Introduction

Low levels of precipitation in the Midwestern USA during the 2012 growing season led to lower yields and other concerns for many soybean producers. One of these areas of concern was the impact of the drought on soil microbial population size, most importantly Bradyrhizobium japonicum, and whether the use of seed-applied inoculants would be warranted for the following growing season. Soybean lives in a symbiotic (mutually beneficial) relationship with B. japonicum. The soybean plant provides nutrients and a protective growing environment for the rhizobia. In turn, the rhizobia “fix” atmospheric nitrogen (N2) into ammonia (NH3), which can then be used by the soybean plant. For this relationship to exist and benefit soybean and B. japonicum, effective N2-fixing bacteria must be present in the soil in relatively high numbers in the vicinity of the planted seed (Conley and Christmas 2005).

Both establishment and activity of the legume-rhizobium symbiosis have been found to be sensitive to drought stress (Kirda et al. 1989; Sprent 1971). However, most of these studies focus on the host plant or on nodule function, but little research has been published on the impact of soil desiccation (e.g., drought) on B. japonicum population size. Pena-Cabriales and Alexander (1979) reported a biphasic decline in Rhizobium japonicum and Rhizobium leguminosarum numbers in soils undergoing drying. The initial rapid phase of R. japonicum and R. leguminosarum loss coincided with major water loss (simulated drought) whereas the secondary and subsequently slower decline in numbers was governed by the soil water content of specific soils. In a parallel study, Osa-Afiana and Alexander (1982) reported similar results when comparing strains of R. japonicum and cowpea rhizobia under desiccation. As one would expect, variation exists among rhizobia as well as individual isolates within species in their response to soil desiccation although significant losses have been observed regardless of species or isolate (Foulds 1971; Trotman and Weaver 1995). This information suggests that soybean growers would expect a reduced B. japonicum population size following a drought year though to what extent have yet to be determined. Quantifying the impact of drought on B. japonicum population size and determining whether use of a seed-applied inoculant would be warranted in a year following a drought year could have a significant impact across millions of hectares of soybean production in the Midwest. With this study, we specifically aimed to answer the following questions: (1) How does drought impact soil population size of B. japonicum? and (2) will drought conditions lower B. japonicum population size to a level which might affect soybean growth and development?

Materials and methods

Soil collection

Two contrasting soil types used in this study were collected in the fall of 2012 from regions of Wisconsin that experienced drought conditions during the summer 2012 (US Drought Monitor 2012). The first was a Plano silt loam (fine-silty, mixed, mesic Typic Argiudoll) collected at the Arlington Agricultural Research Station (43° 18′ N, 89° 20′ W), and the second was a Plainfield sand (mixed, mesic Typic Udipsamments) collected at the Hancock Agricultural Research Station (44° 06′ N, 89° 33′ W). For each soil type, soil was collected only from the top 15 cm of the soil profile. Prior to use in the greenhouse, each soil type was sifted with a 1.3-cm mesh sieve in order to remove large debris and/or rock material.

Inoculum preparation

B. japonicum USDA 110sl101-nptII-lacZ cells were grown on AG medium for 5 days and then suspended in water. A dilution curve previously determined was used to obtain a first estimate of the bacterial concentration using a spectrophotometer (OD 600 nm). Dilutions down to 107 bacteria mL−1 were then prepared. The concentration of these dilutions was confirmed by plate counts. Briefly, cell preparations were diluted to reach an estimated concentration of 103 bacteria mL−1, and 100 μL was plated on AG plates. Three plates were prepared for each inoculum. After 5 days, the number of colonies per plate was quantified, and this confirmed the bacterial concentration in the inoculum.

Greenhouse experiments

Two separate greenhouse experiments were conducted at the University of Wisconsin—Madison during the spring and summer of 2013. In the first experiment, untreated soybean seeds (Channel 2305R2) were sterilized in 0.6 % bleach solution for 5 min and rinsed five separate times with autoclaved Milli-Q™ purified water. The sterilized seeds were germinated for 3 days at 22 °C in wet sterile perlite in a dark covered container.

