Nitrogen-fixing bacterial communities differ between perennial agroecosystem crops

Abstract

Biocrusts, common in natural ecosystems, are specific assemblages of microorganisms at or on the soil surface with associated microorganisms extending into the top centimeter of soil. Agroecosystem biocrusts have similar rates of nitrogen (N) fixation as those in natural ecosystems, but it is unclear how agricultural management influences their composition and function. This study examined the total bacterial and diazotrophic communities of biocrusts in a citrus orchard and a vineyard that shared a similar climate and soil type but differed in management. To contrast climate and soil type, these biocrusts were also compared with those from an apple orchard. Unlike natural ecosystem biocrusts, these agroecosystem biocrusts were dominated by proteobacteria and had a lower abundance of cyanobacteria. All of the examined agroecosystem biocrust diazotroph communities were dominated by N-fixing cyanobacteria from the Nostocales order, similar to natural ecosystem cyanobacterial biocrusts. Lower irrigation and fertilizer in the vineyard compared with the citrus orchard could have contributed to biocrust microbial composition, whereas soil type and climate could have differentiated the apple orchard biocrust. Season did not influence the bacterial and diazotrophic community composition of any of these agroecosystem biocrusts. Overall, agricultural management and climatic and edaphic factors potentially influenced the community composition and function of these biocrusts.

Agroecosystem biocrusts have similar rates of nitrogen-fixation as natural ecosystem biocrusts, but both composition and function are impacted by agricultural practices.

Introduction

Biological soil crusts (biocrusts) are specific assemblages of microorganisms that colonize the soil surface and the top centimeter and bind to the soil matrix to form a crust. Biocrusts often contain cyanobacteria (Yeager et al. 2007), lichens (Kuske et al. 2012) and other non-cyanobacterial organisms (Nunes da Rocha et al. 2015) capable of biological fixation of dinitrogen gas (N-fixation). These organisms fix N across a range of climatic conditions and contribute approximately one-half of the total N fixed in arid lands (Elbert et al. 2012), with rates between 0.08 and 10 kg N ha⁻¹ year⁻¹ (Malam Issa et al. 2001, Belnap 2002a, Billings et al. 2003, Russow et al. 2005, Housman et al. 2006, Holst et al. 2009). Similarly, mesic biocrusts fixed 1.3 kg N ha⁻¹ year⁻¹ in a temperate savannah (Veluci et al. 2006), 5.2 kg N ha⁻¹ year⁻¹ in a seasonally flooded savannah (Williams et al. 2018) and 4 kg N ha⁻¹ year⁻¹ in the seasonally flooded Everglades (Liao and Inglett 2012, 2014). Biocrusts have also been recently described in agroecosystems (Peng and Bruns 2019a, Nevins et al. 2021) and provide comparable rates of N-fixation with natural ecosystem biocrusts with estimates from 8.1 kg N ha⁻¹ year⁻¹ (citrus orchard) to 4.9 kg N ha⁻¹ year⁻¹ (vineyard) (Sorochkina et al. 2022). The N requirements of perennial crops, such as citrus, that range from 22 to 336 kg N ha⁻¹ year⁻¹ in Florida's sandy soils (Husmann and Dearing 1913, Singerman et al. 2017, Kadyampakeni and Morgan 2020, Sorochkina et al. 2022), underscore the great interest of many producers in capitalizing on the potential of naturally occurring N-fixation as a source of crop N.

Biocrust diazotrophs can be subdivided into two major groups, heterocytous cyanobacteria that acquire carbon (C) autotrophically from sunlight and heterotrophic bacteria that acquire C from an exterior food source. Heterocytous cyanobacteria tend to have higher N-fixation activity than the heterotrophic diazotrophs dominating light biocrusts due to their ability to couple N-fixation with C-fixation, allowing for higher available metabolic C (Belnap 2002b, Miller et al. 2003, Housman et al. 2006, Barger et al. 2013, Bentzon-Tilia et al. 2015, Zhou et al. 2016). While heterotrophic bacteria in biocrusts are largely understudied (Maier et al. 2016), some heterotrophic diazotrophs could have a greater association with more disturbed soils than heterocytous cyanobacteria due to their rapid spore activation and growth, allowing for quick recovery after a disturbance (Stringer et al. 2005, Pepe-Ranney et al. 2016).

Overall, pesticides and herbicides have a negative impact on N-fixing cyanobacterial growth; however, organophosphorus herbicides and insecticides can be degraded by certain cyanobacterial and heterocytous diazotrophs or even used for P uptake, resulting in either no effect or an increase in population (Hove-Jensen et al. 2014, Guijarro et al. 2018, Singh Kaushik et al. 2018, Hernández Guijarro et al. 2021). Unfortunately, there is a lack of recent studies that differentiate the influence of agrochemicals on heterotrophic or autotrophic diazotrophs, but earlier work found that heterotrophic and cyanobacterial N-fixation activities are sensitive to herbicide and fungicide addition (Martensson 1993). Therefore, heterotrophic diazotrophs could dominate perennial agroecosystem biocrusts with repeated disturbances from fruit harvest and management activities such as herbicide application.

Anthropogenic factors of fertilization and irrigation could further enhance differences in agroecosystem biocrust community composition across crops and seasons. Because biocrust N-fixation rates appear to be repressed by N additions (Belnap et al. 2008a), Peng and Bruns 2019a), different fertilization management could result in distinct biocrust bacterial and N-fixing community composition in different crops. For example, N fertilizer application to biocrusts could select for organisms with the ability to switch from N-fixation to N heterotrophy (Khumanthem et al. 2007, Fleming and Castenholz 2008) or for obligate N autotrophy (Menge et al. 2015). Irrigation management can also affect the diazotrophic community composition as it can enhance or diminish the climatic seasonal influence on N-fixation activity (Sorochkina et al. 2022). Although N-fixation activity and climatic response differed between a Florida citrus orchard and vineyard (Sorochkina et al. 2022), it is unclear what impacts management imparted on the diazotroph community composition at these sites.

Climate and edaphic conditions, as well as random ecological dispersal (Su et al. 2020), can dictate the bacterial community composition of natural ecosystem biocrusts and could also influence agroecosystem biocrusts. Ecosystems with the largest differences in precipitation exhibited the strongest differences in biocrust bacterial community composition (Su et al. 2020). Furthermore, temperature (Giraldo-Silva et al. 2020) and precipitation (Machado-de-Lima et al. 2019) can provide niches for different biocrust cyanobacteria. Seasonal climatic variation and soil texture (Abed et al. 2010, Yeager et al. 2012, Omari et al. 2022), in addition to soil acidity (Zhang et al. 2022) and soil fertility (Ochoa-Hueso et al. 2016), can all select for distinct bacterial and diazotrophic biocrust communities.

