Rice spikelet rot disease (RSRD), also called pecky rice or kernel spotting, is one of the fastest spreading diseases in China, particularly along the middle and lower reaches of the Yangtze River (Huang et al., 2011a). It was categorized as a minor disease that occurred only sporadically before 2006. After that, RSRD epidemics became more frequent and serious due to the extensive cultivation of japonica varieties and indica-japonica hybrids (Wang L. et al., 2022). Fusarium proliferatum (teleomorph: Gibberella fujikuroi) is the most predominant pathogenic fungus associated with RSRD (Huang et al., 2011b; Wang et al., 2021). Rice is highly susceptible to F. proliferatum during the flowering stage (Sun et al., 2019a). After spore germination, the fungal hyphae enter glumes through the anthers at the pollen-filling stage, proceed to colonize pistils and endosperms, and finally infect the whole floral tissues (Sun et al., 2019b). The initial lesions of spikelet rot symptom are reddish-brown or rust-red on rice spikelets at the flowering and milking stage, eventually become blackish-brown during the ripening stage. High precipitation, relative humidity, and warm temperature during anthesis in rice favor the development of RSRD. The percentage of diseased panicles can reach up to 80% under the severe situation of the epidemic (Li et al., 2015). To date, RSRD has been estimated to occur in one third of rice-growing regions in China, resulting in yield losses of 5–30% (Wang L. et al., 2022). How and why the disease emerged in rice is still a matter of speculation. Examination of the spatial and temporal dynamics of disease progression that is prerequisite to develop risk assessments and provide the effective management approaches.

F. proliferatum is well known for its pathogenicity to a wide range of cereals, such as wheat, rice, barley, maize, rye and oats. Moreover, the pathogen is notorious for producing multiple mycotoxins, including fumonisins, moniliformin, fusaric acid, fusarin C, and beauvericin (Studt et al., 2012). Fumonisins are associated with multiple human and animal diseases, as they are potent inhibitors of sphingolipid metabolism in eukaryotes (Braun and Wink, 2018). Fumonisins induce equine leukoencephalomalacia in horses, pulmonary edema in pigs, as well as liver and kidney toxicities in rats (Riley and Merrill, 2019). Epidemiological evidences indicated a causal relationship between high levels of fumonisins exposure and neural tube defects and esophageal cancer in human beings (Yu et al., 2021). A widespread survey of fumonisins contamination on a global scale in the post-harvest grain foods ranged from 39% in Europe to 95% in North and South America (Lee and Ryu, 2017). Fumonisins comprise a group of 28 analog, which can be separated into four main groups, fumonisin A, B, C, and P (Braun and Wink, 2018). Fumonisin B (FB) analogs include FB1, FB2, and FB3, with FB1 being the most prevalent component and usually being found at the highest levels (Kamle et al., 2019). International Agency for Research on Cancer (IARC) classified FB1 as a class 2B possible carcinogen for human (International Agency for Research on Cancer [IARC], 1993). The European Union had set maximum limits for fumonisins in food and feedstuffs intended for human consumption (European Commission [EC], 2006). In order to prevent the reproduction and contamination of F. proliferatum as efficiently as possible, the control methods should be implemented in dependence on the fungi growth, reproduction, infection and mycotoxins formation in the agricultural fields.

Currently, it is known that the activation of biosynthesis pathways of secondary metabolites (SMs) not only depends on pathway-specific regulators, but also on global transcriptional complexes, signal transduction regulation, and transcriptional and epigenetic manipulation (Brakhage, 2013; Keller, 2019). Fumonisins are a family of polyketide synthases (PKS) derived SMs, consisting of a linear 20-carbon aminopolyhydroxyalkyl chain backbone. The 42-kb long cluster of fumonisin biosynthetic genes (FUM) had been well characterized in Fusarium verticillioides (Brown et al., 2007; Proctor et al., 2013). FUM genes are usually silent or expressed at very low levels under non-inducing conditions. Several external triggers, such as nutritional input (Kim and Woloshuk, 2008), light regulation (Wiemann et al., 2010), ambient pH (Flaherty et al., 2003), and water activity (Cendoya et al., 2018), had been characterized to be involved in fumonisins production. It is therefore necessary to decipher the key molecular players regulating the biosynthesis of fumonisins, as well as the interplay between the lifestyle of toxigenic fungi and genetic constitution.

