Meta-analysis of the effects of Rhizobia inoculants and phosphorus fertilizer on soybean nodulation in Africa

1 National Academy of Agriculture Green Development, Department of Plant Nutrition, College of Resource and Environmental Science, China Agricultural University, Beijing 100193, China. 2 Department of Plant Production Sciences and Technologies, Faculty of Agriculture, Environment and Food Systems, University of Zimbabwe, Zimbabwe. 3 Department of Development Studies, College of Humanities and Development Management, China Agricultural University, Beijing 100193, China.


INTRODUCTION
The African population was expected to double in the next 40 years (Cleland, 2013), raising food insecurity especially in the sub-Saharan region where 239 million people are experiencing dire undernourishment (FAO, 2020). Sustainable intensification and integrated approaches are therefore needed to increase the *Corresponding author. E-mail: tanglx@cau.edu.cn. Tel. +8615198880909 Author(s) agree that this article remain permanently open access under the terms of the Creative Commons Attribution License 4.0 International License agricultural productivity of smallholders, improve household food security, and reduce poverty at the country level (Peoples et al., 1995;McNamara, 2009). Integrating legume crops, especially soybean, is an important approach in many cropping systems, as they can perform biological N 2 fixation (BNF), thus reducing N fertilizer requirements and improving grain yield (Peoples et al., 1995;Giller, 2001). Soybean was first domesticated in China and has been grown in other Asian countries like Japan and Korea for more than 3000 years as a primary source of vital proteins and vegetable oil (Giller, 2001;Herridge et al., 2008;Nishinari et al., 2014). Globally, soybean (Glycine max (L.) Merr.) is the world's largest grown legume crop (Giller, 2001), accounting for 50% of the worldwide legume crop area and 68% of global crop production (Herridge et al., 2008). However, there is no clear evidence of when soybean was first introduced to Africa (Mpepereki et al., 2000), but its nodulation with indigenous Rhizobia in African soils was first ascertained by Corby (1967).
The ability of symbiotic rhizobial bacteria to fix atmospheric nitrogen in legume plants can improve grain yield without applying nitrogen fertilizer (van Heerwaarden et al., 2017). Herridge et al. (2008) reported that soybean can fix more than 16 million tons of N annually, which is 77% of the N fixed by legume crops. Soybean has been reported to fix 80% of its nitrogen requirements (Smaling et al., 2008). Bradyrhizobium strains are commonly used in soybean inoculation worldwide (Chianu et al., 2011;Chen et al., 2015). In Africa, Rhizobia inoculants have been used to control the effects of debilitating soil fertility and high fertilizer costs incurred certainly by smallholders. They became an affordable and effective agronomic approach in improving yield and promoting sustainable agriculture (Dakora and Keya, 1997;Paynel et al., 2008). Ronner et al. (2016) found that Rhizobia inoculants increase soybean yield at a lower financial cost compared to chemical N fertilizers, thus benefiting the resource-poor farmers. Despite a rapid expansion of soybean production in many African countries (Mpepereki et al., 2000) and wide use of inoculants, legume yields in the smallholder farming sector generally remain far below their potential (Ronner et al., 2016). The effectiveness of Rhizobia inoculants can be affected by factors like soil nutrient status, organic matter content, pH, salinity, temperature, drought, and managerial practices (Thilakarathna and Raizada, 2017). However, Ronner et al. (2016) mentioned that the soybean yields in Africa could be improved through the use of adaptive technologies like phosphate fertilizer and improved varieties to aid the Rhizobium inoculation approach. Phosphorus (P) is the second most important macronutrient required by the legume plants in the BNF among other crucial processes (Uchida, 2000). Symbiotic Rhizobium bacteria need P as the energy storage and transfer component (adenosine diphosphate (ATP) and 207 adenosine triphosphate (ATP) for the conversion of free N 2 to ammonium (NH 4 ), a N usable form by legumes (Dakora and Keya, 1997). Furthermore, P increases nodule number and size, and it promotes general root growth. Legumes need optimum P levels for maximum nitrogen fixation and to achieve high grain yield (Bashir et al., 2011). Since 1980, no meta-analysis was conducted to determine the extent to which Rhizobium inoculation and phosphorus fertilizer technologies have influenced soybean productivity under field conditions in Africa. This study aimed to review various researches conducted in Africa to understand the effectiveness of rhizobial inoculants, P-fertilizer, and their interaction on soybean performances. The following conceptual model (Figure 1) was suggested to predict the effects of Rhizobium inoculation, phosphate fertilizer, and other adaptive technologies that could further improve soybean yield.