Although agroecosystem biocrusts could enhance crop growth through N-fixation (Sorochkina et al. 2022) and increase N and moisture retention in the subsurface soil of citrus agroecosystems (Nevins et al. 2020), there is a lack of understanding about their bacterial community diversity and composition. Furthermore, little is known about the influence of climate and management on these agroecosystem biocrust microbial communities. Bacterial community characterization in agroecosystem biocrusts would clarify the influence of management conditions and climatic factors on biocrust microbial communities and identify processes that contribute to the function of these agroecosystems.

In this study, we characterized bacterial and diazotrophic communities seasonally in biocrusts from two differently managed agroecosystems within a similar climate in the Eastern United States (a citrus orchard and a vineyard). Biocrusts were identified in the field with the aid of a visual development scale (Belnap et al. 2008b), where darker textured areas visible on the soil surface and also integrated with the soil particles (i.e. a piece of biocrust removed from the ground had soil particles remain attached to the bottom) were considered to be biocrusts. In contrast to biocrusts, bare soils were identified by the absence of surface roughness or dark coloration (Belnap et al. 2008b). Because of the different crop management requirements, biocrusts located in a vineyard and citrus orchard were expected to have different bacterial and diazotrophic community composition. In addition, we evaluated biocrust samples from an apple orchard in a semi-arid agroecosystem in the Western United States to contrast both climate and soil type among the different cropping systems. The vineyard and citrus orchard biocrust community compositions were expected to differ from the apple orchard because of soil and climatic conditions. Based on the results of Sorochkina et al. (2022), which demonstrated seasonal differences in N-fixation rates in the vineyard but not the citrus orchard, we hypothesized that biocrust bacterial and diazotrophic community composition would also only vary across seasons in the vineyard.

Materials and methods

Sample collection and acetylene reduction assays

Three agricultural sites with different perennial cropping systems known to contain biocrusts were selected for this study. The first two sites, a citrus orchard (“Citrus”, 28.115496, –81.713458) and a vineyard (“Grape”, 29.407195, –82.139980), were located at University of Florida research stations in Florida, United States, and had sandy soils classified as excessively drained Entisols of the Candler series with pH 6.0–6.5 (National Cooperative Soil Survey). The Citrus and Grape sites had a similar climate and soil type but differed in irrigation and fertilization management, with Citrus receiving nine times more water and four times more N than Grape (Sorochkina et al. 2022). The third crop, a commercial apple orchard (“Apple“) located in the steppe climate near Milton-Freewater, Oregon, United States, represented a contrasting climate and soil with cobbly loam soils and a pH of 7.0 (National Cooperative Soil Survey).

Samples were collected from six replicate experimental plots at the Grape and Citrus sites for three seasons. Samples were collected in August (Grape site) or September (Citrus site) 2019 (summer), November 2019 (fall) and May 2020 (spring). At the commercial Apple orchard, samples were collected in July 2020 (summer) and November 2020 (fall).

All samples were collected as described in Sorochkina et al. (2022). Briefly, pairs of samples (an intact biocrust and adjacent bare soil 10 cm away from biocrust) were randomly collected in each plot (or at different areas within the field for the Apple site), defined as an area within 120 cm from either side of the crop trunk or vine (plots n = 6 for a total of six biocrust and six bare soil pairs at each site). For Grape and Citrus, three subreplicates per field plot were collected for the acetylene reduction assay (ARA) in situ, which was used as a proxy for the N-fixation rate. For Apple, the samples for ARA were collected in a similar manner, but instead of being measured in situ, were sealed in Petri dishes and shipped overnight to the Soil Microbiology Laboratory at the UF/IFAS Southwest Florida Research and Education Center (Immokalee, FL, USA) and stored at 4°C until ARA measurement in an incubator. To avoid disturbance of the biocrusts, the three subreplicates collected for ARA were measured separately and the rates averaged. Grape and Citrus soils for ARA measurements were incubated in the field for 2 h, and gas samples were collected for further analysis in the laboratory, as described in Inglett et al. (2004). Apple soils for ARA measurements were shipped overnight on ice to the laboratory, where they were incubated for 2 h at 34°C (average field soil temperature during collection) at 9.97 Klux light intensity in a Percival incubator (Percival, Perry, IA, USA) and gas samples were collected for further analysis in the laboratory as described previously (Sorochkina et al. 2022).

An additional three subreplicates per field plot were collected for DNA analysis and pooled, placed on ice, transported to the Soil Microbiology Laboratory at the UF/IFAS Southwest Florida Research and Education Center (Immokalee, FL, USA), and stored at –80°C until analysis. A total of six paired biocrust and bare soil samples from each site were included for DNA analysis, with three time points collected from the Grape and Citrus sites and two time points collected for the Apple sites, for a total of 96 samples for DNA analysis.

Soil temperature and moisture

Soil surface temperature was measured for each field replicate using a thermocouple attached to a DIGI-SENSE 20250–02 temperature meter (Cole Palmer, Vernon Hills, IL, USA) at the Grape and Citrus sites, and using a Raytek Raynger ST infrared thermometer (Fluke, Everett, WA, USA) at the Apple site. After ARA measurements were conducted, the sub-replicates were pooled to measure gravimetric soil moisture (Sorochkina et al. 2022).

DNA extraction and sequencing

Total genomic DNA was extracted from 0.25 g of each pooled biocrust sample using the DNeasy 96 PowerSoil Pro QIAcube HT Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. Extracted DNA was quantified using the Qubit Fluorometer with the Quant-iT dsDNA HS Assay Kit (ThermoFisher, Wilmington, DE, USA). Briefly, the target genes were amplified in 20 µL of volume reaction containing 2 × GoTaq Master mix product, forward and reverse primers, and 2 µL of DNA, either undiluted or diluted 1:20, depending on the quantified DNA concentration. Primers and PCR thermocycling conditions for 16S rRNA and nifH genes are provided in Table 1. Two-stage PCR amplification using Fluidigm barcoding (Naqib et al. 2018), normalization, amplicon library preparation and sequencing was conducted by the University of Illinois at the Chicago DNA Services Facility (Chicago, IL, USA). Sequencing was performed on a MiSeq sequencer (Illumina, Inc., San Diego, CA, USA), where amplified products were loaded using a 500-cycle sequencing kit and the 2 × 250 paired end cycle sequencing mode. Sequences were submitted to NCBI SRA under BioProject PRJNA1026578.