To date, no reports have been elucidated about the biology and genetics of F. proliferatum population from rice. In this study, we reported the genome sequences of 67 F. proliferatum strains collected from major rice-growing regions in China, in order to characterize its genetic diversity and population structure, and dissect the genetic determinants involved in fumonisins biosynthesis. The findings of this study provided a genome-wide perspective on the population of F. proliferatum and clarified their genetic relationships with fumonisins production.

In the last decade, RSRD has been one of the most widespread diseases in China, causing severe reduction in rice yield and grains contamination by producing toxic metabolites that are hazardous to humans and livestocks. More importantly, the application of fungicides was not fully operative in controlling accumulation of mycotoxins. Few technologies and approaches are available to prevent fumonisins contamination. Given its economic and public health impacts, an efficient way is to gain the perspective of population biology of F. proliferatum before the control is implemented. To the best of our knowledge, this was the first report of the genomic features of F. proliferatum population obtained from the major rice-growing region in China (Figure 1 and Table 1). These findings provided new insights into its epidemiologic characteristics and evolutionary processes of this fungus. Such information could contribute to future prospects in designing disease management for RSRD.

The contemporary patterns of pathogen composition and population structure were shaped by historical evolutionary processes over time and space. However, the early history of F. proliferatum on cultivated rice, especially the impact of rice domestication on pathogen population, remains still unknown. Here, high level of genetic variation and recombinant mode were found in eastern population of F. proliferatum located in the middle and lower valley of the Yangtze River in China, where traditional varieties of indica and japonica rice were domesticated and co-cultivated with a long history. A hypothetical but straightforward scenario was that the eastern population might be the origin center from where F. proliferatum migrate toward the rest of regions, whereas the southern and northeastern populations were the founder populations that may undergo genetic bottlenecks or recent population introduction. There was evidence of gene flow between eastern population and southern or northeastern populations (Table 2), probably as a consequence of human activities like trade and transportation of F. proliferatum–contaminated rice seeds.

Intriguingly, southern and northeast populations of F. proliferatum showed geographically habitat preferences at the regional scale. Northeast population was represented in areas where japonica varieties were cultivated. Conversely, southern population was prevalent in regions where indica varieties were grown. The geographical populations of F. proliferatum followed the unique evolutionary trajectory, which suggested the signature of rice domestication as the major driving force for population divergence. In Microbotryum fungi, species-specific genes corresponded to putative adaptations responding to host-mediated selection (Badouin et al., 2017). Such epidemiological implications, probably accompanied by the host specialization, exemplified that the deployment of resistance cultivars must keep up with co-evolutionary “arm races.” By introducing the genes, such as major resistance genes and defense-responsive or defense-related genes specific for one subspecies into the other subspecies, F. proliferatum might confront selection pressures imposed by host resistance that it has never met before.

