Impact of a conservation agriculture system on soil characteristics, rice yield, and root-parasitic nematodes in a Cambodian lowland rice field

The experimental field was located in Stung Chinith, Santuk district, Kampong Thom province, Cambodia (12°32′55″N and 105°08′47″E). Since April 2011, a study comparing DMC vs CT cropping systems has been implemented. The experiment was managed by agronomists of the General Directorate of Agriculture of Cambodian, the Department of Agricultural Land Resources Management (DALRM), the Conservation Agriculture Service Center (DALRM/CASC), and the AIDA research unit of the French Agricultural Research for Development (CIRAD). The field soil is a sandy podzolic soil (≥70% sand at 0- to 40-cm deep) belonging to the “Prey Khmer group” in the Cambodian agronomic soil classification system (White et al., 1997), equivalent to red-yellow podzols (Crocker, 1962), or fluvisols/arenosols according to the FAO soil taxonomy (Anonymous, 2006).

Annual rainfall in the Kampong Thom province was 1,489 mm in 2014 and 1,154 mm in 2015 (Ministry of Water Resources and Meteorology, Stung Chinit Station). The experiment was conducted during the rainy seasons (from June to November) of 2014 and 2015. During the dry season (from late November to the end of May), the experimental field remained dry before being progressively flooded during the rainy season with most water from the end of August until October due to heavy rains.

The experimental field was about 2.6 ha large, where nine plots (250 m² each) were randomly distributed and managed equitably either in the DMC (3 plots) or in the CT systems (2 × 3 plots) (Supplemental Material 1). In the DMC system, seeds of two leguminous cover crops (Stylosanthes guianensis cv. Nina and Centrosema pascuorum cv. Cavalcade) were broadcasted (4 to 6 kg ha⁻¹ and 6 to 8 kg ha⁻¹, respectively) at the end of the rainy season and before rice harvest (early November). The cover crops were grown during the dry season on the residual soil moisture (until early May). In total, 30 days prior to the sowing of the new rice crop, the cover crops were rolled down using a roller crimper and glyphosate (N-(phosphonomethyl) glycine) at 0.96 kg ha⁻¹ and 2.4 D [2,4-dichlorophenoxyacetic acid] at 0.72 kg ha^-1 were applied. Approximately 15 to 25 Mg ha⁻¹ of fresh biomass was left to decompose on the soil’s surface (F. Tivet, CIRAD, pers. comm.). In early June, rice seeds were directly seeded into the mulch with a NT planter.

The CT system consisted of a practice commonly used by the local rice farmers. After the harvest of the rice crop, rice stubbles were left in the field. One month before sowing rice (in May), the soil was ploughed to control weeds and incorporate the remaining rice stubbles into the soil. Then, soil plowing and harrowing were practiced again on the day of sowing (June). No cover crops were grown between the two rice crop cycles under CT management. Two rice planting/sowing methods were used: transplanting (three plots) and seed broadcasting (three plots). Before transplanting, seedlings were grown in a nursery on an adjacent field during two to three weeks. For seed-broadcasting plots, seeds were manually broadcasted directly onto prepared soil (plowing and harrowing). Then, a harrow was used to incorporate the seeds into the soil. Mineral fertilizer was applied (100 kg ha⁻¹ of diammonium phosphate; 18% N, 46% P₂O₅ and 0% K₂O) at sowing time under the DMC system, and 15 days after rice transplanting and at seed broadcasting under the CT system.

During the rice cycle, weed control was done three weeks after sowing under both treatments and Cyhalofop-butyl (2-4-4-cyano-2-fluorophenoxy phenoxy propanoic acid, butyl ester (R)) at 0.285 kg ha⁻¹ and 2.4 D (2,4-dichlorophenoxyacetic acid) at 0.504 kg ha^-1 were applied under unflooded conditions.

