An assessment of plant growth and N2 fixation in soybean genotypes grown in uninoculated soils collected from different locations in Ethiopia

Achieving food and nutritional security is a major challenge in Ethiopia, especially with increasing human population and low crop productivity. Legumes offer an alternative choice to chemical fertilizers for increasing crop yields. The aim of this study was to assess, under glasshouse conditions, plant growth and symbiotic performance of uninoculated soybean genotypes planted in soils collected from different locations in Ethiopia. The results showed significant differences in plant growth and symbiotic performance among the soybean genotypes planted in different soils. There was a location-specific effect of soil on plant growth and symbiotic N nutrition of soybean. Whole-plant biomass was highest in soil from Amaro, followed by Boricha, Dorebafano, Pawe, and Mambuk. The δ15N values ranged from +0.82‰ for Pawe to +5.11‰ at Dorebafano. However, %Ndfa of soybean was greater in plants grown in Mambuk soil, followed by Pawe with the lowest %Ndfa being in Amaro soil. The amount of N-fixed followed similar pattern as %Ndfa. The significant interaction found between soil type and soybean genotype for plant DM, shoot N concentration, δ15N, %Ndfa, N-fixed and soil N-uptake clearly indicated the effect of soil factors. This study revealed the presence of native rhizobia in Ethiopian soils that are compatible with soybean. The N contribution of the soybean genotypes was variable, and strongly influenced by the soil factors.


Introduction
Industrially, legumes are used to prepare biodegradable plastics, oils, gums, dyes and ink (Graham and Vance 2003). Legumes also have the ability to fix atmospheric N 2 when in symbiosis with soil rhizobia. The annual fixed-N has been estimated to be 21.45 Tg from food legumes, and 12-25 Tg from pasture fodder legumes . In Ethiopia, legume crops such as lentil, faba bean, common bean and chick pea constitute the staple food of the people and are part of the local culinary culture. Additionally, legumes are a valuable source of N-rich feed for livestock and high protein diet for humans. Furthermore, legumes are cultivated as an intercrop, or in rotation with cereal crops (e.g. maize, sorghum and millet) in order to restore soil fertility. Despite their importance, legume crop yields have remained low relative to cereals, especially in developing countries (Graham and Vance 2003;Siddique et al. 2011) due largely to low funding for legume research in Africa.
Soybean is the most important food grain legumes in the world, as the grain contains 40% protein, 26% carbohydrate and 20% oil in addition to being rich in vitamins, fibre and minerals (Wasike et al. 2009). It is estimated that N 2 fixation by soybean alone represents 77% of the N-fixed by crop legumes . About 50-60% or more of the N requirement of soybean is met from BNF (Salvagiotti et al. 2008;Mapope and Dakora, 2016). The high level of N-fixed could be of major benefit to cropping systems in African, where soils are inherently low in nutrients and there is little Presented at the 17th African Association of Biological Nitrogen Fixation (AABNF) Biennial Conference, Gaborone, Botswana, 17-21 October 2016 use of chemical fertilizers (Dakora and Keya 1997;Sinclair et al. 2014). Despite these advantages, soybean production has remained very low in Africa (Maingi et al. 2006;Abate et al. 2012).
Low soil N concentration is a major factor limiting crop production, in Sub-Saharan African countries that includes Ethiopia. In modern agriculture, nitrogen fertilization is widely used to improve crop yields. However, most farmers in Ethiopia cannot afford or access mineral fertilisers to improve crop N nutrition. As a result they practice low N-input agriculture. Hence, the N contribution by the legume-rhizobia symbiosis could be very significant to Ethiopian farmers.
Although soybean is a new crop in Ethiopia, its production is rapidly increasing to meet industry demands and overcome protein-calorie malnutrition. That notwithstanding, Ethiopia is still one of the largest soybean importing countries in Sub-Saharan Africa (Mutegi and Zingore 2014). There is therefore a need to increase soybean production locally and this can be done through identifying soybean genotypes that are highyielding, high N 2 -fixing and can contribute substantially to soil N fertility through effective symbiosis with native rhizobia.
Studies done so far on soybean in Ethiopia have focused on rhizobial inoculation response, genetic variability of soybean for agronomic traits, and intercropping trials with cereals (Bekere et al. 2013;Dilnesaw and Getahun 2013;Zerihun et al. 2014;Jaiswal et al. 2016). Little information currently exists on i) the presence or absence of effective native soybean rhizobia in Ethiopian soils, ii) the response of different soybean genotypes to nodulation and N 2 fixation by indigenous rhizobia in Ethiopia, and iii) the distribution of rhizobia in Ethiopian soils. The aim of this study was i) to assess the ability of native rhizobia to nodulate soybean, as well as ii) to assess plant growth and symbiotic performance of soybean genotypes planted in uninoculated soils collected from different locations in Ethiopia under glasshouse conditions.

