Genetic Analysis of the Aspergillus flavus Vegetative Compatibility Group to Which a Biological Control Agent That Limits Aflatoxin Contamination in U.S. Crops Belongs

Abstract

Some filamentous fungi in Aspergillus section Flavi produce carcinogenic secondary compounds called aflatoxins. Aflatoxin contamination is routinely managed in commercial agriculture with strains of Aspergillus flavus that do not produce aflatoxins. These non-aflatoxin-producing strains competitively exclude aflatoxin producers and reshape fungal communities so that strains with the aflatoxin-producing phenotype are less frequent. This study evaluated the genetic variation within naturally occurring atoxigenic A. flavus strains from the endemic vegetative compatibility group (VCG) YV36. AF36 is a strain of VCG YV36 and was the first fungus used in agriculture for aflatoxin management. Genetic analyses based on mating-type loci, 21 microsatellite loci, and a single nucleotide polymorphism (SNP) in the aflC gene were applied to a set of 237 YV36 isolates collected from 1990 through 2005 from desert legumes and untreated fields and from fields previously treated with AF36 across the southern United States. One haplotype dominated across time and space. No recombination with strains belonging to VCGs other than YV36 was detected. All YV36 isolates carried the SNP in aflC that prevents aflatoxin biosynthesis and the mat1-2 idiomorph at the mating-type locus. These results suggest that VCG YV36 has a clonal population structure maintained across both time and space. These results demonstrate the genetic stability of atoxigenic strains belonging to a broadly distributed endemic VCG in both untreated populations and populations where the short-term frequency of VCG YV36 has increased due to applications of a strain used to competitively exclude aflatoxin producers. This work supports the hypothesis that strains of this VCG are not involved in routine genetic exchange with aflatoxin-producing strains.

INTRODUCTION

Aflatoxins are naturally occurring, highly toxic, carcinogenic polyketides produced by some filamentous fungi belonging to Aspergillus section Flavi (Ascomycota, Eurotiales). Aflatoxin contamination on both domestic and nondomesticated plants (e.g., maize, peanuts, cotton, and mesquite) occurs in most warm regions. The most frequently described causal agents of contamination are Aspergillus flavus Link and A. parasiticus Speare (1). Ingestion of aflatoxin-contaminated food and feed may result in disease, reduced immune function, and even death (2–5). For these reasons, the level of aflatoxin contamination of crops for human and domestic animal consumption is subject to stringent governmental regulation. In the United States, the limit for total aflatoxins in crops intended for human consumption is 20 ppb (6). Regulations can have severe economic repercussions (7), and economic losses due to aflatoxins may exceed $500 million annually in the United States alone (8). However, the effects on human health and the costs of lost markets are the most severe in developing countries (9–12).

Aspergillus flavus is a facultative and opportunistic fungal pathogen of both animals and plants with a worldwide distribution. The vegetative compatibility system of A. flavus is a self-recognition-versus-non-self-recognition system (13, 14) that delimits numerous vegetative compatibility groups (VCGs) within A. flavus populations. A single VCG may consist of few to many A. flavus strains. Strains of the Aspergillus flavus VCGs vary widely in their aflatoxin-producing potential, with many producing no aflatoxins (15–17). Competitive exclusion of aflatoxin-producing strains by indigenous A. flavus strains that do not produce aflatoxin is a biological control strategy that substantially reduces crop aflatoxin levels and has been highly successful in commercial agriculture for over a decade (18, 19). This type of biological control uses well-adapted, atoxigenic strains from endemic VCGs to alter the composition of A. flavus populations associated with developing crops so that the applied atoxigenic strain(s) dominates (17, 20). The atoxigenic strains are applied on a source of nutrients for the fungus, usually a grain, and reductions in contamination are proportional to the percentage of the applied atoxigenic Aspergillus strain(s) associated with the crop at harvest. Changes to the population composition induced by applications extend into environments beyond the treated field and persist to various extents for multiple years.

There are many naturally occurring atoxigenic strains of potential value for biological control, and several of these are active ingredients of biopesticides either already registered for use in various countries or in the process of development. Aspergillus flavus strain AF36 (ATCC 96045) is the active ingredient of the first and longest-registered atoxigenic strain-based biopesticide and is currently registered in the United States for use on cottonseed, maize, and pistachios (17, 20). The biopesticide A. flavus AF36 was first applied to commercial agricultural fields in 1996, and the acreage treated gradually increased from just 120 acres in the first year to over 250,000 acres per year (18, 21, 22). AF36 is a strain of VCG YV36, a VCG with a known natural range across the southern United States from California to Georgia and in northern Mexico near the Rio Grande River south of Texas (16, 21, 23–25). This broad, natural distribution of YV36 suggests that it has evolved with a wide variety of wild and domesticated host plants.

Interest in the population genetics of YV36 stems in part from concern that the deliberate release of AF36 (a strain of VCG YV36) will facilitate the evolution of aflatoxin-producing strains in the population that may be more fit, aggressive, or toxigenic than those already present. However, there has been no evidence of increased aflatoxin production in fields treated with AF36. Genes required for aflatoxin biosynthesis occur in a cluster consisting of over 25 genes and spanning approximately 70 kb (26). A single nucleotide polymorphism (SNP) introduces a stop codon in the polyketide synthase gene, aflC (pksA), and is responsible for the inability of the AF36 strain to produce aflatoxins (26, 27). This SNP also exists in several other VCGs, but the distribution among other strains of YV36 other than AF36 is unknown, and its stability in treated fields has not been examined. Molecular characterization of fungal biocontrol agents is desirable in order to monitor isolates postapplication, especially in the case of biological control agents that may persist for extended periods of time and are expected to contribute to long-term biocontrol strategies (28, 29).