Two-liter pots containing either the silt loam or sandy soil were then weighed. Bacterial inoculum (107 bacteria mL−1, see above) was then added to each pot in order to reach a concentration of 106 colony-forming units (CFU) g−1 soil. The inoculated soil was then mixed. Once the inoculated soil was prepared, three seedlings were transplanted to each pot. Three seedlings were used to ensure growth of at least one plant, and the number of seedlings was reduced to one, where appropriate, 14 days after transplanting. After transplanting, pots were subject to four water regimes for an 8-week period: FC is the maintained field capacity for all 8 weeks, NO means no water added for all 8 weeks, START means no water added during the first 4 weeks then the maintained field capacity during the last 4 weeks, and STOP is the maintained field capacity during the first 4 weeks then no water added during the last 4 weeks. Each water regime included three replications. Greenhouse conditions included a constant temperature at 25 °C with 16 h of light and 8 h of darkness.

Methods in the second experiment were modified based on the results from the first experiment. In the second experiment, soybean seedlings and potting and inoculation methods were followed as outlined in the first experiment. Pots were subject to two water regimes for a 4-week period: (FC) maintained field capacity for all 4 weeks and (NO) no water added for all 4 weeks. Each water regime included three replications. Greenhouse conditions were similar to those described in the first experiment.

Soil sample collection

For both experiments, six samples of approximately 5 g of soil were collected weekly from each pot (total 30 g pot−1). Soil sample collection began at the time of inoculation. A 500-mg subsample was frozen at −20 °C until used for DNA extraction and quantification by quantitative real-time polymerase chain reaction (qPCR). A separate 10-g subsample was used to determine percent water content. In the second experiment, another separate 10-g subsample was used for the most probable number (MPN) technique.

Monitoring drought conditions

Drought conditions for each water regime were monitored by quantifying percent soil water content. Each 10-g soil subsample was weighed, incubated at 65 °C for 24 h, and weighed again. Percent water content for each sample was determined by taking the difference between the initial and dry weights, dividing by the initial weight, and multiplying by 100.

B. japonicum quantitative assessments

The population size of B. japonicum was determined using two different methods. The first method was performed using quantitative PCR, and this method was used for both experiments. Total DNA from the 500-mg subsample of soil from each pot was extracted using MoBio UltraClean®-htp 96 Well Soil DNA isolation kit (Mo Bio Laboratories, Carlsbad, CA). The overall quality and quantity of each DNA sample were checked with a NanoDrop™ 1000 spectrophotometer from Thermo Scientific (Waltham, MA). Absorbance at 260 nm was used as a measure of DNA quantity. DNA quality was checked using the 260:280 absorbance ratio. All quantitative PCR reactions were carried out in 96-well plates using a Bio-Rad CFX96 Touch™ real-time PCR system (Bio-Rad, Hercules, CA). The primer set used (forward primer (5′-GGTTTGGCGACTGTCTGTGGTC-3′); reverse primer (5′-TTCCACCATGTTGGAAAGAATGGTCC-3′)) targets the nodZ specificity gene of soybean-associated rhizobia (Furseth et al. 2010). Reactions consisted of a 20-μL volume, containing 2 μL of soil DNA template, 10 μL of iTaq Universal SYBR green mix (Bio-Rad, Hercules, CA), 1 μL of each primer (from 10-μmol stock), and 6 μL of autoclaved Milli-Q™ purified water. PCR cycling conditions consisting on initial denaturation at 95 °C for 30 s, followed by 40 cycles of 95° for 10 s, 58° for 20 s, and 72° for 30 s. Each sample consisted of three technical replications. Standards (i.e., known amounts of B. japonicum) and a water-negative control (i.e., 2 μL purified water substituted for soil DNA template) for each soil type (Fig. 1) were prepared from methods similar to Mbofung et al. (2011) and were included in each quantitative PCR run. Standards were made beginning with a tenfold dilution series of a B. japonicum suspension starting at 108 CFU mL−1. Two milliliters of dilutions 107 to 101 were added to 2 g of sterilized silt loam and sand in 15-mL Falcon tubes and mixed. Each soil suspension was lyophilized for 24 h. Total DNA was extracted from three 250-mg samples from each dilution using the methods described above. The population size of B. japonicum was calculated from the PCR cycler-calculated linear regression coefficients derived from the standards within each run and adjusted to CFU per gram dry soil. Adjusting the population size to 1 g of dry soil was used to minimize confounding effects between water regimes.