Sexual reproduction plays the fundamental roles in maintaining the genetic diversity and enhancing the adaptation to environments in fungi. Nonetheless, the stage of teleomorph of F. proliferatum has not been well described. One reason for the widespread spread of F. proliferatum was its ability to produce vast numbers of mitotically derived asexual spores, namely conidia. Our results supported that heterothallism was likely to occur frequently in the eastern and southern populations demonstrated by existence of two MAT idiomorphs (Table 1). The opposite mating types were needed for the perithecium formation of F. proliferatum. Epidemic of RSRD was possibly aggravated by the ability of the pathogen to spread through airborne ascospores. The phenomenon of F. proliferatum strains containing both MAT idiomorphs over large regions was congruent to the previous observations of that collected from durum wheat and garlic bulb (Palacios et al., 2015; Gálvez et al., 2017). The evidences implied that there was probability that F. proliferatum propagate asexually by conidia and sexually by ascospores allowing both selfing and outcrossing. The strains with strong competition exhibit dominance in geographical areas, which enhanced its potentiality to respond to disease management, e.g., fungicides modes, resistant cultivars, agronomic practices, or even the climate change. In addition, a single clonal lineage of mating type MAT1-1 was detected in northeast population (Table 1). From the practical standpoint, the finding highlighted the need for persistent vigilance of northeast population, to prevent the invasion of MAT1-2 idiomorph that might initiate sexual component to the rice-F. proliferatum pathosystem.

On a broader scale, the genome-wide level of polymorphism in F. proliferatum was 2.3 variants per kb (Table 1), which was similar to that in wheat powdery mildew Blumeria graminis (1.4–2.0 SNPs/kb) (Wicker et al., 2013), larger than blackleg Leptosphaeria maculans (0.5 SNPs/kb) (Zander et al., 2013), but much lower than poplar rust Melampsora larici-populina (6 SNPs/kb) (Persoons et al., 2014). Naturally, plant pathogens with higher diversity have relative advantages of the ecological fitness. Combining effects of natural selection, sexual recombination and genome-wide polymorphism, F. proliferatum was considered as the high-adaptive pathogen with the ability to occupy ecological niches or infect broad range of hosts during fungal population expansion. In this regard, continued surveillance of pathogen will be critical for evaluation of resistant cultivars. Resistance breeding can benefit from the inclusion of representative strains of F. proliferatum rather than a single strain when screening resistance varieties to RSRD.

There was the remarkable difference in capability of fumonisins biosynthesis among F. proliferatum strains from rice (Table 1). Similar phenotypic differences were observed among F. proliferatum strains from asparagus (von Bargen et al., 2009), pineapple (Stêpieñ et al., 2013), pea (Waśkiewicz et al., 2013), and garlic (Gálvez et al., 2017). It has been proved that the regulatory networks for fungal SMs are modulated by hierarchical interconnection with diverse cellular processes from cluster-specific pathways to global transcriptional complexes (Macheleidt et al., 2016). Biogenesis of fumonisins at the levels of pathway steps and their regulation remains a challenging task. There is a knowledge gap in the genetic diversity and fumonisins production capacity of F. proliferatum. Several attempts have been made to decipher certain interested genes located outside FUM gene cluster, it is not sufficient to elucidate the unexplored biomolecules responsible for fumonisins formation. This work aimed to depict the regulation factors involved in fumonisins production in F. proliferatum using GWAS based on whole-genome resequencing. Currently, GWAS is a preferable method for detection of the genetic determinants associated with phenotypic variation for fungal pathogens at species level, such as cold tolerance in Clonostachys rosea (Broberg et al., 2018), virulence in Zymoseptoria tritici (Hartmann et al., 2017), as well as aggressiveness, deoxynivalenol (DON) production and azole sensitivity in Fusarium graminearum (Talas et al., 2016). As far as we know, this is the first study to apply GWAS to link polymorphism in the genome to functional variability associated with fumonisins production among natural populations of F. proliferatum strains.

