Figure 1. Map showing the locations of the five study sites (black circles) across Serengeti National Park, Tanzania. Each site is labelled with a three-letter code that corresponds to Table 1.

Measurements were made in each of two years (2022-2024) at five sites in Serengeti National Park, Tanzania (Figure 1), selected from a set of eight sites in a Long-Term Grazing Exclosure (LTGE) experiment [29][34]-[36] established in 2001. These five sites exhibit both considerable variation in soil N, P, plant species composition, and above- and belowground biomass [31]. We focused on unfenced areas subject to herbivory to be able to scale N₂-fixation activity assays to area-specific measurements that might apply across the study area.

N₂-fixation measures were made on randomly selected individuals of the target species in the 4 m × 4 m unfenced plots of the LTGE experiment at each site. All other soil and plant community measures were from these same plots. The target grass (Poaceae) species were Digitaria macroblephara, Eragrostis tenuifolia, Panicum maximum,Pennisetum mezianum, Sporobolus fimbreatus, and Themeda triandra. All six genera have broad distributions in the tropics, and T. triandrais common in regions of 600-1000 mm rainfall in Africa, Indonesia and Australia [37]. The six grass species chosen exhibited mean cover > 5% at least one of the five sites, and four species (D. macroblephara, P. maximum, P. mezianum and T. triandra) exceeded that cover in 3 of the 5 sites. In addition, we assayed a ubiquitous legume (Fabaceae), Indigofera volkensii, which had the highest cover compared to other legume species (1% - 2% of total cover) by a factor of > 3 at each of the five sites.

Root biomass was estimated from 3 pooled 10 cm × 40 cm diameter cores within each plot using the flotation method to separate live roots from dead and soil detritus [13] in June 2022 (at the end of the wet season). Cover of each species was estimated visually in April 2023 (during the peak of the wet season) in each of four 2 m × 2 m quadrats within each 4 m × 4 m plot and then averaged [31][32]. Plant and total soil element concentrations were measured for plant leaves and soils sieved from root cores, respectively, with benchtop methods at the Soil Analysis Laboratory at Sokoine University of Agriculture in Morogoro, Tanzania [31]. Total plant and soil N were analyzed using the Kjeldahl method, while plant and soil P were analyzed with standard persulfate digestion. These methods in this laboratory have contributed significant prior published data and reflect similar differences among samples as samples analyzed with combustion and infrared methods that are otherwise not available in Tanzania [31][38].

We used the para-nitrophenyl phosphate (pNPP) colourimetric assay, originally developed by [39], as the standard protocol used to measure both root and soil phosphatase activity. This method measures the enzymatic hydrolysis of pNPP, which releases the yellow-colored compound p-nitrophenol (pNP). The intensity of the yellow colour, proportional to the enzyme activity, was measured using a spectrophotometer at a wavelength of 400 nm - 420 nm.

1. Soil phosphatase activity protocol

Fresh soil was typically used, passed through a 2 mm sieve, and, in some cases, air-dried and stored at 4˚C. One g of soil was mixed with 4 mL of Modified Universal Buffer (MUB), which sets the pH (e.g., pH 6.5 for acid phosphatase, pH 11 for alkaline phosphatase), and 1 mL of 0.115 M pNPP substrate. The mixture was incubated for 1 hour at 37˚C. The reaction was stopped by adding 1 mL of 0.5 M CaCl₂ and 4 mL of 0.5 M NaOH. The soil suspension was centrifuged, and the supernatant was filtered, and the absorbance was read at 405 nm. Activity was expressed as 𝜇mol pNP released per gram of soil per hour (𝜇mol g⁻¹h⁻¹).

2. Root phosphatase activity protocol

Fine roots (≤ 1 mm or first three orders) were carefully washed to remove adhering soil. Between 0.3 - 1.0 g of roots were placed in a vial with 9 mL of 50 mM buffer (e.g., sodium acetate for acid phosphatase, pH 5.0) and 1 mL of 50 mM pNPP. The roots were shaken for 1 hour at 27˚C. 0.5 mL of the sample solution was added to 4.5 mL of 0.11 M NaOH. Absorbance was read at 405 nm and compared against a standard curve prepared from pNP. Activity was expressed per unit of root dry mass (𝜇mol pNP g⁻¹_rooth⁻¹).

