Seedling production

Plant seedlings were produced from a single cultivar with seed provided by the Centro Nacional de Tecnologia Agropecuaria y Forestal (CENTA) in El Salvador. The seeds were sown in plastic trays, 54 by 28 cm, filled with peat moss and covered with 1 cm of vermiculite. Twenty days after sowing, the seedlings were transferred to plastic trays with 72 square cells and placed under controlled conditions until transplant to the field or greenhouse. The settings of the controlled environment were as follows: average temperature of 21 °C during the day, 16 °C at night, and 12 h of light and dark.

Multiplication of the Rhizobia strains

The Rhizobia strains evaluated were Bradyrhizobium sp. (Vigna) (R₁), Bradyrhizobium USDA 3384 (R₂), Rhizobium leguminosarum biovar (R₃) and Bradyrhizobium USDA 2370 (R₄). The strains R₂ and R₄ were multiplied in modified arabinose gluconate (MAG) liquid medium, with the pH adjusted to 6.6. Bacterial growth was induced by thawing a small portion of the frozen bacteria in 100 mL of liquid MAG medium and placing the solution on a rotary shaker at 200 rpm. After sufficient growth of the bacteria was observed, the solution was diluted in four concentrations (1:1, 1:10, 1:100 and 1:1,000) and transferred to individual petri dishes (100 x 15 mm) with solid MAG medium. The whole procedure was performed under a laminar flow hood. The inoculant for strains R₁ and R₃ was obtained from two commercial products developed by INTX microbials-llc, Indiana, USA. The strain R₁ was inoculated with the commercial product N-DURE^® inoculant for cowpea (Vigna unguiculata L.) and the strain R₃ with the N-DURE^® inoculant for beans (Phaseolus spp. L.).

A solution of each of the nitrogen-fixing bacterial strains was prepared with a minimum concentration of 2 x 10⁸ colony-forming units (CFUs). The bacterial colonization of the root system was induced by applying 10 mL of the solution to the substrate in which the seedlings were growing using a plastic syringe.

Greenhouse experiment

A greenhouse trial (Greenhouse experiment 1) was conducted in 2012 to assess the inoculation efficiency of specific Rhizobia strains and determine the interaction of the bacterial colonization with applications of inorganic N. The experiment was carried out at the facilities of the College of Natural Sciences of the University of Massachusetts, Amherst. The general settings of the greenhouse were: 14 h of light with average temperatures of 21 °C during the day and 16 °C at night. The seedlings were transplanted into nursery pots, 18.5 cm in height and 16.2 cm in diameter, filled with a mix of river sand and perlite in 1:1 ratio.

The design of the experiment was a randomized complete block with four replications. In the trial, four Rhizobia strains (R₁, R₂, R₃ and R₄) and a control treatment (R₀) were evaluated in combination with seven concentrations of N (0, 26.25, 52.5, 105, 157.5, 210 and 262.5 mg·L^-1).

The N rates were supplied through a modified 0.5 M Hoagland nutrient solution (^{Hoagland & Arnon, 1950}), using sodium nitrate (NaNO₃) as the main N source. Macro and micronutrients were also included in the Hoagland solution as potassium phosphate (KH₂PO₄), magnesium sulfate (MgSO₄), calcium chloride (CaCl₂) and potassium chloride (KCl) in the following concentrations (mg·L^-1): 234 potassium (K), 31 phosphorus (P), 48 magnesium (Mg), 64 sulfur (S), 200 calcium (Ca) and 525 chloride (Cl). The concentrations of the macro and micronutrients remained constant in all seven nutrient solutions, except for sodium (Na): 0, 44, 87, 173, 259, 345 and 432 mg·L^-1.

The root systems of the plants were carefully washed, and the number of nodules per plant (NN) and the average weight per nodule (NW, in mg) were recorded once the crop cycle was concluded.

Field experiments

A field evaluation (Field experiment 1) was conducted in 2011 to determine the response of the crop to eight doses of inorganic N. In 2012, an additional experimental trial (Field experiment 2) was carried out to assess the inoculation efficiency of three Rhizobia nitrogen fixing bacteria strains under different rates of inorganic N. The two field trials were conducted at the Research Farm of the University of Massachusetts in South Deerfield, Massachusetts. The characteristics of the soil where the experiments were established are as follows: Occum fine sandy loam (coarse-loamy, mixed, mesic Fluventic Dystrudept) with a pH of 6.5, 2.4 % organic matter, 10 ppm P₂O₅, 53 ppm K, 556 ppm Ca, 65 ppm Mg and 91.23 ppm NO₃.

