Introduction

Fungi are ubiquitous and diverse. In nature, fungi evolve into diverse communities residing in such ecological systems as soil or maize as a host plant. The diversity of the communities of fungal taxa may depend on niche partitioning. Through partitioning of ecological niches, living organisms can coexist in one space without competing for the same resource (Chesson 2000). Several fungal genera can infect the same host crop and coexist through niche partitioning. The niche partitioning in fungi can be driven by various abiotic factors including temperature and rainfall as well as carbon source among others (Arroyo et al. 2008; Giorni et al. 2009; Mohale et al. 2013; Perrone et al. 2020). Where the niche partitioning is ‘good,’ fungi will likely coexist and on the contrary, the fungi will likely undergo competition (Chesson 2000). The competition may take the form of reduction in abundance of the less competitive genus or species, relative to another. In quantitative terms, the result can be a negative correlation in relative abundance between the two organisms due to change in abundance of one relative to the other. The deterministic ecological mechanisms related to ‘good’ or ‘poor’ niche partitioning among microorganisms include neutralism (no effect on abundance of the existing populations), mutualism (co-existing population abundances increasing), predation/parasitism (with part of existing population increasing and other reducing), competition (generally the existing populations reducing), etc. (Abdullah et al. 2017; West et al. 2006) (Fig. 1).

Fig. 1
figure 1

Demonstrates idealistic correlations between two fungal genera including mutualism (a) (both populations increasing); competition (b) (one of the two populations reducing compared to the other); predation (c) (one population increasing and the other reducing); and lack of correlation (d); e (one population changing without any relation to the other). The arrows are not a response to timescale, but rather space

Full size image

Resource partitioning may make it possible for different fungi to occupy a single host or differentially infect crops in their demand for different carbon sources (Arroyo et al. 2008; Giorni et al. 2009; Mohale et al. 2013), while finite nutrient availability may trigger competition between microorganisms. Resource partitioning, therefore, makes it possible for crops to be contaminated with an array of fungal genera. Furthermore, resource partitioning may lead to differentiation in genera that will infect a specific crop. For example, of the mycotoxigenic fungi, groundnuts in the field are known to be more prone to metabolite aflatoxin (AF) contamination due to its susceptibility to Aspergillus infection (Hulikunte et al. 2017; Kachapulula et al. 2017; Ndisio et al. 2017) compared to contamination with the metabolite fumonisin (FB) due to Fusarium infection. On the contrary, maize in the field is more prone to the metabolite FB contamination due to Fusarium infection, compared to Aspergillus infection (Lane et al. 2018; Akello et al. 2021; Katati et al. 2023). However, it should be noted that both Fusarium and Aspergillus will infect pre-harvest maize besides genera like Stenocarpella and Penicillium (Lane et al. 2018; Katati et al. 2023).

Lack of resource partitioning will result in niche exclusion due to the competition for specific niche space (Dorner et al. 1999). In competitive exclusion, one fungal genus, or one species, outcompetes another fungus or species effectively reducing the population of the outcompeted fungus. For example, Fusarium outcompeting Ustilago (Alvarado-Serrano and Knowles 2014) may suggest that under the specific conditions Fusarium and Ustilago would be occupying the same niche. Competitive exclusion is one of the bases for biocontrol of mycotoxins including important ones such as aflatoxin due to Aspergilli (Bock and Cotty 1999; Medina et al. 2017; Raksha et al. 2020). Some important antagonisms include the competition between Fusarium and Trichoderma for the competitive exclusion of Fusarium. Trichoderma is a ubiquitous soil dweller (Bacon et al. 2001) and is non-pathogenic to maize (Gromadzka et al. 2019). It is a prospective fungus in the competitive exclusion of wheat and maize pathogenic Fusarium (Palazzini et al. 2018; Filizola et al. 2019; Lu et al. 2020).