We identified that oxidation-reduction processes (GO:0016705, GO:0016491) were connected with fumonisins biosynthesis in F. proliferatum (Table 3). Similarly, the activation roles of cellular redox stress were previously found on aflatoxin biosynthesis in Aspergillus flavus (Fountain et al., 2019) and trichothecene production in F. graminearum (Montibus et al., 2013). The majority of oxidation, reduction and hydrolysis reactions are performed by CYPs. CYPs enzymes have unique ability to catalyze regio-, chemo-, and stereospecific conversions of lipophilic compounds to hydrophilic derivatives by introducing an oxygen atom. Three members from CYPs (FPRO_06718, FPRO_05147, FPRO_05649) were identified to be related with fumonisins production (Supplementary Table 2). Several CYPs participating in SMs biosynthesis pathways had been characterized in other fungi. Among 26 genes located in aflatoxin biosynthesis gene cluster, aflG, aflQ, aflU, and aflV encoded CYPs in A. flavus (Yu et al., 2004). In F. graminearum, three CYPs (TRI1, TRI4, and TRI11) were responsible for trichothecene biosynthesis, and three CYPs encoded ergosterol biosynthetic enzymes (Shin et al., 2017). In Verticillium dahliae, VdCYP1 had a pronounced effect on 14 kinds of fungal metabolites that likely contributed to pathogenic process (Zhang D. D. et al., 2016). Therefore, it was conceivable that the catalytic regulation of fungal CYPs was essential for cell viability and secondary metabolism.

Transmembrane transport (GO:0055085) was found to be involved in fumonisins production (Table 3). Transporters are responsible for nutrient uptake, metabolite extrusion, multidrug resistance and signal exchange. ATP-binding cassette (ABC) transporters are a superfamily of proteins that use the energy from ATP hydrolysis to transport substances across the cell membrane. Our results showed that fumonisins production was affected by an ABC transport (FPRO_03446) (Supplementary Table 2). This seemed to be the case for Bacillus cereus as well because ABC transporters were involved in the biosynthesis of metabolites, such as toxins, antibiotics and siderophores (Gacek-Matthews et al., 2020), but the underlying mechanisms remain unclear. Peptide transporters belonging to the major facilitator superfamily (MFS) transporters use the proton-motive force to translocate di- and tripeptides. Internalized peptides are rapidly hydrolyzed and the resulting amino acids are used for protein synthesis or alternative sources of nitrogen and carbon. We found that a peptide transporter (FPRO_13944) was associated with fumonisins biosynthesis (Supplementary Table 2). In F. graminearum, deletion of peptide transporters (FgPTR2A, FgPTR2C, and FgPTR2D) resulted in a sharp increase in DON and zearalenone and a decrease in fusarielin H. The three mycotoxins fell under the regulation of nitrogen regulator AreA because that the transportation activity of peptide transporters was influenced by the quality of the nitrogen source (Droce et al., 2017). Further studies are required to understand the exact nature of how the peptide transporters modulate the genetic regulatory network underlying the SM biosynthesis.

Carbohydrate metabolic process (GO:0005975) was shown the association with fumonisins production in F. proliferatum (Table 3). The sources of carbon available were reported to play a critical role in induction of SM biosynthesis, such as fumonisins biosynthesis in F. verticillioides (Malapi-Wight et al., 2013), aflatoxin production in A. flavus (Fasoyin et al., 2018), ochratoxin A biosynthesis in Aspergillus ochraceus (Wang et al., 2020). Glucose is the preferred carbon source for the filamentous fungi. Several genes encoding enzymes (e.g., xylanases, cellulases, and arabinases) required for the use of alternative carbon sources, such as lignocellulose, are repressed by carbon catabolite repression (CCR) when glucose is present. From the GWAS, three significant SNPs were located in genes encoding β-glucosidase (FPRO_14041, FPRO_03671, FPRO_15192) (Supplementary Table 2). β-glucosidase mainly hydrolyze β-glucosidic bonds to release glucose from the non-reducing end of β-glucoologosaccharides or glucosides. This effect was mediated in part by sucrose non-fermenting (SNF1) kinase that was regarded as an important regulator in cellulose degradation and sterigmatocystin biosynthesis in Podospora anserina (Li et al., 2020). Several fungal and bacterial β-glucosidases had the abilities to hydrolyze the Fusarium mycotoxins, such as DON, nivalenol and HT-2 toxin (Michlmayr et al., 2015). Thus, this would be reasonable that fungi have evolved mechanisms to balance the operation of substrate feeding (enzyme production) and secondary metabolism during life cycle.