Table 1.

16S rRNA- and nifH gene primers, as well as PCR conditions

Gene	Primer	Sequence (5′–3′)†	Primer conc. (µM)	PCR annealing temp. (°C), time (s)	Cycles	References
16S rRNA‡	515F (forward)	ACACTGACGACATGGTTCTACAGTGYCAGCMGCCGCGGTAA	0.2	50, 60	35	(Quince et al. 2011; Parada et al. 2016)
	926R (reverse)	TACGGTAGCAGAGACTTGGTCTCCGYCAATTYMTTTRAGTTT
nifH	IGK3 (forward)	ACACTGACGACATGGTTCTACAGCIWTHTAYGGIAARGGIGGIATHGGIAA	1	58, 30	32¶	(Ando et al. 2005, Gaby et al. 2017)
	DVV (reverse)	TACGGTAGCAGAGACTTGGTCTATIGCRAAICCICCRCAIACIACRTC

^†Linkers are shown in bold with CS1 linker for forward primer and CS2 linker for reverse primer.

^‡The thermocycling protocol was 94°C for 3 min denaturing followed by the specified number of cycles with 94°C melt for 45 s, specified annealing conditions, 72°C for 90 s and a final extension at 72°C for 10 min.

^§The thermocycling protocol was 94°C for 3 min denaturing followed by the specified number of cycles with 95°C melt for 30 s, specified annealing conditions, 72°C for 1 min and a final extension at 72°C for 10 min.

^¶Cycles were optimized for this study's samples and differ from the cycles used in the reference.

rRNA and nifH sequence processing

Processing of raw sequencing reads for 16S rRNA and nifH was conducted using QIIME2 v2018.8 (Bolyen et al. 2019). Reads were dereplicated with DADA2 using the paired-end option. Chimeras were then removed from the reads and the resulting sequences were trimmed where they fell below a quality score of 25 (Callahan et al. 2016). Rarefaction curves were performed in R using the ranacapa package (Kandlikar et al. 2018).

rRNA and nifH gene taxonomic assignment

For bacterial taxonomic classification of 16S rRNA gene sequences, an amplicon sequence variant (ASV) table was created, and taxonomic classifications were assigned using SILVA database version 138 with naïve Bayes classifier in QIIME2 (Bokulich et al. 2018). Sequences for mitochondria, Archaea, and chloroplasts were removed. For diazotroph taxonomic classification based on the nifH sequences, the phylogenetic tree and ASV sequences were exported and analyzed in ARB version 6.0.6 (Ludwig et al. 2004). Non-nifH sequence groups were identified and removed from downstream analyses using the following steps. First, nifH nucleotide sequences were translated to amino acid sequences, aligned with ClustalW and a new maximum likelihood protein tree with 100 bootstraps based on the GAMMA model was generated using RAxML version 8.2.12 (Stamatakis 2014). Next, sequences belonging to non-nifH clusters were removed. Finally, taxonomic classification of nifH ASVs was performed with BLASTn by aligning high-quality nifH reads with all existing nifH reads present in NCBI GenBank as of 2 June 2021, using BLAST+ and default parameters (Bokulich et al. 2018). Taxonomic assignment was performed to the order level.

Statistical analysis

Acetylene reduction rates in Grape and Citrus biocrusts across seasons and sites were analyzed using general linear mixed models using emmeans (Lenth et al. 2022) and nlme (Pinheiro et al. 2019) R packages. Acetylene reduction rates of biocrusts in Grape and Citrus were compared with those of bare soils within each site separately using emmeans (Lenth et al. 2022) and nlme (Pinheiro et al. 2019) R packages with a random effect for paired bare soil and biocrust samples added to the model. Apple acetylene reduction rates from the fall sampling were compared between biocrusts and bare soils using an unequal variance two-sample t-test in Microsoft 365 Excel (version 2206) using the Analysis ToolPak add-in. Sequence analysis was performed on all bacterial and diazotrophic data separately. R statistical software packages phyloseq (McMurdie and Holmes 2013) and vegan (Oksanen et al. 2013) were used to analyze alpha diversity, beta diversity, and construct taxonomic bar plots.

The relationship between nifH alpha diversity and acetylene reduction rates was analyzed by fitting a linear model to acetylene reduction rates and alpha diversity measures (observed, Shannon, and Simpson) separately for each site with the easynls R package (Arnhold 2017). Although it is possible that a nifH-containing organism whose N-fixation activity was not detected during the specific conditions at sampling could engage in N-fixing activity during other conditions, the goal was to plot the relationship only for samples with a verifiable N-fixation function. Therefore, only biocrust and bare soil samples that both contained nifH and had detectable acetylene reduction rates were included. There was only one biocrust and one bare soil sample with no detectable nifH out of a total of 96 samples. All other samples were removed due to being an outlier, of low quality, or a nifH homolog. In total, four biocrust and three bare soil nifH replicates were removed. No more than one replicate was removed per season, time point, or biocrust/bare soil combination (n = 5). The total number of nifH samples before removal was 96, but was reduced to 89 after filtering. Undetected N-fixation activity of soils was more common for our samples than undetected nifH sequences. Out of 90 samples where N-fixation was measured, 21 had no detectable N-fixation activity, resulting in a total of 69 samples. While an organism could contain the nifH gene, the conditions during the measurement of ARA could have been suboptimal for N-fixation to occur. All acetylene reduction rate values were natural log-transformed to improve normality and 1 was added to prevent negative values.

The effects of biocrust presence and site on alpha diversity (observed, Shannon, and Simpson) were analyzed with Wilcox (two comparisons) or Kruskal–Wallis (more than two comparisons) tests separately within each season. When significant differences were found using Kruskal–Wallis tests, Dunn tests were used for post hoc analyses. Biocrust presence, site, seasonal influence on the community composition and their interactions were analyzed by permutational multivariate analysis of variance (PERMANOVA) (Anderson 2017) with the vegan R package using the weighted Unifrac distance matrix. Bacterial 16S rRNA gene amplicon sequences were analyzed separately from nifH gene amplicon communities. One outlier was removed from Grape biocrust spring samples (n = 5) for alpha diversity analysis. Cyanobacteria and Proteobacteria community composition data subsets were created from the total bacterial community composition dataset by keeping only ASVs that were classified as Cyanobacteria or Proteobacteria. The community composition of biocrusts and bare soils were compared separately and jointly for each site for the total bacterial (16S rRNA genes), Cyanobacteria (16S rRNA genes), Proteobacteria (16S rRNA genes), and diazotroph (nifH genes) composition. Principal coordinate analyses (PCoA) were utilized for visualization of the biocrust community composition for each sequence type. To determine whether there were seasonal or site differences in biocrust community composition, pairwise PERMANOVA was performed using the wrapper function pairwise.adonis for multilevel pairwise comparison with Benjamini–Hochberg P-value correction for multiple comparisons (Martinez Arbizu 2020). Multivariate homogeneity of group variances was tested with the betadisper function with a permutation test and HSD post hoc test using the vegan package (Anderson et al. 2006).

