-promoting growth medium (MSgg minimal medium: 2.5 mM PBS [pH 7.0], 100 mM MOPS [pH 7.0], 50 μM FeCl3, 2 mM MgCl2, 50 μM MnCl2, 1 μM ZnCl2, 2 μM thiamine, 50 mg phenylalanine, 0.5% glycerol, 0.5% glutamate, and 0.7 mM CaCl2) on 24-well microtiter plates and incubated without agitation for 24 hr at 30°C (Branda et al., 2001). The growth medium and planktonic cells were removed, and remaining cells adhered on well walls were stained with 1% crystal violet dissolved in washing buffer (0.15 M (NH4)2SO4, 100 mM K2HPO4 [pH 7], 34 mM sodium citrate, and 1 mM MgSO4) for 20 min at room temperature. Plates were then rinsed with demineralized water to remove excess crystal violet, after the remaining crystal violet bound to well wall biofilms were solubilized in 200 μL of solvent (80% ethanol, 20% acetone). Biofilm formation was defined as the optical density of crystal violet at OD570. Three replicates were used for each mutant.

Pathogen suppression via production of antibiotics was assessed as inhibition of R. solanacearum QL-Rs1115 strain using an agar overlay assay (Parret et al., 2005). Briefly, small volume drops (2 μL) of each B. amyloliquefaciens mutant and wild-type strain were spotted on NA soft agar plates and incubated for 24 hr at 30°C. Next, these plates were chloroform-fumigated to kill all the bacteria (Parret et al., 2005), leaving only the secreted antimicrobials and then fully covered with R. solanacearum suspension (with a final concentration of approximately 107 cells mL–1). The pathogen suppression of each mutant was defined as the area of R. solanacearum inhibition halo around the B. amyloliquefaciens colony (in mm2), which is proportional to antibiotic production (Delignette-Muller and Flandrois, 1994). Three replicates were used for each mutant.

Selecting a representative subset of B. amyloliquefaciens T-5 mutants for greenhouse experiments

Request a detailed protocol

In order to select a representative subset of mutants for greenhouse experiments, we first used K-means clustering (Hartigan and Wong, 1979) to divide the wild-type and 479 phenotyped mutants into clusters based on swarming motility, biomass production, biofilm formation, and pathogen suppression (Supplementary file 1b). Briefly, K-means clustering assigns n observations into k clusters where each observation (in our case mutant) belongs to a cluster with the nearest mean (cluster centers or cluster centroid). According to the gap statistic method (Tibshirani et al., 2001), three clusters was the optimum number (k) for this dataset (Figure 1—figure supplement 2). With this method, each mutant was hence assigned to one of the clusters. Clustering was further visualized using principal component analysis (PCA) based on the first two principal components to show the variety of mutants. We randomly selected approximately 10% of strains from each cluster, resulting in a subset of 47 mutants, which were used for greenhouse experiments (26, 11, and 10 mutants from clusters 1, 2, and 3, respectively, Supplementary file 1c). These 47 mutants were further analyzed to determine the disrupted genes by TnYLB-1 transposon insertion using the inverse polymerase chain reaction (IPCR) method as previously described by Le Breton et al., 2006. First, 5 μg of genomic DNA isolated from each respective transposon mutant was digested with Taq I and circularized using ‘Rapid Ligation’ kit (Fermentas, Germany). IPCR was carried out with ligated DNA (100 ng), using oIPCR1 and oIPCR2 primers (Supplementary file 2h). The cloned sequences were then purified using PCR purification kit (Axygen, UK) and the flanking genomic regions surrounding the transposon insertion sites were sequenced using the primer oIPCR3 (Supplementary file 2h). Obtained DNA sequences were compared against available databases (GenBank and Bacillus Genome Data-base) using the BLASTX and BLASTN available at the NCBI, and against the complete ancestral B. amyloliquefaciens T-5 genome sequence (Accession: CP061168, Figure 1—figure supplement 3A, Supplementary file 1c). The functional classification of disrupted genes for all 47 transposon mutants is summarized in Figure 1—figure supplement 3B.

Assessing the performance of individual B. amyloliquefaciens T-5 mutants in a greenhouse experiment

Request a detailed protocol

All selected 47 mutants and the wild-type strain were individually screened for their ability to colonize tomato rhizosphere and protect plants against infection by R. solanacearum QL-Rs1115 pathogen strain in a 50-day-long greenhouse experiment. Surface-sterilized tomato seeds (Lycopersicum esculentum, cultivar ‘Jiangshu’) were germinated on water agar plates in the dark at 28°C for 2 days, before sowing to sterile pots containing wet vermiculite (Huainong, Huaian Soil and Fertilizer Institute, Huaian, China). Ten-day-old tomato seedlings (at three-leaves stage) were then transplanted to seedling trays containing natural, non-sterile soil collected from a tomato field in Qilin Town, Nanjing, China (Chen et al., 2013). Plants were inoculated with individual B. amyloliquefaciens mutants by drenching, resulting in a final concentration of 107 CFU g–1 soil (Wei et al., 2013). The R. solanacearum strain was inoculated using the same method 1 week later at a final concentration of 106 CFU g–1 soil. Positive control plants were treated only with R. solanacearum, while negative control plants received no bacterial inoculants. Three replicated trays were set up for each treatment, with 20 seedlings (in individual cells) per tray. Each tray was considered as one biological replicate. Tomato plants were grown for 30 dpi with natural temperature (ranging from 25°C to 35°C) and lighting variation (around 16 hr of light and 8 hr of dark). Seedling trays were rearranged randomly every second day and regularly watered with sterile water.

Quantifying B. amyloliquefaciens mutants’ root colonization and plant protection in the rhizosphere

Request a detailed protocol

The root colonization and plant protection of 47 B. amyloliquefaciens T-5 mutants was quantified individually as a change in their population densities in the tomato rhizosphere after 5, 15, and 30 days of R. solanacearum pathogen inoculation (‘dpi’). At each sampling time point, three independent plants per inoculated mutant were randomly selected and sampled destructively by carefully uprooting the plant and gently removing the soil from the root system by shaking. After determining plant fresh weight, the root system of each plant was thoroughly ground in 5 mL of 10 mM MgSO4·7H2O using a mortar, and serial dilutions of root macerates were plated on a semi-selective Bacillus medium consisting of 326 mL L–1 vegetable juice (V8, Campbell Soup Co., USA), 33 g L–1 NaCl, 0.8 g L–1 dextrose, 16 g L–1 agar (pH 5.2, adjusted with NaOH) supplemented with 45 mg L–1 cycloheximide and 22.5 mg L–1 polymyxin B (Kinsella et al., 2009). This media was used to count the densities of B. amyloliquefaciens T-5 wild-type, and the same media supplemented with 10 μg mL–1 kanamycin was used to count the densities of mutant strains. Plates were incubated at 30°C for 30 hr and bacterial densities expressed as CFU per gram of root biomass. The effect of B. amyloliquefaciens wild-type and mutants on plant protection was measured as the reduction of bacterial wilt disease symptoms during the experiment (based on the proportion of plants showing wilting symptoms). The first wilting symptoms appeared 7 dpi and the proportion of diseased plants quantified at 5, 15, and 30 dpi were used in analyses. Plant protection was expressed as the relative reduction in the number of wilted plants compared to the positive control (only R. solanacearum inoculated in the absence of B. amyloliquefaciens T-5 mutants or wild-type).

Assembly of phenotypically dissimilar B. amyloliquefaciens mutant consortia

Request a detailed protocol

To test if the performance of B. amyloliquefaciens T-5 mutants could be improved by using consortia of phenotypically dissimilar mutants, a subset of eight best-performing mutants excelling at different phenotypic traits were selected (Figure 3—figure supplement 1, Supplementary file 2c). Specifically, these included two mutants that showed high swarming motility (M108: pare and M124: DeoR), high biomass production (M59: comQ and M109: hutI), high biofilm formation (M54: hutU and M143: YsnB), and slightly improved pathogen suppression (M38: nhaC and M78: dfnG) relative to the wild-type strain (Figure 3—figure supplement 1, Supplementary file 2c). To test the effect of transposon insertions on potential antagonism between the mutants, we conducted two types of assays: direct growth inhibition by (1) spotting each strain on top of the others using agar overlays and(2) growing each strain in the supernatant of the other strains. With agar overlay assays, 2 μL of each mutant with density OD600 of 0.5 (∼107 cells mL–1) was spotted on the soft agar overlay of the other mutants and direct antagonistic effect was measured as the size of the inhibition halo observed on the soft agar plates (Fields et al., 2022). For the supernatant assay, we first cultured each mutant in liquid LB for 2 days and collected supernatants by using 0.22 μm filters. In the growth assays, 2 μL of each strain with initial concentration of 107 cells mL–1 was mixed with 20 μL of each supernatant and 178 μL of 50% LB. The growth of each strain was measured after 24 hr as optical density (OD600), and inhibition calculated as the relative growth of each strain in its own or other strains’ supernatant compared to strains’ growth in the fresh 50% LB (diluted with sterile water). Here, the reduced growth in other strains’ supernatant relative to the growth in the fresh medium was deemed as inhibition between mutants. A following formula was used where OD600 sup and OD600 LB denote for mutants’ growth in other mutants’ supernatant or in 50% fresh LB after 24 hr:

Relativegrowth=OD600supOD600LBOD600LB×100%

These eight mutants were then used to assemble a total of 29 consortia with 2, 4, or 8 mutants, following a substitutive design where each mutant was equally often present at each community richness level (see left panel key of Figure 3—figure supplement 3 for detailed consortia assembly). Mutants were mixed in equal proportions in each consortium with final total bacterial density OD600 of 0.5 (e.g., 50:50% or 25:25:25:25% in two and four mutant consortia, respectively; ∼ 107 cells mL–1). This design has previously been used to investigate biodiversity-ecosystem functioning relationships in plant-associated bacterial communities (Hu et al., 2016; Becker et al., 2012), allowing disentangling the effects due to consortia richness, composition, and mutant strain identity. In addition, to compare the performance of optimized 8-member consortium (assembled based on phenotypic dissimilarity; see above) with non-optimized 8-mutant consortia, we assembled eight additional 8-mutant consortia randomly from the 479 mutant collection, which were used in in vitro lab and in vivo greenhouse experiments.