Amine metabolic process (GO:0009308) was found to exert influences on fumonisins biosynthesis (Supplementary Table 2). Nitrogen status directly affected the ability of filamentous fungi to biosynthesize SMs (Tudzynski, 2014). Biogenic amines are low molecular weight organic nitrogen compounds in living cells. Amine oxidases catalyze the oxidative deamination of amine substrates to their corresponding aldehydes, including aliphatic and aromatic monoamines, diamines, tertiary amines, polyamines, and amino acids. Amine oxidases were known to be involved in primary amine metabolism, but the effects of amine oxidases on SMs production is rare. Two genes encoding amine oxidase (FPRO_13752, FPRO_06592) were identified to be related to fumonisins production in F. proliferatum (Supplementary Table 2). Amine oxidases of Aspergillus niger and Cochliobolus victoriae were capable of enzymatic conversion of fumonisins and victorin into deaminated and detoxified counterparts, respectively (Garnham et al., 2020; Kessler et al., 2020), which demonstrated relaxed substrate specificity of amine oxidases in mycotoxins biosynthesis.

The gabap gene encoding GABA permease was found to be significantly associated with FB1 production (Figure 5B). The gabap mutant failed to synthesize FB1 and inactivated the expression of FUM1 gene responsible for fumonisin biosynthesis. There were several indications for the connectivity between GABA shunt and the regulation of FB1 production. GABA metabolism had been previously described to link MFS transporters that pumped out DON mycotoxin in F. graminearum (Wang et al., 2018). The addition of GABA-inducible agmatine promoted the production of starting substrate for DON biosynthesis in Fusarium asiaticum (Suzuki et al., 2013). Deletion of gabap impaired GABA utilization in F. proliferatum, which was partially regulated by the global nitrogen regulator AreA. The mutant Δgabap showed downregulation of the key enzymes of the TCA cycle. Thus, we proposed that GABA as a nitrogen source positively regulated FB1 accumulation in F. proliferatum.

We identified a correlation between chsD gene encoding chitin synthase and FB1 production (Figure 6A). Disruption of chsD resulted in reduced accumulation of chitin and enhanced sensitivity against cell wall perturbing agents. In addition, our study found that chitin synthase played an important role in pathogenicity in F. proliferatum as previous studies on Magnaporthe oryzae (Kong et al., 2012) and Botrytis cinerea (Cui et al., 2013). The reduced FB1 production in the ΔchsD deletion mutant further verified the involvement of chitin synthase in aggressiveness since FB1 played an important role in the infection of F. proliferatum in rice tissue (Sun et al., 2019a). The effect of chsD gene on FB1 production in F. proliferatum was similar to that on DON production in F. graminearum (Liu et al., 2016). Our results demonstrated that essentiality of chitin biosynthesis in vegetative growth, virulence and FB1 production in F. proliferatum, which allow the development of novel inhibitory agents in the control of this pathogen by targeting the chitin synthase.

Fungal microorganisms must be able to sense and respond to ambient pH for mycelial growth, host colonization and toxin production. The ambient pH signaling pathway in filamentous fungi was medicated by six Pal proteins (PalA, PalB, PalC, PalF, PalH, and PalI) and transcription factor PacC (Peñalva et al., 2008). It was clear from previous studies that PacC was a key factor for biosynthesis of various SMs, such as ochratoxin A in A. ochraceus (Barda et al., 2020), fumonisins in F. verticillioides (Flaherty et al., 2003), dipicolinic acid in Beauveria bassiana (Luo et al., 2017), patulin in Penicillium expansum (Chen et al., 2018). The activation of PacC by proteolysis required interactions of with PalA and PalB in an alkaline pH dependent manner. In the present study, the palA gene was involved in the association of FB1 production (Figure 7B). Genetic disruption of palA rendered enhanced FB1 accumulation under external pH alkalinization in F. proliferatum. To further characterize the mechanism of the environmental pH-sensing complex on fumonisins, studies are underway to dissect this intriguing phenomenon.