Bacterial and diazotrophic relative abundances at the order level were compared using phyloseq and DESeq2 (Love et al. 2014) R packages. Relative abundances of ASVs were compared across sites (Grape, Citrus, Apple) with the contrast function and Wald test where alpha was equal to 0.01. Cyanobacterial relative abundances at the genus level were compared between sites using a Kruskal–Wallis test with Benjamini–Hochberg P-value correction for multiple comparisons.

The individual effects of environmental factors, nutrient concentrations, and acetylene reduction rates on total bacterial, cyanobacterial, proteobacterial, and diazotrophic community composition were tested separately for each site with Spearman rank correlations using simple Mantel tests (Borcard et al. 2011). Nutrient concentration data were used from a previously published study from the same samples collected in Grape and Citrus (Sorochkina et al. 2022) and included salt extractable carbon (EC), salt extractable N (EN), salt extractable phosphorus (EP), microbial biomass carbon (MBC), microbial biomass N (MBN), and their ratios (EC:EN, EC:EP, EN:EP, MBC:MBN). Variables were transformed by either taking the log and adding 1 or the square root to improve normality and homogeneity and then converted into Euclidean dissimilarity matrices. Bacterial, cyanobacterial, proteobacterial, and diazotrophic abundances of high-quality processed sequences were converted into Bray–Curtis dissimilarity matrices. The Euclidean matrix of each variable was then correlated against the Bray–Curtis matrix of community composition. For the Mantel analysis, summer, fall, and spring seasons were included in Grape and Citrus, while only fall was included in Apple because this was the only time point with measured acetylene reduction rates.

The protein tree generated in ARB was used to create the final nifH tree figure in Microsoft 365 PowerPoint software (version 2206). Rarefaction, alpha diversity, linear regression, beta diversity, and differential abundance plots were created in R statistical software using the ggplot2 package (Wickham 2016). Acetylene reduction rates and standard error bars were plotted using Microsoft 365 Excel (version 2206).

Differential abundance results were considered significant at P < 0.01 because of high sensitivity to differences. All other results were considered significant and of relative importance for the community only when both of the following conditions were satisfied: P < 0.05 and R² ≥ 0.1.

Results

Acetylene reduction rates

Biocrusts at all sites had significantly higher acetylene reduction rates than adjacent bare soils (Fig. 1, general linear mixed model and t-test). Grape biocrusts had the highest average acetylene reduction rates followed by Citrus and Apple. Seasonal differences in acetylene reduction rates within biocrusts were significant only in Grape (Fig. 1, general linear mixed model). Grape biocrust acetylene reduction rates ranged from undetectable to 401 µmol m⁻² h⁻¹, with the highest average rate in summer (260 µmol m⁻² h⁻¹) and the lowest in spring (4.19 µmol m⁻² h⁻¹) (Fig. 1). Acetylene reduction rates of Citrus biocrusts ranged from undetectable to 326 µmol m⁻² h⁻¹, with the highest average rate in spring (137 µmol m⁻² h⁻¹) and the lowest in summer (13 µmol m⁻² h⁻¹). Acetylene reduction rates of Apple biocrusts ranged from 2.7 to 42.3 µmol m⁻² h⁻¹ in the fall.

Figure 1.

Acetylene reduction rates are modified from Sorochkina et al. (2022) to include Apple acetylene reduction rates. Some bare soils bars are not visible due to low or zero values of acetylene reduction rates. An asterisk above the biocrust bar within a season indicates a significant difference between the biocrusts and bare soils for a given sampling date, while letters indicate a significant difference between biocrusts of different seasons (*P < 0.05; n = 6, mean ± SE).

Biocrust soil moisture

During the summer, Grape biocrust average (n = 6) soil moisture was the highest (47.2%, 536 times higher than in Apple), followed by Citrus biocrust (2.61%, 30 times higher than in Apple) and Apple biocrust (0.088%). During the fall, Grape biocrust average (n = 6) soil moisture was the highest (2.49%, 16 times higher than in Apple), followed by Citrus biocrust (0.753%, five times higher than in Apple) and Apple biocrust (0.156%). Soil moisture comparisons were not made for the spring season because Apple biocrusts were not collected during spring (collected only for DNA during fall and collected both for DNA and ARA during summer).

Amplicon sequencing results

Amplicon sequencing of 16S rRNA genes resulted in 4,300,675 reads (an average of 44,337 per sample). Quality trimming resulted in 1,261,820 remaining reads (an average of 13,008 per sample) and 23,219 ASVs. Amplicon sequencing of nifH genes resulted in 2,247,003 nifH reads (an average of 22,929 per sample), with 531,834 reads (an average of 5,581 per sample) after quality filtering. The majority of biocrust nifH sequences were from the aerobic cluster 1, while the majority of nifH homologs were from cluster 5 (Supplementary Figure 1). In biocrusts, cluster 1 was dominated by cyanobacterial nifH sequences, while cluster 5 was dominated by either cyanobacterial or proteobacterial chlorophyll, as nifH homologs can be found in eukaryotic or prokaryotic chlorophyll (Suzuki and Bauer 1992, Fujita et al. 1992, Nomata et al. 2006) (Supplementary Figure 1). After removing nifH homologs, 2,264 nifH ASVs remained. Three samples did not yield high quality sequences and three samples only contained nifH homologs and were therefore removed from further analysis (one Grape biocrust from fall, one Citrus biocrust from spring, one Apple biocrust from summer, and three bare soils). Rarefaction curves confirmed that observed ASVs reached a plateau in all samples (Supplementary Figure 2).

Comparison of biocrust and bare soil bacterial diversity and composition

Total bacterial diversity and composition were not significantly different between biocrusts and bare soils during the summer at all sites (Supplementary Table 1, Kruskal–Wallis and Wilcox, R² = 0.02, P < 0.05). There was no significant interaction between site and biocrust presence (PERMANOVA, R² = 0.03, P < 0.05).

