Growth Response of Crop Legumes to Soil Microbiota Is Linked With Soil Nutrients and Planting History

TL;DR

This study explores how soil microbes affect legume crops, finding that soil nutrients and farming history influence plant growth through microbial interactions.

Contribution

The study identifies how soil microbiota, nutrients, and planting history together influence legume growth, offering a pathway to inexpensively enhance microbial services.

Findings

Resident soil microbiota enhanced cowpea growth, but not soybean or lima bean.

Bioinoculant application altered microbial communities and root nodulation but not crop growth.

Soil nutrients correlated with microbial changes and growth effects, which were microbially mediated.

Abstract

Soil microbiota provide essential services to plants, but predicting or manipulating these benefits is difficult. Here, we investigated microbial benefits to legume crops at a landscape level to uncover factors that predict those services and can be modified by growers. We sampled cultivated soils across a 1000 km transect of production farms and experiment stations with cowpea cultivation. Bioinoculant practices and crop histories were evaluated. Soils were characterized using bacterial metabarcoding and physicochemical analysis, and soil microbial extracts were created to test the capacity of the microbiota to induce root nodulation and growth effects in six legume cultivars, including cowpea, soybean, and lima bean. Resident soil microbiota enhanced cowpea growth, whereas soybean and lima bean experienced negligible benefits. Grower application of bioinoculants was associated with altered microbial communities and enhanced root nodulation but did not affect crop growth. Soil nutrient makeup was correlated with changes in the resident microbial communities and growth benefits to plants, growth effects that were eliminated in sterile soil inoculation treatments, suggesting that they are microbially mediated. Our findings that both planting practices and abiotic soil factors can indirectly affect plant performance, mediated by restructuring of the soil microbial community, suggest how soils could be inexpensively modified to enhance microbial services.

Figures

Field soil collection sites. Soil samples were collected at agronomic sites spanning a ~1000 km transect in California, from Colusa County in the north, to Riverside County in the south. Sampling locations (n = 10) were selected based on having cowpea cultivation up to three years prior to sampling, and were divided into two subsets, those being managed by independent production farms that used bioinoculants (blue circles) and those within agricultural experiment stations (red squares). County borders are drawn in white.

Soil composition and resident microbiome analysis. (A) Principal components analysis summarizes soil chemical composition across the ten sites, with PC1 (X axis) explaining 41.8% of the variance and PC2 (Y axis) explaining 26.4% of the variance. Vectors indicate direction and strength of soil biochemical gradients. The sites clustered into two groups 1 and 2, respectively (purple, green), and the outlier Tulare in blue, based on k‐means clustering. (B) Heatmap shows the relative abundances of the top 50 most common bacterial taxa in the sampled soils. Relative abundance values are based on amplicon sequence variant (ASV) read counts in the 16 s metabarcoding. A phylogenetic tree, showing relationships among ASVs is shown on the left of the heatmap, and taxon names are shown on the right. (C) Relative abundances values are shown for the four rhizobia genera that were present in the soils, including Bradyrhizobium, Ensifer, Mesorhizobium, and Rhizobium. Circle size indicates the percent of total sequence reads encompassed by each genus, and the numbers indicate how many species level groups were present within each detected genus.

Nodule counts per plant by site and host genotype. The six plant lines are indicated in different colors with the cowpeas in blue (CB5), red (CB27) and green (CB46), lima bean is in purple, and the soybean lines are in orange (conventional) and yellow (glyphosate tolerant). Bars and whisker plots indicate means plus and minus one standard error from plants treated with soil microbial extracts (n = 9 per treatment group).

Mean growth response of plants to soil microbial extracts. A relativized measure of host growth was used to analyze the effects of microbial communities in soils (n = 9 per treatment group). Values were log transformed to achieve normality, thus values above zero reflect positive growth effects of microbes and values below zero reflect negative effects of microbes. (A) Mean growth response shows average effects of inoculation with soil microbial extracts across all three cowpea lines combined. Boxes indicate one standard error above and below the mean, and whiskers represent the full range of the data set. (B) Growth response of all treated plants by soil treatment and plant genotype. Growth response values are Log10 transformed. The line at zero represents no growth effects of microbes. The six plant lines are indicated in different colors with the cowpeas in blue (CB5), red (CB27) and green (CB46), lima bean is in purple, and the soybean lines are in orange (conventional) and yellow (glyphosate tolerant). Boxes indicate one standard error above and below the mean, and whiskers represent the full range of the data set.

Tables

TABLE 1. Summary of soil site characteristics and grower practices.

Production farms and research station sites	Date sampled	GPS	Recent planting history (2019–2022)	USDA soil series	Bio‐inoculants used at time of planting?
Colusa, Production farm in Colusa County	6/29/2022	39.21, −121.99	Cowpea unreported rotating crops	Vina	Yes
Paige, Production farm in north Tulare County	6/22/2022	36.18, −119.43	Corn, cowpea, wheat	Colpien	Yes
Stanislaus, Production farm in Stanislaus County	6/29/2022	37.51, −120.81	Cowpea, unreported rotating crops	Handford	Yes
Sutter, Production farm in Sutter County	6/29/2022	39.13, −121.82	Baby lima bean, cowpea	Columbia & Shanghai	Yes
Tulare, Production farm in Tulare County	6/22/2022	36.15, −119.41	Corn, cowpea, sugar beet, triticale	Colpien	Yes
DAVIS, UC Davis Vegetable Crops Field (research station)	6/29/2022	38.53, −121.78	Canola, cowpea, wheat	Yolo	No
Kearney, Kearney Agricultural Research and Extension Station	6/22/2022	36.60, −119.51	Common bean, cowpea, oats, sorghum, wheat	Handford	No
Riverside 10, University of California Riverside Agricultural Experiment Station Field 10F	7/7/2022	33.96, −117.34	Barley, cowpea, wheat	Arlington	No
Riverside 11, University of California Riverside Agricultural Experiment Station Field 11H	7/7/2022	33.96, −117.33	Barley, cowpea, wheat	Arlington	No
Thermal, Coachella Valley Agricultural Experiment Station	7/8/2022	33.52, −116.15	Barley, common bean, cowpea	Coachella	No