The genetic variation in A. flavus may be assessed using a variety of methods. Genome-wide genetic variation in A. flavus populations may be assessed using 24 polymorphic microsatellite loci distributed across the eight chromosomes (30). Historically, A. flavus has been known only from an asexual stage. Recently, a sexual stage was produced following crosses of strains in different VCGs with alternate mating types (i.e., mat1-1, mat1-2) under laboratory conditions (31–33). Whether A. flavus AF36 has a sexual cycle in nature is presently unknown (34). In an earlier study (35) of strains from the three most common A. flavus VCGs on cotton, we found that each VCG was composed of a genetically distinct clonal lineage and that within each VCG the simple sequence repeat (SSR) loci were in linkage equilibrium. This equilibrium may be explained by parasexual recombination within the individual VCGs (35). High levels of genetic differentiation among strains of the three VCGs were also detected (35). Evaluation of the clonality of strains within VCGs and introgression from other VCGs may be accomplished by assessing the genetic variation among many strains that belong to a single VCG.

In the current study, the genetic variation among 237 isolates belonging to VCG YV36 collected from 24 locations from Arizona to Georgia was evaluated (see Tables 1 to 3). Collections were made between 1990 and 2005, and the collection sites included three fields in Texas that were previously treated with AF36. The 24 microsatellite loci (30) evaluated are polymorphic within the three most common VCGs recovered from cotton in Texas and Arizona (35). The mating-type allele (mat1-1, mat1-2) frequencies were determined in each population, and a pyrosequencing assay (36) was used to assess the frequencies of the SNP that produces a stop codon in the aflC gene.

TABLE 1.

State, year, location of sample collection, and sampled substrate for isolates used in this study

State	Yr(s)	General locationa	Substrate
Alabama	1991, 1992	Montgomery (U)	Cottonseed
Arkansas	1991	North Little Rock (V)	Cottonseed
Georgia	1991, 1992, 1993	Albany (S), Macon (T)	Cottonseed
Mississippi	1991, 1992	Armory (O), Greenwood (Q), Greenville (R), Port Gibson (P)	Cottonseed
Arizona	1990	Yuma Valley (E), Dome Valley (F), Gila Bend (G), Maricopa (H)	Cottonseed soil
	1991, 1992, 1993	Phoenix (W), Casa Grande (X)	Cottonseed
	1997, 1998, 1999	Wellton (C), Freman (B), Saguaro National Park (A), Organ Pipe Cactus National Monument (D)	Soil, dung, and wild plants
Texas	1991, 1992, 1993	Rio Grande Valley (I), Harlingen (I), Sweetwater (M), La Mesa (N)	Cottonseed
	1999, 2000, 2001	Upper Coast (K), Coastal Bend (J), Rio Grande Valley (I), Wintergarden (L)	Cottonseed
	2005	Rio Grande Valley (I)	Maize kernels, soil

^aLetters in parentheses correspond to the letters in Fig. 1. Fields treated with strain AF36 were also located in Coastal Bend (J) and the Rio Grande Valley (I) in Texas and were treated in 2003 and 2004 and in 2004, respectively.

TABLE 3.

Characteristics of 123 A. flavus VCG YV36 isolates resident in Texas fields after treatment with the Aspergillus flavus AF36 biocontrol strain and community composition of VCG YV36 pre- and posttreatment^a

Area	Yr		Posttreatment YV36 strains				% YV36b
	Area treated with YV36	YV36 isolated	N	H	H1	Unique haplotypes	Pre treatment	Posttreatment
Edroy	2003	2004	20	4	17	0	1.7	60*
Gregory	2003	2004	11	1	11	0	0.9	58*
	2003 and 2004	2005	44	4	39	2		76*
Hidalgo	2004	2005	48	3	46	0	0.0	25*

^aN, number of isolates; H, number of haplotypes; H1, number of YV36 isolates of haplotype H1. The number of YV36 isolates with unique haplotypes, the number of haplotypes found only in that location, and additional descriptive statistics were not estimated because the number of haplotypes was <5. The mat1-2 allele was found in all YV36 isolates. No functional aflC genes were detected, as all isolates carried the SNP at nt 591 that results in a stop codon at that location.

^bValues are averages from 4 to 12 replicates. Material was applied at the recommended rate of 10 pounds of wheat seed colonized by AF36 per acre. Posttreatment values are for fungi collected from soil samples taken 1 year after treatment from the same field locations as the pretreatment samples. *, posttreatment values differ significantly (P < 0.01) from the corresponding pretreatment value by the paired t test.

The objective of this study was to determine the amount of genetic variation among strains belonging to VCG YV36 by assessing loci known to vary in populations of A. flavus in the southern and western United States. We expected the amount of variation within these strains to be limited and the number of microsatellite haplotypes present in populations from locations treated with AF36 to reduce aflatoxin contamination to be small. The amount of variation in populations from untreated locations but locations where strains of VCG YV36 occur endemically is unknown. The significance of this study is that it shows the genetic stability of strains belonging to this indigenous VCG, whether they occur naturally or have increased in frequency as a result of commercial agriculture practice, and supports the hypothesis that members of this VCG are not involved in routine genetic exchange with strains of at least the most common VCGs that contain aflatoxin producers.