The cyanobacterial and proteobacterial biocrust community composition did not significantly differ from bare soils with all sites combined (Supplementary Figure 3A and 3B; PERMANOVA, R² = 0.04, 0.07, respectively, P < 0.05). However, there was a significant interaction between site and biocrust presence with significant differences cyanobacterial and proteobacterial community composition for Grape (Supplementary Figure 4A and 4B, PERMANOVA, R² = 0.16, 0.19, respectively, P < 0.05), Citrus (Supplementary Figure 4A and 4B, PERMANOVA, R² = 0.11, 0.15, respectively, P < 0.05), and Apple (Supplementary Figure 4A and 4B, PERMANOVA, R² = 0.13, 0.10, respectively, P < 0.05). Although non-homogeneous dispersion could be affecting PERMANOVA biocrust and bare soil comparisons of cyanobacterial and proteobacterial community composition, significant clustering (betadisper, P < 0.05) according to biocrust presence was confirmed by PERMANOVA results using at least two distance matrices (data not shown; unweighted Unifrac, weighted Unifrac, and Bray–Curtis).

Comparison of biocrust and bare soil diazotroph diversity and composition

Diazotroph diversity showed a different trend than bacterial diversity, as biocrust diazotroph diversity was significantly greater than diversity in bare soils for specific sites and seasons. In Citrus, biocrust diazotroph communities were significantly more diverse than bare soils only during fall and spring (Fig. 2A, Supplementary Table 1, Kruskal–Wallis). In Grape, biocrust diazotroph communities were significantly more diverse than bare soils during the summer (Fig. 2A, Kruskal–Wallis). In Apple, there were no significant differences in diversity between apple biocrusts and bare soils during the summer and fall (Fig. 2A, Wilcox). At all sites, biocrust diazotroph community composition was significantly different from bare soils (Supplementary Figure 3C, PERMANOVA, R² = 0.10, P < 0.05) without a significant biocrust/bare soil × season × site interaction. Acetylene reduction rates were significantly and positively correlated with alpha diversity in Citrus (Fig. 2B, linear model), but not Grape (P = 0.075).

Figure 2.

Diazotrophic alpha diversity and scatterplot of alpha diversity plotted against acetylene reduction rates of biocrusts and bare soils. (A) Shannon diversity of biocrust and bare soil bacteria. A red asterisk (*) in alpha diversity indicates a significant difference in alpha diversity between biocrusts and bare soils (P = 0.05), while # indicates a marginal difference (P = 0.056). One outlier was removed from the Grape biocrust spring samples (n = 5). (B) The acetylene reduction rate of each sample was log transformed and 1 was added to eliminate any negative values. The sites are color coded, while shapes represent seasons. A linear model with 95% confidence interval shaded in gray was fitted to each scatterplot. Fitted linear model probability was P < 0.05. Samples with undetected acetylene reduction rates are not included in (B).

Comparison of bacterial biocrust communities across sites and seasons

Biocrust total bacterial community diversity was significantly greater in Citrus compared with Apple during the summer (Supplementary Figure 5A, Supplementary Table 1, Kruskal–Wallis). However, there was no significant difference in biocrust total bacterial community diversity between any of the sites during the fall and spring (Supplementary Table 1, Kruskal–Wallis and Wilcox).

Biocrust total bacterial community composition was significantly different between sites (Fig. 3A, PERMANOVA, P < 0.05; R² = 0.13) without a significant site × season interaction. Grape and Citrus biocrusts had both shared and unique contributing factors to the total community composition, and their total community composition was significantly correlated with acetylene reduction rates (Table 2, Mantel Spearman rank correlation). In addition, soil moisture, EP and MBC:MBN were significantly correlated only with Grape biocrust total community composition, while soil temperature and EC:EP were significantly correlated only with Citrus biocrust total community composition (Table 2, Mantel Spearman rank correlation).

Figure 3.

PCoA of weighted Unifrac distance matrix of biocrust community composition across seasons in Grape, Citrus, and Apple crops. (A) bacterial, (B) cyanobacterial, (C) proteobacterial, and (D) diazotrophic communities significantly differed between all crops (PERMANOVA, P < 0.05, pairwise PERMANOVA, P < 0.05). Each point represents a single replicate (n = 6). High-quality nifH sequences were not detected in three biocrust samples (one in Grape, one in Citrus, and one in Apple).

Table 2.

Mantel Spearman rank correlations (R values) and significance between soil variables and bacterial community composition in Grape, Citrus, and Apple. Soil variables and acetylene reduction rates for Grape and Citrus are published in Sorochkina et al. (2022).

Site	Variable†	Bacteria (16S rRNA)‡	Cyanobacteria (16S rRNA)	Proteobacteria (16S rRNA)	Diazotrophs (nifH)
Grape	Moisture (+)	0.26**	0.09	0.27**	−0.11
	Temperature	0.13	0.24*	0.03	0.22*
	Acetylene reduction rate	0.44**	0.38**	0.33**	−0.10
	EC	0.14	0.09	−0.02	−0.00
	EN	0.02	0.02	−0.12	0.02
	EP (+)	0.23*	0.11	0.28*	−0.21
	EC:EN (−)	0.02	0.16	−0.06	−0.02
	EC:EP	0.06	−0.04	0.05	0.02
	EN:EP (+)	0.08	0.01	−0.04	0.13
	MBC	0.06	0.17	0.12	0.30
	MBN	−0.02	0.04	-0.01	0.26
	MBC:MBN (−)	0.32*	0.28	0.00	0.11
Citrus	Moisture	0.05	0.02	0.15	−0.10
	Temperature (−)	0.29*	0.01	0.17	−0.18
	Acetylene reduction rate	0.36**	0.19*	0.32**	0.09
	EC	0.10	−0.03	0.11	0.08
	EN	0.02	−0.13	−0.02	0.02
	EP	−0.03	0.01	0.01	−0.21
	EC:EN (+)	0.12	0.08	0.22*	−0.10
	EC:EP	0.21*	0.04	0.23*	-0.02
	EN:EP	0.09	0.09	0.08	−0.09
	MBC	0.13	0.01	0.04	0.04
	MBN	0.09	0.01	0.12	0.06
	MBC:MBN (+)	−0.07	0.27	0.04	−0.03
Apple	Moisture	−0.03	−0.15	−0.72	−0.17
	Acetylene reduction rate	−0.18	−0.13	−0.28	−0.15

^†Soil variables identified in Sorochkina et al. (2022) that were either positively (+) or negatively (−) correlated to acetylene reduction rates.