The plots were established on raised beds, with a separation of 1.8 m, covered with degradable black mulch (BioTelo^®). The seedlings were transplanted 30 cm apart in two rows per bed to obtain a plant density of 29,000 plants·ha^-1. Irrigation was provided with a drip system and scheduled according to the information collected from the soil tensiometers (Irrometer^®, USA) placed at 38 and 76 cm deep ground. Phosphorus (33.7 kg·ha^-1 as P₂O₅) and potassium (123.3 kg·ha^-1 as K₂O) were supplied with inorganic fertilizer through drip irrigation according to the general recommendations for kale and collards in the New England Vegetable Management Guide, 2011-2012 (^{Howell, 2011}).

Six rates of inorganic N were evaluated in Field experiment 1 (40, 80, 120, 160, 200 and 240 kg·ha^-1), and two additional N rates were included in Field experiment 2 (0 and 280 kg·ha^-1). Each rate of N was supplied in five applications containing 12.5 % of the total N each. The N source for both experiments was ammonium nitrate (NH₄NO₃). In Field experiment 2, the Rhizobia strains R₁, R₂, R₃ and R₄ were also evaluated in combination with seven concentrations of inorganic N. Both experimental trials were evaluated in a randomized complete block design with five replications.

The fresh shoots (FW, in kg·ha^-1) were harvested five times when the marketable size for chipilín of 15 to 20 cm was reached and samples were dried in an oven for 48 h at 75 °C to obtain the dry weight (DW, in kg·ha^-1). The color, vigor and uniformity of the crop were recorded on a visual scale from 1 to 5, where 1 was the worst and 5 was the best. Plant height (PH, in cm) was also measured. The following traits were examined only in Field experiment 2: leaf chlorophyll content, determined using a Minolta meter (SPAD-502, Mexico) (SPAD reading); total nitrogen content (TN, in mg·L) of dry matter, estimated using the Dumas combustion method in an element analyzer system; and NN and NW, obtained by cleaning the root system within a 35 cm radius around the plant.

Statistical analysis

The FW, DW, SPAD, PH, TN, NN and NW readings were subject to analysis of variance for a randomized completed block experiment.

In the model, Y _ijk is the response variable for nitrogen rate i, with the Rhizobia strain j in the k block; μ is the overall mean; N _i is the effect of the i ^th N rate; R _j is the effect of the j ^th Rhizobia strain; R _j x N _i represents the interaction effect of the j ^th Rhizobia strain with the i ^th N rate; β _k is the k ^th block effect and ε _ijk is the random error effect for the i ^th N treatment and j ^th Rhizobia treatment within the k ^th block. For the statistical analysis of Field experiment 1 data, the terms R _j and R _j x N _i were excluded from the linear model.

Tukey’s honestly significant difference test was applied for the main effects of Rhizobia and N. Additionally, the statistical association of the N rates and the response variables was evaluated by means of orthogonal polynomial contrasts (linear, quadratic and cubic). This analysis was conducted for the greenhouse trial and for both field trials (Field experiment 1 and 2). A regression model was fitted for the FW with the N rates. The whole analysis was conducted in the SAS software version 9.3 (^{SAS Institute Inc., 2011}).

Response of the crop to applications of inorganic nitrogen

Statistical differences were detected among the N treatments for FW and PH, with coefficients of variation of 16.9 and 7.7, respectively (Table 1). The total FW harvested during the crop cycle varied from 5.4 to 6.2 t·ha^-1, with N rates of 240 and 40 kg·ha^-1, respectively. A positive linear association between the BF and the N rate was observed, but only for N applications between 40 and 120 kg·ha^-1. When N applications were high (240 kg·ha^-1), the commercial yield was reduced by 25 %, compared to that at the N application of 40 kg·ha^-1. The average PH of the crop ranged from 12.5 to 13.8 cm, but statistically only the PH of the NR₆ (240 kg·ha^-1) treatment was different from the PH of the other N treatments (Table 1). This suggests that the increase in FW associated with N applications from 40 to 120 kg·ha^-1 (Figure 1) was due to the accumulation of biomass, rather than an increase in PH.

Figure 1 Average shoot fresh weight (FW_average in kg·ha^-1) and total shoot fresh weight (FW_total in kg·ha^-1) for six rates of inorganic nitrogen (N) evaluated. Field experiment 1, 2011.  

A quadratic polynomial trend between the FW of the crop and the rates of inorganic N was observed (Figure 1). If two independent linear functions are fitted for N inputs and FW, a positive slope for the N rates of 40, 80 and 120 kg·ha^-1 is observed, as well as a negative slope for the N rates of 160, 200 and 240 kg·ha^-1. The slopes of these two models were 6.45 and -15.2, respectively. Then, per every kg·ha^-1 of inorganic N applied there was an increase of 6.45 kg·ha^-1 in FW; however, this trend was observed for N applications below 120 kg·ha^-1. In contrast, a decrease of 15.21 kg·ha^-1 in FW was observed per every kg·ha^-1 of inorganic N applied above 120 kg·ha^-1.