Our findings suggested a relationship between hxk1 gene encoding hexokinase and FB1 biosynthesis in F. proliferatum (Figure 8D). The deletion mutant of hxk1 failed to grow on fructose and down-regulated the expression level of glycolytic genes. More importantly, Δhxk1 produced a dramatically low level of FB1, indicating that hxk1 was involved in FB1 biosynthesis in F. proliferatum. Pyruvate, the end-product of glycolysis, was controlled by hexokinase as the main substrate of acetyl-CoA for the biosynthesis of many SMs (Gao et al., 2016), including fumonisins, aflatoxin, trichothecene, DON, and penicillin. HXK1 had been shown to regulate DON production in F. graminearum (Zhang L. et al., 2016) and FB1 production in F. verticillioides (Kim et al., 2011). Our observations demonstrated that HXK1 was required for FB1 biosynthesis by establishing genetic link between primary and secondary metabolism in F. proliferatum.

ISWI catalytic subunit (ISW2) belonging to a subfamily of chromatin remodeling complexes was associated with FB1 production (Figure 9A). Perhaps the more striking observation was that the effects of ISW2 on biosynthesis of multiple SMs were regulated by nitrogen availability in F. proliferatum. The concept of ISW2 triggering SMs biosynthesis in filamentous fungi was almost without precedent. The chromatin remodeling complex RSC1 had been well characterized in Saccharomyces cerevisiae, which was required for TOR protein kinase (Yu et al., 2015). TOR signaling pathway is a central regulatory hub that connects nitrogen availability with fungal development and metabolic processes (Dobrenel et al., 2016). Therefore, it was quite likely that ISW2 played an indirect role for secondary metabolism by establishing the genetic link between nutrient-signaling regulation and SMs production.

As noted above, five candidate genes (gabap, chsD, palA, hxk1, and isw2) from the GWAS that were all required for FB1 production, possibly reflecting pleiotropic effects in coordinating the processes of fumonisins biosynthesis. It was very likely that, in such cases, SM production of filamentous fungi was controlled by a complex regulatory and biosynthesis pathways in a manner related to fungal development or in response to abiotic and biotic stressors. However, it was worthy to note that some associations were poorly annotated with unknown contributions to SM production. Further research should be intensified to identify and validate the physiological functions of the genetic components. The top hit from our association analysis will serve as excellent targets for genetic and biochemical efforts to detoxify or decontaminate mycotoxins of F. proliferatum.

In this study, we firstly characterized the intraspecific genomic variation of F. proliferatum populations from rice in China, to highlight rapid evolution shaped by high genetic diversity and sexual recombination, and elucidate selection pressures imposed by host domestication on population structure. These findings provided the evidences that the host-specific selection left signatures of genetic differentiation in F. proliferatum populations. Additionally, this work indicated the capability of association mapping to pinpoint genetic determinants involved in fumonisins formation in F. proliferatum. The regulatory factors related with secondary metabolism need to be validated for their potential functions. The sustainability of resistant cultivars and fungicides used to control RSRD should be geared toward information on the population characteristics of F. proliferatum, which contribute to monitor the influx of pathogen with novel adaptations, as well as discover pharmaceutical chemicals to impact positively on food safety and security.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found at: https://www.ncbi.nlm.nih.gov/bioproject/?term=PRJNA517364.

LW, ST, and PH contributed to the idea and design of the research. LW, GJ, XW, GS, and LX collected the strains. LW, SG, WHL, WYL, and WL conducted the experiments. QL was responsible for the whole-genome sequencing and analysis work. ZS and SH collected the data and performed statistics. LW wrote the first draft of the manuscript. ST and PH performed the revision and editing of the final manuscript. All authors contributed to revisions of the manuscript, read, and approved the submitted version.