Phenotypic characterization of B. amyloliquefaciens consortia performance in vitro and consortia root colonization and plant protection in the tomato rhizosphere

Request a detailed protocol

The performance of each mutant and assembled consortium was assessed in vitro in the lab by measuring traits as mono- and co-cultures following the same methods as described previously (swarming motility, biomass production, biofilm formation, and pathogen suppression). Mutant strains were prepared individually from frozen stocks by growing overnight in liquid LB, pelleted by centrifugation (4000×g, 3 min), washed three times with 0.85% NaCl and adjusted to OD600 of 0.5 (107 cells mL–1). Consortia were then assembled following the substitutive design describe earlier (Figure 3—figure supplement 3) by mixing mutants in equal proportions for each consortium with total bacterial density OD600 of 0.5 (107 cells mL–1; e.g., 50:50% or 25:25:25:25% in two and four mutant communities, respectively). Consortia traits were characterized as described previously and compared with the ancestral B. amyloliquefaciens wild-type strain. The root colonization and plant protection of B. amyloliquefaciens T-5 consortia were quantified in greenhouse experiments following previously described methods. Predicted performances were calculated following the additive model, equaling the sum of different trait values of each member divided by the richness value of the given consortium. To link the performance of single mutant with functioning of consortia, we used the relative performance measure, which included the magnitude and direction of difference relative to the wild-type strain. The difference and direction in magnitude to the wild-type strain were calculated based on the Euclidean distance and average performance using following formula:

Relative performance=i=1n(Di×APiAPwt|APiAPwt|)n,

Di, Euclidean distance between each consortium member and wild-type based on four traits; APi, average performance of each community member; APwt, average performance of wild-type; n, community richness.

Statistical analyses

Request a detailed protocol

Data were analyzed with a combination of analysis of variance (ANOVA), PCA, linear regression models, unpaired two-sample Wilcoxon tests, and Student’s t-test. Individually measured mutant traits data was normalized between 0 and 1 across the all collection using min-max normalization (Jain et al., 2005). In addition, the different phenotypic traits were combined into a ‘Monoculture average performance’ index, which was calculated as the mean of the four standardized traits for each mutant. Monoculture average performance and consortia traits values were also min-max normalized between 0 and 1 for subsequent analyses. To classify mutants into different functional groups, K-means clustering algorithm (‘fviz_nbclust’ in ‘factoextra’ package and ‘kmeans’ function) was used and clusters were visualized using PCA (‘princomp’ in ‘vegan’ package) based on multivariate trait data. The phenotypic dissimilarity between the same mutants and the wild-type strain was calculated using ‘vegdist’ based on ‘Euclidean’ algorithm. The B. amyloliquefaciens T-5 abundance data measured in root colonization assays were log10-transformed and disease incidence data were arcsine square root-transformed prior the analyses. Linear regression models were used to explain root colonization and plant protection with mutant traits, average performance, consortia richness, and consortia relative performance. Treatment mean differences were analyzed using two-sample Wilcoxon test (‘wilcox.test’ function) or Student’s t-test (‘t.test’ function) depending on the unequal or equal sample sizes, respectively. The temporal effects of four traits on root colonization and plant protection were assessed separately for different time points using ANOVA (‘aov’ function). All statistical analyses were performed using R 3.5.2 (R core Development Team, Vienna, Austria). All code used in this study is available on request from corresponding authors.

Data availability

All data generated or analysed during this study are included in the manuscript and supporting file; Source Data files have been provided in Source data 1.

References

    1. Tilman D
    (1982)
    Resource competition and community structure
    Monographs in Population Biology 17:1–296.

Decision letter

  1. Stéphane Hacquard
    Reviewing Editor; Max Planck Institute for Plant Breeding Research, Germany
  2. Detlef Weigel
    Senior Editor; Max Planck Institute for Biology Tübingen, Germany
  3. Stéphane Hacquard
    Reviewer; Max Planck Institute for Plant Breeding Research, Germany
  4. Miaoxiao Wang
    Reviewer; ETH Zurich, Switzerland
  5. Jacob D Palmer
    Reviewer; University of Oxford, United Kingdom

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Decision letter after peer review:

[Editors’ note: the authors submitted for reconsideration following the decision after peer review. What follows is the decision letter after the first round of review.]

Thank you for submitting the paper "Engineering ecologically complementary rhizosphere probiotics using consortia of specialized bacterial mutants" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by a Senior Editor. The following individual involved in the review of your submission has agreed to reveal their identity: Jacob D Palmer (Reviewer #2).

Comments to the Authors:

We are sorry to say that, after consultation with the reviewers, we have decided that this work will not be considered further for publication by eLife. Note that while we cannot publish the current study, we remain in principle interested in the work. If you are able to seriously revise the work and respond to the concerns, we would be prepared to review the work again, although we would treat it as a new submission.

As you will see, although the reviewers see the work as novel and with the potential to design probiotic consortia, they have also raised a number of issues that will require new experiments, new analyses and important reformatting of the manuscript. After internal discussions with the reviewers, we remain in principle interested in the work but we felt that it is important to give you more time if you wish to thoroughly address the concerns. Particularly, performing pairwise interactions among mutants, testing random sets of 8-member consortia, and reorganizing the manuscript appears to be critical to corroborate the claims and improve clarity, respectively. Furthermore, the reviewers think that the experiments were not designed to test ecological theories regarding ecosystem functioning and stability but rather as a method to design probiotic consortia in agriculture. Many concepts mentioned in the introduction and discussion might not be relevant here and might distract rather than inform.

Reviewer #1 (Recommendations for the authors):

Overall, the concept/methodology is interesting and the results open novel insights towards the engineering of microbial consortia that promote disease resistance. Furthermore, the statistical analyses seem to be conducted and represented adequately. The text is well written and methods are nicely detailed. Despite the novelty, the manuscript has some, yet important weaknesses that will need to be addressed.

Strengths:

Multi-species inoculants often fail to promote plant health under natural conditions because microbe-microbe interactions are not often considered when consortia are designed and assembled. Therefore, the idea of promoting intra-species multi-functionality through the assembly of bacterial mutants derived from a single strain is exciting. This is because it could promote functional diversity whilst at the same time minimizing the risk of competition between strains. The research question is therefore interesting and novel, the combination of methods, which include strain engineering (Tn5 mutant library), large-scale phenotypic screening and assembly of synthetic microbial consortia is adequate to test the relevance of functional complementary and multi-functionality between mutant strains.

The greenhouse experiment using a natural soil is also relevant to test the effects of the inoculants under semi-natural conditions.

The statistical analysis is adequate, the manuscript well written and the figures have been nicely crafted.

The conclusions are largely supported by experimental evidence, although it remains difficult to unambiguously conclude that the method actually works.

Weaknesses: There are some weaknesses.

First, the authors used an initial random pool of 479 mutants for in vitro trait phenotyping. It remains unclear why more mutants have not been tested given the fact that these are high-throughput in vitro screens. The authors did not use a saturated mutant library and might have overlooked many important genes.

Second, the structure of the manuscript is confusing. It remains unclear how the three clusters in Figure 1 have been defined and why they were not considered for the rational selection of the consortia shown in Figure 2. The most confusing part comes from the last paragraph and Figure 3 which shows data based on individual strains and not consortia. This is redundant with the data shown in Figure 1 and this part provides very limited information regarding the "underlying mechanisms between consortia diversity and improved performance". As it is presented this part distracts rather than informs. I do not think that this is the rights way to test the "insurance hypothesis". this hypothesis makes sense if tested in a community context with consortia of different complexities (as in Figure 2).

It remains also unclear whether complementarity between specialized bacterial mutants is really needed for improving plant performance (as stated in the title). The authors showed that the rational selection based on Figure 1 has some limits to predicting community behavior in vitro and in planta. Therefore, a key question is whether microbial consortia composed of random pools of 8 mutants (i.e. not pre-selected based on in vitro phenotypes) would provide similar fitness benefits as the mutant pool selected in Figure 2. In other words, is the first part of the paper really required or could one simply inoculate a pool of Tn5 mutants and observe the same protective effect? Testing the effects of several independent mutant consortia would be needed to draw general conclusions.

Finally, the paper is mechanistically weak. That being said, this is not the core of the manuscript and therefore it is not an issue. Although requesting complementation lines for the 8 mutants used for consortia establishment is beyond the scope of the manuscript, it is important to keep in mind that it is difficult to conclude anything regarding the bacteria genetic determinants identified here without proper functional validations.

To strengthen the conclusions shown in Figure 2, it would be important to test other independent consortia of 8 mutants and validate directionality in their effects depending on whether they were selected randomly or based on rational criteria. The concern is that part 1 and part 2 of the manuscripts might not be connected and it would be important to validate that the rational selection based on in vitro phenotypes is actually required and important.