Labelled N₂ uptake was measured using an in situ incubation method [40], modified for whole roots of I. volkensii and T. triandra and other grasses [13]. Assays were performed during periods following at least 100 mm of rain in the previous month: in November 2022 and March 2023. In June 2022, roots of 5 plants of each of six target grass and one legume species were rinsed thoroughly of soil with distilled water and assayed for N₂-fixation at each of the five sites where we encountered plants with 0.02% and thin 50 m of the permanent plots of the LTGE experiment (Table 1). This resulted in measurements of 25 D. macroblephara, P. mezianum, T. triandra and the legume I. volkensii, plus 15 measurements for P. maximum,10 for S. fimbreatus,and 5 for E. tenuifolia (which was only found at the MSB site). Following these measurements, in March 2023, we re-measured N₂-fixation in the roots of four species for which N₂-fixation was detected (mean N fixed significantly different from zero) in 2022 (D. macroblephara, P. mezianum, T. triandra and the legume I. volkensii) for five new individual plants at each of the five LTGE sites.

For each assay, whole roots of target individual plants of each species were excavated, rinsed of visible soil and divided into control and enriched subsamples of approximately 20 g fresh mass (3 g dry) each in 60 ml chambers (syringes). An additional 3 root samples from each of the seven species were additionally surface-sterilized following [41]: sequential immersion in 99% alcohol, CaOCl (2%), chloramine T (2%) with 2 - 3 drops of Tween^® and an antibiotic solution containing streptomycin sulphate, gentamicin sulphate (0.01%). Surface-sterilized roots and non-sterilized roots differ primarily in the presence of microbial life on their exterior, which affects both microbial analysis and plant behaviour. Surface sterilization removed epiphytic microbes (surface dwellers) to allow for the study of endophytes. This was a critical, yet potentially destructive, step in investigating plant-microbe interactions, as the process must kill surface bacteria without harming the internal endophytes or the root tissue itself. Measurements were then compared to roots receiving only rinsing to determine if N₂-fixation was due to root surface associated bacteria or endophytes (living inside root tissue). The enriched sample received 50% ¹⁵N-enriched atmosphere at 1-atm pressure by injecting 20 ml of 99% ¹⁵N₂ gas. Both control and enriched samples were incubated at 5 cm soil depth for 30 min, after which samples were cooled to < 0˚C in a portable solar-powered box freezer (−20˚C) to stop the reaction. Assayed roots were dried, ground and analyzed for δ¹⁵N to 0.01 ^o/_oo(UC Davis Stable Isotope Laboratory). N₂ uptake (μg N ^.g^−1.hr⁻¹) was estimated as [N] × [(δ¹⁵N_enriched − δ¹⁵N_control) × 2]/(50% × 30 min x g sample), where [N] = μg N/g of roots or soil, δ¹⁵N is the difference between sample and atmospheric ¹⁵N, 50% is the gas enrichment, 30 is the assay time (min) and the factor 2 converts rate 30 min⁻¹ to rate hr⁻¹. Mass-specific rates were obtained by wet-dry mass conversions and the wet mass of roots assayed. To include δ¹⁵N sampling error in the estimates of δ¹⁵N of root and soil tissues, we did not correct the negative δ¹⁵N_enriched − δ¹⁵N_control to zero and evaluated whether mean δ¹⁵N_enriched − δ¹⁵N_control > 0, including any negative values in the distribution for each site x species combination.

To assess whether fixed N contributes significantly to plant tissue N, and therefore whether assayed N₂-fixation is a persistent process in the species tested, we also collected and dried at 45˚C for five days 50 g rhizosphere soil (within 1 cm of roots). Soils were analyzed for δ¹⁵N at the UC Davis Stable Isotope Laboratory (Davis, CA, USA). δ¹⁵N was obtained for control plant roots from those analyzed for the ¹⁵N₂- plant roots atmospheric ¹⁵N₂. Greater proportions of atmospheric N in plant tissue are associated with a greater difference between tissue ¹⁵N and that of soils, which are typically higher in ¹⁵N than the atmosphere [40].