Roots were sampled four times per year in both 2014 and 2015, three to four years after establishment of the experiment. The 1st sampling was done when the rice plants were at the tillering stage (at the end of June in 2014 and mid-July in 2015), the 2nd sampling at the maximum tillering stage (late August), the 3rd sampling at the late reproductive phase or early milky stage (in October) and the 4th sampling at the ripening phase shortly before harvesting (November).

At each sampling date, 20 individual rice plants were randomly uprooted per plot. At the 1st and 2nd sampling dates, whole root systems were collected, whereas at the 3rd and 4th sampling dates, a composite sample of 100 g was collected for each plant due to the large size of the root systems. Composite samples were made after cutting the roots into small pieces and mixing the pieces thoroughly before taking a representative sub-sample of 100 g. The roots were carefully washed in running tap water and placed in labeled plastic bags before being transported in cool iceboxes to the laboratory.

Eggs were recovered from the root systems of 20, individually sampled, plants using a hypochlorite extraction method and a blender (McClure et al., 1973) with minor modifications (Bellafiore et al., 2015). After being filtered through 500 µm and 250 µm sieves, eggs and juveniles were recovered on a third 25 µm sieve. The nematode suspension obtained after extraction was homogenized in 40 ml tap water and 1 ml of this suspension placed in a counting cell chamber to count the number of eggs, second-stage juveniles (J2), and males using a stereomicroscope. Based on the total number of eggs, J2 and males, RPN population density per gram of fresh roots (RPN/g roots) was calculated.

From each plot, two infected rice plants (the infection determined by the presence of RPN after counting) were randomly selected. Using a stereomicroscope, 20 J2 were randomly and individually collected from each plant and put into 1 ml tubes with 0.001% of Tween-20 solution and stored at −80^oC until DNA extraction (Bellafiore et al., 2015). Ten microliters of pre-mix 2X lysis buffer (20 mM Tris-HCl at pH 8.0, 100 mM KCl, 3 mM MgCl₂, 2 mM DTT, 0.9% Tween 20) were added to each 1 ml tube containing a single J2, 10 µl of dH₂O and 0.001% Tween 20. Zero-point five microliter of Proteinase K (100 µg/µl in stock solution, Thermo Scientific (Waltham, MA) and 0.025 µl of DTT (1 M) were added, and the tubes were incubated at 55^oC for 3 hr before being transferred to 98^oC for 15 min. One microliter of RNAse type A (10 mg/ml; Thermo Fisher Scientific, Waltham, MA) was then added to each tube and incubated at 37^oC for 1 hr. The extracted DNA was immediately used for polymerase chain reaction (PCR) and then conserved at −20^oC for future study.

PCR was performed on each J2 DNA extract using M. graminicola’s specific SCAR-MgFw/Rev marker (Sequence Characterized Amplified Regions) (Bellafiore et al., 2015). In the case of no amplification, suggesting the presence of other plant parasitic nematodes, another PCR with the Internal Transcribed Spacer (ITS) primers rDNA2 (5'-TTGATTACGTCCCTGCCCTTT-3') and rDNA1.58 s (5'-ACGAGCCCGAGTGATCCACCG-3') (Vrain et al., 1992; Powers, 2004; Pokharel et al., 2007) was then performed to sequence the amplification product and compare it to the NCBI databases. PCR products were purified after gel electrophoresis using Thermo Scientific GeneJET Gel extraction kit (Thermo Fisher Scientific, Waltham, MA) and sent for sequencing using standard procedures and rDNA2 primers (Macrogen Inc., Seoul, South Korea) (chromatograms and sequences are available on the “Harvard Dataverse”: https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/ZHLQVM).

Using standard nucleotide BLAST, the sequences were compared to a collection of ITS sequences from Hirschmanniella spp. and imported from GenBank database (for H. oryzae accession number DQ309588, EU722286 and for H. mucronata (Das, 1960) accession number KP179331, DQ309589, EU722287). ITS sequences alignment, phylogenetic, and molecular evolutionary analyses were conducted using MEGA version 6 (Tamura et al., 2013).