The results presented in Figure 3 (individual strains) are insufficient to test the insurance effect. One would need to inspect this in the context of the various consortia shown in Figure 2. Here, it would be important to assess 1. total Bacillus load (e.g. using colony counts), 2. disease suppression and 3. "community structure" of the bacilli mutants. The latter can likely easily be conducted using PCRs with pairs of Tn5 transposon specific and random primers similar to classical TnSeq approaches. If the insurance hypothesis holds true, one would expect, that throughout the whole growth cycle, the complex community always performs better than the individual mutants (based on colony counts and disease suppressiveness) and also that relative abundances of Bacilli mutants change across the growth cycle. Such experiments could further reinforce the conclusion regarding insurance effects. Alternatively, authors should consider including this section as Figure 2 and tone-down their claim regarding these insurance effects that are not corroborated by experimental evidence.

Agar overlay data not shown?

Reviewer #2 (Recommendations for the authors):

This study provides a novel, potentially high-throughput method for identifying beneficial microbial consortia towards plant growth and health, which may be applicable towards other biotechnological applications. This study has the potential to be of great interest to all those working in host-microbe interactions. However, it is my opinion that there is an overemphasis on microbial ecology and the inherent value of intraspecies diversity towards ecosystem health, where I think the data of this study does not meet their intended aims.

Strengths: This study provides a novel and unique approach for studying the impact of intraspecies diversity on microbial consortia and understanding host (plant)-microbe interactions. There is a very good introductory section highlighting the field of microbial ecology. The methods used seem applicable to many different potential environments, as highlighted by the authors, and might be of great interest to those working on both applied and fundamental areas of microbiology and/or host-microbe interactions. The justification of the study, the introduction, and the methods used are all very high quality. The stated expectations by the authors are also well-supported by the theoretical literature and clearly explained. The focus on intraspecies interactions and diversity is also very welcomed and appears to me to be a novel approach.

Weaknesses: It is my opinion that the narrative in the results and discussion is overly focused on the impact of intraspecies diversity in microbial ecology, and how this diversity improves ecosystem functioning. I do not think the study provides support for these claims. Instead, the study does provide excellent support for high-performing consortia of mutants, which outperform the WT monoculture across varying metrics, which has both applied and fundamental value regarding agriculture and host-microbe interactions at both the strain and molecular levels. The data provided as apart of this work provides many new avenues of research and I think is an excellent step towards developing beneficial probiotic consortia. Unfortunately, these advances are sometimes overshadowed by claims of ecosystem stability and ecosystem functioning that lack support in the data. In multiple places in the text, it is suggested that intraspecies consortia are not subject to intra-species conflict. I find this to be potentially misleading, as it implies that these genotypes now have neutral (or positive) interactions, which is not appropriately measured nor expected from the data.

Line 44-47 – I am unsure of the support that antagonistic traits often increase invasion success of probiotic bacteria, or what this statement really means. I think it could benefit by being more explicit.

The introduction may benefit from a more thorough explanation of biodiversity-ecosystem functioning theory (line 84), to match the excellent introduction of microbial ecology from lines 50-75.

Line 95: 'benefits' to 'beneficial

Line 96: Expected advances. It does not appear that expectations 2 or 3 are tested. Expectation 3 is also rather unclear. Please consider rewriting this section to better set the reader's expectations for the data and results in the paper.

Line 162 – Please clarify how it is known that OD600 = 0.05 equals ~10^5 cells/mL for B. amyloliquifaciens. Or omit the ~10^5 cells/mL. Additionally, I wonder if OD600 is an appropriate proxy for biomass production in a B. amyloliquifaciens mutant library? Plating and counting CFU/mL would be the much more reliable measurement. Considering the method used for biomass production measurement, did you perform these measurements in a plate reader and have additional time series data beyond the 24hr timepoint? Perhaps maximum growth rate or other factors of the growth curve may also be good indicators of beneficial phenotypes for plant health.

General Methods Comments: Given the more applied aims of this study, maintaining a constant inoculum size and measuring the outputs of the consortium is a very reasonable strategy. However, performing experiments in this way ignores the intraspecies competition that likely defines the system. This makes claims of facilitation, exploitation, or cooperation moot, as they are not being appropriately measured. The authors would need to measure monoculture densities of individual mutants, then add mutants at the same inoculum into the consortia experiments (now doubling the total bacterial inoculum of the monoculture experiments), and then measure individual strain abundance and/or total biomass to accurately test for facilitation or cooperation (the consortia biomass must exceed the sum of the 2 monoculture biomasses, and then it still might be either exploitation, facilitation, or cooperation, based on the strength of the interactions). I am still very impressed and intrigued by this study and the data obtained, but I do think it needs to focus more on the applied aspects: High performing mutant library consortia to improve agricultural yields. The experiments performed, however, don't tell us much about microbial ecology.

I am quite confused by the clusters. I don't understand the methods of this PCA plot and how mutants are assigned into the various clusters. Perhaps isolates that did not fall within the bounds, or intersected multiple groups should be excluded. The K-means clustering needs significantly more explanation in the methods. As a key aspect of the data being provided, the written methods section should make it very clear how this clustering is being performed and why it is a valid method.

Is there precedent for this type of phenotypic clustering for a transposon library? I think there is tremendous value of having a transposon library of this sort. But I don't really understand the purpose of the clustering.

Line 307-315 – Data are provided for the clusters and compared to the WT. However, statistical tests included in Figures 1E, F measure differences between clusters, rather than each cluster compared to the WT. I see that the WT is provided as the solid black line, but I think adding an additional box plot for the WT would be more appropriate.

Line 321 – I worry that a term like 'specialist mutants' could cause confusion, and perhaps imply that these mutants have niche separation relative to the WT, which hasn't been measured. I think simply 'mutants' would be more appropriate.

Line 332 – It is unclear to me how the predictive performance values were calculated. Please be explicit. When you sum the trait values of monocultures, are you first dividing that trait value according to the reduction in biomass added to the inoculum. ie when you do a 2 species consortium and add each species at 50% of their inoculum as compared to the WT, are you halving the expected trait value? Is there an existing precedent for this?

Line 363 – Stability has a specific definition in ecology that I don't think is intended here. Consider not using this term.

Line 366 – I don't understand the difference between high swarming and swarming motility. These are not differentiated in the methods.

Line 356 – 383 – I'm having a hard time following the data and logic provided here.

Discussion:

Line 387 – Did your results show this? Which data/Figure? I think it would be appropriate to remind the reader where to find the evidence for this claim.

Line 392 – The intra-species diversity also introduces conflict. There is likely more inter-genotype competition in the communities assembled for this study as compared to one involved multi-species communities. The competition/conflict just hasn't been measured here.

The discussion seems overly focused on the individual mutations identified and their resulting phenotypes. Perhaps this level of depth provided for these mutants is more suited for the Results section.

Line 462 – "It is thus possible that consortia were together able to overcome the trait trade-offs experienced at the individual mutant level, leading to improved and more stable ecosystem functioning in time." This language is too vague and has the potential to mislead readers. What is meant by improved ecosystem functioning? Regarding stability, this study does not measure the ecological interactions between different bacterial genotypes. The levels of intraspecific competition between microbes is likely very high, which may lead to increased stability. It may also lead to extinction of the weaker strain.

Line 468 – Is a transposon library really an example of synthetic biology? This is not how I understand the limits of the field, but I apologise if I'm uninformed.

Line 469 – I do not agree that this study highlights the importance of intraspecies diversity in ecosystem functioning. Instead, this study does highlight how intraspecies diversity libraries can be an effective mechanism for identifying traits and consortia which can rapidly improve desired measurables within a defined ecosystem. I think an overall refocusing on the applied potential of the study would be appropriate.Reviewer #3 (Recommendations for the authors):

Yang et al. assembled a set of synthetic consortia using functionally specialized mutants generated from a rhizosphere probiotic, Bacillus amyloliquefaciens, and tested whether the consortia can promote the growth of the plant. Because the mutants are derived from the same strain, they should not antagonize their kins. Accordingly, this design potentially reduces within-community antagonism, overcoming a limitation occurring in the previous multi-species consortia.

Although the idea of this paper is of interest to readers in the field of microbiome and plant-microbe interactions, the manuscript is not well-organized. The Introduction section did not adequately discuss the strengths of this design compared to the previous studies. Instead, the author proposed several scientific questions that were not addressed/focused on by this study. In the Results section, several important data are missing, and some claims are not well-supported by the data. The specific comments are listed below.

Introduction:

The author claims that "inoculant design should aim to minimize negative interactions within consortia" (line 46) and the author claimed that three ways (concepts) can be used to achieve this aim: ecological complementarity, division of labor or facilitation between consortia members.

Firstly, the discussion on these concepts (lines 50-75) makes me feel that the author thinks they are three totally segregated concepts (For example, in lines 65-66: "In addition to occupying different niches or performing specialized tasks"). However, there are many overlaps among these three concepts. For example, if different species exploit different niches and such diversification benefits all the individuals, it can be also defined as a "division of labor" (based on definitions in ref. 26). Moreover, many cases of facilitation can be defined as the division of labor (see this ref: https://doi.org/10.1016/j.jmb.2019.06.023). I would suggest that the author use more specific/distinguishable terms throughout the paper, for example, niche complementarity/ specialization within a single niche/ leaky section. In addition, the direct description of how such engineering works (e.g., engineering species to occupy different niches) may be clearer to readers than just giving terms.

Secondly, the authors discussed theories but did not state how the design in this study can benefit from these theories to overcome the proposed limitation. This makes it unclear why constructing such mutant-based synthetic consortia can "minimize negative interactions within consortia". The author proposed a main scientific question "whether this should be based on ecological complementarity, division of labor or facilitation between consortia members. (Lines 74-75)" but this study did not address this question.

Thirdly, following the above point, a logical connection is missing between the last two paragraphs of the Introduction. In other words, does increasing intra-species diversity benefit the ecological complementarity, division of labor, or facilitation, so reduce the within-community antagonism? In addition, I would expect the authors to explain more about why increasing the intra-species diversity of a single bacterium could prevent conflicts between the consortia members. I think this is the main novelty of this study different from the previous design constructing consortia using different species. I saw some reasons were listed in lines 101-103 but expect more explanations/discussions.