TABLE 2. Effects of plant genotype, soil treatment, and soil characteristics on plant performance traits (linear mixed model).

	F v/F m (2 days before harvest)			Leaf count (2 days before harvest)			Host growth response			Number of nodules (per plant)			Total nodule mass (g)
	F‐ratio	df	p	F‐ratio	df	p	F‐ratio	df	p	F‐ratio	df	p	F‐ratio	df	p
Fixed effects
Soil treatment	9.3648	9	< 0.0001	2.3107	9	0.0149	5.7705	9	< 0.0001	12.308	9	< 0.0001	2.6238	9	0.0057
Host genotype	15.7662	5	< 0.0001	171.680	5	< 0.0001	20.188	5	< 0.0001	112.16	5	< 0.0001	20.518	5	< 0.0001
Soil ×	2.4966	45	< 0.0001	1.9827	45	0.0003	5.6228	45	< 0.0001	3.3204	45	< 0.0001	2.2740	45	< 0.0001
PC1 a	3.414	1	0.0652	1.0850	1	0.2980	29.289	1	< 0.0001	8.9118	1	0.0030	2.2410	1	0.1350
Random effects
Block b		1	0.0013		1	0.1097		1	0.0529		1	0.5300		1	0.5255

Full text

Introduction

1

Legumes (Fabaceae) associate with a broad set of soil microbes that can modulate plant growth, but are unique among plant families for their capacity to form a highly beneficial symbiosis with nitrogen‐fixing soil bacteria called rhizobia (Rascovan et al. 2016; Porter et al. 2024). Rhizobia induce tumor‐like structures on roots called nodules, where they fix atmospheric nitrogen into biologically accessible forms in exchange for plant‐derived carbon. In agronomic systems, rhizobia can be leveraged to promote crop growth and yield, in some cases contributing up to 40 kg of nitrogen per metric ton of dry shoot matter (Peoples et al. 2009). Because of potential contributions to plant productivity and soil fertility, biological nitrogen fixation by rhizobia is an attractive alternative to chemical fertilizer applications, the use of which can lead to eutrophication, soil salinification, and increased greenhouse gas emissions (Menegat et al. 2022). Beyond rhizobia, other soil microbiota also modulate legume growth and can interact with rhizobia to enhance nitrogen fixation, suggesting that a broad community of soil microbes could be employed to promote legume production (Rascovan et al. 2016). To date, programs that rely on inoculating legumes with beneficial rhizobia have been unreliable, as the introduced strains are usually outcompeted by resident microbes that provide little or no benefits to the targeted host, a phenomenon referred to as the rhizobia competition problem (Triplett and Sadowsky 1992; Yates et al. 2011; Sinclair and Nogueira 2018).

Agricultural hosts gain unpredictable benefits from soil microbiota, which could be mediated by multiple, poorly understood factors, including the crop planted, its compatibility with available microbes, the soil nutrients present, and interactions among these factors (Heath and Stinchcombe 2014). Plant species, and genotypes within them, naturally vary in their capacity to gain benefits from microbial partners and defend themselves against harmful strains (Heath and Tiffin 2009; Friesen et al. 2011; Simonsen and Stinchcombe 2014; Haney et al. 2015; Wendlandt et al. 2019). The microbial communities that crop plants interact with also vary from field to field and can include plant antagonists, mutualists, and facilitators (Hussain et al. 2018; Li et al. 2019). Abiotic factors, such as soil pH, chemical composition, and particle size, can impact the microbial communities that thrive in soil (Xu et al. 2018), possibly altering the services that hosts receive. Management practices in agriculture, such as tillage, fertilization, and crop rotation, can restructure the microbial community by changing the soil physicochemical properties and enriching for specific microbes through root association (Balachandar et al. 2006; Zhong et al. 2010; Legrand et al. 2018).

Cowpea ( Vigna unguiculata L. Walp) is a legume crop used for food, fodder for livestock, as a cover crop to manage soil erosion, and as a rotation crop to maintain soil health, and is known for its tolerance to heat, drought, and nutrient poor soil, making it ideal for arid and changing landscapes (Sheahan 2012). Cultivated cowpeas are descended from the wild relative Vigna unguiculata subsp. dekindtiana, which is native to Africa (Ali et al. 2015). Modern cultivated cowpeas have a cosmopolitan distribution with hundreds of cultivars and landraces that descend from two main lineages in northern and southern regions of Africa with restricted recombination between these groups (Huynh et al. 2013; Ortiz‐Barbosa et al. 2022). A diversity of rhizobia taxa can instigate root nodule formation on cowpea, though associations with Bradyrhizobium spp. are best studied. Cowpeas that are nodulated with Bradyrhizobium can exhibit increased nitrogen fixation, biomass, grain yield, and water‐use efficiency (Mbah et al. 2022; Mohammed et al. 2022). Some Bradyrhizobium strains have specificity for nitrogen fixation on multiple crop species, including cowpea, lima bean, and soy, demonstrating a degree of symbiotic cross compatibility among these crops (Keyser et al. 1982; da Martins Costa et al. 2017).