As part of competition for niche space, fungi can produce biochemicals. For example, it has been demonstrated in vitro that outcompeting of Ustilago by Fusarium may be through generation of cell wall-degrading secondary metabolites by Fusarium, competitively excluding Ustilago (Alvarado-Serrano and Knowles 2014). With respect to the production of secondary metabolites by fungi, studies have demonstrated the production of the biochemical aflatoxin in Aspergillus as a defence response to invasion of its ecological space (Trienens et al. 2010; Drott et al. 2017). Similar antagonistic mechanisms include the production of antagonistic agents such as bacteriocins in bacteria (Schoustra et al. 2012). In addition to this, the maize pathogen Stenocarpella has been demonstrated to produce antifungal metabolite chaetoglobosin K against Aspergillus flavus and Fusarium verticillioides (Wicklow et al. 2011). Importance of Stenocarpella in maize is that, even though Fusarium is the most common maize pathogen, it is also an important causative genus for ear and stalk rots (Flett and McLaren 1994; Olatinwo et al. 1999; Rossouw et al. 2009; García-Reyes et al. 2022). Significant infection of maize with Stenocarpella leads to yield loss.

More generally, mycotoxins could be important compounds when species compete within an ecological niche. In mixed fungal communities, it is plausible that the levels of mycotoxins such as FB or AF in the fungal environment could be linked to the way fungal genera in an ecological niche correlate either negatively or positively by relative abundance. In this regard, where poor niche partitioning occurs, and thus high levels of competitive exclusion is expected, it would be anticipated that certain genera may negatively correlate with respect to relative abundance with each other. In addition, it would also be expected that competing fungi would secrete anti-competitor toxins under a negative correlation. Specifically, negative correlation in relative abundance between Fusarium and another genus would be expected to lead to higher levels FB in maize. Importance of FB is its carcinogenicity (IARC 2012; Yu et al. 2021). Of the FB variants (FB1, FB2, FB3, FB4), FB1 is known to form the major compound (70–80%) on average of the total unmasked fumonisin in maize grain (Gil-Serna et al. 2014) and has also been found to be the most toxic one (Wentzel et al. 1992; Yu et al. 2020). The identification of fungal genera with negative correlation in relative abundance to Fusarium would provide prospects for combat of FBs, where the Fusarium is able to be competitively excluded. Unlike the biocontrol of Aspergillus to control aflatoxin, the field biological control of Fusarium to control FBs in maize has been elusive, with success limited to laboratory experimentation (Kagot et al. 2019). Hence, this calls for the need to identify additional biological control agents (BCA) naturally adapted to the same natural environment in which Fusarium thrives on maize.

Fungi will exist on crop through niche partitioning. The aim of this study was to elucidate the existence of evidence that niche partitioning has an effect on competitive exclusion in fungi. We investigated negative and positive correlations in relative abundance of fungal genera present on maize kernels in order to make reliable predictions for competitive exclusion of specific genera of interest. Once consistent correlations would be observed, direct experimental studies could be designed to test if antagonism between strains is indeed a mechanism by which competitive exclusion takes place. The study had two specific objectives: (1) To determine correlations in relative abundance among fungal genera present in maize kernels in order to identify potential mutualistic and antagonistic relations between the fungi. We suggest that absence of negative correlation in relative abundance between fungal genera demonstrates ‘good’ niche partitioning in the maize grain whereas negative correlation may indicate ‘poor’ niche partitioning, signifying potential for competition between such negatively correlated genera. We specifically predict that fungal genera with negative correlation with Fusarium exist. Such genera can potentially be used as BCA to competitively exclude Fusarium from a crop. (2) To evaluate the influence of fungi that are negatively correlated with Fusarium on the levels of the mycotoxin FB1. Our specific hypothesis was that the presence of fungal genera negatively correlated with Fusarium in relative abundance will lead to a decrease in FB in maize, owing to the resulting low abundance of Fusarium, which may consequently lead to its reduced competitive edge to produce FB. Conversely, we hypothesize that a high relative abundance of Fusarium in presence of another negatively correlated fungus will still lead to higher level of FB in the maize, because Fusarium in competition with other fungi will secrete anti-competitor toxins (including FB) in order to outcompete the other fungi.

