Impact of an intercropped melon/cowpea system on the coupling between soil bacterial community structure and chemical properties

A greater understanding of the relationship between soil microorganisms and intercropping systems could contribute to the optimization of land use, fostering sustainable and ecient agriculture. This study entails a comparative intercropping assay using cowpea (Vigna unguiculata) and melon (Cucumis melo) under organic management with different patterns and 30% less organic fertilization than that used in monocrops in the rst year. The intercropping system changed the bacterial community structure independently of the intercropping pattern and contributed to an increase in soil nitrogen, phosphorous content and melon crop yield. The intercropped systems were characterized by a higher abundance of Pseudomonas (Proteobacteria), which are related to nutrient cycling, and other benecial microorganisms.


Introduction
Intercropping is a practice involving the simultaneous growing of two or more crops on the same land during the same growing season 1 . Nowadays this practice is becoming increasingly important for maintaining and increasing soil quality and subsequently crop productivity 2 . Intercropping has demonstrated advantages, including e cient nutrient acquisition; reduced pest, disease and weed damage; improved microbial diversity; and improved utilization of land resources 3 . Different types of intercropping and combination systems have been proposed to date. However, not all intercropping systems constitute improvements. It is particularly important to not use crops that compete for physical space, nutrients, water, or sunlight, and the environmental conditions in a given area and the crops or varieties available must be taken into account 4 . Maize is one of the predominant intercrops used, often combined with legume crops 5 . This combination makes it possible to develop an energy-e cient and sustainable system, as the legumes have an N-xing capability and more protein-yielding potential in the form of either grain or forage 6 . In arid environments, the legume crop cowpea (Vigna unguiculata L. Walp) is normally used because it is well adapted to drought and low fertility and can improve legume nitrogen uptake by nodulation 7 . Therefore, it can be intercropped not only with maize, but also millet, sorghum, and some other crops 8,9 . Melon (Cucumis melo L.) is the main export crop in the region of Murcia (57%). Intensive melon cultivation can generate soil and water degradation due to the excessive use of pesticides to reduce the impact of pathogens and the necessary application of synthetic fertilizers due to nutrient depletion 10 .
Intercropping melon and cowpea could provide an important contribution to overcoming the challenges of developing both productive and environmentally friendly agricultural systems for melon cultivation.
Soil microorganisms are key drivers of many soil biological, chemical, and physical processes, such as soil structure formation, nutrient cycle, organic matter turnover, toxin accumulation or removal, and soilborne pathogen suppression 11,12 . Several studies have investigated the changes in the microbial characteristics of soils caused by intercropping 13,14 . However, changes in the soil microbial community resulting from melon-cowpea intercropping have not been studied in depth.
In this paper, we have studied changes in the soil bacterial community resulting from three different types of melon-cowpea intercropping systems and the relationship between these changes and soil chemical properties and crop yield compared to monoculture systems. We hypothesized that intercropping would improve crop yield, increase soil bacterial diversity, and change the soil community structure, and that this change would correlate with soil properties.

Experimental design and sampling
An intercropping experiment with melon and cowpea was performed under organic conditions in La Palma (Cartagena) (37º 41´18´´N 0º 56´60¨ W), a province of Murcia (S.E. Spain), in May-August 2018.
The eld trial was conducted in a soil that had been uncultivated for at least the last ve years prior to the study; the soil was classi ed as Haplic Calcisol (loamic, hypercalcic) IUSS (2015). The climate in the area of sturdy is semiarid Mediterranean, with a mean annual temperature of 18ºC, a mean annual precipitation of 275 mm and an annual potential evapotranspiration of 900 mm.
The assayed treatments were as follows: i) melon (Cucumis melo) monocrop (M); ii) cowpea (Vigna unguiculata) monocrop (C); iii) mixed intercropping, with melon mixed with cowpea in the same row (MC1); iv) row intercropping at a ratio of 1:1 (melon:cowpea), alternating one melon row and one cowpea row (MC2); and v) row intercropping at a ratio of 2:1 (melon:cowpea), alternating two melon rows and one cowpea row (MC3). The eld experiment was a completely randomized design with three plots per treatment, and each plot had a surface area of 150 m 2 . Melon seedlings were planted at a density of 0.4 plants m − 2 , with a spacing of 200 cm between rows and 120 cm between plants in both the monocropped and intercropped systems. The density of cowpea plants was 2.5 plants m − 2 and 1.5 plants m − 2 in the 1:1 row (MC2) and 2:1 row (MC3) systems, respectively. In the intercropped row systems, the cowpea rows were spaced 100 cm from the melon rows, and there were 20 cm between cowpea plants in the same row. In the mixed system (MC1), the cowpea density was 0.4 plants m − 2 with one cowpea plant between melon plants in each row and spacing of 200 cm between rows and 120 cm between plants. The melon density was thus the same in the different treatments, but the cowpea density changed.
All crops were drip irrigated and grown under organic management. The melon plot (M) received the equivalent of 3000 kg ha − 1 of organic fertilizer (N org) (3.2 % N and 7 % K2O), and the cowpea plot (C) received the equivalent of 1875 kg ha − 1 of Norg. The mixed (MC1) and intercropped plots (MC2 and MC3) received 30% less Norg than the melon monocrop to assess the e ciency of the intercropping in reducing external fertilization needs. The melons and cowpeas were simultaneously harvested twice, on July 31, 2018 and August 6, 2018. The harvest was carried out manually, as is the tradition in the area, to avoid damaging the melon fruits.
Five random soil subsamples (0-10 cm depth) were collected with an auger from the plots on August 10, 2018, just after harvest. Soil samples in MC2 and MC3 were only collected from the melon rows. The samples were taken between two adjacent plants in all cases. The soil samples were separated into two aliquots, one of which was kept at ambient temperature for chemical analyses and the other stored in a cool box with ice for biological analysis. All samples were taken to the lab immediately. The soil was airdried for one week for chemical analyses and sieved at < 2mm. Soil for biological analysis was sieved at < 2mm once in the lab and stored at -20ºC.