^‡Asterisks indicate significance at P < 0.01 (**) or P < 0.05 (*).

Similar to the overall total community composition, the biocrust cyanobacterial community composition significantly differed between all sites (Fig. 3B, PERMANOVA, P < 0.05, R² = 0.42), without a significant site × season interaction. Cyanobacterial communities of Grape and Citrus biocrusts were also significantly correlated with acetylene reduction rates (Table 2, Mantel Spearman rank correlation). Soil temperature was significantly correlated with biocrust cyanobacterial community composition in Grape (Table 2, Mantel Spearman rank correlation).

Biocrust proteobacterial community composition significantly differed between sites (Fig. 3C, PERMANOVA, P < 0.05; R² = 0.26). Although non-homogeneous dispersion (betadisper, P < 0.05) could be affecting the PERMANOVA results of biocrust proteobacterial community composition, significant clustering according to site was confirmed by PERMANOVA results with three different distance matrices (unweighted Unifrac, weighted Unifrac, and Bray–Curtis, P < 0.05). Grape and Citrus biocrusts had both shared and unique influences on the proteobacterial community composition and their proteobacterial community compositions were significantly correlated with acetylene reduction rates (Table 2, Mantel Spearman rank correlation). However, EP and soil moisture were significantly correlated only with the Grape biocrust proteobacterial community composition, while EC:EN and EC:EP were significantly correlated only with the Citrus biocrust proteobacterial community composition (Table 2, Mantel Spearman rank correlation). Apple biocrust total community and diazotrophic community compositions were not significantly correlated with acetylene reduction rates or soil moisture (Table 2, Mantel Spearman rank correlation, P > 0.05).

As season did not have a significant influence on biocrust whole-community composition, further comparisons were performed between sites only for the fall season. Grape and Apple biocrusts had significantly different relative abundances of bacteria in 20 classes from nine phyla (Fig. 4A, Wald test, P < 0.01). Citrus and Apple biocrusts had significantly different relative abundances of bacteria in 26 classes from 12 phyla (Fig. 4A, Wald test, P < 0.01). Grape and Citrus biocrusts had significantly different relative abundances of bacteria in 30 classes from 12 phyla (Fig. 4A, Wald test, P < 0.01).

Figure 4.

Differential abundance of bacteria in biocrusts during the fall between Grape, Citrus, and Apple. Each point represents a single ASV that significantly differed between sites. The size of the dot represents the mean (log) relative abundance of the bacterial ASV. Bacteria were identified with (A) 16S rRNA and (B) nifH. Every class or order labeled on the y-axis shows significant log2fold change (P < 0.01). Phyla or classes are color coded, while unidentified classes or orders are in light gray (A) or dark gray (B).

Comparison of diazotrophic biocrust communities across sites and seasons

In comparison with the total bacterial biocrust diversity, which was only significantly different between sites during the summer, biocrust diazotrophic diversity during the fall was significantly greater in Citrus than in Apple (Supplementary Figure 5B, Kruskal–Wallis, P < 0.05). No significant differences in the diazotrophic diversity of biocrusts were detected between sites during the summer and spring. Site had a significant effect on biocrust diazotrophic community composition (Fig. 3D, Kruskal–Wallis, P < 0.05, R² = 0.29), regardless of the season. Soil temperature significantly influenced the diazotrophic community composition in Grape, but none of the measured variables impacted the diazotrophic community composition in Citrus (Table 2, Mantel Spearman rank correlation). Biocrust diazotrophic community composition did not differ significantly between seasons within each site, therefore further comparisons were made within the fall season.

ASVs from the Caryophanales order were the most abundant in Grape biocrusts when compared with Citrus and Apple. However, ASVs from the orders Synechococcales (including Leptolyngbya boryana), Oscillatoriales (including Oscillatoria), Chroococcidiopsidales (including Chroococcidiopsis thermalis PCC 7203), and Hyphomicrobiales had higher relative abundances in Citrus biocrusts compared with Grape and Apple (Fig. 4B, P < 0.01). Orders Pleurocapsales (including Pleurocapsa sp. PCC 7327) and Chroococcales (including Cyanothece sp. PCC 7822) were only present in Citrus (Supplementary Figure 6B). Chroococcidiopsidales and Oscillatoriales were absent in Apple, but present in Grape and Citrus (Supplementary Figure 6B). The biocrust diazotrophic community during the fall season was dominated by Nostocales in Grape (66% relative abundance), Citrus (65%), and Apple (53%), followed by Synechococcales in Grape (2%) and Citrus (24%), and Caryophanales order in Apple (0.86%) (Supplementary Figure 6B). A large portion of diazotrophs was unclassified at the order level in Apple (44%), Grape (29%), and Citrus (7%, Supplementary Figure 6B).

Differences in the relative abundances of biocrust diazotrophic organisms between sites detected with nifH gene were supported by 16S rRNA gene data. Cyanobacterial orders Synechococcales, Oscillatoriales and Nostocales (Fig. 4B, Supplementary Figure 7), Caryophanales order from Bacilli class (Fig. 4B, not shown for 16S rRNA), and Hyphomicrobiales order from Alphaproteobacteria class (Fig. 4B, Kruskal–Wallis, not shown for 16S rRNA) were differentially abundant between sites in both 16S rRNA genes and nifH data. While the potentially diazotrophic Betaproteobacteria class (Pepe-Ranney et al. 2016) was differentially abundant in 16S rRNA data, this class was not differentially abundant in nifH data.

Discussion

Distinct microbial communities were present in biocrusts of a citrus orchard and vineyard, as confirmed by the significantly different bacterial (Supplementary Figure 4A and 4B) and diazotrophic composition (Supplementary Figure 3C), higher diazotrophic diversity, and higher acetylene reduction rates (Figs 1 and 2b) compared with bare soils. The microbial communities in the apple orchard samples were also confirmed to be biocrusts as they had significantly higher acetylene reduction rates and significantly different diazotrophic community composition than bare soils (Fig. 1, Supplementary Figure 4C). These agroecosystem biocrusts were structurally and functionally similar to natural ecosystem dark cyanobacterial biocrusts (Yeager et al. 2004, Belnap et al. 2008b, Colesie et al. 2016), with only 2%–21% of the diazotrophs from non-cyanobacterial taxa, and all diazotrophs (with the exception of two single strains in Grape), belonging to aerobic diazotrophs of cluster I (Supplementary Figure 1). While physical disturbance and stochastic ecological succession in desert ecosystems can reduce acetylene reduction rates and transition dark cyanobacterial biocrusts to light cyanobacterial biocrusts (Belnap 1996, 2002b, Belnap and Eldridge 2001, Yang et al. 2018), the primary disturbances of these perennial agroecosystems (herbicides, fertilizer, and limited foot traffic) may not have been sufficient to result in selection for light biocrusts. In addition, the irrigation of these systems may have allowed for faster recovery than seen in natural ecosystems, as moisture is often a limiting factor in biocrust expansion in dryland ecosystems (Barger et al. 2016, Chamizo et al. 2016).