Site description
The soil samples used in this study were collected from five locations (namely, Amaro, Boricha, Dorebafano, Mambuk and Pawe) within the soybean-growing areas of Ethiopia. maro, Boricha and Dorebafano are located in the Southern Nations Nationalities and People's Region (SNNPR) of Ethiopia, an area found between geographical coordinates 4°4 3′ and 8°58′ N and 34°88′ and 39°14′ E. The SNNPR region experiences annual rainfall ranging from 400 to 2200 mm and daily average temperature of 10 to 27 (http://www.snnprs.gov. et/about.html). However, Mambuk and Pawe are located in the Benshangul Gumuz Region (BGR) with coordinates 10°3 8′ N and 35°43′ E in north-eastern Ethiopia. The BGR is close to the Ethiopia-Sudan border, and is characterized by 500 to 1800 mm of annual rainfall, 21 to 35°C daily temperature and 600 to 2731 m above sea level.

Soil collection
Fields with no history of rhizobial inoculation were particularly targeted for soil collection. Soils were sampled at 20-cm depth, placed in plastic bags and taken to the laboratory, where they were air-dried in a dust-free environment and sieved (5 mm). A sub-sample of the sieved soil was analysed at the Agricultural Research Council laboratory in Pretoria, South Africa, for pH, electrical conductivity (EC), organic carbon, total N, cation exchange capacity (CEC) and plant-available P. At the laboratory, the sieved soil from each location was used to fill 4-L capacity plastic pots that were previously surfacesterilised. All potted soils were supplied with a blanket treatment of 46 kg P ha −1 as triple superphosphate (TSP) before planting.

Seed source, seed sterilisation and planting
Eleven soybean genotypes (namely, AGS-71, AFGAT, Awassa-95, Coker-240, Clark-63 K, Crowford, Gishama, Nova, TGx-3326-44, Wegayen and Williams) obtained from the Awassa and Pawe Agricultural Research Centres in Ethiopia, were used in this study. Seeds of non-legume species included as reference plants were obtained from Hawassa University, Ethiopia. Where there was history of soybean cultivation, the soybean genotypes used by farmers, were planted at those particular locations. The exception was forest soil which had no history of soybean cultivation. Soybean seeds were surface-sterilized (Vincent 1970) and the pre-germinated seedlings planted in potted soil (Somasegaran & Hoben, 1985, Woomer et al., 2011. Three replicate pots were used for each soybean genotype. The 3 non-legume species (Sorghum bicolour (L.), Eragrostis tef, Triticum aestivum) included as reference plants were similarly processed and planted in triplicates to measure soil N uptake. The plants were raised under natural light in the glasshouse.

Plant harvest and processing
At 58 days after planting (DAP) at R1 stage, the soybean plants were harvested and separated into shoots, roots and nodules. The shoots and roots were oven-dried at 60°C for 48 h and weighed. The shoots of reference plants were also harvested, oven-dried and weighed. The shoots of both legume and reference plants were each milled separately to a fine powder (0.85 mm size) and stored prior to 15 N isotopic analysis.

