Mycorrhizal fungi-mediated uptake of tree-derived nitrogen by maize in smallholder farms

Abstract

Trees within farmers’ fields can enhance systems’ longer-term productivity e.g., via nutrient amelioration, which is indispensable to attain sustainable agroecosystems. While arbuscular mycorrhizal fungi (AMF) are known to improve plant access to soil nutrients, the potential of AMF to mediate nutrient uptake of tree-derived N by crops from beyond the crops’ rooting zones is unclear. We hypothesized that AMF quantitatively contribute to the crop uptake of tree-derived N. We set up root and AMF exclusion and control plots around faidherbia trees (Faidherbia albida) and used the ¹⁵N natural abundance technique to determine the magnitude of AMF-mediated uptake of tree-derived N by maize from beyond its rooting zone in smallholder fields. We further tested whether AMF-mediated N uptake decreases with distance-from-tree. We show that within one cropping season, maize obtained approximately 35 kg biologically fixed N ha^-1 from faidherbia. One third of tree-derived N in maize leaves was attributed to AMF-mediated N uptake from beyond the maize rooting zone and two thirds to N from tree leaf litter, regardless of distance-from-tree. As hypothesized, maize grown close (1 m) to faidherbia obtained significantly more tree-derived N than at further distances (4 and 5 m). Thus, the faidherbia-AMF association can enhance agroecosystem functioning.

Sound management practices are essential to attain food security, which is still at continuous risk in sub-Saharan Africa¹. Agroforestry can provide a framework for sustainable farming: trees distributed throughout farmers’ fields can enhance soil fertility via above- and belowground organic matter inputs^2–6. Nutrients of these inputs become available to surrounding crops through various mechanisms such as mineralization^2–4 and mechanisms mediated by arbuscular mycorrhizal fungi (AMF)^6–8.

Arbuscular mycorrhizal fungi, ubiquitous and abundant mutualistic root symbionts that associate with 72 % of terrestrial plants⁹, form extensive mycelia that increase the exploited soil volume¹⁰ and may interconnect plants’ rooting zones. As such, AMF can enable the uptake and transfer of nutrients from root exudates and mediate indirect nutrient transfer between plants^6,10. Moreover, AMF may link different plants via mycelia¹¹ and enable direct interplant nutrient transfer^8,12,13. In return, AMF require plant carbohydrates¹⁰. Nitrogen-fixing trees can provide microdose-fertilization potentially increasing crop yield^14–21 but the role of AMF in N acquisition by crops remains poorly explored. Indigenous AMF could increase the uptake of tree-derived biologically fixed N₂ by crops from outside the crops’ rooting zones, particularly in subsistence farming where soil nutrient inputs are low and farmers need to leverage ecosystem processes to improve food security.

We estimated the total amount of biologically fixed N₂ derived from faidherbia trees (Faidherbia albida; Fabaceae; hereafter referred to as “tree-derived N”) in surrounding maize plants and quantified the potential significance of indigenous AMF in making tree-derived N accessible to maize plants within a season and subsequent effects on maize yield in Malawian farmers’ fields. Faidherbia trees are known for their potential to increase soil fertility and crop yield^14–21 and hosting AMF in topsoil and deep soil layers²². We used the ¹⁵N natural abundance technique to distinguish between tree-derived N and N derived from soil, in combination with root and AMF exclusion and control plots (Fig. 1). Three types of plots were installed around eight faidherbia trees that were distributed in farmers’ fields under maize cultivation to distinguish between three types of interactions between faidherbia and maize: 1) fully restricted belowground interactions, limiting the access of maize to tree-derived N from leaf litter (Litter only plot), 2) belowground interactions restricted to those enabled via mycelia of indigenous AMF (Litter&AMF plot), and 3) unrestricted interaction between tree and maize (Litter&AMF&Roots plot). We note that we neither added tree leaf litter nor AMF inoculum and that the plot designations refer to the sources of or pathways potentially mediating tree-derived N uptake by maize i.e., 1) leaf litter only, 2) leaf litter and indigenous AMF (hereafter simply referred to as “AMF”) potentially exploiting the tree rooting zone, or 3) leaf litter, AMF potentially exploiting the tree rooting zone and direct contact with tree roots. This set-up allowed disentangling the effects of litter, AMF, and tree roots on the amount of tree-derived N in maize across distance. We focused our study on maize-based agroforestry systems of Malawian smallholders because they exemplify a common agroecosystem that must be managed more sustainably to attain food security¹.

Fig. 1. Low-input agroforestry system with maize and Faidherbia albida in central Malawi and visualization of the experimental manipulation used to study the effect of AMF-mediated uptake of tree-derived biologically fixed N₂ by maize.