The bacterial community composition of agroecosystem biocrusts in this study was significantly different between sites, but 87% of taxonomic bacterial groups at the class level were present in all biocrusts (Fig. 4A). The similarity in biocrust composition could be related to their location in agroecosystems, where disturbances due to fertilization and herbicide applications could select for the presence of certain biocrust taxa. However, certain taxa were only found in biocrusts near apple and citrus crops (phylum Chlamydiota, Fig. 4A), or only citrus crops (cyanobacteria Pleurocapsa, Supplementary Figure 6B), which could be due to several factors including climate and edaphic factors such as soil type and agricultural management differences.

Unlike the more arid environment of the Apple site, the Citrus and Grape sites were both located in a mesic environment with a similar soil type. Apple biocrusts also had at least five times lower soil moisture than biocrusts from the Citrus and Grape sites, further supporting the classification of the Apple site as arid. Climatic factors such as temperature (Giraldo-Silva et al. 2020) and precipitation (Machado-de-Lima et al. 2019, Su et al. 2020) can provide niches for different biocrust cyanobacteria within the same biome. The larger pore spaces of coarser soil can prevent cyanobacterial attachment to soil particles leading to slower biocrust growth and more disjointed patches (Rozenstein et al. 2014). In addition to climate, biocrust community composition differences can also be stronger between sites that are geographically distant due to dispersal limitations of filamentous biocrust cyanobacteria (Warren et al. 2019, Su et al. 2020, He et al. 2024). The cyanobacterial genus Leptolyngbya, which tended to be more abundant in Citrus and Grape than in Apple, has increased abundance in natural ecosystem regions with higher precipitation (Machado-de-Lima et al. 2019), thus the significant differences in precipitation between the mesic and arid agroecosystem biocrusts could contribute to these differences in Leptolyngbya and other bacterial groups. However, Leptolyngbya is a ubiquitous organism capable of withstanding disturbance and extreme conditions (Taton et al. 2006, Podda et al. 2014, Roncero-Ramos et al. 2019, 2020) and can also become dominant under higher nutrient concentrations of laboratory cultivation Giraldo-Silva et al. 2019). Thus, it is difficult to pinpoint the reason for the higher abundance of Leptolyngbya in Citrus and Grape rather than Apple. Additional studies of biocrusts in a variety of agroecosystems are necessary to better assess the key factors contributing to differences in microbial community composition.

Crop management also likely impacted biocrust bacterial community composition, as there were significant differences between Citrus and Grape, despite both being from a similar climate and soil type. Differences in irrigation management impacted soil moisture (Sorochkina et al. 2022), which influenced the bacterial community composition of Grape and Citrus. Citrus had a more consistent and higher intensity irrigation than Grape (34 vs 3.8 L per day) (Sorochkina et al. 2022), potentially reducing the influence of moisture fluctuations on the bacterial community composition. By contrast, the lower intensity irrigation in Grape likely increased microbial reliance on precipitation, as suggested by the positive correlation between the Grape bacterial community composition and soil moisture (Table 2). The crop-dependent relationship between nutrients (N:P, EC:EN, EP) and proteobacterial community structure (Table 2) provides evidence that fertilizer composition also has the potential to influence biocrust proteobacterial community structure.

Soil temperature, in addition to other unmeasured variables, could be influencing the diazotrophic community composition. Acetylene reduction rates of Grape and Citrus biocrusts positively correlated with bacterial, proteobacterial, and cyanobacterial community composition, but not diazotrophic community composition (Table 2). Conversely, only soil temperature correlated with the diazotrophic community composition of Grape biocrusts, but not acetylene reduction rates. Temperature could be impacting metabolic processes other than acetylene reduction in Grape, such as photosynthesis or respiration, with increases in microbial biomass being the result of those changes. However, it is not clear why soil temperature and nutrient concentrations that influenced Citrus acetylene reduction rates (Sorochkina et al. 2022) did not also influence diazotrophic community composition. Citrus diazotroph survival could be more resilient than Grape due to the longer time of biocrust development (2 years vs 2 months) and the lack of large moisture and fertilizer fluctuations in Grape. Citrus biocrust diazotrophs might have also been more sensitive to other parameters that were not measured, such as soil pH, nutrients, and metals. In addition, amplicon sequencing does not always reflect changes in the abundance of live organisms because relic DNA also has the potential to be sequenced (Carini et al. 2017)

Crop development time may have contributed to structuring the diazotrophic community composition of Grape and Citrus. While both Grape and Citrus biocrusts can be considered dark biocrusts due to the overall dominance by heterocytous cyanobacterial diazotrophs, they had differentially abundant specific diazotrophic taxa. Grape biocrusts had a higher abundance of heterotrophic diazotrophs from the Caryophanales (Bacillales) order, while Citrus biocrusts had a higher abundance of filamentous Oscillatoriales-affilliated diazotrophs (Fig. 4B). Non-cyanobacterial heterotrophic diazotrophs from the Bacillales order can grow in a biocrust in 1 month with the aid of a single soil binding species (Rozenstein et al. 2014, Mugnai et al. 2020). However, filamentous cyanobacterial diazotrophs are more characteristic of biocrusts that have more time to develop, as filamentous N-fixing cyanobacteria have a slower growth rate (3–4 months to achieve a culture) than heterotrophic diazotrophs (20 days to achieve a culture) (Giraldo-Silva et al. 2019, Nelson et al. 2021). Citrus biocrust diazotrophic community composition, therefore, might differ from that of Grape because of the longer time of crop growth that allowed for a longer duration of biocrust development (2 years vs 2 months).