15 N/ 14 N isotopic analysis
The finely ground (0.85 mm) shoot samples of soybean and reference plants were analysed for 15 N/ 14 N ratio and %N using mass spectrometry. Plant samples were weighed into Al capsules (2.0-2.5 mg/sample) and each fed into a Carlo Erba NA1500 elemental analyser (Fisons Instruments SpA, Strada, Rivoltana, Italy) coupled to a Finnigan MAT252 mass spectrometer via conflo II open-split device. The 15 N abundance, usually expressed in a relative δ (delta) notation, is the ‰ deviation of the 15 N natural abundance of the sample from atmospheric N 2 (0.3667 atom% 15 N). The isotopic composition (δ 15 N) of sample was measured as (Mariotti 1983): Where, 15 N/ 14 N sample and 15 N/ 14 N standard are the abundance ratios of the plant and air as standard, respectively. The percent N derived from atmospheric N 2 (%Ndfa) was quantified as (Shearer & Kohl, 1986;Unkovich et al. 2008): Where δ 15 N ref is the 15 N natural abundance of non-N 2 fixing reference plant, δ 15 N leg is the 15 N natural abundance of the legume (soybean), and B value is 15 N natural abundance of soybean plants which were solely dependent on symbiotic N 2 fixation for their N nutrition. The mean δ 15 N values of the reference plants for each location were used to estimate the percent N derived from the atmosphere in soybean plants.
The amount of N-fixed was calculated as (Maskey et al., 2001): The N content of shoots was determined as the product of %N and sample dry weight (Pausch et al., 1996).

Statistical analysis
Statistical analysis was performed using STATISTICA (StaSoft Inc., Tulsa, OK, USA) package. Plant growth and isotopic data were tested for normality in distribution and then subjected to a one-way analysis of variance (ANOVA) for each location, and a 2-Way ANOVA to compare the means of results between the two regions. Where there was a significant difference, the means were separated using the Duncan's multiple range test at p ≤ 0.05.

Soil chemical properties
The soil samples collected from each of the study sites (Amaro, Boricha, Dorebafano, Mambuk and Pawe) were analysed for pH, organic carbon (OC), electrical conductivity (EC), cation exchange capacity (CEC), total N and plantavailable P ( Table 1). The pH of soils from the study sites ranged from pH 5.46 to 6.18. The EC of soils was low and ranged from 1 to 20 mS m −1 . The organic carbon content ranged from 1.29 to 2.15%. The plant available P ranged from 0.35 to 24.7 mg kg −1 . According to the Bray-1 P test rating, the soil from Dorebafano (24.7 mg kg −1 ) had medium P fertility, Boricha (10.2 mg kg −1 ) low P, Amaro (6.7 mg kg −1 ), Mambuk (0.35 and 0.44 mg kg −1 ) and Pawe (0.60 mg kg −1 ) very low P (Jones, 2002). According to Havlin et al. (1999), soil fertility can be classified according to the soils total nitrogen content (%), as very low (<0.1), low (0.1-0.15) medium (0.15-0.25), and high (>0.25). The soils from all the study sites contained low soil total nitrogen, except for Mambuk forest which had medium N concentration (0.188%). Nitrate levels of the soils were low, except for Mambuk and Amaro with moderate concentrations. The CEC of soils from Pawe (32.67) and Mambuk (37.90) were high, while the rest showed a moderate CEC (cmol (+) kg −1 ) ( Table 1; see Hazelton & Murphy, 2007).

δ 15 N of reference plant species and B value
Estimating the percentage N derived from the atmosphere (%Ndfa) by nodulated legumes using the 15 N natural abundance technique requires a value for soil N uptake by the legume. In this study, Sorghum bicolour (L.), Eragrostis tef and Triticum aestivum were used as reference plants. 3.3 Plant growth, shoot N concentration and content, and symbiotic performance of soybean genotypes at each location

Benshangul Gumuz region
Pawe At Pawe, a 1-Way ANOVA showed significant differences in plant growth and symbiotic performance between and among the six soybean genotypes.
Shoot, root and whole-plant dry matter at Pawe varied significantly among the genotypes (Table 3). Shoot DM ranged from 3.30 g plant −1 for the genotype Coker-240 to 5.66 g plant −1 for TGx-3326-44 (Table 3). Although the root DM was similar for the test soybean genotypes, TGx3326-44, AGS-71 and Wegayen produced the most whole-plant biomass (Table 3).