Top: Faidherbia albida trees distributed throughout smallholder farmers’ maize fields. Bottom: Experimental plots to manipulate the type of belowground interaction between tree and maize plants. Root and AMF exclusion and control plots together with the ¹⁵N natural abundance technique were used to estimate the contribution of AMF-mediated uptake of tree-derived biologically fixed N₂ by maize from beyond its rooting zone.

Results

Faidherbia trees were actively fixing atmospheric N₂ as indicated by the foliar δ¹⁵N of faidherbia which ranged from -0.24 to 1.59 ‰ with an average of 0.92 ± 0.24 ‰ (mean ± SE, n = 8). The fact that the δ¹⁵N values of faidherbia ranged around 0 indicates that the trees acquired almost all of their N from the atmosphere via symbiotic N₂ fixation. Foliar N concentration of faidherbia ranged from 3.62 to 5.38 % with an average of 4.44 ± 0.21 % (mean ± SE, n = 8) suggesting that the trees were well supplied with N.

Plot type and distance effects on maize and soil δ¹⁵N and N

The foliar δ¹⁵N of maize differed among the plot types (i.e., type of interaction between tree and maize) (F2,94 = 10.71, p < 0.001; Table 1). The leaves of maize plants grown in the Litter only plots were enriched in ¹⁵N by 1.3-fold and 1.2-fold i.e., acquired significantly less tree-derived N and more soil N relative to maize grown in the Litter&AMF plots and Litter&AMF&Roots plots, respectively (p < 0.001). But there was no significant difference in foliar δ¹⁵N for maize grown in the Litter&AMF versus the Litter&AMF&Roots plots. Distance from faidherbia also affected maize foliar δ¹⁵N (F4,94 = 5.93, p < 0.001), with maize grown at 4 and 5 m from the tree, being more enriched in ¹⁵N by 1.4-fold and 1.5-fold, respectively, than at 1 m (p = 0.001 and p < 0.001, respectively; Table 1). There was no interaction between the two main factors, i.e., plot type and distance from faidherbia, on foliar δ¹⁵N. Similar results were obtained for the effect of plot type and distance from faidherbia on the proportion of tree-derived N in maize (as determined using equation (1); data not shown). Foliar N concentration in maize was significantly affected by distance from faidherbia (F_4,94 = 5.80, p < 0.001; Table 1) with 1.2-fold greater foliar N concentration in maize grown at 1 and 2 m compared to 5 m (p < 0.001 and p = 0.004, respectively) across all three plot types. Foliar N concentration in maize at 1 m was 1.1-fold greater than at 4 m (p = 0.047). Plot type did not significantly affect foliar N concentration (F_2,94 = 1.50, p = 0.23). On average, foliar N concentration was 1.86 ± 0.07, 1.82 ± 0.06, 1.72 ± 0.05, 1.63 ± 0.07, and 1.54 ± 0.07 at 1, 2, 3, 4, and 5 m, respectively (mean ± SE, n = 24). Nitrogen concentration and δ¹⁵N in paired soil samples were not affected by plot type (F_2,94 = 1.53, p = 0.22 and F_2,94 = 1.34, p = 0.27, respectively) or distance from faidherbia (F_4,94 = 0.91, p = 0.46 and F_4,94 = 1.47, p = 0.22, respectively) and were on average 0.16 ± 0.002 % and 5.74 ± 0.04 ‰, respectively (Table 1).

Table 1. Delta¹⁵N [‰] signatures and N concentrations of maize leaf and paired soil samples at increasing distances of the maize plants from Faidherbia albida.

		Distance from tree [m]
Plot type¶		1	2	3	4	5
Litter only	δ15Nmaize	3.5 ± 0.4A,b	4.1 ± 0.6A,a,b	4.0 ± 0.4A,a,b	4.5 ± 0.5A,a	4.7 ± 0.5A,a
	δ15Nsoil	5.6 ± 0.2A,a	5.8 ± 0.2A,a	5.8 ± 0.2A,a	5.8 ± 0.1A,a	5.7 ± 0.1A,a
	% Nmaize	1.7 ± 0.1A,a	1.7 ± 0.2A,a,b	1.6 ± 0.1A,a,b,c	1.6 ± 0.1A,b,c	1.6 ± 0.1A,c
Litter & AMF	δ15Nmaize	2.5 ± 0.4B,b	3.0 ± 0.2B,a,b	3.4 ± 0.4B,a,b	3.6 ± 0.4B,a	3.7 ± 0.5B,a
	δ15Nsoil	5.7 ± 0.2A,a	5.7 ± 0.2A,a	5.8 ± 0.1A,a	5.8 ± 0.2A,a	5.8 ± 0.1A,a
	% Nmaize	2.1 ± 0.1A,a	1.8 ± 0.1A,a,b	1.8 ± 0.1A,a,b,c	1.5 ± 0.1A,b,c	1.4 ± 0.1A,c
Litter & AMF & Roots	δ15Nmaize	2.4 ± 0.3B,b	3.0 ± 0.3B,a,b	2.8 ± 0.3B,a,b	3.9 ± 0.4B,a	3.7 ± 0.4B,a
	δ15Nsoil	5.5 ± 0.2A,a	5.6 ± 0.2A,a	5.6 ± 0.2A,a	5.8 ± 0.2A,a	5.9 ± 0.2A,a
	% Nmaize	1.8 ± 0.1A,a	1.9 ± 0.1A,a,b	1.8 ± 0.1A,a,b,c	1.8 ± 0.1A,b,c	1.5 ± 0.1A,c