In sum, the author could do a better job linking theory and experimental design.

Results

Line 320-322: I think this "antagonism" experiment is very important for the novelty of this study. Because the mutants are derived from the same strain, they should show less competition/antagonism within consortia, which is different from the situation in the previous multi-species consortia. Therefore, (1) the data of the "agar overlay assays" should be provided in the main text of the paper; (2) It will be beneficial to add more evidence (e.g., liquid co-culture); (3) It is also very important to examine if the mutants within a consortium stable coexist during the community assembly. Accordingly, the structure of the consortium during/after the assays (swarming, biofilm formation, root colonization) should be measured.

Design/analyses of the consortia experiment: in lines 324-326, the authors hypothesized that "consortia could show improved performance due to ecological complementarity where different mutants 'specialize' respective to different traits, overcoming trade-offs experienced at individual strain level (Figure 1B)." However, the experiments were concluded with "mutant consortia diversity was positively linked with consortia performance in vivo, which was associated with consortia mean performance and pathogen suppression measured in vitro. (Lines 354-355)" Obviously, the hypothesis was not well examined using the current design and analyses. For a direct test of the hypothesis, the authors should compare the performance of the consortia with niche specialization (e.g., consortium no.29) with that of the others.

Line 383: I think the conclusion here is less evident. To prove the improvement of the consortia functioning is due to the insurance effects, the performance of the 47-member consortium should be compared to the consortium with fewer members. A more rational design is to build consortia according to the four testing functions. Some consortia contain one mutant specializing in one specific function, while the others contain more mutants for the function (that is, add redundancy). Then compare the root colonization and plant protection ability of the two groups.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for submitting your article "Engineering multifunctional rhizosphere probiotics using consortia of Bacillus amyloliquefaciens transposon insertion mutants" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by myself and Detlef Weigel as the Senior Editor.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

We had substantial internal discussion of the revised manuscript, and we agree that the revisions have been substantial. All referees would – in principle – support publication of the work in eLife.

However, we felt that there is still one important issue that must be carefully considered and addressed. This is regarding the new supernatant experiment, which might not be the best way to infer ecological interactions (please see comments from ref#3) and might lead to misleading interpretations. We agreed that this issue should be clarified before final acceptance of the work. We see two potential options to do so:

1) If you wish to firmly conclude that reduced competition among consortia members exists, please include single strain vs. pairwise co-culture experiments followed by colony counts (see the suggestions below for how to overcome the difficulty that the strains are nearly identical). This might actually require a limited amount of work and would be most appropriate to infer potentially neutral interactions among consortia members.

2) Alternatively, you should remove the supernatant data from the manuscript and adopt a more tempered tone for all assertions linked to how "this study" enhances our understanding of function by mitigating competition between the consortia members.

Reviewer #1 (Recommendations for the authors):

The work is significant because it reports that intra-species phenotypic diversity in bacteria could be an effective way to improve probiotic consortia in agriculture. Authors now provide compelling evidence indicating that near-clonal bacteria assembled into consortia based on diverse functional traits perform better than randomly-assembled consortia. Two solid conclusions emerging from this work are that (1) Communities of near-clonal bacteria can be optimized based on functional traits to promote functional diversity and reduce bacterial competition (2) consortium multifunctionality – but not richness – is key to promoting bacterial colonization in roots and to protect against a plant pathogen.

In my opinion, authors have seriously and thoroughly addressed the previous concerns raised by me and other reviewers. I acknowledge the fact that they have performed additional experiments that have strengthened their conclusions. Particularly, they performed supernatant experiments testing pairwise interactions between the mutants included in their phenotypically-optimized 8-member consortium and performed validation experiments with randomly assembled 8-member consortia. These two novel experiments revealed that:

1. Limited negative or positive interactions were observed in supernatant exposure experiments, indicating that bacterial-bacterial interactions are not extensively contributing to the observed phenotypes.

2. Importantly, they used random, 8-member mutant consortia, and demonstrate that the phenotypically-optimized consortium indeed performs better than what is expected by chance. This experiment convinced me that the strategy used in the manuscript has relevance for optimizing synthetic communities based on (multi)-functional traits.

They have also reorganized the structure of the manuscript and rewrote the introduction and Discussion sections to focus more on the methodological novelty of the work and less on ecological theories.Reviewer #2 (Recommendations for the authors):

I only have a small suggestion optional to the authors. It might be intriguing to discuss the microscale-level mechanisms behind why the consortium outperforms a single strain. Could the four plant growth-promotion traits act as public goods? For instance, could biofilm formation specialists aid pathogen suppression specialists in better colonizing plants? It may be interesting to add some discussions to the paragraph in lines 497-516.

Reviewer #3 (Recommendations for the authors):

Collectively, I am excited to see this manuscript again, as I've been keeping an eye out to see if it had been published in another journal since my first review. It is an intriguing work and I think that many readers in the community will be interested to read it. My primary criticism remains the same as the first review, where ecological interactions are not appropriately measured. I give a detailed explanation of this in Response to R13 (below), along with references that I believe do perform appropriate experiments to measure ecological interactions between strains/genotypes. That said, my other comments appear to be appropriately addressed.

Additionally, because much of the narrative of the discussion has been rewritten to avoid direct claims about ecological interactions between the genotypes throughout the text, I no longer believe that an experiment demonstrating ecological interactions is necessary. Figure S5 should be removed, as this is not an appropriate measure of ecological interactions. The data gathered as part of Figure S5 are likely sufficient to still represent these experiments appropriately. If presented appropriately, I strongly suspect that negative interactions will be the predominant interaction type, so it would be up to the authors if they would like to present this data or not.

So long as reference to ecological interactions (whether neutral or otherwise) and Figure S5 are removed from the text (or ecological interactions are measured according to a method similar to those detailed below), the remainder of this work appears to hold together well, with the text supported by the data.

Response to R13:

Regarding Figure S5. While supernatant assays are not the preferred method to measure ecological interactions, I appreciate the authors performing an experiment to address this major comment from my first review. However, this is not the appropriate experiment and it does not support the claim that interactions between the different genotypes are neutral.

The ideal experiment would be to grow each strain in monoculture and count cfu/mL, then grow each strain in pairwise combination and count cfu/mL of each strain. If it is not possible to selectively plate in order to identify the different genotypes, one could also grow each strain in monoculture and count cfu/mL. Then again repeat growing strains in pairwise co-culture, and count the total cfu/mL. If the interactions are neutral, the prediction is that the total cfu/mL will be the sum of the two strains monoculture cell densities (cfu/mL). One could interpret this as clear niche differentiation, with no interference or exploitative competition. If you are committed to the growth supernatant assay and measuring OD600, you could grow each strain in monoculture, measure final OD600 and then filter the supernatant just as you've done. You can then grow them again in the spend media and measure the final OD600. If the interactions are neutral, then prediction is that growth in monoculture with spent media = growth in monoculture with fresh media. For any strain combination, if growth in monoculture of spent media < growth in monoculture of fresh media, this is a negative interaction.

Unfortunately, it is not satisfactory to infer ecological interactions as a measure of growth in supernatant relative to growth in one's own supernatant. There is full niche overlap when a strain grows in its own supernatant. Using this metric, you could have two different strains with identical nutrient requirements, that will compete with each other for these nutrients, and you would likely measure it as a neutral interaction. It is also challenging when measuring ecological interactions in liquid LB and trying to extrapolate this to interactions in vivo where the host is also present. An experiment like this in liquid LB is probably a good proxy for gauging interactions, but an experiment that is close to the environment of interest would be the most desirable. However, I understand that this can present additional challenges. I also understand that a specific aim of this study is to use in vitro phenotyping in order to optimize consortia functioning. So in that regard, an LB experiment is perhaps appropriate.

I hope that this criticism is clear. For a more comprehensive description of methods in this regard, I suggest Foster and Bell, 2012. DOI: 10.1016/j.cub.2012.08.005. Additionally, Weiss et al. ISME 2022 DOI: 10.1038/s41396-021-01153-z performs nice experiments in this regard, using both monoculture vs coculture experiments (quantifying strain abundances with qPCR) as well as spent supernatant assays.

And yet, after all of this, after reading the revised discussion, I no longer think that this experiment is essential. Because the authors have removed much of the narrative regarding ecological significance and focused more on the applied aspects of this work, I do not think that these ecologically-focussed experiments are completely necessary. So long as the authors avoid claims of the interactions between genotypes (as I would argue these have still not been appropriately measured), then I don't think this experiment is necessary for this manuscript.

https://doi.org/10.7554/eLife.90726.sa1

Author response

[Editors’ note: the authors resubmitted a revised version of the paper for consideration. What follows is the authors’ response to the first round of review.]

Comments to the Authors:

We are sorry to say that, after consultation with the reviewers, we have decided that this work will not be considered further for publication by eLife. Note that while we cannot publish the current study, we remain in principle interested in the work. If you are able to seriously revise the work and respond to the concerns, we would be prepared to review the work again, although we would treat it as a new submission.

As you will see, although the reviewers see the work as novel and with the potential to design probiotic consortia, they have also raised a number of issues that will require new experiments, new analyses and important reformatting of the manuscript. After internal discussions with the reviewers, we remain in principle interested in the work but we felt that it is important to give you more time if you wish to thoroughly address the concerns. Particularly, performing pairwise interactions among mutants, testing random sets of 8-member consortia, and reorganizing the manuscript appears to be critical to corroborate the claims and improve clarity, respectively. Furthermore, the reviewers think that the experiments were not designed to test ecological theories regarding ecosystem functioning and stability but rather as a method to design probiotic consortia in agriculture. Many concepts mentioned in the introduction and discussion might not be relevant here and might distract rather than inform.