Here we investigated cowpea cultivation sites in California. We hypothesized that biotic and abiotic features of the soils, as well as growers' practices, can affect the services that soil microbiota provide to legume crops. Ten agronomic fields were studied, all having a recent history of cowpea cultivation, but varying in grower practices and distributed across a 1000 km transect of California where cowpeas are cultivated. Five sites were managed by experimental research stations, and five were on production farms. The longitudinally oriented sampling allowed for a broad variety of soil types to be tested. Soils were evaluated for physicochemical properties, and 16S rDNA amplicon metabarcoding was used to characterize their resident bacterial communities. Soil microbial extracts were created from each field soil and inoculated onto axenically grown seedlings of six crop cultivars, including cowpea, soybean, and lima bean, selected because each associates with Bradyrhizobium for root nodulation (Burton 1952; da Martins Costa et al. 2017). In greenhouse experiments, inoculated plants were quantified for growth and photosynthetic output at multiple time points. At harvest, plants were dissected to count nodules and dried to quantify root, shoot, and nodule biomass. The goals of the experiment were to (i) characterize biotic and abiotic features of cowpea‐cultivated field soils across California, (ii) examine nodulation and growth effects of soil microbes from those fields on multiple legume cultivars, (iii) investigate the extent of host plant cross‐compatibility for rhizobia communities across agronomic sites, and (iv) uncover the soil characteristics and grower practices that contribute to microbial benefits to plant performance.

Materials and Methods

2

Soil Collection, Analysis, and Preparation of Soil Microbial Extracts

2.1

Soil collections occurred in June and July 2022, before the mid‐summer planting season for cowpea. Soil collection sites were selected based on having cowpea cultivation up to 3 years prior to sampling and included those being managed by independent growers (in Colusa, Sutter, and Stanislaus Counties, and two sites in Tulare County) and those in agricultural experiment stations of the University of California Agriculture and Natural Resources (UCANR; at Davis, Kearney, Thermal, and two sites at UC Riverside; Table 1). The 10 field soils were categorized into soil series based on USDA soil data (SSURGO 2024). Soil was collected from the 0‐30 cm depth at four points which were evenly spaced across each 1 ha field. A total of ~15 L of soil was collected at each field site. From each point sample, 1.2 L was reserved for analysis of chemical composition, including organic matter, nitrogen, phosphorus (weak Bray and Olsen methods), potassium, magnesium, calcium, sodium, soil pH, hydrogen, and cation exchange capacity (Standard Soil Analysis; A&L Western Agricultural Laboratories, Modesto, CA). A 1 kg subsample from each site was used to evaluate content of sand, silt and clay, and available moisture (Soil Texture Analysis; A&L Western Agricultural Laboratories). A separate ~250 g sample from each soil source was frozen at −20°C for metabarcoding of the resident bacterial community. The remaining soil for each site was pooled, sieved, mixed with a 1:1 ratio of sterile water, and strained with cheesecloth to remove organic debris (Unkovich and Pate 1998). The soil slurry was left to settle overnight, and the supernatant was collected into portions of 600 and 250 mL. The larger portion was used as a soil microbial extract and the smaller portion was autoclaved as a sterile control, to be able to dissect biotic effects of the soil on plant hosts (i.e., resident microbial community) from abiotic effects (i.e., soil nutrients). A sample from each soil microbial extract and control (100 μL) per field was spread plated on modified arabinose gluconate agar medium (MAG, Sachs et al. 2009). Plates were incubated at 29°C for 5 days and evaluated to confirm abundance of culturable soil microbiota in soil microbial extracts and the absence of culturable microbiota in controls.

Bacterial Metabarcoding Analysis

2.2

Soil samples from each site (~1 g) were analyzed for resident communities of bacteria, because of their demonstrated associations with legumes and other crops (Rascovan et al. 2016). With materials and services from Zymo Research (Irvine, CA), DNA was extracted using ZymoBIOMICS‐96 MagBead DNA Kit and ZymoBIOMICS DNA Miniprep Kit. To categorize the microbial community taxonomically, the V3‐V4 region of the 16S ribosomal DNA was amplified. Real‐time PCR machines were used to amplify DNA to the log‐linear phase of a PCR to minimize chimera formation (Smyth et al. 2010). Amplified products were quantified using qPCR fluorescence, pooled to equal molarity, cleaned up using the Zymo Select‐a‐Size DNA Clean & Concentrator, then quantified again with a TapeStation and a Qubit high sensitivity assay kit (Thermo‐Fisher Scientific, Waltham, WA). The library of amplified strains was sequenced on an Illumina NextSeq 2000TM with a p1 reagent kit (cat 20075294; 600 cycles). Chimeric amplicons and non‐bacterial reads were removed, and unique sequences were inferred using the Dada2 pipeline (Callahan et al. 2016). The Uclust function in Qiime v.1.9.1. was used for taxonomic assignment against the Zymo 16S Database (updated, March 2024). Absolute microbial abundance was estimated using real‐time PCR with a standard curve made with plasmid DNA containing a single 16S copy. Taxonomic composition and diversity analyses were performed in Qiime v.1.9.1 (Caporaso et al. 2010; Segata et al. 2011).