Southern nations nationalities and People's region (SPNNR)
Amaro At Amaro shoot biomass was similar for the seven soybean genotypes tested, and because the differences in root dry matter were small (though significant), whole plant biomass was also similar for all the genotypes (Table 3). Shoot N concentration and content were also similar for the seven soybean genotypes.
The δ 15 N of soybean shoots was generally high in Amaro soil, and ranged from +3.67‰ to +4.87‰. As a result, percent N derived from fixation was low, and varied from 5.37% for Awassa-95 to 22.2% in TGx-3326-44. The amount of N-fixed by soybean genotypes in Amaro soil was very low, and ranged from 6 mg plant −1 for Awassa-95 to 19.5 mg plant −1 in TGx-3326-44. Due to the generally high N concentration in Amaro soil, the genotypes took up more N from soil than symbiosis. Soil N-uptake was high in the low-fixing Awassa-95 (101.0 mg plant −1 ), and much lower in TGx-3326-44, which fixed the most symbiotic N (Table 3).
Boricha Only five soybean genotypes were evaluated in Boricha soil. The dry matter yield of shoots, roots and whole-plants were similar for all the genotypes, just as shoot N concentration and content were also not different in Boricha soil (Table 3).
As found for Amaro soil, shoot δ 15 N values were also high in Boricha soil, and ranged from +3.99‰ for Wegayen to % m g p l a n t −1 ‰ % m g p l a n t −1 +5.05‰ in Awassa-95. As a result, the %Ndfa values were rather low. Genotype Wegayen recorded the highest %Ndfa value of 25.5%, followed by genotypes AGS-71 (21.5%) and Coker-240 (19.7%). These 3 genotypes fixed relatively more symbiotic N than Awassa-95 and Williams (Table 3). Of the five genotypes, Coker-240 took up more N from soil (Table 3).
Dorebafano Only four soybean genotypes were studied in Dorebafano soil. Plant growth, measured as shoot, root and whole-plant dry matter were similar for all four genotypes. However shoot N concentration and content differed significantly, and were much greater in AFGAT, and lower in Gishama (Table 3). The δ 15 N of soybean shoots was very high for all four genotypes, and ranged from +4.32‰ for TGx-3326-44 to +5.77‰ in Crowford (Table 3). Due to the large δ 15 N values, the %Ndfa of plant shoots was generally very low, and ranged from 7.5% for Crowford to 26.8% in TGx-3326-44. The amounts of N-fixed by the soybean genotypes were low, and ranged from 6.9 mg plant −1 for Crowford to 23.1 mg plant −1 in TGx-3326-44 (Table 3). Relative to N-fixed, soil N-uptake by soybean genotypes was very high, with the highest uptake recorded by AFGAT and the lowest by TGx-3326-44 (Table 3).