We thank the Editor and three anonymous reviewers for highly useful feedback on our manuscript. We have now carefully followed all the suggestions and fully revised our manuscript accordingly. The main changes include:

  • Additional supernatant experiments testing pairwise interactions between the mutants included in ‘phenotypically optimized’, 8-member consortium

  • Additional validation experiments with randomly assembled 8-member consortia

  • Reorganization of the structure of the manuscript following reviewers’ suggestions

  • Rewriting of the introduction and discussion and clarification of the aims of the study

We hope that these changes have improved our manuscript and that it could now be reconsidered as a new submission to eLife.

Reviewer #1 (Recommendations for the authors):

Overall, the concept/methodology is interesting and the results open novel insights towards the engineering of microbial consortia that promote disease resistance. Furthermore, the statistical analyses seem to be conducted and represented adequately. The text is well written and methods are nicely detailed. Despite the novelty, the manuscript has some, yet important weaknesses that will need to be addressed.

Strengths:

Multi-species inoculants often fail to promote plant health under natural conditions because microbe-microbe interactions are not often considered when consortia are designed and assembled. Therefore, the idea of promoting intra-species multi-functionality through the assembly of bacterial mutants derived from a single strain is exciting. This is because it could promote functional diversity whilst at the same time minimizing the risk of competition between strains. The research question is therefore interesting and novel, the combination of methods, which include strain engineering (Tn5 mutant library), large-scale phenotypic screening and assembly of synthetic microbial consortia is adequate to test the relevance of functional complementary and multi-functionality between mutant strains.

The greenhouse experiment using a natural soil is also relevant to test the effects of the inoculants under semi-natural conditions.

The statistical analysis is adequate, the manuscript well written and the figures have been nicely crafted.

The conclusions are largely supported by experimental evidence, although it remains difficult to unambiguously conclude that the method actually works.

We thank reviewer #1 for positive feedback.

Weaknesses: There are some weaknesses.

1. First, the authors used an initial random pool of 479 mutants for in vitro trait phenotyping. It remains unclear why more mutants have not been tested given the fact that these are high-throughput in vitro screens. The authors did not use a saturated mutant library and might have overlooked many important genes.

We have now included more in vitro phenotyping data to the manuscript. The full dataset now includes phenotypic data for 1999 mutants. Of these, we selected 479 phenotypically representative mutants for more detailed analysis and probiotic bioinoculant design. While we likely missed certain mutants with this method, our analysis similarity suggests that this method captured a representative and phenotypically diverse sample of mutants from the original collection. We have now justified this in the manuscript on lines 149-153.

2. Second, the structure of the manuscript is confusing. It remains unclear how the three clusters in Figure 1 have been defined and why they were not considered for the rational selection of the consortia shown in Figure 2. The most confusing part comes from the last paragraph and Figure 3 which shows data based on individual strains and not consortia. This is redundant with the data shown in Figure 1 and this part provides very limited information regarding the "underlying mechanisms between consortia diversity and improved performance". As it is presented this part distracts rather than informs. I do not think that this is the rights way to test the "insurance hypothesis". this hypothesis makes sense if tested in a community context with consortia of different complexities (as in Figure 2).

We have now restructured and clarified these sections. Specifically, we now present all individual strain-based results before the pairwise and consortia data (including new additional data on the effects of randomly assembled 8-mutant consortia). Moreover, we have made the selection of subset of strains from the larger mutant collection clearer (the link between Figures 1 and 2; see also our response to point 1).

3. It remains also unclear whether complementarity between specialized bacterial mutants is really needed for improving plant performance (as stated in the title). The authors showed that the rational selection based on Figure 1 has some limits to predicting community behavior in vitro and in planta. Therefore, a key question is whether microbial consortia composed of random pools of 8 mutants (i.e. not pre-selected based on in vitro phenotypes) would provide similar fitness benefits as the mutant pool selected in Figure 2. In other words, is the first part of the paper really required or could one simply inoculate a pool of Tn5 mutants and observe the same protective effect? Testing the effects of several independent mutant consortia would be needed to draw general conclusions.

This is a very important point. As a result, we have conducted additional experiments where we test the plant protection by randomly assembled 8-mutant consortia that vary in their phenotypic dissimilarity and predicted relative performance. These results confirm that randomly constructed mutant consortia perform worse compared to phenotypically ‘optimized’ consortium, suggesting that improvement was unlikely driven by consortium richness per se.

4. Finally, the paper is mechanistically weak. That being said, this is not the core of the manuscript and therefore it is not an issue. Although requesting complementation lines for the 8 mutants used for consortia establishment is beyond the scope of the manuscript, it is important to keep in mind that it is difficult to conclude anything regarding the bacteria genetic determinants identified here without proper functional validations.

We clearly admit these limitations in the discussion and suggest future work to uncover mechanistic aspects of our approach at the genetic level.

5. To strengthen the conclusions shown in Figure 2, it would be important to test other independent consortia of 8 mutants and validate directionality in their effects depending on whether they were selected randomly or based on rational criteria. The concern is that part 1 and part 2 of the manuscripts might not be connected and it would be important to validate that the rational selection based on in vitro phenotypes is actually required and important.

New data has now been included to strengthen these conclusions; please, also see our response to point 3.

6. The results presented in Figure 3 (individual strains) are insufficient to test the insurance effect. One would need to inspect this in the context of the various consortia shown in Figure 2. Here, it would be important to assess 1. total Bacillus load (e.g. using colony counts), 2. disease suppression and 3. "community structure" of the bacilli mutants. The latter can likely easily be conducted using PCRs with pairs of Tn5 transposon specific and random primers similar to classical TnSeq approaches. If the insurance hypothesis holds true, one would expect, that throughout the whole growth cycle, the complex community always performs better than the individual mutants (based on colony counts and disease suppressiveness) and also that relative abundances of Bacilli mutants change across the growth cycle. Such experiments could further reinforce the conclusion regarding insurance effects. Alternatively, authors should consider including this section as Figure 2 and tone-down their claim regarding these insurance effects that are not corroborated by experimental evidence.

We agree with the reviewer that more detailed measurements and quantification of each mutant frequency would be required to support our conclusions on the insurance hypothesis. As the Tn insertions did not include unique barcodes, quantifying mutant frequencies was not possible (now mentioned in the discussion). We have hence toned down our conclusions on the role of insurance hypothesis and have removed it from the introduction as our hypothesis.

7. Agar overlay data not shown?

Agar overlay data is now included. We have also added new data on the lack of pairwise inhibitory effects between the strains based on supernatant growth assays (see revised methods on lines 375-378).

Reviewer #2 (Recommendations for the authors):

This study provides a novel, potentially high-throughput method for identifying beneficial microbial consortia towards plant growth and health, which may be applicable towards other biotechnological applications. This study has the potential to be of great interest to all those working in host-microbe interactions. However, it is my opinion that there is an overemphasis on microbial ecology and the inherent value of intraspecies diversity towards ecosystem health, where I think the data of this study does not meet their intended aims.

Strengths: This study provides a novel and unique approach for studying the impact of intraspecies diversity on microbial consortia and understanding host (plant)-microbe interactions. There is a very good introductory section highlighting the field of microbial ecology. The methods used seem applicable to many different potential environments, as highlighted by the authors, and might be of great interest to those working on both applied and fundamental areas of microbiology and/or host-microbe interactions. The justification of the study, the introduction, and the methods used are all very high quality. The stated expectations by the authors are also well-supported by the theoretical literature and clearly explained. The focus on intraspecies interactions and diversity is also very welcomed and appears to me to be a novel approach.

We thank reviewer #2 for positive comments.

Weaknesses: It is my opinion that the narrative in the results and discussion is overly focused on the impact of intraspecies diversity in microbial ecology, and how this diversity improves ecosystem functioning. I do not think the study provides support for these claims. Instead, the study does provide excellent support for high-performing consortia of mutants, which outperform the WT monoculture across varying metrics, which has both applied and fundamental value regarding agriculture and host-microbe interactions at both the strain and molecular levels. The data provided as apart of this work provides many new avenues of research and I think is an excellent step towards developing beneficial probiotic consortia. Unfortunately, these advances are sometimes overshadowed by claims of ecosystem stability and ecosystem functioning that lack support in the data. In multiple places in the text, it is suggested that intraspecies consortia are not subject to intra-species conflict. I find this to be potentially misleading, as it implies that these genotypes now have neutral (or positive) interactions, which is not appropriately measured nor expected from the data.

We have now toned down our conclusions on the effects of intra-species diversity for the ecosystem functioning and stability. To provide more support for the lack of inhibitory effects between mutants, we have included new data based on supernatant growth assays, which demonstrates that most of the pairwise interactions were neutral or mildly positive or negative (see updated Figure S5).

Line 44-47 – I am unsure of the support that antagonistic traits often increase invasion success of probiotic bacteria, or what this statement really means. I think it could benefit by being more explicit.

We have now replaced ‘invasiveness’ with ‘competitiveness’ and have expanded the example to be more specific.

The introduction may benefit from a more thorough explanation of biodiversity-ecosystem functioning theory (line 84), to match the excellent introduction of microbial ecology from lines 50-75.

We have now clarified the biodiversity-ecosystem functioning approaches earlier during the introduction to make the key idea clearer (on lines 55-83).

Line 95: 'benefits' to 'beneficial

Revised as suggested.

Line 96: Expected advances. It does not appear that expectations 2 or 3 are tested. Expectation 3 is also rather unclear. Please consider rewriting this section to better set the reader's expectations for the data and results in the paper.

This section has now been simplified and completely rewritten.