In conjunction with the molecular identification of the RPN species, the population dynamics of the RPN were studied using the RPN population densities at four rice development stages: seedling, tillering, milky and harvest. Eggs extracted from infected roots were hatched under controlled conditions (Bellafiore et al., 2015). For the tillering stage in 2014 and the harvest stage in both 2014 and 2015, the hatching was compromised and we were not able to collect J2. The average number of J2 per 100 g of fresh roots was calculated for each plant stage. Based on the percentage of each RPN species from the molecular analyses and taking into account the RPN population density per gram of fresh roots previously formulated, the population density of each RPN species was calculated.

Each plot was subdivided into four sub-plots. Five soil samples were collected from each sub-plot at 0 to 5 cm depth in early December 2014 and these were mixed to create a composite sample. The samples were air-dried at room-temperature, passed through a 2-mm sieve and homogenized. Soil organic C and total N concentrations were determined with the dry combustion method (Nelson and Sommers, 1996) using a Sumigraph NC-80 Analyzer (Sumitomo Chemical Co, LTD, Osaka, Japan). Total organic carbon (TOC) and total nitrogen (TN) were estimated and computed on an equivalent soil mass-depth basis according to Ellert and Bettany (1995). The labile C (LC) fraction was determined using the potassium permanganate oxidizable C (POXC) procedure modified by Culman et al. (2012) at 0.2 M KMnO₄. Briefly, 2.5 g of air-dried soil was transferred into a 50 ml centrifuge tube, to which 18 ml of deionized water and 2 ml of 0.2 M KMnO₄ stock solution were added. The tube was shaken at 120 rpm with a horizontal shaker for 2 min, then left to settle for 10 min. After settling, 0.5 ml of the supernatant was transferred into another 50 ml centrifuge tube containing 49.5 ml of deionized water. The absorbance of the samples was read by a Spectrophotometer UV-1200 (Shimadzu Inc., Kyoto, Japan) at 550 nm.

Soil pH was determined using a 1:2.5 ratio of soil/distilled water and measured with a pH meter D-51 (Horiba Ltd., Kyoto, Japan). Phosphorus content in each sample was determined using the Olsen method to extract the available soil phosphorus (Olsen et al., 1954). The extracted soil phosphorus was measured by the Murphy and Riley (1962) method. Exchangeable base cations Ca²⁺ and Mg²⁺ were extracted with 1 mol/liter KCl and K⁺ with Mehlich-1 solution. Exchangeable Ca²⁺ and Mg²⁺ were determined by titration using 0.025 mol/liter EDTA. K⁺ was determined by flame photometry.

In addition, for each sub-plot the central point was sampled using a core sampler topsoil (0–5 cm depth) to measure soil bulk density (ρ _b ) as assessed by the core method (Blake and Hartge, 1986). These soil samples were oven-dried at 105°C for 24 hr, and dry weights were assessed to calculate the bulk density. In the field, soil conductivity (Cond), redox (Redox) and quantity of salt in parts per million (ppm) were recorded for each plot topsoil (0-5 cm depth) at each sampling date with Hanna probes (Hanna Inc., Woonsocket, Rhode Island).

All statistical analyses were done using the software R (R Core Team, 2015; version 3.3.3). The RPN population density per g of fresh roots did not follow a normal distribution due to the high number of non-infected plants (zero value), and the data was therefore transformed into a binary distribution (0 for non-infected and 1 for infected plants) before performing a generalized linear model (GLM) following a binomial distribution. We tested the effect of the sampling by year (2014 vs 2015), cropping system (DMC vs CT), plot and rice developmental stage on RPN population densities by considering both non-infected and infected plants. The same tests were also performed with only infected plants. Then, the relationship between yield and number of RPN per gram of fresh roots was evaluated using Kendall’s test.