Greenhouse Experiment

2.3

Six plant lines were used, including three California blackeye cowpea cultivars (CB5, CB27, and CB46) (Ehlers et al. 2000), two near‐isogenic soybean ( Glycine max ) cultivars S16‐5503GT (Chen et al. 2023) and S16‐11651C (Chen et al. 2022), and one lima bean ( Phaseolus lunatus ) cultivar UC Luna (Mou 2022). The three cowpea lines were selected from a multi‐parent intercross population used for a genome‐informed cowpea breeding program, with traits including Fusarium and nematode resistance and heat tolerance (Boukar et al. 2016; Ndeve et al. 2018). S16‐5503GT was chosen as a representative herbicide‐tolerant soybean line, S16‐11651C was chosen to represent conventional soybean breeding, and Luna was chosen as a common variety among California growers.

A wetted sand mix (50:50 silica #12 and #30 sizes) was added to one‐gallon nursery pots and autoclaved twice in 90‐min cycles, with > 48 h in between to promote spore germination and ensure sterility. Seeds were surface sterilized in 1% sodium hypochlorite (bleach) for 30 s (lima bean), 2 min (soybean) or 3 min (cowpea), based on pilot trials that tested seed survival and sterility. After sterilization, seeds were rinsed four times in sterile water and planted the same day. To account for different growth rates during germination, lima bean seeds were planted on 7/21/22. Cowpea lines CB5, CB46, and both soybean lines were planted 7/26/22, and cowpea line CB27 was planted on 7/29/22. All pots were seeded in duplicate and then seedlings were thinned or moved to pots so that every pot had one seedling prior to inoculation.

Each plant line and soil source combination had nine replicates treated with a soil microbial extract and three sterilized controls, which were distributed among 12 blocks in the greenhouse. The 9:3 treatment to control ratio was used because controls exhibit little inter‐replicate variance (Manci et al. 2023). Each block contained all treatment combinations in a random placement. Sterilized controls were randomly assigned to three blocks, such that each block contained an assortment of soil microbe and control treatments to minimize confounding block effects (6 legume lines × 10 soil sources = 60 plants per block). The soil microbial extracts were prepared within a three‐day period (8/9/22–8/11/22) and immediately applied to the soil at the base of the stems. All plants were fertilized weekly with 10 mL of sterilized Broughton and Dilworth solution containing 0.4 g/L KNO_3_ as a nitrogen source, a minimal concentration of nitrogen to support cowpea growth (Somasegaran and Hoben 1994; Ortiz‐Barbosa et al. 2022). Greenhouse plants experienced ~12–13 h daylength, 24°C–32°C, and 2× weekly drip irrigation. The greenhouse received pesticide sprays on a weekly basis.

Treatment effects were quantified using leaf counts, photosynthetic output, root and shoot biomass, and nodule counts. We quantified photosynthetic output using fluorometry to measure maximum quantum yield of photosystem II, measured as a ratio of variable fluorescence divided by maximum fluorescence (i.e., F v/F m), using an Opti‐sciences OS30p + device. All plants had two fluorometry readings, one at 4 weeks post inoculation and then again at 6 weeks post inoculation. All readings were taken at least 20 min past the true nighttime start for the day and researchers wore green‐light headlamps to restrict photosystem activation in plants. Harvest of plants occurred by plant genotype, starting at 6 weeks post inoculation with lines that flowered and produced pods earliest.

Data Analysis

3

A unitless relativized measure of host growth response was used to compare plant growth among treatments, calculated by subtracting the mean total biomass (i.e., shoots and roots) of control plants (i.e., sterile) from the total biomass of corresponding plants treated with soil microbial extracts (i.e., shoots, roots, and nodules), and dividing by the mean total biomass of the control plants (Regus et al. 2015). Multiplying this value by 100 determined percent growth effect of microbes on plants. Principal components analysis and k‐means clustering were used to reduce dimensionality and compare soil physicochemical compositions. Linear mixed models were used to examine the fixed effects of soil site, plant genotype, soil physicochemical properties, and soil by host interactions on plant performance and symbiosis traits. Greenhouse block was treated as a random effect. For planned contrasts, ANOVA or T‐tests were used to test for statistical differences in specific nutrient concentrations between field soil collection sites, compare among k‐means clusters, or to compare performance traits of plant genotypes within each site. All correlations were calculated using line of best fit tests with significance level α < 0.05. Growth response data were transformed by Log_10_ to achieve normal distributions to satisfy the assumptions of parametric statistics. Amplicon sequencing read counts were used to estimate total microbial abundance, and relative values were used to analyze abundance of specific taxa. Analyses were conducted using JMP Statistical Software version 16.

Results

4

Sampled Sites Varied in Grower Practices and Physicochemical Properties of Soils

4.1

We quantified how planting practices and soil properties varied among sampled sites, spanning the ~1000 km transect of California (Figure 1). The five production farms had fields with heterogenous histories, with cultivation of a variety of crops, including non‐legumes, in addition to cowpeas, and all applied seed‐coated bioinoculants on cowpeas (strain composition is proprietary, not shared by manufacturers). The other five sites, being managed by the UCANR as experiment stations, were focused on phenotypic and genotypic studies of cowpeas, and did not use bioinoculants. Four of the experiment station fields each had 5–10‐year histories of rotating cowpeas with other crops, except for Riverside 10, which only had one season of cowpea planted the previous year. Nine of the fields were fallow when sampled except the Paige site, which was growing corn, in which case we sampled between rows, ~50 cm from plants (Table 1).