Line 162 – Please clarify how it is known that OD600 = 0.05 equals ~10^5 cells/mL for B. amyloliquifaciens. Or omit the ~10^5 cells/mL. Additionally, I wonder if OD600 is an appropriate proxy for biomass production in a B. amyloliquifaciens mutant library? Plating and counting CFU/mL would be the much more reliable measurement. Considering the method used for biomass production measurement, did you perform these measurements in a plate reader and have additional time series data beyond the 24hr timepoint? Perhaps maximum growth rate or other factors of the growth curve may also be good indicators of beneficial phenotypes for plant health.

The relationship between the OD and cell numbers is based on a calibration curves, which we now present for the wild-type, and low and high biomass producing mutants in Figure S1.

General Methods Comments: Given the more applied aims of this study, maintaining a constant inoculum size and measuring the outputs of the consortium is a very reasonable strategy. However, performing experiments in this way ignores the intraspecies competition that likely defines the system. This makes claims of facilitation, exploitation, or cooperation moot, as they are not being appropriately measured. The authors would need to measure monoculture densities of individual mutants, then add mutants at the same inoculum into the consortia experiments (now doubling the total bacterial inoculum of the monoculture experiments), and then measure individual strain abundance and/or total biomass to accurately test for facilitation or cooperation (the consortia biomass must exceed the sum of the 2 monoculture biomasses, and then it still might be either exploitation, facilitation, or cooperation, based on the strength of the interactions). I am still very impressed and intrigued by this study and the data obtained, but I do think it needs to focus more on the applied aspects: High performing mutant library consortia to improve agricultural yields. The experiments performed, however, don't tell us much about microbial ecology.

As we were interested in separating richness and species identity effects in mutant consortia, we employed substitutive experimental design, which assumes that the biomass of the consortia is kept constant (increasing individual strain density in consortia would make it impossible to disentangle diversity versus abundance effects). We now acknowledge in the discussion that our method also affected the mutant abundances relative to monoculture controls, which could have affected the observed patterns. We have now included additional data for analyzing interactions between a subset of mutants used in the diversity-ecosystem functioning experiment (see response to point 7). The applied aspect of the study has now been emphasized.

I am quite confused by the clusters. I don't understand the methods of this PCA plot and how mutants are assigned into the various clusters. Perhaps isolates that did not fall within the bounds, or intersected multiple groups should be excluded. The K-means clustering needs significantly more explanation in the methods. As a key aspect of the data being provided, the written methods section should make it very clear how this clustering is being performed and why it is a valid method.

Is there precedent for this type of phenotypic clustering for a transposon library? I think there is tremendous value of having a transposon library of this sort. But I don't really understand the purpose of the clustering.

We have now clarified the K-means clustering in the methods (lines 186-189). Briefly, K-means clustering assigns n observations into k clusters where each observation (in our case mutant) belongs to a cluster with the nearest mean (cluster centers or cluster centroid). PCA analysis was then used to visualize the phenotypic clustering where the bounds of the clusters indicate 95% confidence intervals. Each mutant was hence assigned to one of the clusters even though mutants varied in their distance to cluster centroid means. This method is widely used in ecological studies with microbes and other taxa.

Line 307-315 – Data are provided for the clusters and compared to the WT. However, statistical tests included in Figures 1E, F measure differences between clusters, rather than each cluster compared to the WT. I see that the WT is provided as the solid black line, but I think adding an additional box plot for the WT would be more appropriate.

We have now clarified in panels Figure 1E-F and results text that WT was assigned in the cluster 1. As the WT is only one clone, variation between technical replicates is very small. We hence feel that presenting WT performance as solid black line offers better visualization of the WT baseline. We have now included shaded area around WT mean to show variation between technical replicates.

Line 321 – I worry that a term like 'specialist mutants' could cause confusion, and perhaps imply that these mutants have niche separation relative to the WT, which hasn't been measured. I think simply 'mutants' would be more appropriate.

We have now simplified the text as suggested and only use the term ‘Specialist mutant’ when it is related to ecological specialism or generalism. When emphasizing phenotypic differences between mutants, we have now phrased them as ‘phenotypically distinct’ or ‘phenotypically dissimilar’ mutants depending on the context.

Line 332 – It is unclear to me how the predictive performance values were calculated. Please be explicit. When you sum the trait values of monocultures, are you first dividing that trait value according to the reduction in biomass added to the inoculum. ie when you do a 2 species consortium and add each species at 50% of their inoculum as compared to the WT, are you halving the expected trait value? Is there an existing precedent for this?

This has now been clarified in the methods; predicted value equals to the sum of trait values of each member divided by the richness of given consortium based on additive model. As a result, we are not halving the trait values as the chosen growth-based traits. These potential inoculum abundance effects are now briefly mentioned in the discussion.

Line 363 – Stability has a specific definition in ecology that I don't think is intended here. Consider not using this term.

We have removed the references to stability here.

Line 366 – I don't understand the difference between high swarming and swarming motility. These are not differentiated in the methods.

Revised to remove ambiguity – with “high” we just referred to relatively high swarming motility trait values.

Line 356 – 383 – I'm having a hard time following the data and logic provided here.

This section has now been rewritten to improve clarity.

Discussion:

Line 387 – Did your results show this? Which data/Figure? I think it would be appropriate to remind the reader where to find the evidence for this claim.

Reference to specific results now included as suggested.

Line 392 – The intra-species diversity also introduces conflict. There is likely more inter-genotype competition in the communities assembled for this study as compared to one involved multi-species communities. The competition/conflict just hasn't been measured here.

The discussion seems overly focused on the individual mutations identified and their resulting phenotypes. Perhaps this level of depth provided for these mutants is more suited for the Results section.

We now discuss potential conflicts of intra-species diversity in the broader context of bioinoculant design beyond this experiment. We feel that discussing individual mutations is still appropriate to provide some potential mechanistic explanations for our results and have kept them in the discussion.

Line 462 – "It is thus possible that consortia were together able to overcome the trait trade-offs experienced at the individual mutant level, leading to improved and more stable ecosystem functioning in time." This language is too vague and has the potential to mislead readers. What is meant by improved ecosystem functioning? Regarding stability, this study does not measure the ecological interactions between different bacterial genotypes. The levels of intraspecific competition between microbes is likely very high, which may lead to increased stability. It may also lead to extinction of the weaker strain.

Revised to add clarity; with “ecosystem functioning” we refer to “plant protection” in this occasion. With stability we here refer to stability of plant protection – from the microbial point of view this could be mediated by lack or intense competition, which we have now clarified. We have now revised this section to reduce ambiguity of the terms used.

Line 468 – Is a transposon library really an example of synthetic biology? This is not how I understand the limits of the field, but I apologise if I'm uninformed.

Reference to ‘synthetic biology’ removed.

Line 469 – I do not agree that this study highlights the importance of intraspecies diversity in ecosystem functioning. Instead, this study does highlight how intraspecies diversity libraries can be an effective mechanism for identifying traits and consortia which can rapidly improve desired measurables within a defined ecosystem. I think an overall refocusing on the applied potential of the study would be appropriate.

Revised as suggested; we now emphasize our findings in more applied context as suggested and have toned down the significance of intra-species diversity.

Reviewer #3 (Recommendations for the authors):

Yang et al. assembled a set of synthetic consortia using functionally specialized mutants generated from a rhizosphere probiotic, Bacillus amyloliquefaciens, and tested whether the consortia can promote the growth of the plant. Because the mutants are derived from the same strain, they should not antagonize their kins. Accordingly, this design potentially reduces within-community antagonism, overcoming a limitation occurring in the previous multi-species consortia.

Although the idea of this paper is of interest to readers in the field of microbiome and plant-microbe interactions, the manuscript is not well-organized. The Introduction section did not adequately discuss the strengths of this design compared to the previous studies. Instead, the author proposed several scientific questions that were not addressed/focused on by this study. In the Results section, several important data are missing, and some claims are not well-supported by the data. The specific comments are listed below.

We thank Reviewer #3 for highly useful comments.

Introduction:

The author claims that "inoculant design should aim to minimize negative interactions within consortia" (line 46) and the author claimed that three ways (concepts) can be used to achieve this aim: ecological complementarity, division of labor or facilitation between consortia members.

Firstly, the discussion on these concepts (lines 50-75) makes me feel that the author thinks they are three totally segregated concepts (For example, in lines 65-66: "In addition to occupying different niches or performing specialized tasks"). However, there are many overlaps among these three concepts. For example, if different species exploit different niches and such diversification benefits all the individuals, it can be also defined as a "division of labor" (based on definitions in ref. 26). Moreover, many cases of facilitation can be defined as the division of labor (see this ref: https://doi.org/10.1016/j.jmb.2019.06.023). I would suggest that the author use more specific/distinguishable terms throughout the paper, for example, niche complementarity/ specialization within a single niche/ leaky section. In addition, the direct description of how such engineering works (e.g., engineering species to occupy different niches) may be clearer to readers than just giving terms.

These are very good points, and we have now made it clearer in the introduction that these processes are not mutually exclusive. We have also clarified the terminology and rewritten the whole introduction and reduced the emphasis on division of labor and complementarity, which we did not directly quantify in our experiments.

Secondly, the authors discussed theories but did not state how the design in this study can benefit from these theories to overcome the proposed limitation. This makes it unclear why constructing such mutant-based synthetic consortia can "minimize negative interactions within consortia". The author proposed a main scientific question "whether this should be based on ecological complementarity, division of labor or facilitation between consortia members. (Lines 74-75)" but this study did not address this question.

We have now made our research questions clearer at the end of the introduction.