Materials and methods

Determination of fungal relative abundance correlations

Sample collection and preparation

We investigated fungal relative abundance correlations by analysing the internal mycobiome of previously dried maize kernels (Katati et al. 2023) stored at − 35 °C. The maize samples (n = 40) were collected during the 2018/2019 season from four northerly (n = 20 samples) and four southerly (n = 20 samples) districts of Zambia. The northerly districts are located in a high-rainfall agro-ecological zone (AEZ), while the southerly districts lie in a low rainfall AEZ (Bunyolo et al. 1995; Phiri et al. 2013). For each sample, a 150 g homogenised portion was sterilised by dipping the kernels in 10% (v/v) sodium hypochlorite solution (household grade Jik®) for 3 min to remove unwanted surface DNA (Kampmann et al. 2017, 2022; Nilsson et al. 2022) and microbes. The kernels were then rinsed three times in sterile water and dried inside sterilised cotton bags placed in a Forced Draft Oven (Heraeus, model D-6450, Hanau, Germany) for 24 h, set at 42 ± 2 °C to rapidly drive out the moisture. The dried kernels were then milled using an Ultra Centrifugal Mill (model ZM200, Retsch, Haan, Germany) fitted with a 1.0 mm ring sieve and then stored in sterile polyethylene bags. The mill’s pot and ring were cleaned between samples.

DNA extraction

Extraction of internal mycobiome DNA from the milled maize was carried out according to the Qiagen PowerSoil Quick-start Spin Protocol (detailed at https://www.qiagen.com/nl, product catalogue number HB-2495 of 2019) as follows: small portions of homogenised milled maize were carefully scooped from different positions of the sample, collecting about 200 mg into a bead beating tube for DNA extraction. Next, 800 µl of CD1 lysis buffer was added to the tube, then the prescribed PowerSoil Quick-start Spin Protocol was followed. To adapt to the protocol, tube beating was carried out for 12 min continuously using a sideways homogeniser (Model MM400 Retch, Haan, Germany) set at 25 beats per second. Tube centrifuge was done at room temperature in a microcentrifuge (model 5424, Eppendorf, Hamburg, Germany). The concentration and quality of the isolated DNA were read on a spectrometer (Nanodrop model 2000, Thermo Fischer Scientific, Wilmington, DE, USA).

DNA amplicon sequencing and bioinformatic analysis

For amplicon sequencing and bioinformatic analysis, the purified DNA was normalised to 10 ng µl−1. The DNA was sequenced at LGC Genomics (Biosearch Technologies, Berlin, Germany) on Illumina platform, Miseq V3, by paired-end amplicon sequencing (2 × 300 bp). The ITS1 (nuclear ribosomal internal transcribed spacer 1) region of the fungal genome was sequenced, as described in a previous study (Katati et al. 2023). The following primer pairs were used to amplify and sequence the region, partly overlapping into the 5.8S region: ITS1F_Kyo2 (forward) TAGAGGAAGTAAAAGTCGTAA and ITS86R (reverse) TTCAAAGATTCGATGATTCAC. The sequencing output data were received in the form of adaptor- and primer-clipped demultiplexed samples with a sequencing depth of about 105,000 to 158,000 total reads per sample, with reads less than 100 bp discarded. The bioinformatic analysis of the obtained raw amplicon sequence data was processed using the Divisive Amplicon Denoising Algorithm version 2 (DADA2) pipeline (Callahan et al. 2016) as described in a previous study (Katati et al. 2023). The Unite Fungal Database (UNITE Community 2019) was used to assign taxa to the DADA2-generated amplicon sequence variants (ASVs). The assigned ASVs were the measure for genera abundance. Fungal taxon resolution was done up to genus level as this was sufficient for the niche mapping.

Determination of fungal relative abundance correlations

A Spearman correlation matrix (SCM) was used to assess fungal relative abundance correlations. The top 30 genera out of the full internal mycobiome (n ~ 40 genera) were used, given that these were the more abundant members (≥ 12% frequency of field appearance) likely to drive significant correlations. In the SCM, relative abundance correlations between − 1.0 and − 0.3 were considered (strongly) negatively correlated (potential for antagonism). Correlations between + 1.0 and + 0.3 were considered (strongly) positively correlated (potential for mutualism). Values close to or equal to zero (− 0.30 to + 0.30, inclusive) denoted a weak or no correlation.