Soil properties and crop yield
The soil pH and electrical conductivity (EC) were measured in deionized water (1:5 w/v). The total organic carbon (TOC) and total nitrogen (TN) were determined using an elemental CHNS-O analyzer (EA-1108, Carlo Erba). Soil NH 4 + was extracted with 2M KCl in a 1:10 soil:extractant ratio and calorimetrically measured 16,17 . Available P (P) was measured using the Olsen method 18 . Available nutrients were measured using ICP-MS (Agilent 7500CE).

Soil DNA extraction, PCR ampli cation and sequencing
Soil DNA was extracted from 1 g of soil (wet weight) using the DNeasy Power Soil Kit (Qiagen). The quantity and quality of the DNA extracts were quanti ed using a Qubit 3.0 Fluorometer (Invitrogen, Thermo Fisher Scienti c, USA) and a NanoDrop 2000 uorospectrometer (Thermo Fisher Scienti c, Waltham, MA, USA).
The bacterial community was determined via the next-generation sequencing of bacterial 16S hypervariable regions using an Ion Torrent™ Personal Genome Machine™ (PGM) System. Bacterial 16S regions were ampli ed using an Ion 16S™ Metagenomics Kit (Thermo Fisher Scienti c) with two different degenerate primer sets to amplify regions V2-8 and V3-6, V7-9. The ampli ed 16S amplicons were then processed using an Ion Xpress™ Plus Fragment Library Kit in combination with an Ion Xpress™ Barcode Adapter 1-96 Kit (Thermo Fisher Scienti c). All puri cation processes between incubation and ampli cation reactions of the library preparation were processed using DynaMag™-2 magnetic racks (Thermo Fisher Scienti c) and an AMPure XP Puri cation Kit (Beckman Coulter). After library preparation and barcoding, we determined the size and concentration of the nal libraries using an Agilent 2100 Bioanalyzer system and the Agilent High Sensitivity DNA kit. The sequencing templates were prepared using an Ion One Touch 2 System and an Ion PGM™ Hi-Q™ View OT2 Kit (Thermo Fisher Scienti c). The sequencing reaction was performed using Ion Torrent PGM with an Ion PGMTM Hi-QTM View Sequencing Kit (Thermo Fisher Scienti c).

Sequencing data processing
Bacterial raw sequences, barcodes and primers were trimmed according to the BaseCaller application.
The sequences were denoised with ACACIA 19 , and low quality sequences were discarded using the Quantitative Insights into Microbial Ecology (QIIME) pipeline 20 from the Microbiome Helper Virtual Box 21 .
Brie y, bacterial sequences with a Q < 25 were removed, and the retained sequences were then assigned to Operational Taxonomic Units (OTUs) based on 97% similarity with the SILVA reference database after ltering chimeras using VSEARCH 22 with the ribosomal database project (RDP database). Low-con dent OTUs were removed.
The sequences were uploaded to the European Nucleotide Archive (ENA) with the study accession code PRJEB42624.