Inoculation efficiency of Rhizobia strains

As expected, inoculations with Rhizobia strains had a significant effect on NN and NW (P ≤ 0.01), and supplementary applications of inorganic N only affected NN. In contrast, the effect of inorganic N on the inoculation efficiency of Rhizobia was only observed in NN. Table 2 presents the average NN per plant and NW per nodule in response to inoculation with the Rhizobia strains. According to these results, the average NN detected with R₁, R₂, R₃, R₄ and R₀ was 51, 38, 128, 39 and 23, respectively. All four strains of Rhizobia induced a larger NN than the native strains from Central America, for which the average NN was 9.4 (^{Guamán-Díaz et al., 2016}). Thus, Rhizobium leguminosarum biovar induced the development of six times more nodules than the control treatment, with an average NW of 0.8 mg.

As reported by ^{Isidoro and Messier (2009)}, Rhizobium leguminosarum biovar is a superior symbiont for chipilín and induces the production of a larger number of nodules in the root system. The strains R₁, R₂ and R₄ also trigger the development of nodules with significantly larger sizes; however, the presence of nodules in roots does not always ensure an efficient N fixation process. In chipilín, only 20 % of the nodules collected from colonized plants by native bacterial strains in Central America contained leghemoglobin (^{Barnish & Spinelli, 2011}), which buffers the concentration of free oxygen and ensures the proper functioning of root nodules.

Recently, an isolated bacterial strain (S.F3.1) with characteristics of the Rhizobiaceae family was identified and linked to a significant increase in fresh and dry weight, plant height, and root biomass (^{Guamán-Díaz et al., 2016}). Furthermore, a detailed analysis to compare the N fixation efficiency of the isolated S.F3.1 and Rhizobium leguminosarum biovar in an intensive production system is needed.

The presence of nodules in the root system of the control treatment plants was also reported in a recent study (^{Guamán-Díaz et al., 2016}). In our case, we attributed the development of nodules in plants of the control treatment to unknown Rhizobia strains present in the substrate used for producing the seedlings.

Table 3 presents the NN and PN for the N rates, and a negative trend is observed for the NN and the N concentration of the nutrient solution. When we compared the responses of HS₀ and HS₁, a minimum N supply of 26.25 mg·L^-1 seemed to support the development of nodules. Similar to these results, an increase in the NN when 2 mmol·L^-1 of N is supplied was reported in alfalfa (Medicago sativa) (^{Salles-de-Oliviera, Anchão-Oliveira, Corsi, Sanches-Duarte, & Mui-Tsai, 2004}) and Crotalaria juncea (^{Miranda-Mendonça & Aparecida-Schiavinato, 2005}). This minimum N concentration could be enough to support plant growth without negatively affecting the symbiotic and biological N-fixation process. However, when the N rate reaches 260 ppm (HS₆), there is a clear reduction in the NN of approximately 82 % compared to HS₁.

An analysis of the effect of inorganic N on the NN of plants inoculated with the Rhizobia strains is presented in Figure 2. In the graph, the standard concentration of the Hoagland nutrient solution (210 ppm) is represented as 1.0 in the abscissa axis; half of the standard concentration (105 ppm) is represented as 0.5, and 1.25 denotes 262.5 ppm. A clear trend towards a decrease in nodulation and N fixation with high doses of N fertilizer is observed.

Figure 2 Number of nodules per plant in response to inoculation with four Rhizobia strains across the six nitrogen (N) rates of the 0.5 M modified Hoagland nutrient solution. 

Although Rhizobium leguminosarum biovar was affected by a large supply of N, the high NN induced by the inoculation with this strain compensated for this inhibition (Figure 2). With N concentrations as high as 260 ppm, there was a reduction in the NN per plant of approximately 200 nodules. The capacity of Rhizobium leguminosarum biovar to cope with the inhibition of N fixation under high concentrations of inorganic N and to maximize fixation under low concentrations of N highlights the importance of identifying host-specific Rhizobia strains.

Constant increases in the prices of N fertilizers and fuels have a major effect on production decisions and profitability. Thus, emphasis should be placed on developing new production methods that are suitable both agronomically and economically. The capacity of Rhizobium leguminosarum biovar to convert stable N gas from the atmosphere into a biologically useful form is a major advantage in chipilín production by lowering production costs and mitigating the effects on climate change by replacing all or most of the N needed by the crop.

Nitrogen fixation efficiency of Rhizobia strains in commercial production systems

The colonization efficiency of the four Rhizobia strains was validated under field conditions. Table 4 presents the sources of variation N, Rhizobia strains (R) and their interaction (N x R) for all evaluated traits. Supply inorganic N to the crop showed a positive effect (P ≤ 0.01) on primary economic traits: color, vigor and uniformity, total FW and on the green color of leaves. This response coincided with preliminary observations, where a constant and large supply of inorganic N dramatically improved the marketable quality of the crop.