Influence of fungal abundance correlations on fumonisin-B1 (FB1)

Data for FB1 were retrieved from a previous study (Katati et al. 2023) in order to associate fungal genera abundances with FB1 in current study.

Data analysis

All statistical computations were conducted in software R (R Core Team 2023) version 4.3.2 (for the R code see Data Availability) using the following packages and their functions, partly described in a previous investigation (Katati et al. 2023): “phyloseq” (McMurdie and Holmes 2013) for internal mycobiome census; “reshape2” (Wickham 2007) for generation of Spearman correlation matrix (SCM) of fungal relative abundance correlations; “Hmisc” (Harrell and Dupont 2019) for determination of significance of fungal relative abundance correlations; ‘vegan’ (Oksanen et al. 2010) and ‘dplyr’ (Wickham et al. 2021) for MDS (Multi-Dimensional Scaling) data decomposition of agronomic factors; “ggplot2” (Wickham 2016) for visualisation of data as figures (SCM and MDS).

Influence of fungal correlations on FB1

The internal mycobiome was studied for this purpose. This is presuming that this would provide a more discrete relation between the mycobiome colonising the grain and mycotoxin, as the FB is harboured in the grain, even though the phyllosphere would still be expected give a good relation. To associate fungal genera relative abundance with FB1 levels, Spearman rank correlation (rho) was used such that FB1 data and the corresponding fungal relative abundances are ranked in the algorithm, due to the heterogenous distribution of the FB1 datum.

Agronomic (confounding) factors with potential to influence Fusarium and fumonisin-B1 levels

Additional agronomic factors, as confounding elements, with potential to influence Fusarium proliferation (Roucou et al. 2021) as well as FB1 levels in the maize, were appraised. The factors were obtained through a questionnaire (Supplemental Questionnaire S1). The questionnaire was administered with the help of agricultural field extension officers who routinely work with the farmers. To assess the link between the agronomic factors and FB1 levels as a result of Fusarium proliferation, each factor per field was assigned the corresponding levels of Fusarium relative abundance (if response to question was affirmative) or ‘0’ (if response to question was negative). The factors, included seed type (early, medium or late maturing variety), cropping type, and presence or absence of pests. Responses were computed into a dataframe (see Data Availability).

Multi-Dimensional Scaling (MDS) was used to determine if the agronomic factors did not differentially influence Fusarium proliferation as well as FB1 levels in the maize. The measure used in the MDS was Fusarium relative abundance. The FB1 data were derived from a previous study (Katati et al. 2023), whose maize concentration we ordinated in the current study as low (< 50 µg kg−1, n = 17), medium (50–300 µg kg−1, n = 13) and high (> 300 µg kg−1, n = 10) across the 40 sampled fields. Fusarium relative abundance was similarly ordinated into arbitrary levels as high (> 65%, n = 14), medium (21–65%, n = 13) and low (≤ 20%, n = 13) relative abundance. Cutoff point for agronomic factors with influence on Fusarium/FB1 levels was set at a P value of 0.05.

Results

Fungal relative abundance correlations

On average, Sarocladium and Fusarium had the highest internal mycobiome relative amplicon sequence variant (ASV) abundance of 47% and 26%, respectively. Stenocarpella was third highest at 13%. The full table of the ASV relative abundance representation of the internal mycobiome genera is available on a public repository as (https://github.com/bkatati/nichemap/blob/main/InternoBiome.csv). Of the mycotoxin- important genera, and the high relative abundance genera, Fusarium-Sarocladium and Fusarium-Stenocarpella had the strongest negative significant correlations by relative abundance (Fig. 2; see Supplemental Table S1). There was no clear correlation between Sarocladium and Stenocarpella relative abundances (Fig. 2; rho = − 0.15, P = 0.011). Similarly, the correlation between Ustilago and Fusarium was very weak (Fig. 2) although the dry weather pattern showed an inclination for a strong negative correlation between the two genera (rho =− 0.650, P < 0.001) compared to the wet weather pattern with a non-significant correlation (rho = 0.205, P = 0.058).