^¶Plot type refers to the type of interaction between tree and maize. Litter only plot: pond liner to eliminate all belowground interactions between tree and maize, limiting tree-derived N inputs to leaf litter only; Litter&AMF plot: 40 μm mesh to restrict interactions between tree and maize roots to those via extraradical mycelia of AMF only, and eliminate the possibility that tree and maize roots could intermingle; Litter&AMF&Roots plot: no lining for unrestricted interactions between tree and maize, and to control for the potential effect of soil excavation during the setup of the experimental plots. Values represent means ± SE (n = 8). Different letters indicate significant differences (p < 0.05) across plot type (capital letters) and across distance from tree (lower case letters) for each of the three response variables (δ¹⁵N_maize, δ¹⁵N_soil, and % N_maize). There was no significant interaction between plot type and distance from tree for any of the response variables. The average δ¹⁵N_soil across plot type and distance from tree was 5.74 ± 0.04 ‰. The average % N_maize at 1, 2, 3, 4, and 5 m was 1.86 ± 0.07, 1.82 ± 0.06, 1.72 ± 0.05, 1.63 ± 0.07, and 1.51 ± 0.07, respectively. The foliar δ¹⁵N signature of faidherbia ranged from -0.24 to 1.59 ‰ (average: 0.92 ± 0.24 ‰) and the foliar N concentration of faidherbia ranged from 3.62 to 5.38 % (average: 4.44 ± 0.21 %; mean ± SE, n = 8).

Litter-, AMF-, and root-mediated N uptake across distance

The proportion of tree-derived N in maize as a result of litter-, AMF-, and root-mediated processes did not differ with distance from faidherbia trees (F_4,27 = 1.24, p = 0.32; F_4,27 = 0.30, p = 0.87; and F_4,25 = 0.90, p = 0.48, respectively; as determined using equation (2) & (3); Fig. 2). The effect of roots on tree-derived N in maize was negligible (Fig. 2).

Fig. 2. Proportion of tree-derived N in leaves of maize surrounding Faidherbia albida.

The proportion of tree-derived N in maize leaves from tree leaf litter, AMF-mediated uptake by maize, or direct root-to-root contact between tree and maize (litter-, AMF-, and root-mediated, respectively; see Material & Methods for further details, particularly equation (2) & (3)) with distance from F. albida (n = 8). Box-plot elements are defined as follows: the center line represents the median, box limits represent the upper and lower quartiles, whiskers represent 1.5x interquartile range, and the points represent each individual data point.

Yield and total tree-derived N in maize

Shoot biomass, grain yield, N content, and total tree-derived N in maize leaves on a per plot basis did not significantly differ between plot types (F_2,14 = 0.35, p = 0.71; F_2,14 = 0.86, p = 0.44; F_2,14 = 0.32, p = 0.73; and F_2,14 = 2.09, p = 0.16, respectively; Table 2). The total amount of tree-derived N in maize plants grown within 5 m from faidherbia across all three plot types (estimated using average values for maize leaves across distances and plots and using per tree values obtained from stalk, grain and cob subsamples; see methods) summed up to 35 ± 7 kg N ha^-1 of the total amount of 120 ± 7 kg N ha^-1, hence making up about 30 % of total N in maize.

Table 2. Maize shoot biomass, grain yield, leaf N content and total tree-derived biologically fixed N₂ in maize leaves of all maize plants grown within the different experimental plots within a 5-m radius around Faidherbia albida.

Plot type¶	Biomass [t ha-1]	Yield [t ha-1]	N content in leaves [kg ha-1]	Tree-derived N in leaves [kg ha-1]
Litter only	4.0 ± 0.3	4.1 ± 0.3	18.2 ± 1.3	6.3 ± 1.8
Litter & AMF	3.7 ± 0.6	4.3 ± 0.4	17.3 ± 2.4	8.8 ± 1.5
Litter & AMF&Roots	3.7 ± 0.5	3.7 ± 0.4	18.5 ± 2.5	8.9 ± 1.8

^¶Plot type refers to the type of belowground interaction between tree and maize. Litter only plot: pond liner to eliminate all belowground interactions between tree and maize, limiting tree-derived N inputs to leaf litter only; Litter&AMF plot: 40 μm mesh to restrict interactions between tree and maize roots to those via extraradical mycelia of AMF only, and eliminate the possibility that tree and maize roots could intermingle; Litter&AMF&Roots plot: no lining for unrestricted interactions between tree and maize, and to control for the potential effect of soil excavation during the setup of the experimental plots. None of the variables were significantly affected by ‘plot type’ i.e., type of interaction between tree and maize. Values represent means ± SE (n = 8).