Data collection
An extensive literature synthesis was performed based on robust published research articles in 1980-2020 downloaded from the ScienceDirect databases (https://www.sciencedirect.com) and Web of Science (http://apps-webofknowledge-com.vpn.cau.edu.cn). The search terms used as the main topics in both databases were Rhizobium, Phosphorus, Soybean OR Glycine max, Nodulation, and Grain yield. A total of 441 research articles were obtained from ScienceDirect and 170 from Web of Science. Google Scholar provided additional articles. Only 86 articles were retained after excluding duplicates and exploitation of the titles and abstracts' relevance to the subject of the work for further screening. A study had to meet six requirements for its consideration in the dataset. They included: being conducted under rainfed or irrigated field conditions; assessing the effect of any strain of commercial Rhizobium inoculant or chemical phosphorus fertilizer, and/or both on nodulation characteristics and grain yield; presence of a control to either Rhizobium inoculation or phosphorus application; having every treatment being repeated at least three times; being conducted in an African country; having been published between 1980 and 2020.
The different characteristics of the nodulation consisted of nodule number, nodule dry weight, and shoot dry weight. A database with 396 data points extracted from 29 qualified peer-reviewed articles based on the aforementioned criteria was compiled. Studies with a sample size of less than 2 were excluded from analyses because they would have resulted in small size effect (Viechtbauer, 2010). 26 out of the 29 researches retained were conducted in the sub-Saharan-Africa region (SSA) and the other 3 in the Saharan region ( Figure 2). Treatment variances, standard deviations, or standard errors were disregarded as they were only presented in a few studies. In fact, only treatment mean values of nodule number, nodule dry weight, shoot dry weight, and grain yield were collected. Experiment details recorded include location/country, latitude, longitude, annual mean temperature and rainfall, soil type, pH, organic matter content, total available nitrogen, available phosphorus, inoculant strain, nodule number, nodule dry weight, shoot dry weight, grain yield, phosphorus application rates, to name a few. Control and experimental treatments' data were recorded as well as data on the interaction effect that was assessed in 4 studies only. The available data showed that the minimum and maximum  mean annual temperatures were 22 and 29C, minimum and maximum mean annual rainfall were 552 mm and 1300 mm, respectively. Data reported in one study but conducted in more than one country/location or in different years were considered derived from different studies. Tabular and graphical data were collected and in the latter case, Engauge Digitizer software version 12.1 was used for data extraction. Detailed information of the selected peer-reviewed studies is given in Table 1.

Data processing
Data on nodule number, nodule dry weight, shoot dry weight, and grain yield were pooled on per variable basis. Means, and standard Study location N deviations (STDEV.P) were calculated using Microsoft Excel package and the data were exported to R software for statistical analyses. The author's name, publication year, and sample sizes for each study were recorded for corresponding means and STDEV.P.

Statistical analyses
The effects of Rhizobium inoculation and phosphorus fertilizer on soybean performances were estimated using the random-effects model (REM) described by Viechtbauer (2010). The statistical analyses in this study were all performed in R version 4.0.2 (R Core Team, 2020), using the R program Metafor (Viechtbauer, 2010;Schwarzer et al., 2015). The Escalc() function contributed to calculate the standardized mean differences (SMD), which measure effect sizes that allowed comparing treatments to the controls (Viechtbauer, 2010). The REM was used to estimate the average true effect and the total amount of heterogeneity among the true effects (Viechtbauer, 2010;Schwarzer et al., 2015). Results were presented in form of forest plots with the Forest() function (Lewis and Clarke, 2001) and boxplots using the Plot() function (Viechtbauer, 2010). Heterogeneities within-study and betweenstudy were assessed using the I 2 statistic (Higgins and Thompson, 2002). The presence of publication bias and/or heterogeneity was determined by creating Funnel plots for both the inoculation and phosphorus variables (Sterne et al., 2001;Rothstein et al., 2006). A Funnel plot is a simple scatter plot for the study's estimated treatment effects (x-axis) against the measure of study size on the y-axis (Sterne et al.,, 2001). Trim and fill method using the Trimfill() function (Viechtbauer, 2010;Schwarzer et al., 2015) was applied on asymmetrical Funnel plots to determine the effect of missing studies on the overall outcome. The standardized mean difference for the study was plotted on the horizontal axis against the standard error on the vertical axis. A box plot using the Plot() function was realized to show the overall effects of rhizobial inoculation, phosphate fertilizer, and the interaction on soybean grain yield ( Figure 15). Bootstrapping was iterated 1000 times (95% CI) with R package Mosaic (Pruim et al., 2015) to improve the probability that the confidence interval was calculated around the relative yield mean. The frequency distribution plots were then plotted with the Ggplot() function.