Fig. 1
figure 1

Correlation between Bradyrhizobium japonicum population (log10 CFU g−1 soil) and quantitative real-time polymerase chain reaction (qPCR) quantification cycles (Cq) for the silt loam (r = −0.970) and sandy (r = −0.967) soil types

Full size image

The second method performed was the MPN, described by Vincent (1970), and this method was only used in the second experiment. To begin, soybean seedlings were surface sterilized and grown using the methods outlined above. One seedling was aseptically transferred to a 2-cm diameter by 20-cm long glass tube so that the radicle was supported between the glass and an inserted rolled piece of filter paper and the shoot extended outside of the tube. Ten grams of soil, collected from one rep of each water regime for each soil type at the end of the experiment, was mixed with 90 mL of sterile water and placed in a shaker for 30 min. Tenfold serial dilutions (10−1 to 10−10) of each soil sample suspension were made. One milliliter from a dilution was used to inoculate one soybean seedling by directly applying the soil suspension on the roots. This was repeated for all ten dilutions and included four replications for each dilution. An inoculum-free control was also included by using sterile water in place of the soil solution. Each tube was capped with aluminum foil, and tube racks were also wrapped in aluminum foil to keep the roots dark. The tubes were then moved to a transparent plastic container to limit evaporation. The plants were grown for 4 weeks in a growth chamber and watered frequently with N-free Fahraeus medium (Barker et al. 2006). After the 4-week period, the number of plants with one or more nodules was tallied and the MPN of B. japonicum in the soil sample was determined using Table 3.5a from Vincent (1970) with the parameters n = 4 (replications) and s = 10 (dilution steps).

Statistical analyses

B. japonicum population size estimates from the quantitative PCR method from both experiments were used for statistical analysis. Analyses were performed in SAS version 9.3 (SAS Institute Inc., Cary, NC) using PROC MIXED. Population size was subject to a repeated measures analysis of variance (ANOVA) using the first-order autoregressive covariance structure. Models were constructed individually for each soil type, and population size was log transformed to meet model assumptions. Degrees of freedom were calculated using the Kenward-Rogers method (Littell et al. 2006). Comparisons between time points within and between water regimes were made using the differences of least square means, which followed a t-distribution (two-tailed) at α = 0.05.

Based on the MPN population size estimates and standard errors determined from Vincent (1970) at the end of the second experiment, a 95 % confidence interval (CI) was calculated for each water regime within each soil type. The confidence intervals were compared between water regimes for each soil type.

Results

Drought conditions

Percent soil water content responded to each water regime as expected in the first experiment (Figs. 2 and 3). Soil maintained at field capacity displayed average percent water content of 11.2 ± 0.8 % (SE) for the sand and 21.9 ± 0.6 % for the silt loam. Similar soil water content was observed for the START and STOP treatments during appropriate addition of water for each treatment. Water content dropped to <1 % for both the sand and silt loam in the NO treatment. For both soil types, plants under the NO, START, and STOP treatments were visually severely stressed during each time period of no water addition and died within 2 weeks after watering was discontinued for the respective treatment (Fig. S3).

Fig. 2
figure 2

Bradyrhizobium japonicum population response to the FC (upper left), NO (upper right), START (lower left), and STOP (lower right) water regimes for the silt loam soil from the first experiment (circles). Population size was calculated from the qPCR cycler-calculated linear regression coefficients derived from the standards (i.e., known amounts of B. japonicum) within each run. Population size was adjusted to CFU g−1 dry soil, and population data are expressed as log10 CFU g−1 dry soil. Corresponding soil water content (%) for each water regime is also displayed (triangles). Shown are means ± SE. Means for population include nine observations (three reps × three tech reps) and three observations (three reps) for soil water content estimates