Discussion

Incorporating faidherbia trees in agroecosystems can benefit crop yields through microdose-fertilization

N₂-fixing trees provide high-quality above- and belowground organic matter inputs to the soil e.g. in the form of tree leaf litter input, root exudates, and root turn-over^2–4. Faidherbia leaf litter input alone can provide 50 to 80 kg N ha^-1 to the soil under faidherbia trees within a given season^18,19. However, how much of this tree leaf litter-derived N is mineralized and subsequently incorporated into crop biomass is yet unknown. The distinct isotopic N signature i.e., ¹⁵N:¹⁴N ratio of N₂-fixing faidherbia allows distinguishing between tree-derived biologically fixed N₂ and residual soil N^23,24. Because maize does not biologically fix N₂, its isotopic N signature is largely determined by the residual soil N and thus, the ¹⁵N natural abundance technique allows tracing the isotopically distinct tree-derived, biologically fixed N₂ into maize^23,24. Specifically, determining the δ¹⁵N of maize leaf, paired soil samples and faidherbia reference samples and using an isotope mixing model allowed calculating the proportion of tree-derived N in maize. It is to note that the ¹⁵N natural abundance technique does not detect any non-biologically fixed N2 derived from the trees, resulting in an underestimation of the tree-derived N in maize. We found that in total, over the course of one season, tree-derived N accounts for 35 kg N ha^-1 in maize, which corresponds to about 30 % of the total N in maize. Therefore, our results confirm the importance of faidherbia trees in improving the N budget of crops in farmers’ fields. The broad-scale recommended rate of N fertilization in Malawi is 96 kg N ha^-1, but on average only 18 kg N ha^-1 is applied by farmers^25,26 and many farmers, as those in our study region, lack access to fertilizer and thus, do not apply any fertilizer to their field (based on conversations with the farmers we worked with). We demonstrate that faidherbia provides more than one third of the recommended dose of fertilizer and almost twice the amount that is on the average applied by subsistence farmers.

Microdose-fertilization has been shown to result in significant yield increases e.g. microdose-fertilization of 24 kg N ha^-1 resulted in a 64 % increase in maize grain yield relative to an unfertilized control²⁷. We found maize yields within 5 m of faidherbia were approximately 50 % greater compared to yields of maize plants away from faidherbia, the latter was determined in a previous study³². Specifically, maize grain yield was 3.7 ± 0.4 t ha^-1 under faidherbia compared to 2.5 ± 0.6 t ha^-1 at about 35 m away from faidherbia. We note that the yield determined away from faidherbia was based on green cob dry weight³², while yield under faidherbia was determined from mature cobs, after the plants had fully matured, which might have resulted in a slight underestimation of yield away from faidherbia and therefore a slight overestimation of the approximate 50 % increase in yield under relative to away from faidherbia. Nevertheless, our estimates are in line with previous findings²⁷. Furthermore, our results show that foliar N concentration was greater within the immediate vicinity (1-2 m) of faidherbia compared to further distances (4-5 m). The same holds true for tree-derived N in maize leaves as indicated by the increase in δ¹⁵N in maize leaves with distance-from-tree. We conclude that faidherbia trees are effective in providing N microdose-fertilization to crops in subsistence farmers’ fields and, therefore, may contribute to increased yields.

Mycorrhizal fungi increase uptake of tree-derived N by maize

Greater crop yields and soil nutrient contents previously observed around N₂-fixing trees within agricultural fields have been mostly ascribed to high quality organic matter inputs to the soil^14–21. The contribution of AMF in making these inputs available to crops has gained much less attention. There has been some evidence that AMF facilitate N transfer from trees to surrounding plants⁸, but verification of this mechanism under field conditions on smallholder farms has been lacking. Faidherbia has been shown to associate with AMF, both in topsoil and great soil depth²². Our experimental design and methods did not allow to provide evidence for the existence of direct linkages between tree and maize via a mycorrhizal mycelia. Therefore, we cannot differentiate between direct root-to-root and indirect soil-to-root mediated N transfer enabled by AMF. However, combining the ¹⁵N natural abundance technique and root and AMF exclusion plots allowed us to quantify the effect of AMF-mediated uptake of tree-derived N by maize from beyond its rooting zone whether via direct or indirect transfer, in farmers’ fields. While we cannot exclude the possibility that other filamentous fungi may have contributed to the uptake of tree-derived N by maize from beyond its rooting zone^28,29, we assume that most of the uptake was AMF-mediated. We estimated that the AMF-mediated uptake of tree-derived N by maize plants accounted for 28 % of the total tree-derived N in maize leaves within 5-m around faidherbia trees (Table 2). Tree litter was responsible for most, i.e., about two-thirds, of the tree-derived N in maize and the presence of tree roots within the rooting zone of maize (if at all present) had a negligible effect on the uptake of tree-derived N by maize (Fig. 2, Table 1&2).