MATERIALS AND METHODS

Collection and identification of VCG YV36 isolates from biocontrol-free environmental samples.

The Aspergillus flavus VCG YV36 strains used in the current study were originally identified from previous studies that examined A. flavus VCGs from environmental samples (16, 24, 35) or are initially reported here and were from studies that are yet to be published (P. J. Cotty, unpublished data). Environmental samples were collected both from agricultural fields (soil, cottonseed, maize kernels) across several southern U.S. states not treated with biocontrol AF36 and from soil, dung, and wild plants (e.g., fruit and flowers from mesquite [Prosopis L. sp.], ironwood [Olneya tesota A. Gray.], acacia [Acacia Mill. sp.], and palo verde [Cercidium Tul. and Parkinsonia L. spp.]) from natural areas of the Sonoran Desert in Arizona. The geographic location where the substrate was collected, the year that the substrate was collected, and substrate origins are listed in Table 1 and shown in Fig. 1. Regardless of the year, samples were collected from multiple fields in any given area (16, 24, 35). A total of 114 VCG YV36 isolates were isolated from environmental samples from locations not previously treated with biocontrol strain AF36 (Table 2) and were identified as belonging to YV36 through standard protocols for nitrate-nonutilizing auxotroph complementation (15, 37).

FIG 1.

Map of the United States indicating regions where Aspergillus flavus VCG YV36 strains were isolated. Letters correspond to the general sampling locations listed in Table 1. Arizona cottonseed oil mills (W, X) pooled cottonseed from throughout the cotton-growing region in Arizona. The southwest region is represented by Arizona. The south central region is represented by southern Texas. The southeast region is represented by Alabama, Mississippi, Arkansas, and Georgia.

TABLE 2.

Descriptive statistics for 114 isolates of Aspergillus flavus YV36 from locations not previously treated with the biopesticide AF36^a

Location, host source	Yr	N	H	H1	Unique haplotypes	L	D	E	PA	Ar	PAr	LD
Southeast, cottonseed oil mills	1991	10	5	5	3	0.29	0.76	0.63	0	1.01	0.03	0 in 15
	1992	5	2	4	0	0.05	0.40	0.74	0	1.05	0.00
	1993	2	1	2	0				0
Southwest (Arizona)
Soil	1990	17	9	8	5	0.29	0.79	0.43	2	1.14	0.06	11 in 15
Cottonseed oil mills	1991	7	4	4	0	0.14	0.71	0.65	0	1.10	0.00
	1992	7	4	4	0	0.10	0.71	0.65	1	1.10	0.06
	1993	4	2	3	1				0
Arizona desert	1997	5	1	5	0	0	0.0	1.0	0	1.00	0.00
	1998	13	6	8	3	0.24	0.64	0.41	2	1.15	0.04	8 in 10
	1999	2	1	2	0				0
South central (Texas)
Cottonseed	1991	2	2	1	1				1
	1992	1	1	0	0				0
	1993	4	3	3	0				0
	1999	6	4	2	1	0.19	0.87	0.90	0	1.17	0.04
	2000	14	10	3	7	0.38	0.95	0.82	4	1.24	0.07	0 in 27
	2001	9	5	3	3	0.14	0.86	0.85	2	1.19	0.10	3 in 3
Maize kernels, soil	2005	6	3	4	1	0.14	0.60	0.67	0	1.12	0.04
Total		114		61

^aN, number of YV36 isolates isolated from environmental samples; H, number of haplotypes; H1, number of isolates of haplotype H1; Unique haplotypes, number of haplotypes found only in that location; L, proportion of polymorphic loci based on 21 loci; D, genotypic diversity for locations with 5 or more isolates, in which genotypic diversity is the probability that two randomly chosen individuals in the population will have different genotypes and which is estimated by (n/n − 1)(1 − Σp_i²), where n is the number of individuals from the population sampled and p_i is the frequency of the ith genotype (40); E, evenness for locations with 5 or more isolates, in which evenness is an indicator of how evenly genotypes are divided over the population (all genotypes have equal frequencies when E is 1 [39]); PA, number of private alleles across loci; Ar, mean allelic richness; PAr, mean private allelic richness expected for a sample size of five isolates using the rarefaction method in ADZE (version 1.0) (41); LD, locus pairs at a significant linkage disequilibrium (linkage disequilibrium between pairs of polymorphic loci in populations of five or more clone-corrected haplotypes was tested). Southeast United States, 1991, Arkansas, Alabama, Georgia, and Mississippi; southeast United States, 1992, Alabama, Georgia, and Mississippi; southeast United States, 1993, Alabama and Mississippi. The mat1-2 allele was found in all YV36 isolates. No functional aflC genes were detected, as all isolates carried the SNP at nt 591 that results in a stop codon at that location.

Collection and identification of VCG YV36 isolates from agricultural fields previously treated with the biocontrol agent AF36.