Field soil collection sites. Soil samples were collected at agronomic sites spanning a ~1000 km transect in California, from Colusa County in the north, to Riverside County in the south. Sampling locations (n = 10) were selected based on having cowpea cultivation up to three years prior to sampling, and were divided into two subsets, those being managed by independent production farms that used bioinoculants (blue circles) and those within agricultural experiment stations (red squares). County borders are drawn in white.

Soil from the 10 field sites varied substantially in chemical composition, organic matter content, pH, and particle size, factors summarized with a principal components analysis. The first principal component (PC1) explained 41.8% of the variation between sites and was predominantly driven by percent of organic matter, phosphorus, potassium, and calcium. PC2 explained 26.4% of the variation among sites and was driven primarily by sulfur, magnesium, nitrogen, sodium, and soil pH (Figure 2). Using k‐means, the sites were clustered into two groups and one outlier. The first cluster was characterized by field soils with high nitrogen and magnesium, including those in the Riverside, Thermal, Colusa, and Kearney sites, and the second group was characterized by high soil pH and high sulfur, including those at the Davis, Sutter, Paige, and Stanislaus sites. The outlier, Tulare, was distinguished by higher organic matter, phosphorus, sodium, and calcium. Concentrations of primary growth limiting nutrients (i.e., nitrogen, phosphorus, and potassium) varied five to eight‐fold among the 10 sites, respectively (p < 0.0001 in all cases). Potassium concentrations were significantly higher in k‐mean cluster 1 (relative to cluster 2; p < 0.05), while nitrogen and phosphorus did not vary significantly between the clusters.

Soil composition and resident microbiome analysis. (A) Principal components analysis summarizes soil chemical composition across the ten sites, with PC1 (X axis) explaining 41.8% of the variance and PC2 (Y axis) explaining 26.4% of the variance. Vectors indicate direction and strength of soil biochemical gradients. The sites clustered into two groups 1 and 2, respectively (purple, green), and the outlier Tulare in blue, based on k‐means clustering. (B) Heatmap shows the relative abundances of the top 50 most common bacterial taxa in the sampled soils. Relative abundance values are based on amplicon sequence variant (ASV) read counts in the 16 s metabarcoding. A phylogenetic tree, showing relationships among ASVs is shown on the left of the heatmap, and taxon names are shown on the right. (C) Relative abundances values are shown for the four rhizobia genera that were present in the soils, including Bradyrhizobium, Ensifer, Mesorhizobium, and Rhizobium. Circle size indicates the percent of total sequence reads encompassed by each genus, and the numbers indicate how many species level groups were present within each detected genus.

Resident Soil Microbiomes Are Linked With Local Abiotic Factors and Grower Practices

4.2

We next characterized the taxonomic makeup of bacteria in each of the soils and sought to uncover statistical associations between soil physicochemical features and the makeup of their resident bacterial communities. For nine of the 10 sites, bacterial abundance ranged from ~5.2 × 10^5^ to 6.7 × 10^7^ bacteria per gram of soil, with an outlier of 1.5 × 10^10^ for the Stanislaus site (Figure S1 and Table S1). There was a strong positive correlation between PC1 (soil composition; Figure 2A) and microbial abundance with the outlier removed (R ^2^ = 0.811, p < 0.0001), suggesting that soil physicochemical properties shape microbial abundance, or vice versa. More specifically, both potassium and phosphorus content of the soil were positively correlated with bacterial abundance, respectively (R ^2^ = 0.843, p = 0.0005; R ^2^ = 0.676, p = 0.0065; Figure S1 and Table S1). All 10 of the soil microbial communities were exceptionally diverse, with Shannon Index values ranging from 9 to 11 and Simpson's Reciprocal Diversity Index values from ~240 to 800 (Table S1).

Members of the bacterial order Rhizobiales that form root nodule symbioses with legumes, encompassing Bradyrhizobium, Ensifer, Mesorhizobium, and Rhizobium, were common in all 10 soils (Figure 2), with relative abundance ranging from 2.8% at Sutter to 7.2% at Thermal (Table S1; Garrido‐Oter et al. 2018). The relative abundance of Rhizobiales was strongly correlated with soil factors, being positively correlated with soil pH (R ^2^ = 0.456, p = 0.0321) and negatively correlated to soil concentrations of nitrogen and potassium, respectively (R ^2^ = 0.190, p < 0.0001; R ^2^ = 0.195, p < 0.0001; Table S1). The relative abundance of Rhizobiales did not correlate with PC1 (R ^2^ = 0.192, p = 0.2058). Bradyrhizobium, the genus that predominantly nodulates cowpea, was consistently among the most frequent bacterial taxa in the soils and was the most common rhizobia taxon at all 10 sites but Thermal (Figure 2b,c), where Ensifer dominated (the symbiont of Medicago spp. Tampakaki et al. 2017). Members from the family Bacillaceae and Phyllobacteriaceae, known as plant growth‐promoting bacteria (Bresson et al. 2014; de Marcons Souza et al. 2014; Jessberger et al. 2024), were also found in most sites (only Phyllobacteriaceae was absent at Sutter; Table S1).