The experimental plots were located under the tree crown (average crown radius of 5 ± 0.4 m) and therefore, it was expected that the proportion of tree-derived N obtained by maize from tree litter input was the same within the 5 m radius around faidherbia (Fig. 2). Similarly, the δ¹⁵N and total N concentration of the surface soil must have been mostly affected by tree leaf litter input which, given homogenous tree leaf litter input across the plots, may explain why we observed no increase of soil δ¹⁵N and total N concentration with distance-from-tree. The contribution of AMF-mediated N transfer from tree to crops (direct or indirect) versus uptake via root-to-root contact and direct uptake of N from tree root exudates by crop roots to the crops’ N budget likely depends on the architecture of the tree root system and the distance-from-tree. Specifically, we expected a decrease in the contribution of AMF- and root-mediated uptake of tree-derived N by maize with distance-from-tree due to an increasing distance between the maize and tree rooting zones. Indeed, given the maize foliar δ¹⁵N but not the soil δ¹⁵N increased with distance-from-tree, our data provides some evidence for this hypothesis. However, the proportion of tree-derived N obtained by maize via litter-, AMF-, and root-mediated processes was not significantly affected by distance-from-tree (Fig. 2). We did not examine the tree root system, but observed no fine tree roots within a radius of 5 m from faidherbia (at a depth of 0 to 50 cm) which is in line with previous findings about faidherbia’s deep taproot development³⁰. Even if maize roots usually grow deeper than 50 cm in the absence of our experimental plots, the lack of fine tree roots within the top 50 cm and across the 5 m from the base of the tree suggests that root-to-root contact between faidherbia and maize is typically minimal and explains why we found no additional increase in tree-derived N obtained by maize plants grown in the Litter&AMF&Roots plots relative to those grown in the Litter&AMF plots (Table 2). Therefore, given the apparent separation of faidherbia and maize roots, our results highlight the potential importance of AMF in connecting the soil volume between the rooting zone of faidherbia and maize for maize to gain access to a larger pool of tree-derived N.

Despite the contribution of AMF to the proportion of tree-derived N in maize, maize shoot biomass and grain yield were not significantly increased (Table 2). This is probably linked to the fact that total foliar N content was not affected by plot type, i.e., type of interaction, between tree and maize (Table 2). Total tree-derived N in maize leaves was also not significantly different between plot types but the data follow the same trend as the proportion of tree-derived N (Table 2). While AMF-mediated uptake of tree-derived N by maize from beyond its rooting zone may not increase the N content in maize, an AMF-mediated uptake of tree-derived N from beyond the rooting zone of maize may improve internal N cycling within the agroecosystem.

In conclusion, this study provides insight into the underlying ecological process through which N input from trees may be made accessible to crops. It appears that AMF connect the space between the rooting zones of trees and crops via mycelia and as such, increase the amount of tree-derived N accessible to crops. Especially in low-input cropping systems such as those in Malawi, N microdose-fertilization by faidherbia trees and AMF-mediated uptake of tree-derived N by crops could enhance sustainability of agroecosystems in the longer-term.

Methods

The study site was located in central Malawi, in the lowlands of Dedza district. The fields are distributed within an area of approx. 5 km² around the village of Ndindi in the Golomoti Extension Planning Area (lat. -14.3, long. 34.6). Soil types are Fluvisols³¹ with mostly sandy clay loam textures. At 1 and 4 m from faidherbia, respectively mean soil total C was 28.7 ± 0.44 g kg^-1 and 26.8 ± 0.55 g kg^-1, mean soil total N was 2.2 ± 0.05 g kg^-1 and 2.0 ± 0.04 g kg^-1, mean soil total P was 2.1 ± 0.11 g kg^-1 and 2.1 ± 0.06 g kg^-1, mean soil resin-P was 33.2 ± 6.01 mg kg^-1 and 46.5 ± 6.73 mg kg^-1, and mean weight diameter (a measure of soil aggregate stability) was 3.15 ± 0.07 mm and 3.17 ± 0.08 mm, as determined in a previous study³² (mean ± SE; n = 10). At 35 m from faidherbia, mean soil total C was 17.8 ± 2.14 g kg^-1, mean soil total N was 1.5 ± 0.28 g kg^-1, mean soil total P was 1.9 ± 0.09 g kg^-1, and mean soil resin-P was 24.8 ± 6.66 mg kg^-1 and mean weight diameter of the soil aggregates was 2.01 ± 0.22 mm (mean ± SE; n = 10). The climate is sub-tropical, humid with a unimodal precipitation pattern. Most rain falls from November to March and the average annual precipitation is 884 mm³³. Farmers prepare their fields by hand-hoeing (i.e., ~ 15 cm deep). Maize is planted at the onset of the rainy season around December and typically harvested in April/May. Farmers’ fields within the study site are not amended with chemical fertilizer and weeding is done with a hand-hoe.

