Vielle-Calzada et al. reply

Previously, early seed formation was thought to involve transcription from both parental copies immediately following fertilization, and maternal effects were considered rare or non-existent2. The time at which paternal activity can first be detected, however, is likely to vary from embryo to embryo and from gene to gene in different nuclei, as in Drosophila3. Weijers et al. report paternal expression of AtRPS5A::GUS as early as the two-cell stage, confirming that transcription in the zygote is not the rule for paternally inherited alleles, whereas transcription from maternal alleles has been demonstrated immediately after fertilization of the central cell4. We do not know what percentage of embryos show early AtRPS5A::GUS expression, nor the relative paternal and maternal activity, but there may also be less pronounced parent-of-origin differences.

New evidence supporting the non-equivalence of maternal and paternal genomes during early seed development is based on experiments with reporter genes5,6,7,8 and genetic assays revealing maternal effects of genes thought to act purely zygotically6 (S. Gilmore and C. Somerville, personal communication; J. Moore and U. G., unpublished results). Whether and at what stage expression of the paternal allele is sufficient for normal development will depend on the level of activity required for gene function. In a two-component transactivation system, no paternal activity was found during early seed development using pOp::GUS reporter lines with several activator lines8. Some early defects were evident with a pOp::BARNASE reporter, however, suggesting that paternal transcription is very low but is sufficient to cause BARNASE-induced defects in some embryos8. These results confirm the non-equivalence of maternal and paternal contributions to early seed development. Like imprinted genes in mammals, this difference is probably not absolute and may be due to different levels of maternal and paternal transcripts.

Our titration experiments indicated a difference in transcript levels of at least 80-fold for genes we tested by PCR. Weijers et al. report an expression difference in reciprocal crosses with UAS:GUS at the heart to torpedo stage (Fig. 1d), when we showed that both parental alleles are active at other loci we tested; indeed, this differential expression translates into an absence of detectable paternal activity at earlier stages using the pOp::GUS reporter system8. For some genes, such as KEULE or KNOLLE, low paternal expression may be sufficient for normal development, although very early defects (such as developmental delay) that are rescued by a paternal wild-type allele may be difficult to detect by scoring multinucleate embryos. Moreover, rescue of an early embryonic phenotype by a paternal wild-type allele provides no evidence against differences in parental transcript levels.

Although the exact time of paternal activation was not central to our report, most evidence so far suggests that no consistent paternal gene activity can be detected in the embryo or endosperm for several cell divisions. The results of Weijers et al. do not contradict our findings, but instead represent possible exceptions to a general rule. Specific genes that are important during early development (for example, those involved in cytokinesis that are distinctly regulated in the female gametophyte and the zygote9) may be under selection for earlier expression and be specifically activated early in development. Further investigation is required into how common early-expressing paternal genes are, and how maternal and paternal expression differs quantitatively.