1USDA, Agricultural Research Service, Soybean and Alfalfa Research Laboratory and 2 Biocontrol of Plant Diseases Laboratory, Beltsville, MD 20705, USA

Background and objectives
Phytopathogenic fungi are usually identified to species and race based on morphology, host specialization, cultivar specificity, and mode of parasitism. The application of molecular analysis methods have helped clarify the genetic relationships of fungal taxa which are not clearly distinguished by their morphology. A significant improvement in identification and classification of phytopathogenic fungi has come from RFLP, PCR-based RAPD and rDNA sequence analyses. Limitations to these techniques include difficulty in standardization,time and effort required, the relatively small numbers of polymorphisms generated, small and very selective sampling of the genome, and reproducibility. Amplified restriction fragment length polymorphism (AFLP) is a PCR-based fingerprinting technique that has been used with plants, bacteria, nematodes and, recently, fungi. This technique has significant advantages over other procedures. AFLP DNA analysis was applied to fungal genera and species pathogenic to three hosts to determine if it can be used to determine taxonomic relationships within these groups. Colletotrichum isolates unusually aggressive to anthracnose-resistant Medicago sativa cultivars were compared with isolates of C. trifolii races 1 and 2 and with C. gloeosporioides. Molecular genetic differences were evaluated between two groups of morphologically distinct isolates of Dendryphion penicillatum (telomorph Pleosporapapaveracea ) pathogenic to Papaver somniferum. AFLP and RAPD analysis was used to determine genetic relatedness among isolates of Fusarium udum representing four race combinations pathogenic to pigeonpea (Cajanus cajan).

Materials and methods
For AFLP assay, purified genomic DNA was digested with restriction endonucleases, ligated to EcoRI and MseI adapters, and amplified by PCR, using primers that contain the common sequences of the adapters and two nucleotides as selective sequences. Selective nucleotides were EcoRI+AG, +AT, or +TA. Each pair was evaluated in combination with selective primer MseI+C. PCR products were amplified and separated by efectrophoresis. Autoradiographs were scored using Proscore and similarity dendrograms were constructed using Unweighted Pair Group Cluster analysis Using Arithmetic Means (UPGMA) constructed from matrices analysed with the Numerical Taxonomy and Multivariate Analysis System (NTSYS-pc). Results from analyses using simple matching coefficients were compared with analysis using Jaccard's coefficient for pairwise comparisons, and by the parsimony method using the Dollop and polymorphism parsimony program.

Results and conclusions
AFLP DNA analysis was effective in establishing taxonomic relationships within species. Complex AFLP patterns were obtained using three different primer pairs and genomic similarity analyses derived from qualitative data enabled us to identify two isolates (C. trifolii isolate Ari-NW and C. gloeosporioides isolate 57RR) whose taxa had been uncertain based on morphological criteria. The similarity matrices generated by each of three primer pairs were highly correlated and were combined to determine the genetic relationships among the fungal species and isolates. The genetic diversity detected among and within Colletotrichum species from alfalfa and other crops corroborated morphology, RDNA, and RAPD based taxonomy. Less that 50% similarity was found between two groups of Dendryphion penicillatum isolates, and correlated with differences in production of microscierotia, virulence, and size of conidia. Fusarium udum isolates belonging to one VCG were closely related by RAPD and AFLP analysis (>90% similar) and were not distinguishable for race or regional location. There was less than 70% similarity between F. udum and F. oxysporum formae speciales pathogenic to coca, cowpea, and tomato. AFLP represents a practical advance in fungal DNA fingerprinting because of greater resolution and the collection of more information than is possible by RAPD and RFLP techniques. AFLP analysis will be useful to identify genetic diversity and analyse population structure within complex genera such as Colletotrichum andFusarium.