Tree selection

We focused our study on faidherbia (Faidherbia albida) because this leguminous tree has been highly promoted as an agroforestry species due to its “reverse phenology”. The trees’ foliage is shed with the onset of the rainy season¹⁸ resulting in minimized light competition between trees and crops at crop establishment, and in high-quality litter inputs at a time when soil moisture conditions are favorable for rapid mineralization^15,17. A total of eight single-standing faidherbia trees distributed throughout farmers’ maize fields were selected for this study. All trees were well-established trees of a similar size (on average diameter at breast height 53 ± 3.5 cm, height 16 ± 1.0 m, crown radius 5 ± 0.4 m), single-standing (at least 40 m away from the base of any neighboring tree), and with a recent cropping history of maize. Within our study system, farmers do not commonly prune faidherbia trees and none of the selected study trees were pruned. Hence, the N derived from the trees originated from leaf litter and/or roots.

Experimental set-up (year 1)

At the beginning of the growing season of 2016/2017, we excavated three rectangular plots (1 m x 5 m, 0.5 m deep), around each tree within farmers’ maize fields (Fig. 1). Plots started at a distance of 0.7 m from the base of the tree and were oriented towards 0° (North), 120° (Southeast), and 240° (Southwest). Each plot was fitted with a different bottom and sidewall lining, resulting in three types of experimental plots, i.e., three types of interactions between tree and maize. The types of interactions were randomly assigned to the experimental plots to account for possible differences in microclimate. The three types of lining were:

• 1)pond liner (AlfaFol PVC pond liner 0.5 mm thick; Oase Living Water, Hörstel, Germany) to eliminate all belowground interactions between tree and maize, limiting tree-derived N inputs to leaf litter only (Litter only plot),
• 2)40 μm mesh (SEFAR Petex 07-40/25, Sefar AG, Heiden, Switzerland) to restrict interactions between tree and maize to those via the extraradical mycelia of AMF only, and eliminate the possibility that tree and maize roots could intermingle (Litter&AMF plot), and
• 3)no lining for unrestricted interactions between tree and maize, and to control for the potential effect of soil excavation during the setup of the experimental plots (Litter&AMF&Roots plot).

Soils host by default AMF³⁴, which when no mineral fertilizer is applied are usually particularly abundant and beneficial to the mineral nutrition of plants³⁵. Hence, the ‘&AMF’ in the plot type designation does not refer to the addition of AMF but indicates the presence of extraradical mycelia of indigenous AMF connecting the soil volume exploited by the roots of the maize plants and the soil volume used by the tree. We note that our experimental design does not exclude the possibility that filamentous fungi other than AMF may have also contributed to the uptake of tree-derived N from beyond its rooting zone^28,29. We further note that we did not add any tree leaf litter to the plots but that the plot names refer to the source of tree-derived N that maize could access. Leaf litter that had accumulated on the soil surface was removed prior to excavating the plots to prevent leaf litter N input to deeper soil. The excavated soil was piled up next to the plot and after the lining had been put in place, the soil was placed back into the plot, beginning with the soil from the top of the pile to that on the bottom to keep the original soil depth position. After completion of the experimental plots, maize was sown along two rows within each plot and across farmers’ fields (at least within a 10-m radius circle around each study tree). Farmers weeded, harvested, and eventually prepared the fields (including our experimental plots) for the next growing season, following their common practices (i.e., using a hand-hoe, planting around December, and harvesting around April/May, as described above). No measurements were taken in the year of the experimental plot set-up to let the soil and AMF mycelia recover from the experimental plot installation. Leaving the system to recover for one year minimized the risk of potentially altered mineralization of soil organic matter resulting from the experimental plot installation to influence our results.