Response of nodule dry weight to inoculation
The effects of inoculation on nodule dry weight are shown in Figure 3. The heterogeneity (I 2 ) of the combined studies was 92% (P=0.0001). The study by Argaw (2014) clearly showed that inoculation significantly favored nodulation contrasting seven other studies, which had 95% confidence interval (CI) lines touching or crossing no effect line, indicating that inoculation did not influence nodulation. The position of the diamond symbol on the graph, which showed the overall effect of inoculation, testified that inoculation significantly favored nodulation. Furthermore, the overall effect of Rhizobium inoculation had 95% CI of -2.02[-2.95, -1.08] (Figure 3). The purpose of the funnel plot was to indicate the level of bias in this Kanomanyanga et al. 209 study. Twenty studies showed an asymmetric distribution pattern, indicating the presence of bias ( Figure 4). However, after trim and fill (white dots), the overall result was not significantly affected.

Soybean nodule number
The studies had an I 2 of 75 % (P=0.0001) indicating that they were heterogenous. Rhizobium inoculation favored high nodule numbers (Argaw, 2014) ( Figure 5). The overall effect of soybean inoculation is shown by the diamond which is on the left side of the no effect line, indicating that inoculation significantly favored a high nodule number compared to no inoculation ( Figure 5). The overall effect of inoculation had 95% CI of -1.62[-2.17, -1.07] ( Figure 5). There was a little publication bias in 23 studies analyzed as shown by an asymmetric distribution pattern on the Funnel plot ( Figure 6). Trim and fill did not result in a significant change to the overall result.

Shoot dry weight response to rhizobial inoculants
An I 2 of 54 % (P=0.0001) resulted from the 13 studies showing the presence of heterogeneity. Two studies clearly demonstrated that Rhizobium inoculation increased shoot weight (Pulver et al., 1982;Okereke et al., 2001) (Figure 7). The overall effect of soybean inoculation significantly increased shoot dry weight compared to the control treatment, according to the position of the diamond, which is on the left side of no effect line. A 95% CI of -1.31[-1.74, -0.88] was produced for the overall effect of inoculation. The studies showed a slightly asymmetric distribution pattern on the Funnel plot, indicating very limited bias in the 13 studies ( Figure 8). Trim and fill of the missing studies did not bring meaningful change to the overall result.

Grain yield
The analyzed studies were heterogenous with an I 2 of 64 % (P=0.0001). Rhizobium inoculation significantly favored high grain yield compared to non-inoculated control with a symmetrical distribution pattern, indicating the absence of publication bias ( Figure 10). a 95% CI of -1.05 [-1.39, 0.72] (Figure 9).

Nodule number
The heterogeneity of the all the analyzed studies (I 2 ) was Table 1. Summary of the 29 studies in the meta-analysis on the effects of Rhizobium inoculation and P-fertilizer on soybean.
-Grain yield was significantly high in inoculant plus phosphorus fertilizer and phosphorus fertilizer only treatments (7.61 t/ha and 7.30 t/ha respectively) and lowest in Rhizobia inoculation only and control treatments (4.41 t/ha and 3.80 t/ha respectively).

Grain yield
The 10 studies were less heterogenous (I 2 = 20%) (P=0.0001), and the overall effect of phosphorus application on standardized differences had 95% CI of -1.55[-2.14, -0.96] (Figure 13). The overall addition of phosphorus increased the grain yield compared to no phosphorus controls. The Funnel plot showed an asymmetrical distribution of the 10 studies (Figure 14), hence the presence of bias. The overall result remained unchanged after trim and fill.

Grain yield variations as influenced by Rhizobium inoculation and P fertilizer
The interaction of inoculation and phosphorus resulted in high grain yield (2.51 t ha -1 ) compared to the main effects of inoculation and phosphorus (1.67 t ha -1 and 1.95 t ha -1 , respectively) ( Figure 15). However, the main effects of phosphorus and inoculation also contributed to high grain yields.

Relative yield increase
The relative yield increase of the inoculated treatments over non-inoculated controls for the combined studies ranged from 74 to 87% (mean = 80%). The median, first and third quartiles were 80, 78.6, and 81.4%, respectively ( Figure 16a). Phosphorus-treated plants had a mean relative yield increase of 73.4% over control treatments with median, first and third quartiles of 73, 71.5, and 75%, respectively (Figure 16b).