VCG YV36 isolates (n = 123) were initially isolated from soil collected from Texas cotton fields in 2004 and/or 2005 (Table 3). These fields had been treated with the biocontrol agent AF36 in 2003 and/or 2004 (Table 3) (Cotty, unpublished). Soil collection followed a standard laboratory protocol that was slightly modified from that of Cotty (16). Briefly, in each field a 30- to 40-m transect was sampled at 3- to 4-m intervals. At each point, four small scoops of soil (1 to 3 g each) were collected and combined with samples from other points. These pooled samples were homogenized, and isolates of A. flavus were collected by the dilution plate technique on modified Rose Bengal agar (38). Cottonseed from agricultural fields was collected and processed as previously described (35). VCG YV36 strains were distinguished from strains of other VCGs through the use of nitrate-nonutilizing auxotroph complementation as described by Bayman and Cotty (15) and as outlined by Grubisha and Cotty (35). Isolates were stored as plugs of sporulating cultures that were grown on 5/2 agar (5% V8 juice, 2% agar, pH 5.2) (37) and submerged in 5 ml of sterile distilled water in 15-ml vials at 8°C.

Biocontrol AF36 strain isolates.

In order to compare the genetic composition of strain AF36 used for biocontrol formulations to that of the strains of VCG YV36 collected posttreatment, isolates from two subcultures of strain AF36 from each of 7 years (2002 through 2008) and one subculture from 2001 were isolated from the formulations used to treat commercial fields and, along with the isolate from the originating mother culture (ATCC 96045), were genotyped. This resulted in an additional 16 isolates.

DNA isolation and microsatellite genotyping.

Erlenmeyer flasks (250 ml) containing 70 ml sterile potato-dextrose broth (Difco, Becton, Dickinson and Company, Sparks, MD) were inoculated with 100 μl of a conidial spore suspension from water vials and incubated on a rotary shaker that was shaken at 160 rpm and 32°C in the dark for 48 h. Mycelium was harvested onto Miracloth material by vacuum filtration and immediately placed into DNA isolation buffer (MP Biomedicals, Santa Ana, CA). DNA was isolated from the mycelial cultures using a FastDNA Spin kit (MP Biomedicals).

PCR using 24 microsatellite loci, fragment analysis, and genotyping were performed by previously published protocols (30). To verify the reproducibility of the results, approximately 10% of the total number of isolates was independently amplified by multiplex PCR and genotyped at least three times.

Microsatellite data analyses.

Descriptive statistics were estimated for each population, defined as sample location by state and year. Samples from Mississippi, Georgia, Alabama, and Arkansas were grouped together as the southeast population due to low sample numbers and geographic proximity. Identification of haplotypes (multilocus, haploid genotypes), allele frequencies, and evenness were estimated using the GenoDive program (version 2.0b11) (39). Evenness is a measure of how evenly haplotypes are distributed across populations. Genotypic diversity, the probability that two randomly chosen individuals in a population will have different genotypes, was estimated by use of the Multilocus program (version 1.3) (40). Allelic richness (the mean number of alleles across loci) and private allelic richness (the mean number of private alleles across loci) were estimated on the basis of a sample size of 5, corresponding to the smallest sample size, using the ADZE program (version 1.0) (41). Linkage disequilibrium (LD) for all pairs of polymorphic loci for populations with a clone-corrected sample size of ≥5 was estimated using Multilocus (version 1.3) (40), and significance was assessed following a Bonferroni correction (initial α = 0.05) (42).

To determine how genetic variance was partitioned among the three regions, i.e., the southeast (Mississippi, Alabama, Georgia, Arkansas), south central (Texas), and southwest (Arizona) regions, and among years within regions, analysis of molecular variance (AMOVA) (43) was conducted using the Arlequin program (version 3.5) (44). Significance was based on 10,100 permutations.

Genetic relationships among haplotypes were examined by computing a genetic distance matrix using the method of Bruvo et al. (45) in GenoDive, which takes into account microsatellite mutation events. The distance matrix was converted to a minimum spanning network (MSN) using the Minspnet program (46) for each location and year and for the entire set of haplotypes. The MSN was visualized using the Neato program (for networks by location and year) or the fdp and gvedit programs (all haplotypes) in Graphviz graph visualization software (versions 2.26.3 and 2.38, respectively) (47). The MSN for all haplotypes was edited for presentation in Inkscape graphics software (version 0.91) (48).

In order to further explore the genetic relationships among VCGs, the data generated in this study were combined with those from a previous study that examined the genetic diversity within and among the three most common A. flavus VCGs found on cotton in Texas and Arizona (35). Principal coordinate analysis (PCoA) was conducted using 21 loci and data from strains belonging to VCGs OD02, MR17, and CG136 (30, 35) and, from the current study, strains from VCG YV36 and the AF36 biocontrol strain in the GenAlEx package (version 6.5) (49, 50).

aflC SNP detection.