Not surprisingly, N-fixation activity was more sensitive to season than the diazotrophic community composition. Diazotroph N-fixation activity is energetically costly, requires specific environmental conditions, and is strongly suppressed when not needed (Vitousek et al. 2013). For example, insufficient moisture (Sharma and Singh 2020), nutrient deficiency, low N demand (Liao et al. 2014), or ammonium inhibition (Yan et al. 2010) could all contribute to the seasonal decline in acetylene reduction rates of the vineyard, without being reflected by the diazotrophic community composition. Diazotrophs can utilize available N and survive major environmental changes that could exist in a managed agroecosystem with fluctuating fertilizer inputs (Menge et al. 2015). As we previously mentioned, amplicon sequencing does not always reflect changes in the abundance of live organisms because relic DNA also has the potential to be sequenced (Carini et al. 2017).

As bacterial community composition can be sensitive to nutrient addition, fertilization was expected to be a potential influencing factor of biocrust total bacterial community composition. Nitrogen additions significantly decreased cyanobacterial relative abundance by ∼2%–30% (Ayuso et al. 2017) and biomass by ∼25% (Zhou et al. 2016) in hot deserts. As expected, the average relative abundance of cyanobacteria in the agroecosystems of our study (14%) was lower than in unfertilized natural ecosystem biocrusts from drylands (25%–70%) (Ayuso et al. 2017, Pombubpa et al. 2020, Fernandes et al. 2022). Conversely, while Alphaproteobacteria, Betaproteobacteria, and Gammaproteobacteria were the most dominant in agroecosystem biocrusts, they are usually the second most dominant phylum in natural ecosystem biocrusts from an arid climate (Nunes da Rocha et al. 2015, Moreira-Grez et al. 2019). Nitrogen fertilizers also increased the relative abundance of proteobacteria in agricultural soils, cultivated biocrust, and temperate forest (Ayuso et al. 2017, Castellano-Hinojosa et al. 2020, Xia et al. 2020). The addition of P has also been shown to increase proteobacteria in biocrusts (Ayuso et al. 2017, Xia et al. 2020). Thus, proteobacteria may have outcompeted cyanobacteria if they were more suited to take advantage of fertilization. Proteobacteria contain a diverse set of organisms that might have distinct responses to the different fertilizer regimes of Citrus and Grape, and we hypothesize that this could explain why proteobacterial community composition significantly differed between Grape and Citrus.

Our results reflected that identification of diazotrophs using nifH gene sequencing still lags behind bacterial identification using 16S rRNA gene sequencing. While all bacterial taxa in the 16S rRNA gene data could be classified at least at the phylum level, 7%–44% of diazotrophs could not be classified to the phylum level in the nifH data (Supplementary Figure 6B). The difficulty of diazotrophic classification and comparison with other studies stems from a lack of protocol standardization like the Earth Microbiome Project for 16S rRNA genes (Gilbert et al. 2018) and lack of primers that encompass the large natural diversity of nifH genes, hence, many primers have biases towards particular phylogenetic lineages (Gaby et al. 2017, Angel et al. 2018). We chose a potentially universal nifH primer set (IGK3/DVV) with the least amount of documented bias, whose wide coverage detects a large number of thus far unclassified nifH genes, including those in cyanobacteria (Gaby and Buckley 2012). However, the broader coverage of diazotrophs in the universal primers also captures nifH homologs, which can potentially lead to overestimation of diazotroph diversity and incorrect representation of community structure. Additional sequence data processing steps are required to remove nifH homologs (Gaby et al. 2017, Angel et al. 2018). In our study, approximately one-half of the nifH amplicon sequences were homologs affiliated with cluster 5 (Supplementary Figure 1). These included genes for photosynthetic machinery in eukaryotes, cyanobacteria, and proteobacteria (Young 2005), highlighting the importance of careful nifH sequence postprocessing. The high number of unclassified diazotrophs in our agroecosystem biocrusts detected in our study could therefore be due to improved recovery of nifH variants of previously unidentified diazotrophs using more inclusive primer sets. Thus, these biocrusts are good candidates for the additional cultivation, isolation, and identification of potentially novel N-fixing cyanobacteria.

Conclusions

Although eutrophic conditions and agricultural disturbances may have resulted in compositional differences, the agroecosystem biocrusts investigated here potentially share functional and taxonomic characteristics with previously published descriptions of natural ecosystem biocrusts. The studied agroecosystem biocrusts had a lower percentage of cyanobacteria (14%) than natural ecosystem dark cyanobacterial biocrusts (> 25%), most likely due to fertilization and disturbance. However, agroecosystem biocrust diazotroph communities were dominated by Nostocales, as is typical for natural ecosystem biocrusts. The dominance of phototrophic diazotrophs is in line with previous N input estimates from N-fixation measurements of 8.1 kg N ha⁻¹ year⁻¹ (Citrus) and 4.9 kg N ha⁻¹ year⁻¹(Grape) in Sorochkina et al. (2022) that were measured during the day when only phototrophic diazotrophs fix N.

A combination of climatic, edaphic and management factors may have shaped bacterial and diazotrophic biocrust communities in these three distinct agroecosystems. Differences in fertilization and irrigation management appear to have selected for different microbial communities in the Grape and Citrus biocrusts. Climate, management, and soil type contributed to distinguishing Apple biocrust microbial communities from those of Grape and Citrus. However, season did not influence the bacterial and diazotrophic community composition of Grape, Citrus, or Apple biocrusts, suggesting that agricultural management practices could be dampening the effects of seasonal temperature and rainfall changes. Further diazotroph culturing and sequencing studies are needed to fully characterize the agroecosystem diazotrophic community composition. In addition, field studies of biocrusts located in other crops and agroecosystems, as well as controlled laboratory or greenhouse experiments, should be conducted to tease apart the influence of crop type, irrigation, fertilization, and herbicide application on the microbial community and N-fixation activity of agroecosystem biocrusts. Such efforts would help synchronize crop management with biocrust's ability to enhance soil fertility and mitigate N fertilizer costs.

Author contributions

Kira Sorochkina (Conceptualization, Formal analysis, Investigation, Methodology, Software, Visualization, Writing – original draft, Writing – review & editing), Willm Martens-Habbena (Data curation, Formal analysis, Resources, Software, Supervision, Validation, Writing – review & editing), Catherine L. Reardon (Conceptualization, Formal analysis, Investigation, Methodology, Project administration, Resources, Validation, Writing – review & editing), Patrick W. Inglett (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Writing – review & editing), and Sarah L. Strauss (Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Writing – review & editing)

Conflict of interest

We declare no conflicts of interest.

Funding

This work was supported by the United States Department of Agriculture National Institute of Food and Agriculture and Food Research Initiative grant (2018-67019-27707).