Response of soybean to rhizobial inoculation
The results of the meta-analysis confirmed that the inoculation of soybean with Rhizobia strains in African soils has a highly significant influence on nodule number, nodule dry weight, shoot dry weight, and yield. The performance of Rhizobia inoculants varies with strain species/isolates (Bradyrhizobium/ Sinorhizobium) and/or indigenous/introduced), soybean genotype, and soil underlying characteristics (pH, soil organic matter, nutrients, salinity, temperature) (Mapope and Dakora, 2016;Thilakarathna and Raizada, 2017). Many of the studies meta-analysed concluded that rhizobial inoculation effectively increased nodule number per plant. However, Thilakarathna and Raizada (2017) Pulver et al., 1982bOkereke et al., 2001Tarekegn et al., 2017Rechiatu et al., 2015Khalid et al., 2011Muhammad, 2010Waswa et al., 2014Rurangwa et al., 2018Mukhtar et al., 1987Okogun et al., 2005Pulver et al., 1982aSavala, 2020Youseif et al., 2014Solomon et al., 2012Fituma et al., 2018van Heerwaarden, 2017Lamptey et al., 2014Ahiabor et al., 2016Gyogluu et al., 2016Chowdhury et al., 1983Zengeni & Giller, 2007Okereke et al., 2000Argaw, 2014Ronner et al., 2016Okogun et al., 2003   occurs in soils where indigenous Rhizobia are absent or extremely low. This could be probably due to less competition between the commercial Rhizobia and indigenous strains. A recent field research conducted across three sites found that inoculation increased nodulation of different soybean genotypes ranging from 37-95% against the non-inoculated treatments (Savala and Kyei-Boahen, 2020). Okereke et al. (2000) also found that soybean inoculation with Bradyrhizobia strains significantly increased nodule number but with huge variability at 84 days after planting (DAP); and this was attributed to the variations in the ability to nodulate the   soybean variety used (TGX 536-02D). On the other hand, Rhizobium inoculation has failed to significantly increase nodule number as demonstrated by Ahiabor et al. 2016). These results implied that N might not be always the limiting factor to lack of nodulation but other nutrients like low phosphorus and molybdenum may impede the inoculation response; and also, indigenous Rhizobia could prevent the introduced Rhizobia from forming   nodules on soybean (Ahiabor et al., 2016). The significant increase in the nodule dry weight was not surprising given the effective response of the nodule number to inoculation. This result concurs with the recent works conducted in Ethiopia and Kenya which demonstrated that rhizobial inoculation resulted in increased nodule dry weight per plant (Fituma et al., 2018;Mulambula et al., 2019). Different Rhizobia strains also showed significant effects on nodule dry weight, ranging from 0.33 to 0.44 g plant -1 in contrast to the noninoculated controls (Argaw, 2014). Similarly, a field research demonstrated a prolific nodule dry response to Rhizobium inoculation, which ranged from 0.27-0.36 g plant -1 versus 0.02-0.03 g plant -1 for non-inoculated treatments (Youseif et al., 2014). The results indicated the importance of soybean Rhizobium inoculation in African soils. Soybean shoot dry weight increase was highly significant in response to inoculation with Rhizobia strains as compared to the non-inoculated control. This result concurs with Ibrahim et al. (2011) who reported a significant increase of shoot dry weight in inoculated treatment over non-inoculated treatment. However, Lamptey et al. (2014) also reported the highest shoot fresh and dry weight following application of 30 kg P ha -1 . Inoculation of soybean with Rhizobia strains improves nodulation leading to higher nitrogen fixation which subsequently increases the vegetative growth as well as dry matter formation from the inoculated soybean (Lamptey et al., 2014). Interestingly, the results showed a highly significant grain yield of inoculated soybean increased by 25.5% over non-inoculated, confirming the benefits from Rhizobium inoculation to soybean in Africa. A global meta-analysis also reported the efficacy of various Rhizobia inoculants on soybean yield ranging from -34% to +109% over non-inoculated controls (Thilakarathna and Raizada, 2017). Another farmermanaged field research conducted across 10 sub-Sahara African countries estimated mean grain yield at 1.34 t ha -1 and 1.23 t ha -1 for inoculated and non-inoculated treatments, respectively, indicating a very narrow margin (van Heerwaarden et al., 2017). They mentioned huge varietal and spatial variations across the region as major contributing factors to their results. Ulzen et al. (2018) also found that Rhizobium inoculation increases soybean yield, hence improving the livelihood of smallholders.

