To identify associations and putative functional relationships between DEGs, cluster analysis was performed and
each cluster analyzed for over-represented GO terms. The
DEGs were grouped into 18 clusters by k-means clustering,
seven of which yielded statistically over-represented GO
terms (P < 0.05 Hochberg false discovery rate; Figure S1;
Table S10). Although many GO categories overlapped with
those identified in our time-point comparison (Table S9),
several interesting new groups were highlighted. These
included: lipase activity (cluster 5), nuclear protein import
(cluster 6) and, endo-1,4-b-xylanase activity (cluster 7).
Endo-1,4-b-xylanases are associated with cell expansion
and shape changes, and inclusion of three (of five on the ATH1 microarray) within cluster 7 suggests WUS, or
dependent processes, reduce these activities.
After 30 h of treatment with 2iP, we hoped to identify
early events in WUS-dependent LRP ? SM conversion. To
provide greater resolution of WUS-related DEGs at this
time point, we examined the transcriptome of another wus
mutant allele (GABI_870H12). GABI-KAT constructs were
designed for activation tagging, but in this line an intragenic insertion appears to drive elevated expression of a
truncated non-functional product (Methods S1). Heterozygotes yield loss-of-function wus phenotypes in approximately 25% of progeny, and these homozygous mutants
are unable to undergo LRP ? SM conversion. For comparison of the two alleles with the WT, processed data were
filtered to remove genes absent in one wus mutant allele,
but not the other. Using the LIMMA package of BIOCONDUCTOR,
a P-value threshold of