Plant growth and symbiotic N nutrition across soil types
Combined analysis was performed to compare plant performance across soil types. The data showed marked differences in organ and whole-plant biomass as well as symbiotic performance in soybean planted in the different soils. Amaro soil supported greater shoot growth and hence higher whole-plant biomass (Table 4). Plants in Mambuk soil showed the least organ growth, and hence lower whole-plant dry matter yield. Although shoot N concentration was greater in Boricha soil, N content was similar in all six soils.
Shoot δ 15 N values ranged from +0.82‰ for Pawe to +5.11‰ in Dorebafano. Shoot δ 15 N was highest in the soil from Dorebafano, followed by Boricha, Amaro, and Mambuk, and lowest in soil from Pawe (Table 4). As a result, higher %Ndfa values were obtained in plants grown in Mambuk, Mambuk forest and Pawe soils. The lowest shoot %Ndfa was found in Amaro soil (Table 4). The amount of N-fixed in soybean genotypes was higher in soil from Mambuk, Mambuk forest and Pawe, while the least Nfixed was in Amaro soil. Soil N-uptake varied from 33 mg plant −1 in Mambuk soil to 87 mg plant −1 in Boricha soil (Table 4).
3.5 2-way ANOVA comparison of growth and symbiotic performance of soybean planted in soils from two regions A combined 2-Way ANOVA, comparison of the 11 soybean genotypes grown in soils collected from the two regions (BGR and SNNPR) was done (Table 5).
Soybean plants grown in soil from SNNPR exhibited greater shoot (4.90 g plant −1 ), root (1.78 g plant −1 ) and whole-plant (6.68 g plant −1 ) biomass when compared to soil from BGR (Table 5). There were however no significant differences in shoot N concentration and content. But shoot δ 15 N was lower in the BGR than SNNPR region. As a result, %Ndfa was much higher in soil from BGR region ( Table 5). Amount of N-fixed was expectedly greater at BGR than SNNPR region, which had higher soil N uptake (Table 5).
Of the 11 soybean genotypes tested in soils from the two regions, Awassa-95 produced the most shoot biomass and hence higher whole-plant dry matter. However, Gishama and Crowford recorded the lowest shoot biomass, and Clark-63 K the least whole-plant dry matter (Table 5). Genotypes AFGAT and Coker-240 showed much higher shoot N concentration. As a result, AFGAT recorded the most N content, followed by Awassa-95.
Shoot δ 15 N was higher in Awassa-95 and Crowford, and lowest in TGx-3326-44 and AGS-71. The %Ndfa values of soybean shoots ranged from 21.9 for Crowford to 43.8% in TGX-3326-44 (Table 5). The amounts of N-fixed also differed among the 11 soybean genotypes, and were much greater in genotypes AFGAT, AGS-71, TGx-3326-44 and Wegayen. The lowest amount of N-fixed was in Crowford, followed by Clark-63 K (Table 5). N-uptake from soil was higher than N obtained from symbiosis in all the genotypes tested (Table 4). But Awassa-95 took up more soil N, followed by AFGAT, with Gishama and TGx-3326-44 as the least ( Table 5).
The region x genotype interaction was significant for shoot and whole-plant biomass, shoot N concentration, δ 15 N, %Ndfa, N-fixed and soil N-uptake (Table 5). In general, shoot biomass of Clark-63 K, Coker-240, Nova and Williams was greater when planted in soils from SNNPR than BGR. Conversely, AGS-71, Awassa-95 and Wegayen produced more shoot biomass in soils from BGR than SNNPR (Fig. 1A). Shoot biomass of the remaining genotypes was similar in the two regions. Due to differences in shoot biomass, whole-plant dry matter was higher in six of the 11 genotypes grown in soils from SNNPR when compared to BGR (Fig. 1B). Only Awassa-95 recorded higher whole-plant biomass in soil from BGR over SNNPR. Shoot N concentrations were generally similar for the two regions, except for genotypes Clark-63 K and Nova, which showed higher shoot N concentration in soils from BGR than SNNPR (Fig. 1C).
Shoot δ 15 N of soybean genotypes was markedly greater in soils from SNNPR than BGR for all 11 genotypes (Fig. 1D). As a result, %Ndfa values were significantly lower in soybean genotypes planted in soils from SNNPR than BGR (Fig. 1E). The amount of N-fixed was also expectedly lower in all soybean genotypes raised in soils from SNNPR than BGR (Fig.  1F). The lower N 2 fixation in soils from SNNPR region was due largely to higher soil N uptake. Nine out of the 11 genotypes took up more N from SNNPR soil when compared to BGR (Fig. 1G).

Correlation analysis
Whether dealing with soil from Pawe, Mambuk forest, Amaro, or Dorebafano, shoot δ 15 N was significantly correlated with soil N uptake (Fig. 2 A, B, C and D). As a result, percent N derived from fixation was markedly correlated with soil N uptake by soybean grown in Pawe, Mambuk forest, Amaro and Dorebafano soils (Fig. 3 A, B, C and D).