Statistical analysis
All tests were performed using R language 23 . Normality and homogeneity of variance assumptions were assayed by Shapiro-Wilk and Levene's tests using the car 24  Bacterial alpha diversity [Chao1 as richness and Shannon (H') as diversity index] was estimated on rare ed microbial data using the vegan package 26 .
A linear discriminant analysis (LDA) effect size (LEfSe) pipeline 27 , available at http://huttenhower.sph.harvard.edu/galaxy/, was used with the default parameters at all taxonomic levels to identify genera that were differentially abundant among the cultivation systems. Three different steps were performed using the following algorithm: i) a nonparametric Kruskal-Wallis test to detect the statistical differences between abundances; ii) a pairwise among subclasses using the Wilcoxon ranksum test to evaluate biological consistency; and iii) an LDA to estimate the effect size between abundances.
Principal coordinates analysis (PCoA) was used to visualize the variation in community composition by cultivation system based on the Bray-Curtis distance. To evaluate differences between the cropping systems, a Permutational Multivariate Analysis of Variance (PERMANOVA) was conducted using the 'betadisper' and 'adonis' functions with 999 permutations from the vegan package, followed by the 'pairwise.adonis' function with Benjamini-Hochberg corrections for multiple comparisons between speci c cultivation systems from the pairwiseAdonis package 28 when the homogeneity of variance assumption was met. In the cases in which homoscedasticity was not ful lled, an Analysis of Similarities (ANOSIM) was carried out instead. Relationships between the bacterial community and the rest of the parameters were determined using the 'bioenv' function from the vegan package to nd the best subset of parameters (using Euclidean distance) that had a maximum correlation with the community dissimilarity matrix 29 . Redundancy analysis (RDA) was performed through the vegan package to visualize the correlation between OTUs and physico-chemical, biological and harvest parameters. The OTU abundance was Hellinger transformed prior to analysis with the retained variables from the bioenv procedure 30 , which was performed via the 'bioenv' function based on Spearman's rank correlation coe cient. To equalize the number of replicates for 'bioenv' and 'rda', the function 'sample_n' in the dplyr package 31 was used.

Effects of intercropping on melon crop yield
The intercropped melon systems showed a signi cantly higher melon yield (MC1 74%, MC3 64%) and number of melons (MC3 52%, MC1 40% and MC2 33%) than the monocrop (M) ( Table 1). The cowpea yield, on the other hand, was higher in the monocrop system than in the intercropping systems (Table 1).

Effects of intercropping on bacterial community diversity and composition
After ltering, 821,795 reads were yielded and 6,676 OTUs were identi ed with 97% similarity for the bacterial community. No signi cant differences were found in the Shannon or Chao1 diversity indexes between cropping systems ( Supplementary Fig. 1). Sequence analyses at the phylum and genus taxonomic levels are shown in Fig. 1A and Fig. 1B.
Bacterial community structures were distinctly grouped by cropping system on a PCoA plot (Fig. 2). Moreover, the bacterial community structure in the monocrop systems (M and C) differed signi cantly (F = 2.7262; P = 0.001) from that in the intercropping systems (MC1, MC2 and MC3). This difference was con rmed by pairwise comparison (Supplementary Table 3).
Concerning the speci c gene community related to N cycles, strong differences were found in AMOA, NARG (p < 0.05) and NIRK (p < 0.01) genes. In general, the log copies of these three genes were higher in monocropping systems (M and C) than in intercropped systems (MC1, MC2 and MC3). Among the three intercropping patterns, MC2 showed the lowest values ( Fig. 3; Supplementary

Effect of intercropping on soil properties
Signi cant differences were found in some of the physicochemical and chemical soil properties (Table 1). Compared to the melon monocrop (M), TN was signi cantly (p < 0.05) higher in all the intercropped systems (MC1, MC2 and MC3) and the cowpea monocrop (C). NH 4 + was also higher in all the intercropped systems assayed than in the monocrops, but it was only signi cantly higher for the MC3 treatment (p < 0.05). The TOC content was signi cantly (p < 0.05) higher in MC3 and C than in the other two melon intercropping systems (MC1, MC2) and the melon monocrop (M). The available P content was signi cantly (p < 0.001) higher in the intercropped systems (MC1, MC2 and MC3) than in both monocrops (M and C). Available Mg and Na were signi cantly higher (p < 0.05) in MC1 than in the other treatments, and available Ca was signi cantly higher in MC1, M and C. No signi cant differences were observed in available K (Table 1).

Relationships between soil properties and the bacterial community
Redundancy analysis (RDA) (Fig. 5) revealed a relationship between the bacterial community structure, soil properties and crop yield. The TN, AmoA, available Na and P and melon yield appeared to be strongly correlated with the bacterial community. Namely, the P content, melon crop yield, available Na, and AmoA abundance were correlated with the rst axis. The TN was correlated with the second axis. Monocrops showed clear divergence from the intercropping systems, while the latter could not be easily separated.
Signi cant correlation was observed between different genera and the chemical and harvest parameters of the intercropping systems: Pseudomonas (r 0.69 P < 0.01; r 0.70, P < 0.01; r 0.68, P < 0.05) with P, TN and M. yield respectively.

