In the MS-AFLP analysis, polymorphisms were analyzed by comparing the presence
or absence of bands between two sets of digestion/ligation reactions. Band presence in only
one digestion set and absence in the other set were recorded as methylation polymorphism in
order to eliminate nucleotide polymorphism. The polymorphic bands were scored as binary
data and analyzed using the same procedure as for standard AFLP. From nine primer combinations,
60 methylation polymorphisms were detected. Estimates of the PIC value had a
similar range as the AFLP data ranging from 0.03 to 0.46 with an average of 0.179. Genetic
similarity estimated by Sneath and Sokal (1973) among samples studied was moderately high,
ranging from 0.70 to 1.00. Results from cluster analysis using UPGMA separated 56 samples
into five groups (Figure 1B). The cophenetic correlation coefficient revealed a good fit of
cluster analysis, with a value of r = 0.84 for the dendrogram. The first group comprised only
one sample, the hybrid from Anna21 and E-L23 (JCL23). The second group was all non-toxic
JCL. The third group was mutation-induced JCL (JCL25 and JCL26). The fourth group was
the JCL sample from an area with saline soil (JCL8). The fifth group consisted of the rest of the
samples including mostly Thai and foreign accessions, as well as some mutants and hybrids.
PCA showed the same pattern as the UPGMA tree. Principal components 1, 2 and 3 accounted
for 14.5, 9.06 and 8.40% of the total variation, respectively. MS-AFLP analysis revealed the
alteration in the DNA methylation level among the samples, which possibly was caused by
different environmental effects.