The use of bacterial inoculants—unique to the production farms—was associated with changes in the local microbial communities. The soils from production farms had higher overall bacterial abundance (t = 8.26, p < 0.0001). Moreover, Bradyrhizobium spp. were more diverse in the production farm soils relative to the experiment station sites, with twice the number of strains present on average (Figure 2). One B. canariense strain was ~6× enriched at the production farms relative to the experiment station sites (Supporting Information). No such enrichment of strains or diversity was observed for the other taxa of rhizobia, including Ensifer, Mesorhizobium, and Rhizobium. These differences could be driven by the establishment of the inoculant strains and/or shaped by other management practices.

Root Nodulation Patterns Were Shaped by Grower Practices

4.3

Soil microbial extracts from each soil sample were used to test their ability to nodulate legume cultivars. On the basis of plating of the extracts, all were confirmed to have bacteria at high density (i.e., > 10^7^ mL^−1^) and controls were confirmed to be sterile. All 10 soil sources consistently induced root nodules on all three cowpea cultivars, ranging in mean nodule counts among treatments from 12 to 68 per plant, whereas fewer or no nodules formed on the soybean or lima bean cultivars (Figure 3). Among the sites, the Riverside 10 field had the lowest mean nodule count, which might be explained by its brief cowpea history, having only one season of cowpea planting, compared to the recurrent planting at the other experiment station sites. The soils from growers' fields instigated higher nodule counts across all host genotypes compared to research station sites (p < 0.0001; Figure 3). Only soils from the Thermal, Sutter, and Tulare sites induced any root nodules on lima bean, while the remaining soil treatments did not induce any nodules (Figure 3), and among these, only the Sutter site reported recent planting of lima beans (Table 1). Soil treatments induced nodules on at least one soybean plant in every plant–soil treatment apart from the glyphosate tolerant soybean treated with soil from Davis and Riverside 10. Among the cowpea cultivars, CB5 had significantly fewer nodules per plant on average than at least one other cowpea cultivar within four soil treatments, including two grower sites (Colusa, Stanislaus) and two research stations (Kearney, Thermal; Figure 3). Among the soybeans, the glyphosate tolerant line had a slightly lower proportion of plants with nodules relative to conventional soybean (i.e., 64% vs. 70%), but significantly higher nodule counts per plant (p = 0.0130; Table S2).

Nodule counts per plant by site and host genotype. The six plant lines are indicated in different colors with the cowpeas in blue (CB5), red (CB27) and green (CB46), lima bean is in purple, and the soybean lines are in orange (conventional) and yellow (glyphosate tolerant). Bars and whisker plots indicate means plus and minus one standard error from plants treated with soil microbial extracts (n = 9 per treatment group).

Nodule counts were positively correlated with PC1 of the soil chemical composition analysis (R ^2^ = 0.016, p = 0.0030), suggesting that soil chemistry is linked with nodulation rate, albeit weakly. There was also a weak negative correlation between nodule count and relative abundance of the order Rhizobiales (R ^2^ = 0.016, p = 0.0028). Although this is counterintuitive, since Rhizobiales are the main taxa that induce nodules, the symbiotic taxa only represent a small subset of the diversity (Table S1). While omitting Stanislaus as an outlier, there was no correlation between absolute abundance and nodule count (R ^2^ = 0.005, p = 0.105), suggesting that total microbial abundance is not correlated with nodule formation.

Plant Performance Is Primarily Shaped by Biotic Aspects of Source Soils

4.4

Nondestructive measures were used to examine how soil inoculation affected plant performance at two timepoints (Table S3). Plants were first measured 4 weeks after inoculation, during which growth effects of nodulation typically commence in cowpea (Ortiz‐Barbosa et al. 2022; Manci et al. 2023). At this timepoint, we confirmed that plants that received soil microbial communities had significantly higher leaf counts (t = 6.74, p < 0.0001), and higher F v/F m (i.e., indicating lower levels of stress; t = 1.892, p = 0.0300) compared to control plants, both consistent with benefits from the soil microbial communities. Plants were measured again during harvest two to 3 weeks later, when cowpeas typically flower and nodules begin to senesce (Ortiz‐Barbosa et al. 2022; Manci et al. 2023). At this timepoint, plants treated with soil microbial extracts had significantly higher leaf counts (t = 14.040, p < 0.0001), but lower F v/F m values than control plants (t = −11.211, p < 0.0001). In addition to effects of the soil microbial treatments, both plant line used, and the soil treatment × line interactions had significant effects on F v/F m (Table 2). PC1 from the soil analysis did not influence pre‐harvest plant traits (Table 2).

After harvest, plants were weighed for dry biomass to compare growth across treatments. The soil treatments, host line, and soil treatment × host line interaction each had significant effects on host growth response (Table 2). Relative abundance of Bradyrhizobiaceae was positively correlated with growth response across all plants, albeit the relationship is weak (R ^2^ = 0.032, p < 0.0001). Focusing on the cowpeas, the three plant lines that were consistently nodulated, inoculation with the soil microbial extracts provided a positive growth response, with mean effects ranging from 9% to 136% growth increase relative to control plants (Figure 4 and Table S4). Cowpea growth response to soils was positively correlated with total nodule mass per plant (R ^2^ = 0.2514; p < 0.0001) and with mean nodule counts per plant (R ^2^ = 0.0387, p = 0.0012). These results suggest that fields planted in cowpeas consistently had rhizobia that were beneficial to cowpea, though effects varied widely dependent on the field and the cowpea cultivar planted (Table S5). There was no significant difference between research stations and growers' fields for plant growth response on cowpeas (t = 1.610, p = 0.1087), despite the growers' soils being associated with more nodules.