Full size image
Fig. 3
figure 3

Bradyrhizobium japonicum population response to the FC (upper left), NO (upper right), START (lower left), and STOP (lower right) water regimes for the sandy soil from the first experiment (circles). Population size was calculated from the qPCR cycler-calculated linear regression coefficients derived from the standards (i.e., known amounts of B. japonicum) within each run. Population size was adjusted to CFU g−1 dry soil, and population data are expressed as log10 CFU g−1 dry soil. Corresponding soil water content (%) for each water regime is also displayed (triangles). Shown are means ± SE. Means for population include nine observations (three reps × three tech reps) and three observations (three reps) for soil water content estimates

Full size image

Percent soil water content in the second experiment responded again as expected to each water regime (Figs. S1 and S2). Soil maintained at field capacity displayed average percent water content of 17.0 ± 0.6 % for the sand and 22.3 ± 0.8 % for the silt loam. Soil water content again dropped to <1 % for both the sand and silt loam in the NO treatment. For both soil types, plants under the NO treatment were severely stressed and died by the end of the second experiment.

Evaluation by quantitative PCR

A strong linear relationship was found as expected between B. japonicum population size (log10 CFU g−1 soil) and quantitative PCR quantification cycles (Cq) for the silt loam (r = −0.970) and sandy (r = −0.967) soil (Fig. 1).

While the repeated measures analysis allowed us to compare the difference between any two time points, we placed emphasis on how the population size responded from inoculation to week 8 postinoculation (PI) for all four water regimes and how the population size responded from inoculation to week 4 PI, week 4 to week 5 PI, and week 4 to week 8 PI for the START and STOP water regimes (Figs. 2 and 3). Population size response to water regime was mixed for each soil type.

For the silt loam (Fig. 2), a significant decrease in population size was found between inoculation and week 8 PI for the START (0.532 log10 CFU g−1 dry soil, P = 0.02) and NO (0.526 log10 CFU g−1 dry soil, P = 0.02) treatments, while no difference for the same comparison was found for the FC (P = 0.89) and STOP (P = 0.81) treatments. From inoculation to week 4 PI in the START and STOP treatments, no difference was found for both treatments (P = 0.27; P = 0.71, respectively). Population size from week 4 to week 5 PI significantly decreased for the START treatment (0.459 log10 CFU g−1 dry soil, P = 0.04) but did not change for the STOP treatment (P = 0.30). From week 4 to week 8 PI, population size significantly decreased for the START treatment (0.771 log10 CFU g−1 dry soil, P < 0.01) and did not change for the STOP treatment (P = 0.89).

For the sandy soil (Fig. 3), a significant decrease in population size was found between inoculation and week 8 PI for the FC (0.430 log10 CFU g−1 dry soil, P = 0.04), START (0.512 log10 CFU g−1 dry soil, P = 0.01), and STOP (0.788 log10 CFU g−1 dry soil, P < 0.01) treatments, but no difference for the same comparison was found for the NO treatment (P = 0.34). From inoculation to week 4 PI, no change was found for the START treatment (P = 0.99), but a significant decrease was found for the STOP treatment (0.838 log10 CFU g−1 dry soil, P < 0.01). Population size did not change from week 4 to week 5 PI for both the START (P = 0.15) and STOP (P = 0.54) water regimes. From week 4 to week 8 PI, the population size decreased (0.515 log10 CFU g−1 dry soil, P = 0.01) for the START treatment, but no change was found for the STOP treatment (P = 0.81).

Evaluation by the MPN method

Nodules were observed on soybean plants from both water regimes for each soil type and were not found on the inoculum-free control (Table 1). The 95 % CI for water regimes was 5.37–6.53 (FC) and 4.37–5.53 (NO) in silt loam and 6.14–7.30 (FC) and 5.60–6.76 (NO) in sandy soil. The overlap between FC and NO water regimes in both soils suggested no decrease in B. japonicum population size after 4 weeks of simulated drought (Table 1). Quantitative PCR estimates from the same time point (week 4 PI) also showed no statistical difference between the FC and NO water regimes for the silt loam (P = 0.07) and sandy soils (P = 0.37) (Figs. S1 and S2).