Discussion
This study assessed the performance of soybean symbiosis with native rhizobial populations in soils from different locations in Ethiopia, using the 15 N natural abundance technique. The accuracy of the 15 N natural abundance technique depends on the levels and uniformity of the 15 N isotope in the soil and the choice of B value. δ 15 N values of reference plants > +6.0‰ are preferred, even though values as low as +2‰ can still be useful, as the δ 15 N is less affected by temporal and depth variations in agricultural soils (Herridge and Rose 2000;Okito et al. 2004;Unkovich et al. 2008). It has been recommended that more than one non-N 2 -fixing reference plants be used for estimation of N 2 fixation using the 15 N natural abundance method (Unkovich et al. 2008). Therefore, 3 nonlegume reference plants [Sorghum bicolour (L.), Eragrostis tef and Triticum aestivum] were used in this study, with average δ 15 N values that varied from +2.72 to +6.33‰, a range suitable for estimating soil N-uptake by legumes (Unkovich et al. 2008).
The 1-Way ANOVA found significant differences among the soybean genotypes planted in different soils (Table 3). While this variation could be attributed to genetic differences among the genotypes, soil N uptake and/or rhizobial strain effect is equally important (Herridge and Rose 2000). Other studies have similarly found variations in plant growth, nodulation and N 2 fixation of soybean genotypes due to differences in N 2 -fixing efficiency of the microsymbiont (Pule-Meulenberg et al. 2011;Salvucci et al. 2012).
There was a location-specific effect of soil on plant growth and symbiotic N nutrition in soybean. For example, wholeplant biomass was greater in soybean planted in soil collected from Amaro, followed by Boricha, Dorebafano, Pawe, and % m g p l a n t −1 ‰ % m g p l a n t −1 mg plant −1 Values (Mean ± SE) with dissimilar letters in a column are significantly different at p ≤ 0.05 (*), p ≤ 0.01 (**), p ≤ 0.001 (***) % m g p l a n t −1 ‰ % m g p l a n t −1 Mambuk (Table 4). The soil properties of these locations were variable due largely to differences in nutrient concentrations (Table 1). Soybean plants grown in soil from Boricha had the highest shoot N concentration, with the lowest in soil from Amaro. Shoot N content was comparable across soil types although plants grown in soils from Pawe, Boricha and Mambuk forest had relatively higher shoot N content when compared to those from the other soils (Table 4). Shoot δ15N values ranged from +0.82‰ for Pawe to +5.11‰ in Dorebafano. Consequently, greater %Ndfa of soybean was obtained in plants grown in Mambuk soil, followed by Pawe. The lowest %Ndfa value was in Amaro soil (Table 4). The amount of N-fixed followed a similar trend. Though soils from Mambuk had medium nitrate levels they supported high levels of N 2 fixation in soybean (Table 4). It has been shown that some soybean genotypes are capable of developing nitrate-tolerant symbioses especially when the rhizobial population is high (Herridge and Rose 2000). Soybean plants grown in soils from the BGR region recorded the lowest δ 15 N values, and as a result derived the highest N from fixation when compared to SNNPR (Table 5). This could be attributed to the presence of high population of compatible very effective indigenous soil rhizobia that formed functional N 2 -fixing root nodules. Though plant biomass was greater for soybean grown in SNNPR soils, they recorded low %Ndfa. This could be due to the fact that the plants were meeting their N requirements from soil since that location (Amaro) showed moderate levels of soil nitrate and high organic C concentration (Amaro) ( Table 1). In high nitrate soils, most legumes including soybean, tend to take up moderate to An assessment of plant growth and N 2 fixation in soybean genotypes grown in uninoculated soils collected... large amounts of soil N at the expense of fixation, often resulting in low %Ndfa and reduced amounts of N-fixed (Herridge and Rose 2000). A recent study also found nitrate inhibition of nodulation and N 2 fixation in nodulated soybean receiving 5 mM NO 3 − (Mbah and Dakora 2017). It has