Mean growth response of plants to soil microbial extracts. A relativized measure of host growth was used to analyze the effects of microbial communities in soils (n = 9 per treatment group). Values were log transformed to achieve normality, thus values above zero reflect positive growth effects of microbes and values below zero reflect negative effects of microbes. (A) Mean growth response shows average effects of inoculation with soil microbial extracts across all three cowpea lines combined. Boxes indicate one standard error above and below the mean, and whiskers represent the full range of the data set. (B) Growth response of all treated plants by soil treatment and plant genotype. Growth response values are Log10 transformed. The line at zero represents no growth effects of microbes. The six plant lines are indicated in different colors with the cowpeas in blue (CB5), red (CB27) and green (CB46), lima bean is in purple, and the soybean lines are in orange (conventional) and yellow (glyphosate tolerant). Boxes indicate one standard error above and below the mean, and whiskers represent the full range of the data set.

Growth effects on soybean and lima bean cultivars varied by soil source (Figure 4). Notably, soils from the Kearney and Thermal sites caused significant deficits to lima bean with mean growth responses of −71% and −74% respectively (Figure 4), and most of these plants had no nodules. When comparing soybean genotypes, the glyphosate‐tolerant cultivar had a significantly higher growth response overall (p = 0.0001).

Abiotic soil factors had significant effects on plant benefits from soil inoculation. These appear to be indirect effects, mediated through changes in microbial communities, as there were no growth differences among soil sources when considering the sterilized soil rinsates that lacked microbes (F = 1.040, p = 0.413). PC1 from the soil analysis had significant effects both on root nodulation and on host growth response to inoculation (Table 2). Concentrations of both phosphorus (R ^2^ = 0.013, p = 0.0087) and potassium (R ^2^ = 0.045, p < 0.0001) were positively correlated with growth response, albeit weakly. Moreover, biotic soil factors were linked with the composition of microbial communities recovered from the soils. Shannon alpha diversity was positively correlated with PC1 from the soil analysis (R ^2^ = 0.060, p < 0.0001). The growth response to inoculation was positively correlated with both absolute microbial abundance (R ^2^ = 0.055, p < 0.0001) and with Shannon diversity (R ^2^ = 0.149, p < 0.0001).

Discussion

5

To harness the services that soil microbes can provide to plants, a predictive framework is needed to model the conditions that support beneficial microbial communities in cultivated soils, including crop selection, rotation design, soil characteristics, and other factors that can be selected or modified by growers. Our results suggest four important patterns about beneficial microbial communities in agriculture and their potential to enhance crop performance. Firstly, the grower practices at each site were associated with significant changes in the soil microbial community. In the five production farms studied, we found significantly higher abundance of bacteria in the soil, a higher diversity of Bradyrhizobium present, and greater rates of legume root nodulation. Although there are likely many differences between the experiment stations and the production farms, the latter were the only sites where bioinoculants were applied. These data suggest that inoculation—or associated practices unique to production farms—had important impacts on the soil microbial community. Past work indicated that introduced strains do not benefit target plants because they are outcompeted by resident microbes (Triplett and Sadowsky 1992; Yates et al. 2011; Sinclair and Nogueira 2018). However, our data suggest that inoculation can impact the bacterial community, in particular to enrich soils with Bradyrhizobium strains, suggesting that inoculant strains are persisting in soils.

Second, our data showed that soil abiotic factors are correlated with benefits that microbial communities provide to plant hosts. The soil chemical composition at the sampled field sites, summarized by PC1 of the soil analysis, is statistically associated with the benefits provided by microbial communities in those soils (Figures 2, 4 and Table 2). One interpretation is that soil chemistry can restructure microbial communities in soil, in some cases promoting microbial taxa that provide services to plant hosts. Consistent with this hypothesis, both potassium and phosphorus concentrations were correlated with growth benefits provided by the soil microbial community, and soil pH was positively correlated with the relative abundance of rhizobia. Previous work also found a link between soil chemistry and microbial growth benefits. A meta‐analysis found that soils deficient in phosphorus and potassium support smaller plants, with fewer root nodules and lower nodule mass (Divito and Sadras 2014). In our study, the grower's field that had the highest cowpea growth responses, and also the highest phosphorus levels, reported using gypsum fertilizer, which has been shown to prevent phosphorus depletion in soils by binding to soluble phosphorus molecules and selecting for beneficial bacteria strains (Watts and Torbert 2009; Abd and Alkurtany 2023). It might be that altering these abiotic soil factors is more effective at promoting plant growth than the current practice of using bioinoculants, given that the latter has been so unreliable (Triplett and Sadowsky 1992; Yates et al. 2011; Sinclair and Nogueira 2018).