Table 1 Quantification of Bradyrhizobium japonicum from the second experiment by most probable number (MPN)
Full size table

Discussion

The high frequency of soybean used in crop rotations throughout the Midwest has increased rhizobia population size to high levels in most soils (Furseth et al. 2012). Starting B. japonicum population size used in this study was consistent with soil rhizobia population estimates found across soybean-growing regions of the Midwest (Berg et al. 1988; Furseth et al. 2011, 2012). In the first experiment when comparing the population size estimates at inoculation to the end of the experiment (week 8 PI), the results showed that 4 weeks of simulated drought could have the same effect as 8 weeks of drought (i.e., NO compared to START) for the silt loam soil, but this was not exactly the same for the sandy soil. Four weeks of drought (i.e., START and STOP water regimes) for the sandy soil had more impact on B. japonicum population size than 8 weeks of drought. While the second experiment only used 4 weeks of drought, population sizes did not statistically differ between the FC and NO treatments for both soil types for both quantification methods. Although we observed statistically significant responses, these differences were less than 1 log10 CFU g−1 dry soil and therefore can be considered a minimal response, biologically. We hypothesized dramatic decreases (i.e., greater than 1 log10 CFU g−1 dry soil) in B. japonicum population size due to drought, with decreases likely dependent on the length of the drought. In addition, we observed in some cases where increases in soil water negatively impacted population size. This result was unexpected, and we feel that the addition of water potentially leached the bacteria from the system. At first glance, our results seem to contradict older studies examining soil rhizobia under desiccation (Osa-Afiana and Alexander 1982; Pena-Cabriales and Alexander 1979). However, the conditions used in these studies are different from those we used and those in soybean fields in the Midwest during the summer of 2012. Both studies reported that R. japonicum and R. leguminosarum numbers declined rapidly at first, coinciding with major water loss; after that, numbers declined much slower and were governed by the soil water content of soils. However, these studies were using very small soil samples (10 g) inoculated with very high inoculum doses (more than 9 log10 CFU g−1 soil) and dried in a 30 °C incubator. These treatments were probably harsher than conditions experienced by most rhizobia in a soybean field. In addition, even after 2 weeks of such a treatment, bacterial counts did not drop below 7 log10 CFU g−1 soil. Moreover, in both of those studies, rhizobia population size was monitored in the absence of a host plant. A host plant was used in the current study to represent a field scenario as closely as possible, and this may explain why population size did not dramatically decline. Plants release in their exudates a large repertoire of carbohydrates and other small molecules (Carvalhais et al. 2011). These molecules, along with the roots and nodules themselves, may have physically protected the bacteria, allowing for increased survival. The effect of environmental stress on B. japonicum population size has also been examined. Experiments conducted on land in Southern Wisconsin which had been flooded for an extended period of time in 2008 that found initial rhizobia population size at the time of planting in 2009 were sufficient for optimum soybean growth and development (Furseth et al. 2011). According to Furseth et al. (2012), additional environmental factors such as soil pH (Keyser and Munns 1979), soil nitrate levels (Gibson and Harper 1985), and soil phosphorous content (Keyser and Munns 1979) can also influence soil rhizobia population size and may also serve as potential explanations as to why we did not observe dramatic decreases in populations due to drought.

One question we aimed to answer in this study was to determine if drought conditions would lower B. japonicum population size to a level which might affect soybean growth and development. Although no formal soil rhizobia thresholds have been developed, we compared our results to previous studies which have examined soybean yield response from inoculant use to speculate whether a possible positive yield response from inoculant use would occur with our rhizobia population size estimates after experiencing drought. Research using MPN found little to no yield response from rhizobia inoculant use when rhizobia population size was above 2.0 log10 cells g−1 soil (Thies et al. 1991). Using quantitative PCR methods, Furseth et al. (2011) reported that initial average rhizobia population size at planting (10,000 rhizobia g−1 soil) was adequate for optimum soybean growth, and the use of a rhizobia inoculant would not be required. None of our population size estimates at the end of the first and second experiments were decreased to levels lower than 10,000 rhizobia g−1 soil (i.e., 4.0 log10 CFU g−1 dry soil) for either of our quantification methods. However, drought conditions extending longer than 8 weeks may warrant inoculant use, but these were not tested in our study. Based on the results from this study, we feel soybean growers would not be required to use a rhizobia inoculant in a year following a drought year. Use of an inoculant following drought conditions would certainly not hurt soybean production, but for growers trying to minimize input costs, use of an inoculant would not be required under those conditions.