been found that large amounts symbiotic N is produced if the legume demand for N exceeds soil N supply, under optimal conditions (Salvagiotti et al. 2008). The significant interaction between soil and soybean genotype for plant DM, shoot N concentration, δ 15 N, %Ndfa, Nfixed and soil N-uptake clearly indicate the modulating effect of soil factors on legume N 2 fixation (Table 5). Soybean genotypes Coker-240, Clark-63 K, Williams, Nova, TGx-3326-44 and Gishama had greater whole-plant biomass values in SNNPR than BGR soils (Fig. 1 B). All genotypes studied also showed significantly greater δ 15 N values in SNNPR compared to BGR soils ( Fig. 1 D), which led to higher N derived from fixation in soils from the BGR. For example, 8 of the 11 soybean genotypes recorded higher %Ndfa values in BGR compared to SNNPR soils, a clear indication of more effective rhizobia in the former location. Soybean is estimated to have %Ndfa ranging from 0 to 97% and N contribution of 0 to 450 kg per hectare, depending on the host-rhizobia symbiosis, the strength of the symbiotic N sink, the concentration of endogenous soil N, and other soil factors (van Kessel and Hartley 2000;Unkovich and Pate 2000). Thus, the wide variation in %Ndfa of soybean in this study could be linked to these factors. Not only was soil N uptake greater in SNNPR region, but N uptake by individual soybean genotypes was also large, ranging from 53 mg plant −1 in Nova to 96 mg plant −1 in Awassa-95. In fact, the negative correlation obtained between soil N uptake and %Ndfa (Fig. 3 D) clearly indicate that soil N was a major factor affecting soybean N 2 fixation in the soils used. It has also been reported that N 2 fixation is often underestimated by about 24% when below-ground N is ignored in nodulated roots (Unkovich and Pate 2000; Salvagiotti et al. 2008). As this study did not include N in nodulated roots, the amount of N-fixed in the soybean genotypes could have been underestimated.
Although this study did not assess the existing cropping systems at the sites of soil collection, both legume-based and cereal-based rotations have been reported to alter the populations of soybean rhizobia in soil, with the former stimulating more rhizobial growth than the latter (Kumar et al. 2017). Thus, some of the differences in nodulation and N 2 fixation could possibly be attributed to the type of rotations used by farmers in the BGR and SNNPR regions of Ethiopia, especially where there were location-specific effects of soil on symbiotic performance. Interestingly, even where there was no history of soybean cultivation as found at Mambuk forest soil, there were still native rhizobia that could nodulate and fix N 2 in soybean, probably suggesting alternate host plants that haboured rhizobia in their rhizospheres in the absence of soybean, the homologous host. In a separate study, rhizobia isolated from root nodules of soybean planted in these soils were found to differ in their symbiotic performance (data not shown), and this could also help to explain the variations found in soybean nodulation and N 2 fixation in this study.
Although soil factors such as P, CEC, and EC revealed no correlation with soybean symbiotic performance, shoot δ 15 N was significantly correlated with soil N uptake by soybean in soil from Pawe, Mambuk forest, Amaro, or Dorebafano ( Fig.  2 A, B, C and D), suggesting greater dependence on soil N than symbiosis for N nutrition in those soils. As a result, percent N derived from fixation was reduced by uptake of soil N from all four locations through inhibition of N 2 fixation (Fig. 3 A, B, C and D; see Ayisi et al. 2000).
Taken together, the differences in symbiotic functioning among soybean genotypes could be attributed to a range of factors, which include i) the presence and/or competitiveness of native soybean rhizobia in the soil, ii) the level of strain symbiotic efficacy, iii) mineral concentrations in soil, especially N (Dakora and Keya 1997), and iv) the presence of bacteriophages in soil (Msimbira et al. 2016). There was also clear evidence of mineral N inhibition of N 2 fixation in soybean planted in soils from various sites.