We’ve know for a long time that 24-nt siRNAs are more abundant in plant reproductive tissue and a common model is that endosperm produces siRNAs that move into the embryo. But data supporting this model are pretty thin! /2
There is also debate about the extent to which siRNAs are produced from both matrigenic and patrigenic chromosomes in the endosperm. Is there imprinting of siRNA production? /3
My lab has been using Brassica rapa as a model to study siRNAs in seeds. Because it’s a crop, because it has an interesting mutant phenotype, but mostly because we like big seeds. /4
@groverj3 looked at siRNA production in each component of the Brassica seed and discovered that seeds are dominated by a few SUPER highly expressed siRNA loci. We call these sirens. /5
(The name comes this paper on rice siRNAs in endosperm, but we think they’re the same thing, so we’re keeping the name.) /6
pubmed.ncbi.nlm.nih.gov/23613580-impri…
pubmed.ncbi.nlm.nih.gov/23613580-impri…
Siren siRNAs express before fertilization (>90% of siRNAs in ovules come from ~1.5% of loci) and accumulate highly throughout seed development. Basically these loci are the 1 percenters of seed siRNAs. /7
In most ways, siren loci look like most siRNA loci: they make 24-nt siRNAs that start with an A, they require Pol IV and RDR2, and they trigger increased methylation (especially CHH). /8
One difference is that siren loci tend to be unique. 90% of siren siRNAs map to a single genomic location, compared to ~70% for other 24-nt siRNA. Consistently, siren loci have a single strong BLAST match in the genome and are depleted for most transposons. /9
In addition to rice, we find siren-like loci in Arabidopsis, suggesting sirens are a general feature of seed development. But they evolve FAST - we found almost no sequence conservation, although sometimes siRNA production was conserved. /10
We were surprised to see that siren siRNAs are most abundant in the developing seed coat, although we can also find them in the endosperm and at much lower levels in the embryo. /11 
So is strong siren expression in the (diploid, maternal) seed coat the reason seed siRNAs appear to be maternal-specific? Maybe, but that’s not the whole answer. /12
We also looked at parental bias of siRNAs in developing embryos and endosperms. It turns out that siren siRNA accumulation is entirely maternal in endosperm, but bialleleic in embryo. /13
So maybe siren loci are imprinted in endosperm and only matrigenic alleles express siRNAs. Alternatively, sirens might express in the seed coat and move into the endosperm. Either way, seeds are *swimming* in maternal-specific siRNAs from a couple hundred loci! /14
So what are these siren siRNAs doing? That’s a story for another day! The parallels with piRNAs and tapetum-specific phasiRNAs are intriguing. We’re also fascinated with how this specific expression pattern is established. There are lots of interesting new questions! /15