Fig. 2
figure 2

Spearman correlation matrix, showing top 30 genera in descending abundance. The correlations are the overall combining internal mycobiome fungal relative abundances for both dry (southerly) and wet (northerly) agro-ecological zones of Zambia. Negative values (blue) are negative correlations and positive (red) are positive correlations in fungal relative abundances. Individual correlations of the genera per district are presented in Supplemental Table S1. Of the total number of correlations appraised (n = 435 pairs out of 30 genera), most correlations were either strongly positive (13.6%) or were weak in either direction (82.7%). Strong negative correlations were 3.7%. Note: Fusarium (equiseti) is detected as a distinct anamorph of the former Gibberella (intricans) based on the UNITE fungal database (UNITE Community 2019). Talaromyces is assessed as a distinct clade from its asexual anamorph Penicillium

Full size image

Influence of fungal genus abundance on FB1 levels

From the internal mycobiome, there was a negative correlation between relative abundance of Stenocarpella and levels of FB1 (rho = − 0.33, P = 0.04). The observed negative correlation between the levels of Sarocladium and FB1 was not significant (rho = − 0.10, P = 0.52). However, fields with lower relative abundance of Sarocladium (< 10%) had a higher frequency of higher FB1 levels when compared to fields with medium or higher levels of Sarocladium (Table 1). Similarly, there was no correlation between Ustilago relative abundance and levels of FB1 (rho = − 0.21, P = 0.20). Fusarium relative abundance positively correlated with levels of FB1 (rho = 0.39, P = 0.01).

Table 1 Frequency of FB1 (µg kg−1) levels in maize fields in relation with Sarocladium levels
Full size table

Assessment of influence of confounding (agronomic) factors on Fusarium and fumonisin-B1 levels

The additional agronomic factors, as confounding elements, did not differentially influence Fusarium and FB1 levels (Fig. 3). The factors with significance to be able to influence FB1/Fusarium proliferation included seed type by maturity (early: P = 0.001, R = 0.37; medium: P = 0.001, R = 0.46; late maturing: P = 0.001, R = 0.39), and pest incidence during cropping (P = 0.006, R = 0.27). Factors with no significance to influence FB1/Fusarium proliferation were field burning (P = 0.166) and monocropping (P = 0.937).

Fig. 3
figure 3

In the MDS, arrows are orientation of agronomic factors based on relative abundance of Fusarium. For example, the arrow ‘med’ refers to a medium maturing seed variety. Dots are fields (n = 40), and each field has a value of FB1 level (μg kg−1) and Fusarium relative abundance (%). Ellipses link fields of similar arbitrary levels of Fusarium (a) or FB1 (b). The dots have been scattered by MDS based on relative abundance of Fusarium. Levels of Fusarium are ordinated as ‘High’ (> 65%, n = 14), ‘Med’ (21–65%, n = 13) and ‘Low’ (≤ 20%, n = 13). Similarly, FB1 values are ordinated as ‘High’ (FB1 > 300 µg kg−1, n = 10), and ‘Med’ (FB1 = 50–300 µg kg−1, n = 13) and ‘Low’ (FB1 < 50 µg kg−1, n = 17). Only arrows of the significantly (P < 0.05) important agronomic factors on Fusarium proliferation are shown (n = 4 out of six), namely seed type by maturity (early: P = 0.001, R = 0.37; medium: P = 0.001, R = 0.46; late maturing: P = 0.001, R = 0.39) and pest incidence during cropping (P = 0.006, R = 0.27). Thus, monocropping (P = 0.93) and burning (P = 0.18) are excluded. The arrows (agronomic factors) subsequently orient towards ellipses for Fusarium/FB1 levels. The ellipse overlaps in the MDS for both Fusarium and FB1 and the absence of clear orientation of the arrows to a particular ellipse, even at reduced confidence level (67%), indicate a lack of correlation of the appraised agronomic factors with Fusarium proliferation or FB1 levels in the pre-harvest maize