Thirdly, following the above point, a logical connection is missing between the last two paragraphs of the Introduction. In other words, does increasing intra-species diversity benefit the ecological complementarity, division of labor, or facilitation, so reduce the within-community antagonism? In addition, I would expect the authors to explain more about why increasing the intra-species diversity of a single bacterium could prevent conflicts between the consortia members. I think this is the main novelty of this study different from the previous design constructing consortia using different species. I saw some reasons were listed in lines 101-103 but expect more explanations/discussions.

We have now rewritten this section of the introduction and give a better justification why we would expect more closely related communities to be less competitive based on kin selection theory (with appropriate references).

In sum, the author could do a better job linking theory and experimental design.

This is a fair point, and we hope that our revised version makes this link more obvious.

Results

Line 320-322: I think this "antagonism" experiment is very important for the novelty of this study. Because the mutants are derived from the same strain, they should show less competition/antagonism within consortia, which is different from the situation in the previous multi-species consortia. Therefore, (1) the data of the "agar overlay assays" should be provided in the main text of the paper; (2) It will be beneficial to add more evidence (e.g., liquid co-culture); (3) It is also very important to examine if the mutants within a consortium stable coexist during the community assembly. Accordingly, the structure of the consortium during/after the assays (swarming, biofilm formation, root colonization) should be measured.

More data has now been included (agar overlay and liquid supernatant assays; see response to point 7). Unfortunately, we were not able to quantify changes in mutant frequencies during the experiments as transposon insertions did not include barcodes. We have now made this limitation clear in the discussion.

Design/analyses of the consortia experiment: in lines 324-326, the authors hypothesized that "consortia could show improved performance due to ecological complementarity where different mutants 'specialize' respective to different traits, overcoming trade-offs experienced at individual strain level (Figure 1B)." However, the experiments were concluded with "mutant consortia diversity was positively linked with consortia performance in vivo, which was associated with consortia mean performance and pathogen suppression measured in vitro. (Lines 354-355)" Obviously, the hypothesis was not well examined using the current design and analyses. For a direct test of the hypothesis, the authors should compare the performance of the consortia with niche specialization (e.g., consortium no.29) with that of the others.

Please, see our response to point 3. Briefly, we have now conducted additional experiments where we test the plant protection by randomly assembled 8-mutant consortia. These results confirm that randomly constructed mutant consortia perform worse compared to phenotypically ‘optimized’ consortium.

Line 383: I think the conclusion here is less evident. To prove the improvement of the consortia functioning is due to the insurance effects, the performance of the 47-member consortium should be compared to the consortium with fewer members. A more rational design is to build consortia according to the four testing functions. Some consortia contain one mutant specializing in one specific function, while the others contain more mutants for the function (that is, add redundancy). Then compare the root colonization and plant protection ability of the two groups.

Please, see our response to point 3 and above.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

We had substantial internal discussion of the revised manuscript, and we agree that the revisions have been substantial.

However, we felt that there is still one important issue that must be carefully considered and addressed. This is regarding the new supernatant experiment, which might not be the best way to infer ecological interactions (please see comments from ref#3) and might lead to misleading interpretations. We agreed that this issue should be clarified before final acceptance of the work. We see two potential options to do so:

1) If you wish to firmly conclude that reduced competition among consortia members exists, please include single strain vs. pairwise co-culture experiments followed by colony counts (see the suggestions below for how to overcome the difficulty that the strains are nearly identical). This might actually require a limited amount of work and would be most appropriate to infer potentially neutral interactions among consortia members.

2) Alternatively, you should remove the supernatant data from the manuscript and adopt a more tempered tone for all assertions linked to how "this study" enhances our understanding of function by mitigating competition between the consortia members.

We thank the Editor and three anonymous reviewers for highly useful feedback on our manuscript. We have now carefully followed all the suggestions and fully revised our manuscript accordingly. While we have not been able to collect more data using CFU counts, we have re-analyzed the data using 50% LB media as control baseline as suggested by the reviewer #3.

The main changes include:

  • Discussion behind the reasons why consortia outperform single strains.

  • Reanalysis of pairwise mutant interactions using 50% LB media as the baseline control.

  • Toning down our interpretation on the competition between consortia members throughout the manuscript; supernatant data is used to test if mutants inhibit each other’s growth and references to competition have been removed.

Reviewer #2 (Recommendations for the authors):

I only have a small suggestion optional to the authors. It might be intriguing to discuss the microscale-level mechanisms behind why the consortium outperforms a single strain. Could the four plant growth-promotion traits act as public goods? For instance, could biofilm formation specialists aid pathogen suppression specialists in better colonizing plants? It may be interesting to add some discussions to the paragraph in lines 497-516.

We thank reviewer #2 for positive and very helpful comments. The production of public goods could indeed play a role here and we now mention this on lines 524-529.

Reviewer #3 (Recommendations for the authors):

Collectively, I am excited to see this manuscript again, as I've been keeping an eye out to see if it had been published in another journal since my first review. It is an intriguing work and I think that many readers in the community will be interested to read it. My primary criticism remains the same as the first review, where ecological interactions are not appropriately measured. I give a detailed explanation of this in Response to R13 (below), along with references that I believe do perform appropriate experiments to measure ecological interactions between strains/genotypes. That said, my other comments appear to be appropriately addressed.

Additionally, because much of the narrative of the discussion has been rewritten to avoid direct claims about ecological interactions between the genotypes throughout the text, I no longer believe that an experiment demonstrating ecological interactions is necessary. Figure S5 should be removed, as this is not an appropriate measure of ecological interactions. The data gathered as part of Figure S5 are likely sufficient to still represent these experiments appropriately. If presented appropriately, I strongly suspect that negative interactions will be the predominant interaction type, so it would be up to the authors if they would like to present this data or not.

We thank reviewer #3 for positive and very helpful comments. We have now replaced the Figure 3 —figure supplement 2 with a revised one, where we use 50% LB as the baseline control to calculate the relative growth as following formula:

Relative growth=(OD600 supOD600 LB)OD600 LB×100%;

OD600 sup and OD600 LB denote for mutants’ growth in other mutants’ supernatant or in 50% LB after 24h of growth (on lines 691-694).

The results show that the biomass production of most strains decreased in the supernatant compared to their growth in 50% LB. While most of these effects were negative (up to 12.7% in magnitude), also some positive effects were observed (up to 7.1% in magnitude). This data has been kept in the manuscript to simply test if insertions affected how mutants shape each other’s growth and references to competition have been removed as suggested.

So long as reference to ecological interactions (whether neutral or otherwise) and Figure S5 are removed from the text (or ecological interactions are measured according to a method similar to those detailed below), the remainder of this work appears to hold together well, with the text supported by the data.

Response to R13:

Regarding Figure S5. While supernatant assays are not the preferred method to measure ecological interactions, I appreciate the authors performing an experiment to address this major comment from my first review. However, this is not the appropriate experiment and it does not support the claim that interactions between the different genotypes are neutral.

The ideal experiment would be to grow each strain in monoculture and count cfu/mL, then grow each strain in pairwise combination and count cfu/mL of each strain. If it is not possible to selectively plate in order to identify the different genotypes, one could also grow each strain in monoculture and count cfu/mL. Then again repeat growing strains in pairwise co-culture, and count the total cfu/mL. If the interactions are neutral, the prediction is that the total cfu/mL will be the sum of the two strains monoculture cell densities (cfu/mL). One could interpret this as clear niche differentiation, with no interference or exploitative competition. If you are committed to the growth supernatant assay and measuring OD600, you could grow each strain in monoculture, measure final OD600 and then filter the supernatant just as you've done. You can then grow them again in the spend media and measure the final OD600. If the interactions are neutral, then prediction is that growth in monoculture with spent media = growth in monoculture with fresh media. For any strain combination, if growth in monoculture of spent media < growth in monoculture of fresh media, this is a negative interaction.

We have now used 50% LB as the baseline control to determine how mutants’ supernatants affect each other’s growth in Figure 3 —figure supplement 2.

Unfortunately, it is not satisfactory to infer ecological interactions as a measure of growth in supernatant relative to growth in one's own supernatant. There is full niche overlap when a strain grows in its own supernatant. Using this metric, you could have two different strains with identical nutrient requirements, that will compete with each other for these nutrients, and you would likely measure it as a neutral interaction. It is also challenging when measuring ecological interactions in liquid LB and trying to extrapolate this to interactions in vivo where the host is also present. An experiment like this in liquid LB is probably a good proxy for gauging interactions, but an experiment that is close to the environment of interest would be the most desirable. However, I understand that this can present additional challenges. I also understand that a specific aim of this study is to use in vitro phenotyping in order to optimize consortia functioning. So in that regard, an LB experiment is perhaps appropriate.

We fully agree with the reviewer that in vivo environment would be the most appropriate way to test the effect of competition between mutants. Hence, we have toned down our interpretation on the competition between consortia members throughout the manuscript.

I hope that this criticism is clear. For a more comprehensive description of methods in this regard, I suggest Foster and Bell, 2012. DOI: 10.1016/j.cub.2012.08.005. Additionally, Weiss et al. ISME 2022 DOI: 10.1038/s41396-021-01153-z performs nice experiments in this regard, using both monoculture vs coculture experiments (quantifying strain abundances with qPCR) as well as spent supernatant assays.

And yet, after all of this, after reading the revised discussion, I no longer think that this experiment is essential. Because the authors have removed much of the narrative regarding ecological significance and focused more on the applied aspects of this work, I do not think that these ecologically-focussed experiments are completely necessary. So long as the authors avoid claims of the interactions between genotypes (as I would argue these have still not been appropriately measured), then I don't think this experiment is necessary for this manuscript.

