The Tomato Wilt Fungus Fusarium oxysporum f. sp. lycopersici shares Common Ancestors with Nonpathogenic F. oxysporum isolated from Wild Tomatoes in the Peruvian Andes

Fusarium oxysporum is an ascomycetous fungus that is well-known as a soilborne plant pathogen. In addition, a large population of nonpathogenic F. oxysporum (NPF) inhabits various environmental niches, including the phytosphere. To obtain an insight into the origin of plant pathogenic F. oxysporum, we focused on the tomato (Solanum lycopersicum) and its pathogenic F. oxysporum f. sp. lycopersici (FOL). We collected F. oxysporum from wild and transition Solanum spp. and modern cultivars of tomato in Chile, Ecuador, Peru, Mexico, Afghanistan, Italy, and Japan, evaluated the fungal isolates for pathogenicity, VCG, mating type, and distribution of SIX genes related to the pathogenicity of FOL, and constructed phylogenies based on ribosomal DNA intergenic spacer sequences. All F. oxysporum isolates sampled were genetically more diverse than FOL. They were not pathogenic to the tomato and did not carry SIX genes. Certain NPF isolates including those from wild Solanum spp. in Peru were grouped in FOL clades, whereas most of the NPF isolates were not. Our results suggested that the population of NPF isolates in FOL clades gave rise to FOL by gaining pathogenicity.

Fusarium oxysporum Schlecht. emend. Snyd. et Hans. is an ascomycetous fungus that inhabits various environments including the phytosphere, which includes both plant tissues and the rhizosphere. Most isolates from asymptomatic plants do not cause disease on any plants, and are referred to as nonpathogenic F. oxysporum (12).
On the other hand, plant pathogenic forms, formae speciales (f. spp.), are recognized in the species, and each form is defined by its strict host specificity (4,5). F. oxysporum f. sp. lycopersici Snyd. et Hans. (FOL) is a pathogenic form that causes soilborne vascular wilt disease in the tomato (Solanum lycopersicum L.). Moreover, each of the three FOL pathogenic races (1,2,5) has been defined based on the possession of different combinations of SIX (secreted in xylem) protein genes, SIX4, SIX3, and SIX1 (16,17,41), and determined by their specificities to particular tomato cultivars (2,13,53). These SIX genes are recognized to be pathogenic determinants and can be useful tools for race determination (18,30).
"When, where, and how did plant pathogenic F. oxysporum emerge?" This is a very fundamental, but difficult question to address. Several phylogenetic studies have examined other plant pathogenic fungi using isolates from the places of origin and domestication of plants, for example, rice blast fungus Pyricularia oryzae Cavara [synonym, Magnaporthe oryzae (Hebert) Barr], late blight pathogen Phytophthora infestans (Mont.) de Bary, wheat fungal leaf blotch pathogen Mycosphaerella graminicola (Fückel) Schrot, and corn smut fungus Ustilago maydis (DC.) Corda (8,14,33,48). To date, phylogenetic studies have also been extensively performed on F. oxysporum isolates (9, 12, 21-23, 29, 32, 34). For example, FOL is considered to be polyphyletic because it is composed of isolates involved in three clades (19,23), and the pathogen of Fusarium wilt of melon (f. sp. melonis) has also been shown to be polyphyletic (12), whereas the cabbage yellows fungus (f. sp. conglutinans) is composed of one cluster and appears to be monophyletic (22). Studies on pathogenic isolates are generally limited, and very little is known about the relationship between pathogenic and nonpathogenic isolates. Therefore, we focused on the coevolution of the tomato wilt pathogen and tomato.
The tomato (S. lycopersicum) is thought to have originated in South America, which is now occupied by Peru, Chile, Ecuador, and Bolivia. This region continues to sustain wild species of Solanum L. section Lycopersicon  (39).
A wild Solanum sp., possibly S. pimpinellifolium, spread prehistorically from South America to Central America (Mexico) in which the tomato was domesticated (20). S. lycopersicum var. cerasiforme, an apparent intermediate between wild and cultivated tomatoes (42), is currently found as a natively grown ("silvestre" in Spanish) tomato in some rural areas of Mexico. Traditional tomato cultivars, so-called "jitomate criollo" in Spanish, have been handed down by generations of peasants in mountain villages, and are considered the archetype of modern tomatoes due to their diverse morphologies (20). S. lycopersicum var. cerasiforme and jitomate criollo were designated transition tomatoes in this study. Tomatoes were transported to European countries, such as Italy and Spain, in which modern tomato breeding started, during the Spanish conquest in the 16th century (20,39,50).
In the present study, we 1) collected F. oxysporum isolates from tissues and the rhizosphere of asymptomatic Solanum biotypes: wild tomatoes in Chile, Ecuador, and Peru; transition tomatoes in Mexico; and modern tomatoes worldwide, 2) evaluated the pathogenicity of each isolate by an inoculation test using tomato tester cultivars, 3) evaluated the susceptibility of each Solanum biotype to FOL by the inoculation test, 4) determined the mating type and VCG of each isolate, 5) performed phylogenetic analyses based on sequences of the ribosomal DNA intergenic spacer (rDNA-IGS) region of the F. oxysporum isolates, together with FOL and other f. spp. collected worldwide, and 6) detected SIX genes in the F. oxysporum isolates collected. Based on the results obtained, we attempted to determine when, where, and how the plant pathogenic forms of F. oxysporum emerged.

Plant tissues and rhizosphere soil samples
We sampled the leaves, flowers, stems, fruits, roots, and rhizosphere soils of asymptomatic Solanum (sect. Lycopersicon) spp. in Chile, Peru, Ecuador, Mexico, Italy, Afghanistan, and Japan between 2002 and 2011 (Table 1). Here, rhizosphere soil refers to soil sampled from an area ca. 5 cm from the plant base and the surface at a depth of ca. 5 cm.

Isolation of F. oxysporum from plant tissues and rhizosphere soil
Fungal isolations were prepared within 10 d of collecting the Solanum tissues. Small pieces (ca. ≤ 9 mm 2 ) from individual tissue samples were cut and placed on Fusarium-selective media (25,35) and potato sucrose agar (PSA) medium in a Petri dish, and incubated at 28°C in the dark.
Fungal isolations from the rhizosphere were prepared by the soil-plate method (54) using Fusarium-selective media. Briefly, approximately 0.5 g of a soil sample was dispersed in 15 mL molten medium in a Petri dish and then incubated at 28°C under dark.
Fungal colonies that emerged after the 2-4-d incubation were transferred onto fresh medium and purified by repeated single hyphal tip isolation. Each established isolate was maintained on a PSA plate at 28°C, and isolates identified as F. oxysporum based on morphological characteristics (28) were subjected to further studies. All the isolates were stored in 25% glycerol solution at −150°C.

Inoculation test
The pathogenicity of each F. oxysporum isolate was evaluated using tomato tester cultivars. To prepare the inoculum, each isolate was cultured for 5 d on 3 mL potato dextrose broth (PDB; Becton and Dickinson, MA, USA) in a 15-mL screw cap test-tube at 25°C on a reciprocal shaker (Taitec, Saitama, Japan) at 200 strokes min −1 . Budding cells were collected by centrifugation (3000×g, 15 min) and adjusted to ≥1.0×10 7 cells mL −1 . FOL MAFF 305121 (race 1), JCM 12575 (race 2), and Chz1-A (race 3) were used as positive controls in this assay.
Three tomato standard tester cvs. Ponderosa (i i2 i3, susceptible to all FOL races; Takayama Seed, Kyoto, Japan), Momotaro (I i2 i3, resistant to race 1 and susceptible to races 2 and 3; Takii seeds, Kyoto, Japan), and Walter (I I2 i3, resistant to races 1 and 2 and susceptible to race 3; gift from the National Institute of Vegetable and Tea Science, Mie, Japan) were used (3). Two seeds were sown for each test in sterilized soil (andosol) in a plastic pot (7 cm in diameter) and were grown in a greenhouse at 28°C.
Prior to the inoculation, the roots of 2-3-week-old plants were injured by repeatedly inserting a plastic peg into the soil. The inoculum (2 mL pot −1 ) was poured on the soil surface and allowed to soak into the rhizosphere. After a month, the external symptoms of each plant were evaluated as follows: 0, no wilt or yellowing; 1, lower leaves yellowing; 2, lower and upper leaves yellowing; 3, lower leaves yellowing and wilting, and upper leaves yellowing; 4, all leaves wilting and yellowing, or dead.

Fungal DNA extraction
Genomic DNA (gDNA) was extracted from fungal mycelia following a protocol modified from the original method (45). Briefly, a small amount of mycelia on PSA medium (≤25 mm 2 ) was placed in 500 µL lysis buffer (50 mM EDTA, 200 mM NaCl, 1% n-lauroylsarcosine sodium salt, 200 mM Tris-HCl pH 8.0) in a microtube, incubated for 10 min at room temperature, centrifuged at 20,000×g for 5 min at 4°C after the addition of 150 µL of 3 M potassium acetate. The supernatant was then transferred to a fresh microtube. gDNA in the supernatant was concentrated by ethanol precipitation and resuspended in TE buffer (10 mM Tris-HCl pH 8.0, 1 mM EDTA).
ML phylogenies were estimated using RAxML implemented in raxmlGUI 1.0 (46). MrModeltest v2.3 (36) deter-  mined the appropriate substitution model as the HKY+G model from the model of the hierarchical likelihood ratio test (hLRT). Although the HKY+G model was not implemented in raxmlGUI, the HKY+G model was displaceable by the GTR model (A. Stamatakis, pers. comm.); therefore, the analysis was performed with the GTRGAMMA model and rapid bootstrap option (47) with 1,000 bootstrap replicates.
In the MP analysis using PAUP* 4.0b10 (49), searches of trees included 1,000 random additions, heuristic replicates with tree bisection, and reconnection (TBR) branch-swapping. One thousand bootstrap replicates were performed with the heuristic search option.
BI phylogenies were estimated using MrBayes 3.1.2 (43) based on the HKY+G model. In the BI analysis, the Markov Chain Monte Carlo (MCMC) iterations with four chains were started from a random tree topology and lasted 500,000,000 generations. When the average standard deviation of the split frequencies was below 0.01, the MCMC iterations were stopped automatically. Trees were saved at each 100-generation interval, and 12,500 trees were discarded as burn-in. Finally, the posterior probabilities of each branch were calculated.

Vegetative compatibility group (VCG) typing
VCG reflects genetic variations among fungal isolates (40). Four VCGs (0030+0032, 0031, 0033 and 0035) have been reported previously in FOL (6), and these have correlated with phylogeny (23,32). The following FOL tester isolates: OSU-451B (VCG 0031), MN-66 (VCG 0030+0032), and H-1-4 (VCG 0033) were used to determine the VCG of each isolate. The basis of the VCG test was as follows; by a selection on MMC (minimal agar medium with 1.5% chlorate), a mutation (at either nit1 or NitM) causing nitrate nonutilization was introduced into each collected isolate to be tested and into each of the three tester strains. The mutation in each tester was assessed using hypoxanthine medium (0.2 g L −1 of hypoxanthine plus minimal agar medium without NaNO 3 ; nit1 +, NitM −) and nitrite medium (0.5 g L −1 of NaNO 2 plus minimal agar medium without NaNO 3 ; nit1 +, NitM +). To assess VCGs, a part of the collected isolates was paired on MM (minimal medium) with nit-complementary testers; nit-complementary testers were paired with each other as positive controls. Vigorous growth on MM reflected heterokaryon formation, which indicating that the paired isolates belonged to the same VCG of the tester (7).

Sampling of Solanum spp. and isolation of fungi from plant tissue and rhizosphere soil
Among the wild tomatoes, S. chilense was sampled in Chile and Peru, S. habrochaites was sampled in Peru, S. pennellii was sampled in Peru, S. peruvianum was sampled in Chile and Peru, and S. pimpinellifolium was sampled in Peru and Ecuador. Transition tomatoes were sampled in Mexico. The Mexican transition tomatoes were morphologically diverse; the colors of mature fruits were red, orange, or yellow. In addition, jitomate criollo fruits had irregular multiloculated shapes and were heterogeneous in size (Fig.  S1i, j). Modern tomatoes cultivated in farmlands were sampled in Chile, Mexico, Italy, Afghanistan, and Japan. None of the plants exhibited wilt symptoms at the time of collection. The precise locations (latitude, longitude, and altitude) of each collection field and plant sample are presented in Table 1 and Fig. S1a-j. Approximately 2,500 fungal isolates were obtained from the plant and rhizosphere soil samples. Based on the morphological characteristics and nucleotide sequences of IGS regions, 433 of these isolates were identified as F. oxysporum; 42 were from plant tissues and 391 were from rhizosphere soils. F. oxysporum was not isolated from the tissues of S. chilense. A multitude of other fungi were also recovered from plant tissues and rhizosphere soils, e.g. mitosporic ascomycetes such as Fusarium spp., Trichoderma spp., Penicillium spp., Cladosporium spp., Alternaria spp., and Phoma spp., and zygomycetes such as Mucor spp.

F. oxysporum pathogenicity assay
None of the 433 F. oxysporum isolates, except for CE-391s, caused wilt disease when inoculated on the three tomato tester cultivars. We designated the F. oxysporum isolates that did not cause wilt on the tomato as NPF in this study (Table 1). CE4-391s was isolated from the rhizosphere soil of a modern tomato cultivar in a Chilean tomato farmland, and caused crown and root rot symptoms (27) on all three tester cultivars ( Table 3). The IGS sequence of CE4-391s was identical to that of F. oxysporum Schlecht. f. sp. radicis-lycopersici Jarvis et Shoem. (FORL) strain Saitamarly (Fig. 1, Table 3), a known crown and root rot pathogen of the tomato. These results, along with the finding that CE4-391s lacked SIX genes that are unique to FOL (52), led us to conclude that CE4-391s was neither NPF nor FOL, but rather FORL.

Phylogenetic analyses
Among the 432 NPFs identified, several isolates from the same sample and carrying identical rDNA-IGS sequences, were considered clonal, and one of them was used as their representative for phylogenetic studies. Therefore, phylogenetic trees were estimated using 233 NPFs (Table 1), together with 18 FOL isolates, 18 isolates of other formae speciales, and 5 NPFs isolated in previous studies (Table 3).
Maximum likelihood (ML), maximum parsimony (MP), and Bayesian inference (BI) methods were used to construct phylogenetic trees, and the ML tree was shown in Fig. 1. The topology of the ML tree was nearly identical to those of the MP and BI trees (data not shown). Each branch was statistically estimated by a bootstrap (BS) test in ML and MP analyses, and posterior probability (PP) in BI analysis. The parameter of the ML tree (-ln L = 3419.861497) was as follows; base frequencies = (A = 0.159040, C = 0.175477, G = 0.363070, T = 0.302413). MP analysis yielded 1,000 equally parsimonious trees (tree length = 413 steps; consistency index = 0.741; retention index = 0.929; rescaled consistency index = 0.688; homoplasy index = 0.259).
These 16 NPF isolates in the FOL clades were obtained from Peruvian wild species of tomatoes, Mexican transitional tomatoes and modern tomato cultivars worldwide, while none of the NPF isolates were obtained from wild species in Chile and Ecuador.
We tested vegetative compatibility between FOL and the subset of 16 NPF isolates from our fungal collection that fell into the three FOL clades (Fig. 1). Although each of the NPFs was paired with the VCG 0031, 0030+0032, and 0033 tester strains, none were compatible.
Tests for SIX genes PCR analyses indicated that the 16 NPF isolates that grouped into the FOL clades did not carry SIX1, SIX3, or SIX4. These genes were readily amplified from the authentic FOL strains.

Solanum spp. susceptibility assay
The Mexican transition tomatoes, S. lycopersicum var. cerasiforme and S. lycopersicum (jitomate criollo), showed an almost equivalent degree of susceptibility to that of cv. Ponderosa (a modern tomato cultivar carrying no resistance) to FOL races 1-3. Among the wild species of tomatoes, S. chilense and S. peruvianum showed resistance to FOL races 1-3 (Table 4). On the other hand, the resistance of all S. pimpinellifolium collections from Ecuador was less than that of the above two wild species (Table 4), although they presented no external symptoms.

Discussion
It has generally been assumed that a plant pathogen emerged from a nonpathogenic strain during the domestication and breeding of its host plants. Several previous studies (8,14,33,48) suggested a relationship between the origin of pathogens and domestication of host plants. However, such studies have not yet been performed on Fusarium oxysporum.
In the present study, we isolated F. oxysporum from the tissues and rhizosphere soils of asymptomatic Solanum spp. sect. Lycopersicon and found that all the F. oxysporum  isolates recovered were nonpathogenic F. oxysporum (NPF), except for one isolate (CE4-391s) from a modern tomato field in Chile, which was considered to be FORL. This result was consistent with the findings of previous studies (12), which showed that NPFs were frequently isolated from plants and, therefore, are part of the normal field population. In our phylogeny, FOL isolates were distributed in any of the three clades (A1, A2, and A3; Fig. 1), suggesting that FOL has at least three origins (polyphyletic), which is consistent with the findings of previous studies (23,37,38). We also found that 16 NPFs were grouped in the three FOL clades (3 for the A2 clade, 8 for the A1 clade and 5 for the A3 clade), and that they are more closely related to FOL (99.8 to 100% nucleotide identity of rDNA-IGS) than to other NPFs and isolates of other forms (82.0 to 99.5% nucleotide identity). These 16 NPFs were isolated from Peruvian wild species, transition tomatoes, or modern cultivars. This result suggests that these NPFs share common ancestors with FOL and that the possible origin of FOL existed with the wild Solanum spp. in the Andes, possibly in Peru. How did FOL acquire pathogenicity to the tomato? Kistler proposed a horizontal gene transfer (HGT) to explain the evolution of pathogenicity in F. oxysporum (24). HGT or horizontal chromosomal transfer (HCT) has been reported in other plant pathogenic fungi, such as Nectria haematococca (15), Cochliobolus heterostrophus (44), and Alternaria alternata (1). A small (ca. 2.0 Mb) chromosome, designated chromosome 14 (Ch14), was recently detected on FOL (31), and was found to carry effector genes, such as SIX1, SIX3, SIX4 and other genes presumably related to pathogenicity (19,51). FOL isolates belonging to each distinct FOL clade in the phylogeny shared genes (Fig. 1). These results suggest that FOL had a polyphyletic origin, and that the original NPF may have acquired the small chromosome involved in pathogenicity and/or host specificity of FOL by HCT.

S. chilense
The detailed mechanisms underlying HCT and HGT in fungi are unclear (51). However, Ma and co-workers demonstrated detected HCT in F. oxysporum in vitro (31). They co-incubated the pathogenic FOL strain Fol007 (possessing Ch14) with the NPF strain Fo-47 (lacking Ch14), and recovered a Fo-47 bearing Ch14 that presented pathogenicity to the tomato. Ch14 could only be transferred to strain Fo-47, but not to F. oxysporum f. sp. melonis or F. oxysporum f. sp. cubense, by the same manner. This experiment suggested that HGT or HCT may not occur randomly among strains, but rather depends on particular strains or environmental conditions. To test this foregoing hypothesis, it will be necessary to demonstrate that the 16 NPF isolates in the FOL clades (Fig.  1) have a greater capacity to acquire the small chromosome carrying effector genes than other more distantly related isolates.
The results of this study suggest that the nonpathogenic ancestors of FOL were in Peru, and a part of their progenitors gained effector genes or the small chromosome later, which resulted in the emergence of FOL. The origin(s) of the effector genes carried by the small chromosome are of interest. Mexican transitional tomatoes and modern cultivars are less resistant to FOL than wild species (Table 4); therefore, clear damage by FOL may have appeared during/after tomato domestication in Mexico.
Our study represents an initial step in an investigation to discover the origin of FOL. We are now interested in examining the origin of the pathogenicity determinants/Ch14 in FOL (31). Studies on the distribution of resistance genes (I-I3) among tomatoes, Solanum section Lycopersicon, are also warranted. Our goal is to advance our understanding on the molecular mechanisms underlying host-parasite co-evolution.