Full size image

Discussion

Fungal relative abundance correlations defined by niche partitioning

In this study, we associate the fungal genera relative abundance correlations with niche partitioning. Sarocladium and Fusarium had the highest relative abundance by amplicon sequence variant representation in the maize internal mycobiome (47% and 26%, respectively) followed by Stenocarpella (13%). The strong negative relative abundance correlation between Fusarium and Sarocladium suggests poor niche partitioning between the two genera. The poor niche partitioning is characteristic of competition, a likely consequence of similarity in niche utilisation. Similar niche antagonism between Fusarium and Sarocladium has been demonstrated by previous solid substrate plating in vitro studies (Comby et al. 2017; Kemp et al. 2020). Similarly, our simulation data of 2018/2019 cropping season between Sarocladium and Fusarium external mycobiome relative abundances from a previous study (Katati et al. 2023) reveals a consistent negative correlation (rho = − 0.70, P < 0.001) (Supplemental Data S1 available as https://github.com/bkatati/nichemap/blob/main/S1_ExternoBiome.csv).

A similar negative correlation in relative abundance was observed between Stenocarpella and the mycotoxin-important genera Fusarium and Aspergillus (Fig. 2; see Supplemental Table S1). These observations are consistent with a past study done by solid substrate plating that demonstrated a negative correlation between Stenocarpella and two species, Fusarium verticillioides and Aspergillus flavus (Wicklow et al. 2011). Our findings suggest that even in natural ecosystems, such negative or positive correlations in abundance between certain specific genera are likely to exist, even at higher taxonomic ranking (genus level in the current study).

There was no demonstrable overall correlation in relative abundance between Fusarium and Ustilago (rho = 0.08, P < 0.001). This is contrary to findings by Alvarado-Serrano and Knowles (2014) who demonstrated growth ability competition between the two genera in vitro. The variance between our findings and those by Alvarado-Serrano and Knowles (2014) may be attributed to our in situ abiotic conditions. For example, a strong negative correlation, similar to the report by Alvarado-Serrano and Knowles (2014), is observed in the current study between the Fusarium and Ustilago under the drier conditions (rho < − 0.60, P < 0.001), whereas under the wetter conditions the positive correlation was not significant (P = 0.058; see Supplemental Table S1).

While other studies have demonstrated that Trichoderma is a ubiquitous dweller of soil (Bacon et al. 2001; Egidi et al. 2019), which is the natural reservoir for fungi translocating to crop, we did not detect this fungus in the maize internal mycobiome. Similarly, in previous Zambian studies, Trichoderma was only detected on maize in one of 11 districts (Mukanga et al. 2010) and not detected at all on maize grain phyllosphere (Katati et al. 2023). We may attribute the absence of Trichoderma on maize to its poor colonisation of this crop. It would be worth testing for Trichoderma presence or absence in the soil mycobiome from the sampled areas of the current study to better understand if the fungus’ poor or non-colonisation of the maize mycobiome was due to its exclusion in the environment (soil) or on the maize. Importance of Trichoderma to the maize mycobiome would be its non-pathogenicity to maize (Gromadzka et al. 2019), which would make it the desirable endophyte on maize compared to maize-pathogenic Fusarium.

Influence of genera abundance correlations on FB1 levels

With respect to antagonism that may be mediated by presence of mycotoxins, it would have been expected that higher levels of FB1 would have correlated with higher levels of Stenocarpella. This is taking into account that production of secondary metabolites in fungi can be triggered as a response by a fungus to defend its ecological niche due to competition with other fungi. For instance, Fusarium is reported to produce cell-degrading metabolites when in competition with Ustilago (Alvarado-Serrano and Knowles 2014). Similarly, Stenocarpella produces the metabolite chaetoglobosin K, which inhibits the proliferation of Fusarium verticillioides and Aspergillus flavus (Wicklow et al. 2011). In the current study, a negative correlation was detected between Stenocarpella and FB1 besides its negative correlation with Fusarium relative abundance. We attribute this to the reduced growth competitiveness of Fusarium when the relative abundance of Stenocarpella is high, as seen from the negative correlation between the two genera (Fig. 2; see Supplemental Table S1), leading to a diminutive effect on the FB-producing ability of Fusarium. The observed relationship between Stenocarpella and levels of FB demonstrates for the first time the (indirect) negative influence of Stenocarpella on levels of FB1 in pre-harvest maize.