Discussion
Intercropping is considered to be an environmentally friendly cropping system that can improve crop yield as well as water and nutrient-use e ciency 32,33 . Crops have different needs, so it is especially important to combine them in the right way to obtain yield improvements. As far as we know, the melon-cowpea intercropping system has not been studied in depth, nor have intercropping patterns between these two crops. However, this combination could be an important choice for sustainable agriculture management, given that cowpea is a legume, which xes atmospheric nitrogen and thus supplies it to companion plants like watermelon or melon that at the same time provide soil shading to conserve water moisture 34 .
The intercropping system assayed was found to increase melon yield and the number of melons produced compared to melon monocrops starting from the rst year of experimentation. Moreover, the intercrops used 30% less fertilization than the monocrop. This increase in melon yield could be due to a higher nitrogen disposal from cowpea rhizosphere, which, it is should be higher in soils with low N fertilization addition 35 . This fact has previously been observed in other cowpea intercrop relationships, such as cowpea-maize 36 , cowpea-sorghum 37 and cowpea-cassava 38 . The cropping patterns and N fertilization rates can alter soil conditions, which subsequently in uence the abundance of functional Ncycling genes 39 . In our study, we also observed a decreasing trend in nitri cation and denitri cation processes in the three intercropped systems compared to the monocrops. This decrease in the intercropping systems could allow for sustainable nutrient use, diminishing nitrate losses due to leaching and N oxide emissions 40 .
In general, legume crops included in intercropping systems improve P availability and soil organic carbon 41 , mostly through root exudates, nodules, and the sloughing off of root cells and root turnover during the growing season 42 . Roots excrete larger amounts of protons and carboxylates (malonate, malate, and citrate), which would facilitate root-borne phosphatases to hydrolyze organic P 43 . This could be supported by a high abundance of phosphate-solubilizing bacteria like Pseudomonas, which are more abundant in intercropped soils, as observed previously by 44 , that correlated with available P, TN and melon yield. Moreover, the presence of several phosphate-solubilizing bacteria like Bacillus in both the monocrops and intercropping systems could also in uence in this behavior, as has been observed by Chen et al. (2006) 45  It is important to note that soil microbial community composition is signi cantly correlated with changes in soil chemical properties 47,48 . In this study, the TN, available Na and P content, as AmoA abundance and melon crop yield play important roles in changes in the microbial community structure. Our ndings could indicate that nutrient changes subsequently affect the carbon-and nitrogen-use e ciency of bacteria. Generally, an increase in soil microbial diversity is bene cial to soil function and health, but no differences were detected through diversity or richness estimators, indicating that our hypothesis was not validated. Until now, there is no consensus about changes in alpha diversity caused by intercropping systems, since some researchers have reported that some intercropping systems can increase diversity 1,49 , while others have reported no signi cant changes 50,51 .
On the other hand, we found signi cant differences in the bacterial community structure between intercropping and monocrop systems, although not between the different intercropping patterns. These differences showed the in uence of cowpea on the bacterial structure of the melon crop, suggesting that cowpea could play an important role in maintaining agricultural ecosystem stability and improving crop growth 52 . The dominant taxonomic groups identi ed in the soils assayed were Proteobacteria, Actinobacteria, Acidobacteria, Firmicutes, Gemmatimonadetes, Planctomycetes, Chloro exi, Bacteroidetes and Nitrospirae, all depicted as common inhabitants of soil 53 58,59 . On the other hand, only ve bacteria were associated with the intercropping system, which indicates that changes are occurring. Moreover, these changes do not depend too much on the speci c intercropping pattern. The fact that these bacteria did not have a greater abundance than in the monocrops is likely due to the high resilience of bacterial community to changes 60 . These results indicate that one year of intercropping, which has been studied here, is not enough to result in a signi cant amount of associated microorganisms. It would be expected that repetitive intercropping in the same soils would increase a different microbial pattern.

Conclusion
The introduction of cowpea in an intercropping system with melon in a short-term experiment resulted in a sustainable cropping system with decreased external input use and an increase in melon yield. The intercropping system changed the bacterial community structure, which contributed to an increase in soil TN and P concentrations and melon crop yield. The intercropped systems were characterized by a higher abundance of Pseudomonas (Proteobacteria), which are related to nutrient cycling and bene cial microorganisms. Further long-term analysis in these intercropping systems will be needed to reinforce ndings on the positive interaction between cowpea and melon and to study more in depth which intercropping pattern farmers can choose according to their necessities.

Declarations Competing Interests
The author(s) declare no competing interests.