Inoculations with the Rhizobia strains also had a positive effect (P ≤ 0.01) on the FW of the plants. Specific strains induced the increase in the total biomass of the plant but have no effect on the intensity of the green color of the leaves. Although the colonization with the bacterial strains increased the NN, there was no effect on the total N content of FW.

High doses of inorganic N decreased N fixation (Table 2). This negative effect on symbiosis was also reported for Sinorhizobium meliloti in alfalfa (Medicago sativa) when the annual N supply reached 450 kg-ha^-1 (^{Salles-de Oliviera et al., 2004}) and for Rhizobia in hemp (^{Miranda-Mendonça & Aparecida-Schiavinato, 2005}). However, the NN was not affected by the continuous increase in N dose in Field Experiment 2 (Table 5). This discrepancy might be caused by differences in the number of times the crop was harvested, and the method employed for the application of the N treatments. Under field conditions, inorganic N was applied every time the crop was harvested, while in the greenhouse, the N was applied every week, and the crop was harvested only once. The constant regrowth of the plants in the field trial might have mined the soil N to the point where it did not affect the formation of nodules or N fixation.

The response of the economic yield of FW to N rates is presented in Table 5. If we analyze the tradeoff between the N application and the FW in Field experiment 2, we determine that per every kg of N supplied in the range from 0 to 80 kg·ha^-1, there is an increase in FW of 30 kg·ha^-1 (P ≤ 0.01). Additional applications above 80 kg·ha^-1 are less cost-effective, with a maximum increase in FW of 1.12 kg·ha^-1 per kg of N applied. Similar to the FW, a constant supply of N in excess of 80 kg·ha^-1 provides an increase of only 0.175 kg·ha^-1 (P ≤ 0.01) in DW per unit of N. The null supply of inorganic N produced only 0.61 t·ha^-1 of DW per season, while in tropical areas, the total DW can reach 3.4 t·ha^-1 per year (^{Sosa-Rubio, Cabrera-Torres, Rodriguez, & Ortega-Reyes, 2008}). A simultaneous increase in the chlorophyll content and the TN content of leaves was observed with N rates as high as 280 kg·ha^-1.

The strong association between the chlorophyll content and the absolute N accumulation in leaves (r = 0.95, P ≤ 0.01) validates the use of the SPAD meter for in situ estimations of the plant N content. In rice, corn, barley and wheat, a similar association was determined between the SPAD readings and the concentration of N in leafs (^{Dwyer et al., 1995}; ^{Lin et al., 2010}; ^{Peltonen, Virtanen, & Haggrèn, 2008}; ^{Reeves, Mask, Wood, & Delaney, 2008}; ^{Shukla et al., 2004}).

The importance of inorganic N in the expression of marketing quality traits for vegetable production was examined. Differences in the color, vigor and uniformity of the crop were detected. When N was not applied, the average color, vigor and uniformity were 2.3, 2.5 and 2.3, respectively, and with 280 kg·ha^-1 of N, the corresponding scores were 3.5, 3.5 and 3.3. Further addition N more than 80 kg·ha^-1, and up to 280 kg·ha^-1, result in a minimum enhancement in the color of 0.2, in vigor of 0.3 and in uniformity of 0.4. The analysis to determine the effect of Rhizobia inoculations on these three quality characteristics did not show statistical significance.

Table 6 shows the statistical association between the Rhizobia strains and the FW, DW, SPAD reading, TN, PH, NN and NW. A non-significant increase of 158 kg·ha^-1 in FW was linked to the inoculation of plants with Rhizobium leguminosarum biovar. On the other hand, plants colonized with Bradyrhizobium USDA 3384 showed a minimal increase in economic yield but significant improvement in the green color of leaves; however, the same strain significantly decreased NN. On average, Bradyrhizobium sp. (Vigna) and Rhizobium leguminosarum biovar induced the development of six more nodules per plant compared to Bradyrhizobium USDA 3384. Although colonization with the Rhizobia strains modified the NN, this factor did not affect the weight per nodule.

The plant-microbe association of chipilín and Rhizobium leguminosarum biovar and Bradyrhizobium USDA 3384 was an efficient N-fixing system, but further efforts are needed to develop cultivars adapted to specific inoculants and to evaluate how biotic and abiotic factors can affect their fixation efficiency. Additionally, the presence of nodules in the root system of plants in the control treatment suggests potential competition between the native Rhizobia in the soil and the introduced strains. When competitors in the soil are poor N fixers, N fixation rates and plant productivity can decrease (^{Olivares et al., 2013}). Therefore, it is important to consider multiple inoculations to ensure adequate colonization by the inoculant.