A previously designed pyrosequencing assay (36) was used to detect the SNP in the aflC gene that produces a premature stop codon and thus prevents aflatoxin production in the AF36 strain and strains of several other VCGs (27). This SNP occurs at nucleotide (nt) 591 (the position with respect to the translational start site) of the polyketide synthase gene (GenBank accession number AY501905 for strain AF36 with an A residue and GenBank accession number AF441403 for a VCG AF13 strain with a G residue [27]). The protocol of Das et al. (36) for pyrosequencing on a PyroMark Q24 pyrosequencer (Qiagen, Valencia, CA) was followed with minor modifications. The sensitivity of the PyroMark Q24 pyrosequencer permitted a reduction in the concentrations of DNA and primers. PCR amplification was carried out in 20-μl reaction mixtures consisting of AccuPower HotStart PCR premix (1× PCR buffer with 1.5 mM MgCl₂, 1.0 U HotStart Taq DNA polymerase, and 250 μM each deoxynucleoside triphosphate; Bioneer, Inc., Alameda, CA), 5 to 10 ng DNA, and 0.1 μM each primer (forward primer, 5′-TGT AGC GAT GAT CGA GGT GA-3′; reverse primer, 5′-TGT GGT CTG ACA GAA CAC TT-3′). The reverse primer was biotinylated at the 5′ end and purified using high-performance liquid chromatography (HPLC). The thermal cycler conditions were those of Das et al. (36), except that the annealing temperature was 56°C and 40 cycles were run. The resulting 97-bp PCR amplicon, visualized on a 1% agarose gel stained with SYBR Gold (Life Technologies, Grand Island, NY), formed a strong band with no visible unincorporated primers.

Pyrosequencing was conducted using a PyroMark Q24 pyrosequencer according to the manufacturer's instructions (36, 51) and the sequencing primer 5′-TTG GTC TAC CAT TGT TTG-3′. Pyrosequencing with the PyroMark Q24 pyrosequencer precisely quantifies the ratio of bases at a specific location without generating sequences in the traditional sense and should not be confused with 454 pyrosequencing. The pyrosequencing assays used in the current study (36) quantified the bases at the location known to harbor the SNP that prevents strain AF36 from producing aflatoxins (27). The following pyrosequencing controls were confirmed for the assay: (i) PCR with no DNA, (ii) PCR with the sequencing primer alone, and (iii) PCR with the template but without the sequencing primer. The aflC pyrosequencing assay was run on all 237 YV36 isolates from environmental samples and isolates from the 16 AF36 subcultures. Each set of 24 PCRs included an aflatoxin-producing strain from VCG EB11 that has a G residue and not an A residue at the SNP position and a negative PCR control.

Mating-type idiomorph frequencies.

In A. flavus, the mating-type locus (mat) has two idiomorphs, mat1-1 and mat1-2 (52, 53). The frequency of the mating-type idiomorphs present in the YV36 strains was estimated by multiplex PCR amplification of portions of mat1-1 and mat1-2 from all 237 YV36 strains and isolates from the 16 AF36 subcultures. Ramirez-Prado et al. (53) developed a multiplex PCR amplification protocol that can be used to identify both mating-type loci in A. flavus strains from a single PCR (GenBank accession numbers EU357934 and EU357936 for A. flavus mat1-1 and A. flavus mat1-2, respectively). The mat1-1 and mat1-2 loci can be distinguished by PCR product sizes of 390 and 270 bp (bp), respectively, when the PCR product is run on an agarose gel (52). Multiplex PCR amplification using primers M1F and M1R (mat1-1) and primers M2F and M2R (mat1-2) was performed as described by Ramirez-Prado et al. (53) with minor modifications. Here the multiplex PCR was carried out in 20-μl reaction mixtures, using AccuPower HotStart PCR premix (Bioneer, Inc.; see above), the concentration of each primer was reduced to 0.3 μM, and the number of thermal cycler runs was reduced to 30.

RESULTS

YV36 strains used in this study.

A total of 253 VCG YV36 isolates were used in this study; 237 were isolated from environmental samples, and 16 were from subcultures of strain AF36 used in biopesticide formulations. The total included (i) 114 VCG YV36 isolates from environmental samples (e.g., soil, plant material) from locations not treated with biocontrol strain AF36 (Table 2), (ii) 123 VCG YV36 isolates from cotton fields previously treated with biocontrol strain AF36 (Table 3), and (iii) isolates from the 16 subcultures of biocontrol strain AF36. Each year biocontrol strain AF36 is subcultured before it is used for biocontrol formulations and applied to agricultural fields. Two subcultures for each of 7 years, one subculture for 2001, and the original culture were used for a total of 16 AF36 subcultures. This gives a total of 237 isolates of VCG YV36 isolated from field samples over 15 years and isolates from an additional 16 subcultures of biocontrol strain AF36 (which was originally identified as being in VCG YV36) for a grand total of 253 isolates representing VCG YV36.

Genetic data analyses.

VCG YV36 strains were identified by their multilocus, haploid microsatellite genotypes (referred to as “haplotypes” from here on). These were determined from polymorphic loci that had unambiguous, clearly identifiable electropherogram peaks when genotyped. Locus AF48 produced stutter patterns that were difficult to score and was not included in the analyses. Of the remaining 23 loci, 13 were polymorphic and 10 were monomorphic. The AF18 and AF64 loci had more alleles (14 and 13, respectively) than the other 21 loci and were removed from the analyses to avoid introducing bias associated with hypervariable loci that may be under selection (54). A final set of 11 polymorphic microsatellite loci was used for all analyses except PCoA, as described above. Redundant haplotypes from each geographic environmental sample site (i.e., from each population) were not included in the analyses.

Allelic variation at the 11 polymorphic loci was low and ranged from one to seven alleles per locus, and not all loci in each population were polymorphic. The proportion of polymorphic loci in the populations by year and location ranged from 0.00 to 0.38 (Table 2). Average allelic richness based on a sample size of 5 was consistently low across the populations (mean allelic richness, 1.0 to 1.24; Table 2). Private allelic richness, based on a sample size of 5, was negligible (mean private allelic richness, 0.0 to 0.10; Table 2). Pairwise linkage disequilibrium (LD) between polymorphic loci was estimated in seven populations that had ≥5 clone-corrected haplotypes (Table 2). LD could be rejected in two populations, SE1991 and TX2000, out of the five that were tested for LD (Table 2).