As a measure of N₂-fixation potential, we measured copy number for the nifH gene coding for the Fe-protein in the enzyme nitrogenase in the ¹⁵N₂ uptake assayed roots using a customized DNA extraction followed by quantitative PCR [13]. Copy number of nifH genes was measured from standard molecular methods of DNA extraction [42] followed by assessment with standard real-time PCR methods [43] of the number per g sample of the nitrogenase genes, nifH. Frozen soil and roots were transported in sealed plastic bags at −20˚C by express courier, and DNA was extracted in duplicate from soil and plant samples essentially as described previously [42] with an optimized extraction [13]. To conduct real-time PCR, we designed nifH gene-specific degenerate primers following standard methods [43] to maximise the amplification of most of the nifH genes in different prokaryotic taxa [44][45] (forward primer: CSATCAACTTCCTBGARGA, reverse primer: GCCATCATBTCRCCGGA).

Community weighted mean (+ s.e.m.) area-specific N₂-fixation, NF_A (µg N^.m^−2.hr⁻¹) at each site was estimated as the sum of directly measured mean mass-specific N₂-fixation of plant species i(NF_{M, i}, μg^.g^−1.hr⁻¹) multiplied by estimated total community root biomass (RB, g/m²) and proportional cover of species i, COVER_i [13] (Equation 1). This assumes that root biomass for a species is proportional to its aboveground cover. Annual area-specific N₂-fixation, ANF_A,i for each species was estimated by further multiplying NF_A,iby the mean number of “wet days” (WD) (Table 1) where soil moisture exceeded a 10% threshold associated with root microbial activity [29] and by 12 hr^.day⁻¹ (as N₂-fixation can occur at night) and by appropriate area and mass conversions. WD represents the number of “wet days” (days with suitable moisture for microbial activity) over a specific period, which drives nitrogen fixation in that area. Soil moisture was measured using the Gravimetric Method [46], which measures water mass relative to dry soil mass (Wet Weight − Dry Weight)/Dry Weight).

Total NF_A is estimated by summing ANF_A,I for the three diazotrophic grass species and I. volkensii.

All analyses were performed with SPSS 27 (IBM Corp., Released 2020. IBM SPSS Statistics for Windows, Version 27.0. Armonk, NY) Generalized Linear Models using maximum likelihood estimates. The Likelihood Ratio (LR) test statistic was used to ensure optimal statistical reproducibility. Calculated N₂-fixation activity, which was approximately lognormally distributed but featured some negative values due to sampling error in tissue ¹⁵N, was performed on an adjusted logarithmic transformation N F log , = ln ( 1 + Y − min ( Y ) ) . Copy numbers of nifHtranscripts were log-transformed before analysis as well.

Figure 2. N₂-fixation sampled across two wet seasons in roots of six grass species (Dig mac, Digitariamacroblephara; Era ten, Eragrostis tenuifolia; Pan max, Panicum maximum; Pen mez, Pennisetum mezianum; Spo fim, Sporobolus fimbreatus and The tri, Themedatriandra across five sites (ordered by increasing mean annual rainfall in Serengeti National Park, Tanzania). A. Mean (+ s.e.m.) copy number (PCR units) per g root of the nifHgene for the nitrogenase enzyme for each species across all sites. B. Mean (+ s.e.m.) root mass-specific uptake of N by each species.

Mass-specific copy numbers of nifH genes varied significantly, and by over four orders of magnitude, among sites, species and site x species combinations (Table2, Figure 2(A)). Across all sites combined, mean root mass-specific nifH varied significantly among species (Table 2), with two species (D. macroblephara and P. mezianum) exhibiting 100 - 150 times higher copy numbers than three grass species (E. tenuifolia, S. fimbreatus, P. maximum). The legume I. volkensii and the grass T. triandracontained similar mean copy numbers to these three grass species but exhibited a much greater range (Figure 2(A)). Mean copy numbers for these species also varied significantly among sites (Table 2), and the sites with the highest copy numbers differed among species (Site x Species Interaction) (Table 2, Figure 3(A)). Variation in nifHcopy number was very weakly correlated with mass-specific N₂-fixation (R² = 0.02, N= 215, P= 0.04).