Third, our data corroborate past work showing the powerful effects that planting specific crops, such as legumes, can have on the resident community of rhizobia and other plant‐associated microbes in field soils (Miethling et al. 2000; Marschner et al. 2001; Kowalchuk et al. 2002). The soils that we studied—all with a history of cowpea cultivation—had a significant net growth benefit on cowpea relative to sterilized versions of the same soil (Figure 4 and Table 1), suggesting that when crops are planted in fields, they have the ability to selectively enrich those soils with beneficial microbes for that same crop. Although multiple plantings of a crop might enrich beneficial microbial communities to a greater degree (Mueller and Sachs 2015; Quides et al. 1951), our data suggest that even a single planting of cowpea is beneficial for future plantings, as the Riverside 10 soil enhanced growth of cowpea with only one season of cowpea cultivation in its record. Effects on cowpea varied widely in magnitude, including soil × plant line combinations that resulted in negligible growth benefits; hence, benefits from the soil microbial communities were not universal and depended both on the plant genotype and the abiotic conditions of the field. An important caveat is that our soil microbial extracts only approximate the microbial diversity found in each field soil, as not all strains would be able to establish under the experimental conditions with the sterile sand soil.

Finally, our experiments indicated that enrichment of beneficial microbial communities can be crop specific. While planting of cowpea enriched our study fields with microbes beneficial to cowpea, these benefits did not extend to other tested legumes, lima bean and soybean, even though these crops also interact with and benefit from root nodulating Bradyrhizobium. The effects of soil microbe treatments on soybean and lima bean genotypes varied by site (Figure 4). Cross‐compatibility of rhizobia symbionts on legume crops is not well understood, but there appears to be an important host species effect. For instance, previous work on soybeans suggested that their benefits from rhizobia are not influenced by cropping history because the researchers found no statistical difference between the biomass of plants inoculated with soils that had a history of soybean planting and those that had no such history (Elkins et al. 1976). Lima bean appears to be more selective for symbionts than cowpea (Burton 1952), and some work suggests that there might be few compatible rhizobia for lima beans in native California soils (de Araujo et al. 2017). Our data are consistent with this, in that only 4% of lima bean plants formed nodules when inoculated by soils from across our 1000 km transect, and that the soils that induced nodulation in lima bean were the ones that had a history of planting lima bean (Figure 3 and Table 1). Moreover, some researchers have suggested that cowpea is relatively nonspecific for rhizobia symbionts (Lewin et al. 1987). Our results uncovered significant differences among the cowpea lines in their benefits from whole microbial communities, including the root nodulating rhizobia (Figure 3), which differed from previous experiments evaluating single strain rhizobia inoculations (Burton 1952; Kanonge‐Mafaune et al. 2018; Abd and Alkurtany 2023; Altai and Muhee 2023). Thus, a broader aspect of host specificity should be considered and further studied which quantifies how different plant genotypes respond to whole microbial communities. Understanding host specificity at this broader level is key to improving sustainability in agriculture, as researchers have been relatively unsuccessful in enhancing crop growth by adding specific beneficial strains (Triplett and Sadowsky 1992; Yates et al. 2011). Future work should select on these plant traits for improved benefits and work to understand the genetic bases for these traits (Sinclair and Nogueira 2018).

Conclusions

6

Our results suggest that growers can make simple and inexpensive changes to improve productivity of cowpea crops. One compelling take‐away is that any previous cowpea planting could benefit future cowpea planting, especially when legumes are rotated with other crops as occurred in the sampled fields. Since soil composition and plant genotype influenced plant growth, growers should consider choosing their cowpea cultivars with the soil type of their fields in mind. Our experiment did not test soils with no history of cowpea, but work on soybeans suggests that soil with no soybean history is more beneficial than sterilized soil, and yields may be increased through seed coat inoculation, but only in the case of a first crop (Elkins et al. 1976; Zilli et al. 2021). Our findings that phosphorus or potassium in soil were correlated with enhanced growth effects suggest that research should experimentally examine effects of phosphorus or potassium fertilizers, to quantify if they enrich the overall soil microbiome for increased plant benefits. In contrast, soil nitrogen content had no significant effects in our analysis, and thus does not appear to enrich the microbiome. Growers who have planted cowpea in the past may not see the same success when planting other legume species, and growers planting cowpea for the first time in California may see limited growth benefits from rhizobia for the first season. Growers would benefit from development of simple tests for the presence of compatible symbionts in their soils, which could guide them on the utility of inoculating their plants. However, since bioinoculants have historically been ineffective in infecting legume crops (Triplett and Sadowsky 1992; Sinclair and Nogueira 2018) future studies could attempt to further understand the drivers of selection for beneficial rhizobia specific to cowpea in California soils.

Funding

J.L.S. was supported by a NIFA‐USDA Award 2022‐67019‐36500, a USDA Hatch Grant CA‐R‐EEOB‐5200‐H, and an NSF Award, Division of Environmental Biology 1738009.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: Summary results of bacteria metabarcoding analysis. Table S2: Mean and standard deviation of nodule counts for plant × soil treatment combinations. Table S3: Chlorophyll fluorometry and leaf count data by soil and plant genotype. 4WAI = 4 weeks after inoculation; 2DBH = 2 Days before harvest. Table S4: Host growth response by soil and plant genotype. Table S5: Linear Mixed Model analysis for cowpea cultivars only.

Figure S1: Primary soil nutrients, microbial abundance, and diversity. (A) Bar graphs show mean concentrations of nitrogen, phosphorus, and potassium content for each soil site (n = 4 soil samples per site). (B) Bar graphs show absolute microbial abundance, relative abundance of the order Rhizobiales, Shannon diversity index, and Simpson reciprocal diversity index for each soil site (n = 1 soil sample per site).