Effect of applied P on soybean productivity
Phosphorus is one of the irreplaceable nutrients (Giller and Cadisch, 1995), and its deficiency in many tropical regions is limiting legume performance (George et al., 1995). The BNF process in legumes is substantially driven by phosphorus, which functions as the energy storage and transfer component for the symbiotic bacteria (Dakora and Keya,1997), and increases tissue-%N as well as uptake of N derived from fertilizer (Thomas, 1995; cited by Giller and Cadisch, 1995). This study demonstrated that the supplementation of Pfertilizer on soybean at rates between 20 to 60 kg P ha -1 across African soils has a higher significant effect on nodule number. In their findings, Ahiabor et al. (2016) found that applying 22.5 kg and 45 kg P 2 O 5 ha -1 also effectively increased the number of nodules in soybean by 12 and 22%, respectively, as compared to untreated control. Despite the benefits of this technology, about 24% of farmers were reportedly applying P-fertilizer on soybean in Western Kenya (Franke and Wolf, 2011), which remains true for the majority of smallholders especially, in the SSA (Sheahan and Barrett, 2017). The significant response of grain yield to applied P was not surprising because 11 out of 12 studies that reported on phosphorus fertilizer in this meta-analysis found concurring results. The yield was increased by 36.4% as Relative yield increase of inoculated treatments over uninoculated controls (%) compared to the control treatment ( Figure 15). A field experiment conducted in Nigeria demonstrated that P supplementation increased soybean yield by 452 kg ha -1 under smallholder farming (Ronner et al., 2016). Another recent study concluded that 23-46 kg P 2 O 5 ha -1 of Pfertilizer applied together with a lower level of N (11.5 kg N ha -1 ) as starter fertilizer potentially increases yield (Tarekegn and Kibret, 2017).

Rhizobium inoculation and P-fertilizer interaction effect on soybean grain yield
The combined application of Rhizobia inoculants and Pfertilizer on soybean has resulted in 50.3% and 28.7% yield increase, respective of the independent effects of the two technologies. Servani et al. (2014) reported that P plays a critical role in nodulation processes in legumes; hence its deficiency can limit the yield. Supplementing Pfertilizer will however enhance the BNF process in soybean through improved nodulation processes by the rhizobial bacteria. Ekeleme et al. (2009) mentioned that phosphorus is mostly deficient in many soils, and its optimum application improves the shoot weight and yield of legumes. Another field study also found that applying 30 kg P ha -1 of phosphorus together with Rhizobium inoculant significantly increased soybean grain yield (Lamptey et al., 2014). Relative yield increase from inoculated treatments over non-inoculated controls was 80%, on average, attesting the effectiveness of rhizobial inoculation in soybean grain yield's improvement. In this regard, van Heerwaarden et al. (2017) obtained an average yield response of 88 kg ha -1 from inoculated plants over non-inoculated controls. Phosphorus-treated plants had also a high relative yield increase over control plants, averaging 73.4%. Eleven studies reported that supplementation of P fertilizer resulted in improved BNF with direct impacts on grain yield. However, it was hard to conclude that rhizobial inoculation and P fertilizer result in higher relative yield increase given the huge variability in agronomic, climatic, and edaphic factors across African countries which could affect the biological nitrogen fixation process of soybean.
Although the meta-analysis ascertained the general effect of rhizobial inoculation on soybean, it should be noted that the efficacy of the commercial inoculants varies with the underlying soil factors like indigenous rhizobial level, soil available nutrients, soil pH, organic matter content, temperature, and precipitation. Unfortunately, the effects of the above-mentioned factors were not analyzed due to the huge variation of the data and scarcity of valid studies. It was also clear that Pfertilizer notably between 20 kg and above 60 kg P ha -1 had varying effects on soybean's nodulation and grain yield. Furthermore, it was not established that the rates between 20 or above 60 kg P ha -1 could not affect soybean productivity.

Conclusion
Meta-analyses of the effects of Rhizobia inoculants and phosphorus fertilizer on soybean nodulation in Africa revealed rhizobial inoculation has, in absolute terms, highly significant effects on nodulation characteristics, shoot dry weight, and grain yield of soybean on African soils that may vary with the underlying soil characteristics, Rhizobium strain, climatic conditions, to name a few. Application of phosphate fertilizer at rates of 20-60 kg P ha -1 proved to increase the nodule number per plant and most importantly, soybean grain yield. Phosphorus showed a slightly higher effect on grain yield as compared to rhizobial inoculation, in absolute terms. Finally, the application of both inoculant and P-fertilizer on soybean greatly increased grain yield by 50.3% compared to a simple Rhizobium inoculation and 28.7% compared to P application alone. Therefore, it was recommended to African farmers to adopt this sustainable approach of combined application of both Rhizobium inoculants and phosphate fertilizer for reduced financial costs of production and increased yield.