A total of 38 haplotypes (haplotypes H1 to H38) were detected, and of these, 68% were private haplotypes found in only one location in 1 year (Tables 2 and 3; Fig. 2 and 3). One dominant haplotype, H1, was found in all years and at all locations except in Arkansas in 1991 (n = 2; part of the southeast region) and Texas in 1992 (n = 1; Table 2; Fig. 2). Haplotypes were generally genetically similar and frequently differed from each other by one microsatellite repeat at one or two loci, as shown by minimum spanning networks (Fig. 2 and 3). The AF36 strain used in biopesticide formulations from 2001 to 2008 (n = 15) and the original AF36 strain were haplotype H1 (Table 4). The majority of YV36 isolates from treated fields were also haplotype H1 (Tables 3 and 4). In fields previously treated with AF36, the haplotypes differed from haplotype H1 by one or two repeats, and eight out of nine allelic differences had a frequency of ≤0.05 (Table 4).

FIG 2.

MSNs for Aspergillus flavus VCG YV36 strains by geographic region, source, and year. (A) Arizona; (B) southeast region; (C) Texas. Circles represent haplotypes, and haplotypes occurring more than once are labeled. Circles with bold borders represent haplotypes found only in that location. The circle size reflects haplotype frequency, where the smallest circle is equal to one haplotype (e.g., haplotype H1 occurred once in Texas cottonseed oil mills in 1991), except for haplotype H1 in panel C for Texas soils from fields previously treated with AF36. In this case, the circle size was shrunk to fit in the figure. Branch lengths are proportional to the genetic distance calculated by the method of Bruvo et al. (45) and are equivalent across all MSNs. Bold branches indicate a genetic distance based on a difference of one repeat size at one locus. Multiple tied minimum spanning trees are shown as loops. Sampling locations that had only one haplotype were not included (Table 2).

FIG 3.

Minimum spanning network showing the relationships for all 38 A. flavus VCG YV36 haplotypes sampled over 15 years across the southern United States. Geometric shapes containing haplotype names indicate the number of the 237 YV36 isolates belonging to each haplotype. The large circle represents haplotype H1 (174 YV36 isolates). H1 was detected in all collections except those from Arkansas in 1991 (n = 2) and Texas in 1992 (n = 1). Squares contain haplotypes detected at only one location in only 1 year. For the remaining three shapes, haplotypes were detected multiple times at the same or different locations and from the same or different years: diamonds, haplotypes detected twice; double circles, haplotypes detected four times; septagons, haplotypes detected six times. In order to allow all haplotypes to be viewable, branch lengths were not drawn in proportion to the genetic distance calculated by the method of Bruvo et al. (45). Thick branches indicate haplotypes that differ by one microsatellite repeat unit at one locus. Thin branches represent either two, three, four, five, or six repeat differences at one locus. The dotted branch indicates four or five repeat differences at two loci.

TABLE 4.

Comparison of polymorphic microsatellite allele frequencies for AF36 biocontrol formulations and for VCG YV36 strains after application of AF36

Locusa	Allele	Microsatellite allele frequencyb
		Biocontrol formulations (n = 16)	Fields in Texas previously treated
			Edroy, 2004 (n = 20)	Gregory, 2004 (n = 11)	Hildago, 2005 (n = 48)	Gregory, 2005 (n = 44)
AF8	177	1.0	1.0	1.0	1.0	0.98
	183					0.02
AF10	285	1.0	1.0	1.0	1.0	0.98
	288					0.02
AF11	162	1.0	0.95	1.0	0.98	0.98
	165		0.05			0.02
	168				0.02
AF13	161	1.0	0.95	1.0	1.0	0.91
	164		0.05			0.09
AF43	385	1.0	0.95	1.0	0.98	1.0
	388		0.05		0.02

^aThe six loci not included here were polymorphic in VCG YV36 isolates at locations not treated with AF36 but monomorphic in YV36 strains posttreatment.

^bn, number of isolates of VCG YV36.

Analysis of molecular variance did not find significant genetic variation partitioned among the three regions, the southeast (Mississippi, Alabama, Georgia, Arkansas), south central (Texas), and southwest (Arizona), or among years within regions, regardless of whether or not data were clone corrected (Table 5). Over 98% of the genetic variance was partitioned within populations by year in both cases. Over 95% of the genetic variance was partitioned within years when regions were analyzed separately, using the F_ST analogue (43, 44) as follows: for the southeast, Φ_ST = −0.04 and P = 0.57; for the south central region, Φ_ST = 0.04 and P = 0.08; for the southwest region, Φ_ST = −0.05 and P = 0.97.

TABLE 5.

Results from analysis of molecular variance for Aspergillus flavus YV36 populations^a

Source of variation	df	Sum of squares	% of molecular variationb	P valuec	Φ statisticb
Among regions	2 (2)	1.93 (2.30)	1.52 (1.98)	0.23 (0.22)	ΦCT = 0.02 (0.02)
Among years within regions	13 (10)	8.10 (8.50)	0.10 (−4.91)	0.45 (0.93)	ΦSC = 0.00 (−0.05)
Within regions and years	101 (45)	62.44 (47.70)	98.38 (102.93)	0.27 (0.88)	ΦST = 0.02 (−0.03)

^aThe regions are southeast (Mississippi, Alabama, Georgia, Arkansas), south central (Texas), and southwest (Arizona). Data from the fields treated with YV36 were not included. Results based on clone-corrected data are in parentheses. df, number of degrees of freedom; Φ_CT and Φ_SC are F-statistic analogues (43, 44).

^bNegative estimates should be considered zero.

^cSignificance is based on 10,100 permutations.

Haplotype H1 was the most frequent and central haplotype in all minimum spanning networks by sample location and year (Fig. 2). Networks were generally simple and compact, reflecting similarity among the small number of haplotypes. Three of the nine networks contained one loop (Fig. 2) that may indicate either homoplasy or recombination (55). In each case, the three haplotypes in the loop differed by one microsatellite repeat at one or two loci between adjacent haplotypes.

In the MSN formed from all 38 haplotypes sampled across 15 years, haplotype H1 was also central (Fig. 3). Branch lengths in this network were relaxed so that Fig. 3 would be visually informative. Thus, branch lengths are not proportional to the genetic distance obtained by the method of Bruvo et al. (45). The majority of branches indicated a one microsatellite repeat size difference between adjacent haplotypes. The loops are most likely a result of homoplasy, since the majority of haplotypes occurred in only one location in 1 year and adjacent haplotypes may be from different years and/or distant locations. However, parasexual recombination may explain loops where the adjacent haplotypes were from the same year and locations were in close proximity.

Results from PCoA identified four distinct clusters of similar genotypes (Fig. 4) in which the first two axes explained 60% of the variation among isolates. These genotypes represented each of the four VCGs analyzed. Nested within VCG YV36 is the multilocus haplotype of the biocontrol AF36 mother culture.

FIG 4.

Results from principal coordinates analyses indicate four genetic groups which correspond to each VCG: OD02, MR17, CG136, and YV36. AF36 indicates the mother culture used to produce the biopesticide. The biological control agent AF36 is nested within the VCG from which it originated. The two coordinates accounted for 60% of the genetic variation.

aflC SNP detection.

All 237 YV36 isolates from environmental samples (as well as the isolates from the 16 biocontrol AF36 subcultures) had the aflC A-residue SNP, which is responsible for producing a premature stop codon in the aflC gene. All EB11 isolates had the G nucleotide, and the results for all PCR negative controls were negative.

Mating-type identification.

Use of PCR to amplify portions of mat1-1 (390 bp) and mat1-2 (270 bp) (53) revealed that all 237 YV36 isolates (as well as the isolates from the 16 AF36 subcultures) yielded a single PCR product of approximately 270 bp in length, consistent with the presence of mat1-2 and the absence of mat1-1.

DISCUSSION

Aspergillus flavus populations may contain significant proportions (∼10% up to ∼90%) (56–59) of isolates that do not produce aflatoxins. Strains of A. flavus VCG YV36 do not produce aflatoxins and are widely distributed in the warm latitudes of North America (27). Strain AF36, a member of VCG YV36, has been used as a biopesticide since 1996 and is registered with the U.S. EPA to limit aflatoxin contamination of cotton, corn, and pistachio (18). During the 15-year sampling period of this study, strains of VCG YV36 have maintained (i) the SNP in the aflC gene responsible for the atoxigenic phenotype, (ii) the same mating-type idiomorph, and (iii) low levels of genetic diversity at microsatellite loci with small but detectable levels of mutation and genetic drift. All three types of molecular markers suggest that VCG YV36 strains reproduce clonally in nature. These results also support the hypothesis of a lack of sexual recombination of YV36 strains with members of other VCGs.

Proteins involved in aflatoxin biosynthesis are encoded by over 25 contiguous genes contained in a single 66- to 70-kb cluster (60). AF36 has many SNP-derived nonsynonymous amino acid substitutions in the aflatoxin gene cluster that could impact aflatoxin production (27). However, one SNP in aflC inserts a stop codon that prevents translation of a functional aflatoxin-polyketide synthase and is sufficient to prevent the formation of both aflatoxins and all toxic aflatoxin precursors, including sterigmatocystin and versicolorin. The aflC SNP is the primary lesion to which the atoxigenicity of AF36 has been ascribed (27). The current work shows that the aflC SNP was fixed in YV36 populations across a broad geographic range over the 15 years in which samples were collected. This speaks both to the stability of the aflC SNP and to the lack of importance of aflatoxin biosynthesis to the YV36 life strategy. In order for this fixation to occur, the aflC SNP and the associated loss of aflatoxin-producing ability must be selectively neutral or beneficial. The success of AF36 as a biocontrol agent (61) and the lack of a benefit from aflatoxins during plant parasitism (37) also suggest adaptive neutrality in crop-associated niches.

Analysis of genetic differentiation, allelic diversity, and linkage disequilibrium among SSR loci previously revealed that in agroecosystems containing A. flavus populations, gene flow occurs within but not between each of the three most common Aspergillus flavus VCGs associated with cotton production in Texas and Arizona (35). There was sufficient diversity at microsatellite loci among members of these VCGs to assign strains to VCGs on the basis of their microsatellite haplotypes. The microsatellite haplotypes of the YV36 strains were also sufficiently distinct (Fig. 4) to differentiate YV36 strains from those in the previously studied VCGs (OD02, MR17, and CG136).