Sample-collection (year 2)

One year after the experimental plot set-up, at the beginning of the growing season of 2017/2018, maize was sown into each experimental plot and across farmers’ fields (at least within a 10-m radius around each study tree). In each plot, maize was sown along two rows (0.6 m apart) at every meter, from 1 to 5 m from the base of the trees. The experimental plots were continuously hand-weeded by the farmers. Seventeen weeks after sowing, at the time of harvest, maize leaf samples and paired soil samples (0-15 cm) were collected at every meter from the tree trunk, in each plot. The maize leaf samples and soil samples from the same distance of the same plot were pooled into a composite sample per distance from the tree, resulting in five leaf and five soil samples per plot per tree and a total of 120 leaf and soil samples each, for physiochemical analyses. Faidherbia leaf samples were collected to obtain reference values of the δ¹⁵N signature and N concentration to estimate the amount of tree-derived N in maize. Dry, homogenized plant and soil samples were analyzed for δ¹⁵N and total N with an elemental analyzer (Vario PyroCube, Elementar, Langenselbod, Germany) connected to an isotope ratio mass spectrometer (Isoprime 100, Elementar) in continuous flow mode.

The fraction, i.e., proportion of tree-derived N (frac_N(tree)) in the leaves of the maize plants was determined for all five distances (1-5 m) from the trees, according to the following equation:

$$fra{c}_{N(tree)}=\frac{{\delta }^{15}{N}_{maize}-{\delta }^{15}{N}_{soil}}{{\delta }^{15}{N}_{tree}-{\delta }^{15}{N}_{soil}}$$ (1)

where δ¹⁵N_maize and δ¹⁵N_soil are the δ¹⁵N [‰] of maize leaf and soil samples at each distance, respectively, and δ¹⁵N_tree are the δ¹⁵N [‰] of the corresponding faidherbia tree leaf samples. We determined AMF-mediated and root-mediated uptake of tree-derived N by maize as follows:

$${\text{frac}}_{\text{AMF}\_\text{med}}={\text{frac}}_{\text{Litter}\&\text{AMF}}-{\text{frac}}_{\text{Litter}}$$ (2)

$${\text{frac}}_{\text{Root}\_\text{med}}={\text{frac}}_{\text{Litter}\&\text{AMF}\&\text{Roots}}-{\text{frac}}_{\text{Litter}\&\text{AMF}}$$ (3)

where frac_{AMF_med} and frac_{Root_med} are the proportion of tree-derived N in maize leaves obtained via AMF- and root-mediated processes, respectively, frac_Litter is the proportion of tree-derived N in maize leaves in the Litter only plots, i.e., where belowground interactions between tree and maize were fully restricted and tree-N inputs were restricted to leaf litter only, frac_Litter&AMF is the proportion of tree-derived N in maize leaves in the Litter&AMF plots, i.e., where belowground interactions between maize plants and the tree were restricted to those possible via extraradical mycelia of AMF connecting the tree and maize rooting zone, and frac_{Litter&AMF&Roots} is the proportion of tree-derived N in maize leaves in the Litter&AMF&Roots plots, i.e., where belowground interactions between maize plants and the tree were not restricted. We inferred the effect of tree roots in absence of AMF by the ‘difference calculation method’ because it is impossible to establish an “AMF-free control treatment” in farmers’ fields. While it would have been ideal to have an “AMF-free soil”, we note that there is a plethora of literature discussing the challenge of establishing an “AMF-free control soil”^36–39. AMF are ubiquitous in field soil and the hitherto best “AMF-free control soil” is using reduced mycorrhizal mutants (rmc)³⁷. However, even the rmc varieties are being colonized by AMF^37,40 and hence, for the purpose of our study would not be suitable for a + legume root / AMF-free treatment. Furthermore, there is no rmc mutant for faidherbia and even if there was, establishing a new tree in the field would result in a generation-time experiment. The use of fungicides, e.g., benomyl to create an “AMF-free soil” is highly debated because of side effects on the saprotrophic ascomycetes³⁶. Not only would the use of fungicides have caused such major experimental artifacts⁴¹ but moreover, the use of fungicides in farmers’ fields would have been a very invasive approach. Also, given that we would have had to treat the soil down to a very deep depth⁴² to treat the soil around the rooting zone of maize made this method unfeasible.

Total maize biomass and mature cob fresh weight per experimental plot were determined in the field and composited, homogenized subsamples per tree were oven-dried to determine dry weights. Dry weight of each fraction i.e., leaves, stalks, grain, and cob per tree were determined from the total shoot and the cob weights, using known proportions for maize⁴³. The oven-dried, composited biomass and cob samples obtained per tree were subsequently separated into stalk, grain and cob subsamples, which were analyzed for total N concentration and δ¹⁵N signature, as described above. For maize leaves, we used the average total N concentration and δ¹⁵N signature across distance and plots to obtain a per tree estimate. We used the total N concentration and δ¹⁵N signature of each fraction (stalks, grain, cob and leaves) to determine the N content of each fraction per tree and to estimate the total N content in maize and the total amount of tree-derived N in maize grown within 5 m from faidherbia (see below). The total amount of tree-derived N (total_N(tree); [kg N ha^-1]) in maize leaves was calculated on a per plot basis as follows:

$${\text{total}}_{\text{N}(\text{tree})}=({\text{N}}_{\text{maize}})({\text{frac}}_{\text{N}(\text{tree})})$$ (4)

where N_maize is the total amount of N in maize leaves per hectare [kg N ha^-1], and fracN(tree) the proportion of tree-derived N in maize leaves as calculated by equation (1). To obtain total tree-derived N in maize on a per tree basis, we estimated total tree-derived N per fraction and subsequently, determined the sum of total tree-derived N per fraction for all fractions (leaves, stalks, grain, and cobs) combined. Hence, we adjusted equation (4) to have N_maize and frac_N(tree) represent the total N content and the proportion of tree-derived N in maize stalk, grain, or cob to determine the total amount of tree-derived N (total_N(tree)) in maize stalk, grain, or cob, respectively for each tree. For maize leaves, we used the average obtained across plots. The proportion of tree-derived N (frac_N)tree)) in maize stalk, grain, and cob on a per tree basis were calculated using equation (1) with δ¹⁵N_maize as the δ¹⁵N [‰] of maize stalks, grain, or cob per tree, respectively, δ¹⁵N_soil the average δ¹⁵N [‰] of soil determined per tree, and δ¹⁵N_tree the δ¹⁵N [‰] of the corresponding faidherbia tree leaf samples. For maize leaves, we used the average obtained across distance and plots. Similarly, we calculated total N in maize [kg ha^-1] by determining N content per plant fraction based on N concentration obtained per fraction per tree and dry weight per fraction per tree (for leaves an average across distance and plots was used) and taking the sum of the N content of all fractions.

Maize root colonization by AMF

Given the relatively low number of plants per plot and the fact that making the plots larger to allow for more plants per plot was financially unfeasible, we decided against assessing root colonization by AMF to increase the chances of obtaining reliable maize yield and δ¹⁵N data at the time of harvest. We refer to a previous study³² in which we assessed root colonization by AMF within 15 m around the same study trees and found no difference in root colonization of maize plants grown close to versus at further distances (i.e., at 1 versus 15 m) from the trees. Percent colonization by hyphae and arbuscules was on average 33 ± 1.2 % and 31 ± 0.8 % within the 15 m radius around faidherbia trees, respectively and the average number of vesicles was 92 ± 25 (mean ± SE, n = 10). Furthermore, in a greenhouse experiment, we determined a high AMF infection potential of the Malawian field soil leading to highly colonized maize roots regardless of whether or not the soil could have been colonized by AMF originating from the trees (unpubl. data, J. Dierks 2017). Hence, we feel confident in assuming that maize root colonization was approximately the same across all plots and did not affect the outcome of our results.

Maize yield under versus away from Faidherbia albida

To assess the effect of tree on maize grain yield, we compared the average grain yield within 5 m from faidherbia with grain yield estimates derived from green cob dry weights of maize plants that grew about 35 m from faidherbia trees in the 2015/2016 cropping season.

Statistical analyses

All statistical analyses were performed in the software environment R (R 4.0.3 GUI 1.73)⁴⁴. We used linear mixed effects models to assess the effect of plot type, i.e., type of interaction between tree and maize, and/or distance of the maize plants from the tree on the various response variables (see below) while accounting for variation caused by inherent differences across individual trees (i.e., fields) by including ‘tree’ as a random effect variable in the models in addition to the study factors that were tested as fixed effects⁴⁵. Specifically, we used linear mixed effects models with individual tree included in the models as a random effect variable and plot type and distance as fixed effect variables to analyze the effect of plot type and distance on the δ¹⁵N signature of maize leaves and soil, foliar and soil N concentrations, and the proportion of tree-derived N in maize leaves (frac_N(tree)). Further, we used linear mixed effects models with individual tree included in the model as a random effect variable and plot type only as a fixed effect variable to analyze the effect of plot type on maize biomass and yield, N content and tree-derived N in maize leaves (total_N(tree)). Similarly, the impact of distance from faidherbia on the proportion of tree-derived N in maize leaves from litter-, AMF-, and root-mediated processes was assessed with linear mixed-effects models with distance included in the model as a fixed effect variable and individual tree as a random effect variable. The Cook’s distance measure was used to detect outliers. There was no significance found for the interaction between plot type and distance from faidherbia trees. Post hoc pairwise means comparisons were made using Tukey’s test to calculate least-squares means (function ‘lsmeans’ in R package ‘lsmeans’) in those cases in which the main effects were significant.

Data availability

The analyzed datasets and R scripts used to analyze the data are available in the Zenodo repository: https://doi.org/10.5281/zenodo.5275322