Figure 3. Box and whisker representations of mass-specific N₂-fixation for A. a legume and B-D three grass species across five sites in Serengeti National Park, Tanzania. Asterisks indicate means for species x site combinations that are significantly different from zero.

Table 2. Outcomes of Generalized Linear Models to assess variation in nitrogenase gene copy number, root mass-specific N₂-fixation and phosphatase activity by site, species, and their interaction.

As suggested by the pattern for nitrogenase gene copy number, over all sites, plant species varied significantly in mean root mass-specific N₂-fixation (diazotroph activity) (Figure 2(B), Table 2). Four species’ 95% confidence intervals did not overlap zero, indicating evidence for consistent N₂-fixation across individuals in these species, which included three grasses, D. macroblephara, P. mezianum, T. triandra and the legume I. volkensii (Figure 2(B)). Mean N₂-fixation in these three grass species and the legume ranged from 10 - 25 μg^.g^−1.hr⁻¹, well above the means for each of the three inactive grass species (−0.5 to 0.02 μg^.g^−1.hr⁻¹), whose 95% CI each included zero. The nifH gene copy number was not correlated with measured N₂-fixation at the level of individual root samples (r = 0.04, N =221, P= 0.550), but mean N₂-fixation rate for a species was correlated with its mean nifHgene copy number (r = 0.744, N = 7, P = 0.022, one-tailed test for an expected increasing relationship).

Figure 4. Persistent mass-specific N₂-fixation (diazotroph activity) inferred from the difference between plant root and soil ¹⁵N for three grasses, D. macroblephara, P. mezianum, T. triandra and the legume I. volkensii across 5 sites in Serengeti National Park.

For the four species that occurred within the LTGE experiment permanent plots, we found that, as in the previous analysis with all seven species, the four species differed significantly in mean mass-specific N₂-fixation (X²_4,200 = 18.42, P = 0.013), but the mean across species did not differ among sites (X²_4,200 = 3.83, P = 0.53). There was a significant interaction between site and species (X²_16,200 = 29.07, P = 0.023), which reflected the fact that different species’ peak mass-specific N₂-fixation occurred at different sites (Figure 4) and the confidence interval for the peak rates for each species did not include zero. D. macroblephara roots exhibited peak diazotroph activity at the driest site with the lowest soil N (SOT) and at a site with higher rainfall but the lowest P (KCW). P. mezianum roots exhibited peak activity also at the drier, low N site (SOT) but were largely inactive at the other sites. In contrast, legume I. volkensii roots displayed peak activity at a site with the highest soil N but intermediate rainfall and soil P, while T. triandra roots exhibited peak activity at the lowest soil P site (Figure 4(D)).

Figure 5. For all seven plant species tested, across all sites, relationship between mean short-term N₂-fixation activity and the difference in leaf versus soil ¹⁵N, an indicator of persistent N₂-fixation through the season.

Across all seven species, N₂-fixation was strongly correlated with the difference between root tissue ¹⁵N and rhizosphere soil ¹⁵N (X²_1,223 = 61.59, P < 0.001) (Figure 5), suggesting persistent N₂-fixation in species and at sites with high measured N₂-fixation activity. This correlation was strong for species with high mean diazotroph activity, such as the legume I. volkensii, D. macroblepharaand P. mezianum, but weaker for species with lower activity and nifH copy number.

Soil (rhizosphere) and root phosphatase activity was associated with each other (Figure 6(A)), and phosphatase activity on each substrate was associated significantly (P < 0.0001) with that plant’s N₂-fixation activity (Figure 6(B), Figure 6(C)) across all four N₂-fixing species.

Figure 6. Relationship between A. phosphatase activity in roots versus soil, B. Root phosphatase activity and N₂-fixation activity, and C. soil phosphatase activity and N₂-fixation activity for all four N₂-fixing plant species across all sites in Serengeti National Park.

Area-specific N₂-fixation was greatest at low soil N:P, consistent with a response to N-limitation, and at high soil N:P, where its association with higher rhizosphere soil phosphatases are consistent with a response to P-limitation. Cover of our target species also varied substantially across sites (Table 3), as did total aboveground and root biomass.

Table 3. Mean (s.e.m.) plant above- and belowground biomass, cover for each species and for all N₂-fixing species at each of the five study sites.