With respect to Sarocladium, another high relative abundance genus in this study, we did not detect a significant correlation between levels of FB1 and Sarocladium abundance. However, it was observed from ordinal data that, when Sarocladium was in very low abundance (< 10%), the frequency of detection of high FB1 levels (> 100 µg kg−1) increased (Table 1). Similarly, we did not detect any correlation between Ustilago abundance and FB1 levels in maize (rho = − 0.21, P = 0.20), which resonates with absence of an overall correlation in relative abundance between Ustilago and Fusarium. We note, however, that, where confrontation between Fusarium and Ustilago is demonstrated, the upregulation of the excretion of secondary metabolites and cell wall-degrading proteins by Fusarium may occur (Alvarado-Serrano and Knowles 2014). Although Sarocladium is a reported rice pathogen (Sakthivel et al. 2002), our current findings and those from a previous study (Katati et al. 2023) indicate its capability to naturally and extensively infect maize, satisfying preliminary conditions as a genus with prospects for the competitive exclusion of Fusarium and subsequently FB in maize.

Prospects in biocontrol of fumonisin (FB)

Considering that the biocontrol of mycotoxins such as AF, FB and deoxynivalenol is premised on competitive exclusion of one fungus by another (Bandyopadhyay et al. 2016; Tian et al. 2016; Błaszczyk et al. 2017; Filizola et al. 2019), the observed negative correlations in fungal relative abundances are a useful attribute in prospecting for fungal genera that could be used for the competitive exclusion of Fusarium and its mycotoxin FB. In this regard, the observed negative correlation between Fusarium and Sarocladium relative abundances helps to define prospects for the utilisation of such a genus as biological control agent (BCA) for the biocontrol of FB. It should also be noted that one of the attributes for the success of biocontrol is that the BCA should be able to effectively and widely contaminate the crop and thrive in the natural ecosystem in which the target pathogen thrives. In this investigation, this attribute has been demonstrated by Sarocladium (Fig. 2; see Supplemental Table S1). This then suggests the identification of strains of Sarocladium that are non-pathogenic to humans and maize for the prospective competitive exclusion of Fusarium. For example, non-pathogenic Sarocladium zeae is prospective candidate for the biocontrol of Fusarium in wheat (Kemp et al. 2020). Although Stenocarpella showed a negative correlation with Fusarium relative abundance and FB1 levels, the fungus is largely a pathogen to maize and produces harmful metabolites that can induce diplodiosis in livestock (Snyman et al. 2011; Masango et al. 2015). Hence, unlike genera such as Stenocarpella and Ustilago, non-pathogenic strains of Sarocladium may be good candidates for the control of FB1 contamination of pre-harvest maize.

A critical next step in this work is to design directed experiments to assess if these relative abundance correlations are indeed caused by niche partitioning and that mycotoxins play a critical mechanistic role. Experiments using pairs of strains grown under various levels of niche partitioning (e.g., different types of nutrients) could be used to test direct and specific predictions on what strains may outcompete others under what levels of niche partitioning, and whether or not mycotoxins such as FBs would be important antagonistic agents.

In conclusion, we studied correlations between fungal genera relative abundances, under the assumption that these correlations may be driven by niche partitioning. We found evidence for these correlations using internal mycobiome fungal populations on sampled maize, as the natural ecosystem model. We have demonstrated that niche partitioning may affect the potential for competitive exclusion of fungi on maize. In this regard, the overall hypothesis that lower levels of Fusarium and FB1 would ensue due to fungus with a negative correlation in abundance with Fusarium, and vice versa, was subsequently not rejected.