Plot parameters were further investigated using a principal component analysis (PCA) to summarize the information conveyed by all the variables to a reduced number of dimensions. The analysis was carried out with the library ade4 (Dray and Dufour, 2007). A multivariate analysis was conducted to further analyze the contribution of the different plot parameters to the a priori classification of plots according to the agricultural system (DMC vs CT), since our main objective was to compare the effects of these contrasted practices on RPN. We compared the three DMC plots to the six CT plots. In order to have all the parameters, we conducted the PCA using plot means of both years. A between-class analysis was performed to investigate whether families cluster according to the agricultural system (DMC vs CT). The between-till structure was tested using permutations tests (Monte–Carlo tests; ade4, Dray and Dufour, 2007). Pairwise comparison of the number of RPN and the rice yield in a DMC system vs two CT systems was performed using the Kruskal–Wallis non-parametric test due to non-normal data distribution.

The PCA showed an impact of the DMC and CT cropping systems on soil nutrient concentrations, TOC, TN, pH, labile soil organic C and bulk density (Fig. 1). The two axes explained 63.6% of the variance (first: 42.7%; second: 20.9%). The significance of this grouping (DMC vs CT) was significant (Monte–Carlo test, p = 0.016).

Figure 1:

The DMC cropping system resulted in higher TOC concentration, LC, TN, Mg and K concentrations (Table 1). In contrast, the pH was lower in the DMC cropping system compared with the CT cropping system.

For all cropping systems, the yield was significantly higher in 2014 (2,572 kg ha^-1) compared to 2015 (2,299 kg ha^-1, p < 0.001) (Fig. 2). When both evaluated CT systems are combined, the PCA emphasized higher yields under the DMC system compared to the CT system (Fig. 1). Still merging both CT systems, when one considers the rice yield it is not significantly different (p = 0.671) in 2014 between DMC and CT but in 2015, it becomes significantly higher under DMC (p < 0.001) (Fig. 2). Using the plot means of all plots combined (DMC and CT), a low negative correlation was found between the rice yield and the number of RPN, but it was not significant (correlation −0.25; p = 0.364). Using only the plot means of the DMC system, a significant negative correlation was found between the rice yield and the number of RPN (correlation -0.89; P = 0.039). When observed individually, the conventional tillage-seed broadcasting method (CT-B) gave the lowest rice yield for both studied years (Fig. 2).

Figure 2:

GLM analysis showed the effects of cropping system, rice developmental stage, year of sampling, and plot on RPN. Interactions were revealed between all variables measured (Table 2). When only the nematode-infected plants were considered, GLM analysis showed no interaction between the variables measured. All variables had an effect on RPN, except the sampling year (Table 2). The average number of RPN per gram of roots was significantly higher in 2014 than in 2015 for all cropping systems (p < 0.001; Fig. 3). When comparing the three methods (DMC system, hand-broadcasting and transplanting in the CT systems), higher numbers of RPN were observed following DMC (p <  0.001). Within the CT systems, the number of RPN was not different between hand-broadcasting or transplanting. Using rDNA barcoding, 360 and 480 J2 were analyzed in 2014 and 2015, respectively. This molecular identification revealed that the main nematodes associated with the rice roots were either root-knot nematodes (Meloidogyne spp.) or Hirschmanniella spp, albeit with lower abundance. Sequence alignment using Blastn revealed that only two RPN species, M. graminicola and H. mucronata, were present and M. graminicola was dominating. The occurrence of M. graminicola was higher than H. mucronata in both years and in both cropping systems (Fig. 4). The frequency of occurrence of M. graminicola decreased from 2014 to 2015 in both cropping systems, while the frequency of occurrence of H. mucronata, instead increased (Fig. 4). M. graminicola was present during all plant developmental stages, whereas H. mucronata was only found at the tillering and milky stages. In both 2014 and 2015, the total number of RPN per gram of roots was highest at the seedling stage when the plants were 2 weeks old, and lowest at harvest (Fig. 5A). For both years, only M. graminicola was found in the roots of the young rice plants two weeks after sowing (Fig. 5B). No data for 2014 are available but in 2015, the RPN population at the tillering stage consisted of M. graminicola (73%) and H. mucronata (27%). At the milky stage, H. mucronata was present in 2014 (15% of the RPN population), but not in 2015.

Figure 3:

Figure 4:

Figure 5: