Genetic variation and population structure of Fusarium oxysporum f.sp. vasinfectum in Australia




Genetic variation among 348 isolates of Fusarium oxysporum f.sp. vasinfectum (Fov) collected from diseased cotton plants in 31 fields in six cotton-growing regions in New South Wales and Queensland in 2002 and 2004 was analysed using amplified fragment length polymorphisms (AFLPs). Twenty-eight haplotypes were identified based on 146 polymorphic bands generated with four EcoRI and MseI and four HindIII and MseI primer combinations. The haplotypes separated into two distinct groups (37% similarity), with 21 in group I and seven in group II. The two unique vegetative compatibility groups of Fov known to occur in Australia (VCG 01111 and VCG 01112) were correlated to the two AFLP groups, with both VCG 01111 reference isolates being included in group I and both VCG 01112 reference isolates in group II. Group I was widespread, occurring in all regions sampled and all but one of the fields, while group II was limited to three fields in the Boggabilla region. Group I was further divided into two subgroups. The two haplotypes in subgroup I-B (I-20 and I-21) may represent the emergence of a new form of Fov based on their marked genetic discrimination from the subgroup I-A haplotypes. No spatial population differentiation was discernible at the national level, as only 3·9% of total genetic variation was attributed to differences among regions (P = 0·4868). When each region was analysed separately, clear differentiation was found in the Boggabilla region, with 86·3% of total genetic variation resulting from differences among fields (P < 0·0001).


Fusarium oxysporum f.sp. vasinfectum (Fov) causes fusarium wilt in all the major cotton (Gossypium hirsutum) growing countries of the world (Hillocks, 1992). This pathogen is soilborne and characterized by a parasitic phase within the vascular tissue of cotton plants and a saprophytic phase in the soil or plant residue after the host's death. It can be spread in soil, seed, plant material, etc. (Jeffers et al., 1984; Hillocks & Kibani, 2002). Once introduced, Fov can survive for long periods in fields, even in the absence of cotton (Smith & Snyder, 1975).

The use of resistant cotton cultivars is the most effective measure to control the disease (Hillocks, 1992). However, the success of resistance breeding programmes depends largely on a better understanding of the genetic variation and population structure of Fov. Furthermore, understanding how the distribution of Fov genotypes is changing over time will allow evaluation of whether current phytosanitary measures are effective or whether more stringent internal quarantine protocols should be applied to slow the spread of the pathogen. In other words, if distinct genotypes remain regionally restricted, that is evidence that current sanitary measures are working; but if genotypes unique to one region begin to appear elsewhere, then this will be evidence that current measures are not sufficient. In addition, more detailed knowledge about population structure will provide clues to the evolutionary potential of the pathogen (McDonald & Linde, 2002).

Global genetic variation of Fov has been intensively studied. Eight geographically distributed races were identified based on pathogenicity on differential hosts (Chen et al., 1985; Hillocks, 1992). Twelve vegetative compatibility groups (VCGs) were characterized, each representing a clonal lineage within the races (Fernandez et al., 1994; Bentley et al., 2000). Variation was also observed for aesculin utilization patterns (Rutherford et al., 1993) and fatty acid profiles (Hering et al., 1999). In addition, Assigbetse et al. (1994) described three groups revealed using random amplified polymorphic DNA (RAPD) and Fernandez et al. (1994) reported four rDNA and seven mtDNA haplotypes identified using restriction fragment length polymorphisms (RFLPs). However, all this work assesses genetic variation on a worldwide scale, with less being known of variation in Fov on regional or national scales, although such information is considerably more important to local disease management (Abd-Elsalam et al., 2004; Abo et al., 2005).

Sources and patterns of genetic variation in Fov may differ between global and local scales. Polyphyletic evolutionary origins (Skovgaard et al., 2001) and possible parasexual recombination (Molnar et al., 1990) are important to global variation, but may have less impact on local variation than factors like mutation, gene flow, genetic drift and selection. The lack of knowledge about local genetic variation may well reflect the fact that Fov is a haploid asexual pathogen and hence most of the markers effective in studies of a global collection of isolates are of limited value in measuring the genetic variation of local populations. Indeed, RAPDs were not successful in distinguishing among Californian isolates of Fov (Smith et al., 2001). Similarly, no variation was found when a number of isolates of F. oxysporum f.sp. dianthi causing fusarium wilt of carnation in Argentina were studied using VCG typing and rDNA intergenic spacer (IGS) sequences (Lori et al., 2004).

In Australia, fusarium wilt of cotton was first diagnosed in the Brookstead and Cecil Plains regions of Queensland in 1993 and the Boggabilla region of New South Wales in 1994 (Kochman, 1995). It has subsequently spread to most of the major cotton-growing regions of Australia and causes substantial economic losses (Kochman et al., 2002). The current assumption is that the occurrence of this disease outside the original sites of diagnosis is attributable to the spread of Fov by water movement within catchments and human-mediated dispersal (e.g. movement of dirty equipment). Stringent containment measures were quickly adopted, but the incidence of this disease continues to increase. This probably reflects the fact that the pathogen was spread prior to the institution of containment measures and the populations of Fov are only now reaching sufficient levels to cause notable disease.

Knowledge about population genetics of Fov in Australia is limited. Previous studies showed that the Australian isolates, which have similar levels of aggressiveness against cotton, behaved similarly to race 6 on differential hosts (Davis et al., 1996). They belong to two VCGs (01111 and 01112) that are both vegetatively incompatible with other forms of Fov found elsewhere in the world. In initial studies, representatives of the two VCGs were shown to be genetically distinct using DNA amplification fingerprints, but no further variation was detected within each VCG (Bentley et al., 2000). More recent work demonstrated that Fov in Australia has arisen indigenously, as it is genetically related to a lineage of indigenous F. oxysporum found in both the rhizosphere soil of wild native cottons (Gossypium spp.) and uncultivated soil from cotton-growing regions (BW, unpublished data). However, little is known about the degree and distribution of genetic variation within and among populations of Fov in Australia.

Amplified fragment length polymorphisms (AFLPs) are effective and powerful markers for determining intraspecific genetic variation in fungi. They have the capacity to generate a large number of markers that span the whole genome without prior knowledge of the genetic background (Majer et al., 1996). AFLP markers have been used to detect genetic variation among isolates of F. oxysporum responsible for various diseases (Baayen et al., 2000; Bao et al., 2002; Sivaramakrishnan et al., 2002). There appear to have been only a few instances where AFLPs were used to examine genetic relationships among isolates of Fov (Abd-Elsalam et al., 2004). However, there are no reports using AFLPs or other markers to assess population-level variation of Fov in Australia.

The history of fusarium wilt of cotton in Australia is short, and therefore the Australian situation provides an ideal system to chart the evolutionary processes operating on a newly emergent pathogen as it spreads locally and regionally. The objective of the current study was to generate a picture of the genetic variation and population structure of Fov across the major cotton-growing regions in Australia. This will provide information relevant to attempts to control the disease through breeding of resistant cultivars and development of improved phytosanitary strategies.

Materials and methods

Fungal isolates

A total of 348 isolates of Fov were recovered from cotton plants in 31 fields across the Warren, Moree, Boggabilla and Bourke regions of New South Wales and the Cecil Plains and St George regions of Queensland in 2002 and 2004 (Fig. 1; Table 1). The average distance between fields within regions was 20 km (ranging from 1 to 67 km), with an average of 200 km (range 57–361 km) between regions. Ten plants with symptoms, separated by a minimum distance of 20 m, were sampled from all but one field (only eight plants were sampled from field 21 in 2002). Stem sections (5–10 cm) were cut from the main shoot of plants approximately 10 cm above the soil surface. The stem segments were surface-disinfected in 0·5% sodium hypochlorite for 5 min and peeled under aseptic conditions. Small pieces of discoloured vascular tissue were placed on Peptone PCNB agar plates (Burgess et al., 1994) and incubated at 25°C for 1 week, after which a single isolate was recovered from each sample by subculturing fungal hyphae growing out of the tissue pieces on small slants of 10% potato dextrose agar (Difco). The slants were incubated at 25°C for 1 week and conidia were harvested by adding 1·5 mL of sterile 15% glycerol to each tube and pipetting the liquid several times. The resulting conidial suspensions were stored in 2·0 mL cryogenic vials at −80°C until use.

Figure 1.

Map of a section of southeastern Australia showing the locations of the sampled fields (x) and geographic locations of the six regions sampled in this study. The St George samples were supplied without precise locality information, and are not indicated.

Table 1.  Provenance information for the Fusarium oxysporum f.sp. vasinfectum isolates analysed
RegionField codeFarm (field)Latitude (S)Longitude (E)Year of collectionNumber of isolates
Moree03Norwood (field 12)29°24′15′149°45′02′200210
04Norwood (field 25)29°23′28′149°44′40′200210
05Rivergum (field 41)29°25′15′149°55′37′200210
06Red Mill (field 6)29°23′51′149°59′51′200210
07Red Mill29°24′07′149°57′27′200410
Boggabilla09Korolea (field 7)28°38′07′150°16′14′200210
09Korolea (field 7)28°38′19′150°16′10′200410
10Korolea (field 5)28°37′41′150°16′17′200210
10Korolea (field 5)28°37′31′150°16′16′200410
11Korolea (field 2)28°36′54′150°16′20′200210
12Morella (field E2)28°37′25′150°17′41′200210
12Morella (field E2)28°37′21′150°17′47′200410
13Mullala (field 4)28°45′44′150°26′50′200210
14Mullala (field 2)28°44′25′150°27′10′200210
15Carbucky (field 17 N)28°37′46′149°58′05′200210
16Carbucky (field 18)28°38′18′149°57′47′200210
17Carbucky (field 22)28°39′25′149°58′31′200210
Bourke18Janbeth (field 2)30°03′59′145°47′07′200210
19Darling farms (field 63)30°03′30′145°49′59′200210
20Rumleigh (field 4)29°53′36′146°52′36′200210
Cecil Plains21Keeley's farm27°41′24′151°22′22′2002 8
21Keeley's farm27°41′11′151°22′18′200410
23Kooyong (field K2)27°15′59′151°09′28′200210
24Riverlae (field 6 N)27°17′37′151°10′54′200210
25Elsen's farm27°46′32′151°25′51′200210
26Rainbow Valley (circle 1)27°22′44′151°37′33′200410
St George29Bubbymaur (field 4)Not recordedNot recorded200210
30Aspen (field 1)Not recordedNot recorded200210
31Redgen (field 1)Not recordedNot recorded200210

Two isolates each of VCG 01111 (24500 and 24595) and VCG 01112 (24492 and B/96/02), provided by Dr Natalie Moore and Mr Wayne O’Neil (Queensland Department of Primary Industries, Indooroopilly, Australia), were included as references on all AFLP gels to maintain consistency of gel scorings in this study. They are available upon request.

DNA extraction

Isolates were grown in 12 mL of 80% potato dextrose broth (Difco) in 15 mL sterile test tubes at 25°C for 3 days. Mycelia were harvested by centrifuging cultures at 2800 g for 15 min, decanting the liquid, and transferring the pellet onto Whatman no.1 filter paper to remove extra water. Genomic DNA was extracted from lyophilized mycelia using DNeasy Plant kits (Qiagen) according to the manufacturer's instructions. DNA concentrations were determined using a GeneQuant II spectrophotometer (Pharmacia Biotech) and adjusted to 50 ng µL−1.

AFLP genotyping

Isolates were genotyped using a modification of the procedures described by Vos et al. (1995) and eight selective primer combinations (Table 2). DNA (250 ng) was codigested with either EcoRI and MseI or HindIII and MseI at 37°C for 2 h and oligo-adapters were ligated to DNA fragments at 37°C for 3 h in 40 µL of digestion-ligation buffer. Preselective amplification was performed with 5 µL of digestion-ligation reaction mixture in 50 µL of PCR buffer containing primers with no selective nucleotides (20 cycles of 30 s at 94°C, 60 s at 56°C and 60 s at 72°C). Selective amplification was performed with 5 µL of diluted (1:30) preselective amplification reaction mixture in 20 µL of PCR buffer containing a 33P-labelled EcoRI or HindIII primer with three selective nucleotides and a MseI primer with one selective nucleotide (one cycle at 94°C for 30 s, 65°C for 30 s and 72°C for 60 s; 12 cycles with a 0·7°C decrease in annealing temperature in each cycle, from 65 to 57·3°C; 23 cycles at 94°C for 30 s, 56°C for 30 s and 72°C for 60 s). Products were separated on 6% polyacrylamide gels electrophoresed at 50 W for 2·5 h flanked by a 30–330 bp AFLP DNA ladder. Autoradiographs were obtained by exposing Kodak BioMax MR films to dried gels.

Table 2.  Primer combinations used for amplified fragment length polymorphism (AFLP) analysis of Fusarium oxysporum f.sp. vasinfectum, the number of amplified bands of molecular size ranging from 80 to 800 bp, the number and percentage of polymorphic bands, and the number of VCG-specific bands
PrimersAmplified bandsPolymorphic bandsPolymorphic bands (%)VCG-specific bands
EcoRI-AGG/MseI-A 66 2639·415
EcoRI-AGC/MseI-A 65 1929·215
EcoRI-ACC/MseI-A 75 2026·710
EcoRI-AAC/MseI-A 73 1621·9 7
Sub-total279 8129·047
HindIII-AGG/MseI-A 48 1122·9 4
HindIII-AGC/MseI-A 64 1320·3 7
HindIII-ACC/MseI-A 86 2731·416
HindIII-ACG/MseI-A 54 1425·9 9
Sub-total252 6525·836

Data analyses

Amplified fragment length polymorphism bands ranging from 80 to approximately 800 bp were scored as biallelic loci (present or absent). Fragments between 330 and 800 bp were scored with the aid of the four reference isolates included on each gel to ensure that all bands ascribed to a locus were homologous. Only bands that were consistently spaced from monomorphic bands were scored.

All polymorphic bands, i.e. those that were absent in at least one isolate, were included in the analysis. Genetic relationships among isolates were determined using ntsyspc 2·02j (Exeter Software). The minimum number of haplotypes was determined following Saleh et al. (2003). A representative of each haplotype was used to create a Dice similarity matrix, which was used to generate a upgma dendrogram of haplotypes. Bootstrap values (10 000 replicates) for each branch were calculated using winboot (International Rice Research Institute).

Population structure and differentiation were estimated using both total data and clone-corrected data (each haplotype was counted only once in a field). Total genetic variation was partitioned among regions, among fields within regions, and among haplotypes within fields, by performing a hierarchical analysis of molecular variance (amova) using arlequin 2·000 (Schneider et al., 2000). Gene diversity in the total population (Ht) and within subpopulations (Hs) (Nei, 1987), coefficient of genetic differentiation [Gst = (Ht − Hs)/Ht] (Nei, 1987) and the amount of gene flow among populations (subpopulations) [Nm = 0·5 × (1 − Gst)/Gst] (Slatkin, 1987) were calculated using popgene 1·31 (Yeh et al., 1999). In addition, for regions where significant differentiation occurred, pairwise differences (Nei, 1987) among subpopulations were determined and exact tests (Raymond & Rousset, 1995) were conducted.

Linkage disequilibrium was estimated using popgene. To avoid repeated scoring of the same DNA fragments, only data produced with one primer combination (EcoRI-AGG/MseI-A) were used in this analysis (Koenig et al., 1997).


AFLP genotyping

A total of 531 unambiguous bands were generated with the eight selective primer combinations, of which 146 (27·5%) were polymorphic as a result of their absence from at least one isolate (Table 2). Thirteen alleles occurred only once: six bands that were absent in only one isolate and seven bands that were absent in all but one isolate. These rare alleles were validated in replicated gels using independent DNA extractions and PCR reactions. The number of polymorphic bands ranged from 11 to 27 per primer combination. Overall, the level of polymorphism revealed by the EcoRI and MseI primer combinations (29·0%) was marginally greater than that revealed by the HindIII and MseI primer combinations (25·8%).

The reference isolates were unambiguously distinguished by AFLP markers, with 83 (56·8%) of the polymorphic bands discriminating between the two VCGs. The number of VCG-specific bands varied from four to 16 among the eight primer combinations (Table 2).

Genetic variation

Twenty-eight genetically distinct haplotypes were identified among the 348 isolates and they separated into two distinct groups (I and II, similarity 37%). Group I comprised 21 haplotypes with a minimum similarity of 80% (designated I-01 to I-21; Fig. 2). Group I was further divided into two well-supported subgroups (I-A and I-B). Subgroup I-A contained 19 haplotypes from all six regions. Subgroup I-B contained two haplotypes (three isolates) which occurred in the Warren, Moree and Bourke regions of New South Wales (Table 3). Group II comprised seven haplotypes with a minimum similarity of 97% (designated II-01 to II-07; Fig. 2) and was found only in the Boggabilla region (Table 3). The two VCG 01111 reference isolates nested in group I and the two VCG 01112 reference isolates nested in group II (Fig. 2).

Figure 2.

Dendrogram based on the Dice coefficient constructed by upgma cluster analysis of 146 polymorphic bands from 28 haplotype representatives of Fusarium oxysporum f.sp. vasinfectum generated with eight amplified fragment length polymorphism (AFLP) primer combinations. Bootstrap values (> 50%) from analysis of 10 000 replicates are shown above nodes. Haplotype codes and clusters of haplotypes (main groups and subgroups) are also indicated. Ref11 and Ref12 are two reference isolates of the two vegetative compatibility groups, VCG 01111 and 01112, known to occur in Australia.

Table 3.  Distribution of haplotypes of Fusarium oxysporum f.sp. vasinfectum across six cotton-growing regions in Australia
RegionSample sizeNo. of haplotypesSubgroup I-ASubgroup I-BGroup II
Warren 20 3 181                 1        
Moree 60 6 53  11   1  3        1        
Boggabilla12015 71110  1    1 1  51    16136111
Bourke 30 2 29                   1       
Cecil Plains 8811 70  13 31  121  4  11         
St George 30 4 27   1   1    1              
All regions3482826821241411116111451112116136111

Haplotype structure and distribution

The number of haplotypes found in different regions varied, with 15 in the Boggabilla region, 11 in the Cecil Plains region and two to six in the remaining four regions (Table 3; Fig. 3). The number of isolates per haplotype ranged from one to 268 (Table 3). Haplotype I-01 was the most prevalent genotype, represented by 268 isolates (77·0%) and occurring in all regions and all but one of the sampled fields (field 12, Boggabilla). Eight haplotypes (I-02, I-03, I-04, I-06, I−11, I-20, II-01 and II-04), represented by a total of 52 isolates (14·9%; two to 16 isolates per haplotype) were recovered from at least two fields. Three haplotypes (I-15, I-16 and II-03), represented by a total of 12 isolates (3·4%; three to five isolates per haplotype), were found in only one field. The remaining 16 haplotypes were rare, each being represented by a single isolate only (Table 3).

Figure 3.

Diagrammatic representation of the distribution of the 28 haplotypes (I-01 to I-21 and II-01 to II-07) of Fusarium oxysporum f.sp. vasinfectum among the six cotton-growing regions. I-A and I-B, the two subgroups in group I; II, group II. Haplotypes in the shaded area (only I-01) indicate they occur in all six regions; those in circles indicate they are unique to the particular regions; and those in connecting bars indicate they occur in all relevant regions.

Isolates from field 12 in the Boggabilla region were the most diverse, as they belonged to six haplotypes, followed by those from field 09 (Boggabilla), with five haplotypes, and fields 11 (Boggabilla) and 24 (Cecil Plains), with four haplotypes each (Table 3). In contrast, isolates from nine fields (03, 07, 13, 14, 17, 19, 20, 23 and 29) were homogeneous as they all belonged to haplotype I-01. The number of haplotypes in other fields ranged from two to three (Table 3). Within group I, besides haplotype I-01, two haplotypes (I-03 and I-11) were also widespread, being detected in the Moree and Boggabilla regions of New South Wales and the Cecil Plains region of Queensland. However, within group II, no haplotypes were shared by different fields except haplotypes II-01 and II-04, which were found in three and two fields, respectively (Table 3).

Population structure

When total genetic variation of all polymorphic loci (clone-corrected data produced congruent results, data not presented) was partitioned by amova, 79·8% of the variation was attributed to differences within regions (P < 0·0001), with only 3·9% of the variation resulting from differences among regions (P = 0·4868) (Table 4). This indicates that the populations were not differentiated among regions. Similar results were obtained from the analysis of genetic diversity, where the gene diversities in the total population and within subpopulations were comparable (0·0638 and 0·0522, respectively), the extent of differentiation was low (Gst = 0·1825) and the level of gene flow was relatively high (Nm = 2·2392) (Table 5).

Table 4.  Analysis of molecular variance (amova) for populations of Fusarium oxysporum f.sp. vasinfectum from 31 fields across six regions (a) and nine fields in the Boggabilla region (b). Both based on 146 amplified fragment length polymorphism (AFLP) polymorphic loci generated with eight primer combinations
AnalysisSource of variationd.f.Sum of squaresVariance componentsPercentage of total varianceP-valuea
  • a

    Based on 1023 random permutations.

aAmong regions  5 478·3 0·3391 3·9  0·4868
Within regions 251999·4 6·979379·8< 0·0001
Within fields317 252·2 1·426716·3 
Total3472929·9 8·7451  
bAmong fields  81983·518·663586·3< 0·0001
Within fields111 330·0 2·973013·7 
Table 5.  Genetic diversity analyses of populations of Fusarium oxysporum f.sp. vasinfectum from six sampled regions
SubpopulationNumber of populationsSample sizeHtaHsbGstcNmd
  • a

    Gene diversity in the total population as described by Nei (1987).

  • b

    Gene diversity within subpopulations as described by Nei (1987).

  • c

    Coefficient of genetic differentiation Gst = (Ht − Hs)/Ht as described by Nei (1987).

  • d

    The amount of gene flow between subpopulations Nm = 0·5 × (1 − Gst)/Gst as described by Slatkin (1987).

All regions313480·06380·05220. 18252·2392
Warren 2 200·01950·01850·05269·0000
Moree 6 600·00870·00770·11143·9882
Boggabilla 91200·24010·03280·86330·0792
Bourke 3 300·01190·01110·06906·7500
Cecil Plains 8 880·00630·00500·20651·9211
St George 3 300·00260·00240·09304·8750

When each region was analysed separately, differences within fields were responsible for all of the total genetic variation in the Warren and Bourke regions, as well as 85·7, 96·3 and 96·6% of that in the Cecil Plains, St George and Moree regions, respectively. However, clear genetic differentiation was observed in the Boggabilla region, where the total genetic variation was primarily attributed to differences among rather than within fields (86·3 vs. 13·7%, respectively; P < 0·0001) (Table 4). Similar results were obtained from the analysis of genetic diversity (Table 5). Gene diversity in the total population (0·2401) was much greater than that within subpopulations (0·0328) in the Boggabilla region, while gene diversities in the total population and within subpopulations were comparable in other regions. The extent of genetic differentiation was higher and the level of gene flow lower in the Boggabilla region (Gst = 0·8633, Nm = 0·0792) than in other regions (Gst 0·0526–0·2065; Nm 1·9211–9·0000).

The strong genetic differentiation in the Boggabilla region was also verified by Nei's pairwise differences and results from exact tests (Table 6), with the field 16 population showing significant differentiation from those in all other fields (P < 0·001). Populations in fields 11 and 12 were similar to each other but different from all others (P < 0·001), while populations in fields 09 and 10 showed slight differentiation (P < 0·05). This was consistent with the haplotype structures of these fields: field 16 was dominated by haplotype I-03, field 11 by haplotypes II-01 and II-04, and field 12 by haplotypes II-01 and II-03 (Table 3).

Table 6.  Nei's pairwise differences among nine subpopulations of Fusarium oxysporum f.sp. vasinfectum from the Boggabilla region, Australiaa
  • *

    P < 0·05,

  • **

    P < 0·01.


There is some question as to how to treat VCGs for population genetic analyses. Given that sexual reproduction is very rare or absent in Fusarium species (Kistler, 1997), it is clear that VCGs probably represent reproductively isolated clonal lineages. Therefore, it could be argued that each VCG should be treated as a separate ‘species’. However, all the isolates are related genetically and evolutionarily, even though they may be different morphologically (i.e. in different VCGs). For this reason, a traditional approach was adopted in this study and all F. oxysporum isolates were treated as individuals within a single species, and populations defined by geography rather than morphology.

Linkage disequilibrium

Linkage disequilibrium was observed for alleles at 25 of 26 loci identified by the EcoRI-AGG/MseI-A primer combination. Of 325 possible locus pairs, 55·7% (181) were in significant disequilibrium (P < 0·05), supporting non-random association between alleles of different loci.


Genetic evidence has shown that Fov is of polyphyletic evolutionary origin (Skovgaard et al., 2001). In Australia, the close genetic relationship between Fov isolates found in cotton fields and indigenous F. oxysporum associated with wild native cottons (Gossypium spp.) indicates that the pathogen is indigenous and that VCGs 01111 and 01112 of Fov evolved locally (BW, unpublished data). In this study, great genetic variation was revealed among isolates of Fov derived from infected cotton plants, with haplotypes separating into two distinct genetically defined groups (Fig. 2). Given the low genetic similarity (only 37%) between the two groups and their nearly simultaneous appearance (they were both diagnosed within a 12-month period) in cotton fields separated by ∼200 km, these groups probably evolved from different local ancestral strains, implying the presence of multiple inoculum sources of Fov in Australia.

Similar results have been observed in other F. oxysporum pathogens. Populations of F. oxysporum f.sp. radicis-lycopersici causing crown and root rot disease of tomato in Florida were highly diverse and at least two VCGs of the pathogen appeared to have originated from that region (Rosewich et al., 1999). In addition, the present results coincide with evidence that pathogenic isolates of F. oxysporum are widely associated with wild native cottons in Australia, and these isolates were able to cause mild but typical fusarium wilt symptoms when used to inoculate plants of a susceptible cotton cultivar in glasshouse trials (Wang et al., 2004). The high genetic diversity among indigenous pathogenic F. oxysporum isolates suggests that they have considerable evolutionary potential, and it is possible that new forms of Fov could arise in future.

There is also a possible correlation between vegetative compatibility groupings and the two primary genetic lineages of Fov in Australia. Group I includes the VCG 01111 reference isolates, while group II includes the VCG 01112 reference isolates. While the other isolates have not been assayed for vegetative compatibility, over the past 10 years all isolates of Fov obtained from diseased cotton plants in Australia have been assigned to either VCG 01111 or VCG 01112 (J. Kochman, Plant Science, Queensland Department of Primary Industries and Fisheries, PO Box 102, Toowoomba, Qld 4350, Australia, personal communication). Nonpathogenic forms of F. oxysporum from native soils probably represent a diverse range of VCGs, but the current data suggest that cotton pathotypes all belong to VCG 01111 or VCG 01112. The average genetic similarities among isolates within group I and group II (> 93 and > 96%, respectively; Fig. 2) are consistent with this hypothesis.

The exception to this pattern is subgroup I-B, represented by isolates 021310 and 161105. These two isolates represent a distinct genetic lineage within group I (Fig. 2) and, based on the average genetic distance between these isolates and the subgroup I-A isolates (only 80%; Fig. 2), it is possible that they are the first known representatives of a third form of Fov, and possibly a new VCG (L. Smith, Queensland Department of Primary Industries and Fisheries, 80 Meiers Road, Indooroopilly, Qld 4068, Australia, personal communication), in Australia. Observations from other fusarium wilt pathogens demonstrated that VCGs can be further divided into subgroups (Katan et al., 1991; Katan & Katan, 1999) and new VCGs or races can evolve from existing ones (Elias & Schneider, 1991; Katan et al., 1994; Rosewich et al., 1999; Cai et al., 2003). Previous studies showed that the level of genetic similarity among VCGs and races of Fov is approximately 80%; RFLP data were used to distinguish six VCGs from each other at an 83% similarity level, while races 1, 2 and 6 of Fov were differentiated at an 85% similarity level using RAPDs (Assigbetse et al., 1994; Fernandez et al., 1994). Similar results were also observed in F. oxysporum f.sp. cubense using DNA amplification fingerprinting (Bentley et al., 1998; Gerlach et al., 2000). Indeed, nonpathogenic isolates of F. oxysporum, F. oxysporum f.sp. lycopersici and F. oxysporum f.sp. radicis-lycopersici showed 90% genetic (AFLP) similarity (Bao et al., 2002), and genetic dissimilarities revealed with RFLPs were used to predict a new VCG prior to formal complementation testing (Rosewich et al., 1999). Because of differences in marker systems and genetic distance estimators, 80% similarity cannot be considered to be an absolute criterion. Nonetheless, the genetic distance between the subgroup I-A and subgroup I-B isolates does indicate that the subgroup I-B isolates are worthy of further investigation, and subgroup I-B is the focus of ongoing investigations to determine whether these isolates are vegetatively compatible with the VCG 01111 and 01112 isolates.

The data reported here may also provide an indication of where Fov arose in Australia, although it must be noted that site of original diagnosis is not necessarily the site of origin. The presence of two VCGs that are genetically distinct and that appeared nearly simultaneously suggests that the group I and group II isolates emerged as cotton pathogens independently (Davis et al., 1996). The centre of genetic diversity for group II is the Boggabilla region, and while centres of diversity do not necessarily correspond to centres of origin, there is often a strong correlation between the two (McDonald & Linde, 2002). Following this reasoning, the Boggabilla region is the most likely site of origin for the group II isolates, an inference that is supported by the observation that group II haplotypes were not recovered outside this region. The origins of the group I isolates are more complex. VCG 01111 isolates from this lineage were first reported in the Cecil Plains region, but the level of genetic diversity for group I isolates is also quite high in the Boggabilla region (Table 3; Fig. 3). Based on the reasoning outlined above, both regions would be natural starting points for a detailed investigation into the origins of the group I lineage.

A typical feature of asexual plant pathogens, such a F. oxysporum, is the spread of a limited number of clonal lineages on a large geographic scale (Gordon & Martyn, 1997). In this respect, Fov in Australia is no exception. A single haplotype, I-01, dominates all regions and almost all fields, accounting for 77% of isolates recovered (Table 3). The only exceptions are field 16 (Boggabilla), where I-03 dominates the population, and field 24 (Cecil Plains), where the number of I-01 isolates is equal to that of other isolates. Nonetheless, these observations, in conjunction with the observation that the frequencies of haplotypes in fields 09, 10, 12 and 21 changed over 2 years, suggest that while a single haplotype may dominate at one point in time, the dominant haplotype can shift over time.

The over-representation of a single haplotype reflects extensive clonal reproduction of Fov in fields. The predominance of asexual reproduction is also evident in the high levels of linkage disequilibrium; over half the loci pairs among 26 AFLP loci exhibited significant linkage disequilibrium. Similar results were observed in other F. oxysporum pathogens (Skovgaard et al., 2002). Further evidence is the absence of any obvious recombinant genotypes, although the two VCGs of Fov in Australia belong to different mating types (CLB, unpublished data).

As is evident in Fig. 3, the distributions of Fov haplotypes in Australia fit three patterns: a single haplotype of high frequency (I-01) was the most common in all six regions sampled in this study; a suite of haplotypes of moderate frequency (I-02, I-03, I-06, I−11 and I-20) were shared across two or more regions; and a pool of haplotypes were unique to single regions.

The appearance of those haplotypes of high and moderate frequencies suggests that (assuming a single point of origin) they have moved across regions at different times (although the relative frequencies of these haplotypes in the original pool will bias the dispersal). It cannot be ascertained whether this movement occurred before or after the institution of containment protocols, given the long lag phase between contamination and the diagnosable occurrence of the disease. However, the results from this study establish a baseline distribution, and continued monitoring of the geographic distribution of these haplotypes will allow evaluation of the efficacy of current containment protocols. If the protocols are effective, the number of regionally restricted haplotypes is likely to increase while the number of haplotypes shared across regions remains static. In fact, the genetic distinctiveness of the haplotypes in subgroup I-B and their restricted occurrence in the three most southerly cotton-growing regions sampled in this study indicate that new regional lineages may be emerging. This possibility is also supported by the large number of haplotypes specific to five of the six regions (Fig. 3). If region-specific pathotypes do emerge, it may require the development of region-specific cotton cultivars. Alternatively, if future surveys indicate that haplotypes are continuing to move among catchments and regions, more stringent quarantine should be implemented in future disease-management strategies.


This research was supported by the Cotton Research and Development Cooperation of Australia. We thank Dr Stephen Allen (Cotton Seed Distributors), Dr David Nehl (New South Wales Department of Agriculture), Dr Joe Kochman and Mr Greg Simon (Queensland Department of Primary Industry) and Dr Augusto Becerra and Ms Janelle Scown (CSIRO Plant Industry) for their assistance in sample collection.