We hope that our revisions have now resolved this final outstanding issue. We also want to thank the reviewer for raising this important point and taking time to thoroughly revise our manuscript.

https://doi.org/10.7554/eLife.90726.sa2

Article and author information

Author details

  1. Jingxuan Li

    Key Lab of Organic-based Fertilizers of China and Jiangsu Provincial Key Lab for Solid Organic Waste Utilization, Nanjing Agricultural University, Nanjing, China
    Contribution
    Formal analysis, Investigation, Visualization, Writing – original draft
    Contributed equally with
    Chunlan Yang
    Competing interests
    No competing interests declared
  2. Chunlan Yang

    Key Lab of Organic-based Fertilizers of China and Jiangsu Provincial Key Lab for Solid Organic Waste Utilization, Nanjing Agricultural University, Nanjing, China
    Contribution
    Formal analysis, Investigation, Visualization, Writing – original draft
    Contributed equally with
    Jingxuan Li
    Competing interests
    No competing interests declared
  3. Alexandre Jousset

    Key Lab of Organic-based Fertilizers of China and Jiangsu Provincial Key Lab for Solid Organic Waste Utilization, Nanjing Agricultural University, Nanjing, China
    Contribution
    Conceptualization, Writing – review and editing
    Competing interests
    No competing interests declared
  4. Keming Yang

    Key Lab of Organic-based Fertilizers of China and Jiangsu Provincial Key Lab for Solid Organic Waste Utilization, Nanjing Agricultural University, Nanjing, China
    Contribution
    Formal analysis, Investigation, Visualization
    Competing interests
    No competing interests declared
  5. Xiaofang Wang

    Key Lab of Organic-based Fertilizers of China and Jiangsu Provincial Key Lab for Solid Organic Waste Utilization, Nanjing Agricultural University, Nanjing, China
    Contribution
    Formal analysis, Investigation
    Competing interests
    No competing interests declared
  6. Zhihui Xu

    Key Lab of Organic-based Fertilizers of China and Jiangsu Provincial Key Lab for Solid Organic Waste Utilization, Nanjing Agricultural University, Nanjing, China
    Contribution
    Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-3987-8836
  7. Tianjie Yang

    Key Lab of Organic-based Fertilizers of China and Jiangsu Provincial Key Lab for Solid Organic Waste Utilization, Nanjing Agricultural University, Nanjing, China
    Contribution
    Formal analysis, Investigation
    Competing interests
    No competing interests declared
  8. Xinlan Mei

    Key Lab of Organic-based Fertilizers of China and Jiangsu Provincial Key Lab for Solid Organic Waste Utilization, Nanjing Agricultural University, Nanjing, China
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  9. Zengtao Zhong

    College of Life Science, Nanjing Agricultural University, Nanjing, China
    Contribution
    Resources, Methodology, Writing – review and editing
    Competing interests
    No competing interests declared
  10. Yangchun Xu

    Key Lab of Organic-based Fertilizers of China and Jiangsu Provincial Key Lab for Solid Organic Waste Utilization, Nanjing Agricultural University, Nanjing, China
    Contribution
    Conceptualization, Supervision, Project administration
    Competing interests
    No competing interests declared
  11. Qirong Shen

    Key Lab of Organic-based Fertilizers of China and Jiangsu Provincial Key Lab for Solid Organic Waste Utilization, Nanjing Agricultural University, Nanjing, China
    Contribution
    Supervision, Project administration
    Competing interests
    No competing interests declared
  12. Ville-Petri Friman

    1. Key Lab of Organic-based Fertilizers of China and Jiangsu Provincial Key Lab for Solid Organic Waste Utilization, Nanjing Agricultural University, Nanjing, China
    2. Department of Microbiology, University of Helsinki, Helsinki, Finland
    Contribution
    Conceptualization, Formal analysis, Visualization, Writing – review and editing
    For correspondence
    ville-petri.friman@helsinki.fi
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-1592-157X
  13. Zhong Wei

    Key Lab of Organic-based Fertilizers of China and Jiangsu Provincial Key Lab for Solid Organic Waste Utilization, Nanjing Agricultural University, Nanjing, China
    Contribution
    Conceptualization, Data curation, Supervision, Funding acquisition, Project administration
    For correspondence
    weizhong@njau.edu.cn
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-7967-4897

Funding

National Key Research and Development Program of China (2021YFD1900100)

  • Zhong Wei

National Key Research and Development Program of China (2022YFF1001804)

  • Xiaofang Wang

National Key Research and Development Program of China (2022YFD1500202)

  • Tianjie Yang

National Natural Science Foundation of China (42325704)

  • Zhong Wei

National Natural Science Foundation of China (42090064)

  • Qirong Shen

National Natural Science Foundation of China (41922053)

  • Zhong Wei

National Natural Science Foundation of China (31972504)

  • Yangchun Xu

Fundamental Research Funds for the Central Universities (KYT2023001)

  • Zhong Wei

Royal Society Research Grants (RSG\R1\180213)

  • Ville-Petri Friman

Royal Society Research Grants (CHL\R1\180031)

  • Ville-Petri Friman

Strategic Priorities Fund Plant Bacterial Diseases programme (BB/T010606/1)

  • Ville-Petri Friman

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Daniel Rozen, Benoit Stenuit, and Zheren Zhang for valuable comments on the manuscript. This research was supported by the National Key Research and Development Program of China (2021YFD1900100, 2022YFF1001804, 2022YFD1500202), the National Natural Science Foundation of China (42325704, 42090064, 41922053, and 31972504), the Fundamental Research Funds for the Central Universities (KYT2023001). V-PF is supported by the Royal Society Research Grants (RSG\R1\180213 and CHL\R1\180031) and jointly by a grant from UKRI, Defra, and the Scottish Government, under the Strategic Priorities Fund Plant Bacterial Diseases programme (BB/T010606/1).

Senior Editor

  1. Detlef Weigel, Max Planck Institute for Biology Tübingen, Germany

Reviewing Editor

  1. Stéphane Hacquard, Max Planck Institute for Plant Breeding Research, Germany

Reviewers

  1. Stéphane Hacquard, Max Planck Institute for Plant Breeding Research, Germany
  2. Miaoxiao Wang, ETH Zurich, Switzerland
  3. Jacob D Palmer, University of Oxford, United Kingdom

Version history

  1. Preprint posted: March 18, 2022 (view preprint)
  2. Received: July 4, 2023
  3. Accepted: September 13, 2023
  4. Accepted Manuscript published: September 14, 2023 (version 1)
  5. Version of Record published: September 25, 2023 (version 2)

Copyright

© 2023, Li, Yang et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 519
    Page views
  • 134
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Jingxuan Li
  2. Chunlan Yang
  3. Alexandre Jousset
  4. Keming Yang
  5. Xiaofang Wang
  6. Zhihui Xu
  7. Tianjie Yang
  8. Xinlan Mei
  9. Zengtao Zhong
  10. Yangchun Xu
  11. Qirong Shen
  12. Ville-Petri Friman
  13. Zhong Wei
(2023)
Engineering multifunctional rhizosphere probiotics using consortia of Bacillus amyloliquefaciens transposon insertion mutants
eLife 12:e90726.
https://doi.org/10.7554/eLife.90726

Further reading

    1. Ecology
    Alan V Rincon, Bridget M Waller ... Jérôme Micheletta
    Research Article

    The social complexity hypothesis for communicative complexity posits that animal societies with more complex social systems require more complex communication systems. We tested the social complexity hypothesis on three macaque species that vary in their degree of social tolerance and complexity. We coded facial behavior in >3000 social interactions across three social contexts (aggressive, submissive, affiliative) in 389 animals, using the Facial Action Coding System for macaques (MaqFACS). We quantified communicative complexity using three measures of uncertainty: entropy, specificity, and prediction error. We found that the relative entropy of facial behavior was higher for the more tolerant crested macaques as compared to the less tolerant Barbary and rhesus macaques across all social contexts, indicating that crested macaques more frequently use a higher diversity of facial behavior. The context specificity of facial behavior was higher in rhesus as compared to Barbary and crested macaques, demonstrating that Barbary and crested macaques used facial behavior more flexibly across different social contexts. Finally, a random forest classifier predicted social context from facial behavior with highest accuracy for rhesus and lowest for crested, indicating there is higher uncertainty and complexity in the facial behavior of crested macaques. Overall, our results support the social complexity hypothesis.

    1. Ecology
    2. Genetics and Genomics
    Franziska Grathwol, Christian Roos ... Gisela H Kopp
    Research Advance

    Adulis, located on the Red Sea coast in present-day Eritrea, was a bustling trading centre between the first and seventh centuries CE. Several classical geographers--Agatharchides of Cnidus, Pliny the Elder, Strabo-noted the value of Adulis to Greco--Roman Egypt, particularly as an emporium for living animals, including baboons (Papio spp.). Though fragmentary, these accounts predict the Adulite origins of mummified baboons in Ptolemaic catacombs, while inviting questions on the geoprovenance of older (Late Period) baboons recovered from Gabbanat el-Qurud ('Valley of the Monkeys'), Egypt. Dated to ca. 800-540 BCE, these animals could extend the antiquity of Egyptian-Adulite trade by as much as five centuries. Previously, Dominy et al. (2020) used stable istope analysis to show that two New Kingdom specimens of P. hamadryas originate from the Horn of Africa. Here, we report the complete mitochondrial genomes from a mummified baboon from Gabbanat el-Qurud and 14 museum specimens with known provenance together with published georeferenced mitochondrial sequence data. Phylogenetic assignment connects the mummified baboon to modern populations of Papio hamadryas in Eritrea, Ethiopia, and eastern Sudan. This result, assuming geographical stability of phylogenetic clades, corroborates Greco-Roman historiographies by pointing toward present-day Eritrea, and by extension Adulis, as a source of baboons for Late Period Egyptians. It also establishes geographic continuity with baboons from the fabled Land of Punt (Dominy et al., 2020), giving weight to speculation that Punt and Adulis were essentially the same trading centres separated by a thousand years of history.