The maintenance of one mating-type idiomorph over the entire 15 years, the low levels of microsatellite locus allelic diversity, and the significant pairwise linkage disequilibrium in three of five populations tested are consistent with clonal reproduction by strains in YV36. Of the 24 polymorphic microsatellite loci tested in previous studies (30, 35), only 11 were polymorphic in the current study, and even these loci were strongly biased toward the presence of a single allele at each locus (Tables 2 and 3). One haplotype, haplotype H1, dominated both spatially and temporally in populations of VCG YV36 (Fig. 2). Meiotic recombination with an isolate from a VCG other than YV36 may result in F1 progeny belonging to VCGs different from those of either parent, if a sufficient number of segregating loci governing vegetative incompatibility (vic) are involved in the cross. However, strains with parental VCGs should reform, at least in rare cases, from subsequent meiotic recombination of strains within the derivative VCGs. During the process of recovering parental VCGs, introgression of additional alleles and increased diversity in microsatellite haplotypes would be expected. Likewise, if vic loci are not shuffled during a recombination event, the progeny retain the parental VCG membership but should result in increased allelic diversity. Because strains of the same VCG are in linkage equilibrium (36), over successive generations, the additional alleles would be distributed across the strains within the VCG. Diversity would thus be expected to increase as the number of meiotic partners and the number of successful meiotic events increase. The levels of diversity detected here do not support meiosis between members of VCG YV36 and members of other VCGs.

Changes in allele frequency and the occurrence of new alleles over time could be the result of mutation and genetic drift. Fluctuations in population size, alleles present in the population at low frequencies, migration from unsampled populations, and parasexual recombination may all contribute to genetic drift. The compact minimum spanning networks for both individual populations by year and all haplotypes identified from 15 years demonstrated that the haplotypes were closely related to each other and to the most frequent haplotype, haplotype H1 (Fig. 2 and 3).

Since the publication of work identifying a sexual state of A. flavus and the close relative A. parasiticus from laboratory crosses (31, 32), there has been speculation that sexual reproduction is an important portion of the A. flavus life cycle and that meiosis may influence the use of isolates belonging to YV36 as biological control agents for the prevention of aflatoxin contamination. The ability of some isolates of alternate mating types to undergo sexual reproduction may primarily be a remnant from an ancestral sexual cycle that has largely been lost in natural A. flavus populations (62). Kwon-Chung and Sugui (63) suggested that these laboratory sexual crosses may be considered demonstrations of a general lack of fertility rather than proof of the occurrence of an active sexual cycle in nature. This conclusion is based on (i) the lengthy incubation (6 to 9 months for A. parasiticus and 6 to 11 months for A. flavus) required for the production of the sexual stage (31, 33), (ii) the highly specific media and growth conditions required to induce a sexual state, and (iii) the extremely low numbers of stroma and ascospores produced (61). However, the results from the current study suggest that meiotic recombination does not occur in populations of VCG YV36 associated with cotton crops in regions where AF36 has been used to reduce aflatoxin contamination for more than a decade.

Biological control with atoxigenic strains alters the frequency of genotypes within local A. flavus gene pools (Table 3) through founder effects, competitive displacement, and multiyear residual influences on fungal community structures. Critics of the use of atoxigenic A. flavus strains to reduce aflatoxin-producing fungi in crop environments have argued that a recombination event between the AF36 strain and other endemic strains might result in more fit aflatoxin-producing progeny. However, AF36 frequencies decrease over time as other endemic A. flavus strains outcompete it. As a result, it is recommended that AF36 be applied every year in order to maintain effective biological control. Furthermore, there is little variation in virulence among A. flavus genotypes (64–66), and increased competitiveness on a host is not associated with increased virulence (66, 67). Thus, putative recombination events between AF36 and other endemic strains are unlikely to result in aflatoxin-producing strains with increased fitness. The existing data are consistent with this hypothesis, since after more than a decade of commercial use of AF36, more fit recombinants have not been observed in the field. In the current study, we detected no evidence of recombination between AF36 and strains belonging to other VCGs.

The VCGs of A. flavus examined to date have coexisted for thousands of years (35), frequently interacting during coinfection of hosts and coexistence in soils, and there have been numerous opportunities to exchange genetic material throughout this period. Application of strains from endemic, widely distributed VCGs, like strain AF36, does not fundamentally change the opportunity for rare recombination events, as members of these VCGs also are a part of the indigenous fungal community. The AF36 strain is used as a biopesticide in areas where it is already resident and has coevolved with both other VCGs in the A. flavus population and native host plants. Results from this study demonstrate that the variation of the genetic composition of strains of VCG YV36 across the wide geographic area where the VCG occurs naturally is limited. However, some VCGs of A. flavus have relatively limited distributions (Cotty, unpublished). The selection of a strain for biological control from a widely distributed VCG was deliberate, as it reduces the risk to nontarget hosts (68) and the risks of a nonnative introduced organism having unforeseen ecological results (68–72). As more biological control programs to reduce aflatoxin contamination are pursued globally, the strains used as biopesticides should be derived from local fungal populations and should not be based on atoxigenic strains that are not indigenous to the targeted region. Current efforts in Nigeria (73, 74) and Kenya, where atoxigenic strains of A. flavus are being developed for commercial applications to maize (17), adhere to the precept of identifying endemic fungi for use in aflatoxin control.