Building pangenome graphs.

Pangenome graphs can represent all variation between multiple genomes, but existing methods for constructing them are biased due to reference-guided approaches. In response, we have developed PanGenome Graph Builder (PGGB), a reference-free pipeline for constructing unbi-ased pangenome graphs. PGGB uses all-to-all whole-genome alignments and learned graph embeddings to build and iteratively refine a model in which we can identify variation, measure conservation, detect recombination events, and infer phylogenetic relationships.

Structural and functional analysis of viral siRNAs.

A large amount of short interfering RNA (vsiRNA) is generated from plant viruses during infection, but the function, structure and biogenesis of these is not understood. We profiled vsiRNAs using two different high-throughput sequencing platforms and also developed a hybridisation based array approach. The profiles obtained through the Solexa platform and by hybridisation were very similar to each other but different from the 454 profile. Both deep sequencing techniques revealed a strong bias in vsiRNAs for the positive strand of the virus and identified regions on the viral genome that produced vsiRNA in much higher abundance than other regions. The hybridisation approach also showed that the position of highly abundant vsiRNAs was the same in different plant species and in the absence of RDR6. We used the Terminator 5'-Phosphate-Dependent Exonuclease to study the 5' end of vsiRNAs and showed that a perfect control duplex was not digested by the enzyme without denaturation and that the efficiency of the Terminator was strongly affected by the concentration of the substrate. We found that most vsiRNAs have 5' monophosphates, which was also confirmed by profiling short RNA libraries following either direct ligation of adapters to the 5' end of short RNAs or after replacing any potential 5' ends with monophosphates. The Terminator experiments also showed that vsiRNAs were not perfect duplexes. Using a sensor construct we also found that regions from the viral genome that were complementary to non-abundant vsiRNAs were targeted in planta just as efficiently as regions recognised by abundant vsiRNAs. Different high-throughput sequencing techniques have different reproducible sequence bias and generate different profiles of short RNAs. The Terminator exonuclease does not process double stranded RNA, and because short RNAs can quickly re-anneal at high concentration, this assay can be misleading if the substrate is not denatured and not analysed in a dilution series. The sequence profiles and Terminator digests suggest that CymRSV siRNAs are produced from the structured positive strand rather than from perfect double stranded RNA or by RNA dependent RNA polymerase.

Deep sequencing of tomato short RNAs identifies microRNAs targeting genes involved in fruit ripening.

In plants there are several classes of 21-24-nt short RNAs that regulate gene expression. The most conserved class is the microRNAs (miRNAs), although some miRNAs are found only in specific species. We used high-throughput pyrosequencing to identify conserved and nonconserved miRNAs and other short RNAs in tomato fruit and leaf. Several conserved miRNAs showed tissue-specific expression, which, combined with target gene validation results, suggests that miRNAs may play a role in fleshy fruit development. We also identified four new nonconserved miRNAs. One of the validated targets of a novel miRNA is a member of the CTR family involved in fruit ripening. However, 62 predicted targets showing near perfect complementarity to potential new miRNAs did not validate experimentally. This suggests that target prediction of plant short RNAs could have a high false-positive rate and must therefore be validated experimentally. We also found short RNAs from a Solanaceae-specific foldback transposon, which showed a miRNA/miRNA*-like distribution, suggesting that this element may function as a miRNA gene progenitor. The other Solanaceae-specific class of short RNA was derived from an endogenous pararetrovirus sequence inserted into the tomato chromosomes. This study opens a new avenue in the field of fleshy fruit biology by raising the possibility that fruit development and ripening may be under miRNA regulation.

Evidence for GC preference by monocot Dicer-like proteins.

Dicot Dicer-like (DCL) enzymes operate preferably on GC rich regions when producing small interfering (si)RNA and micro (mi)RNA. This GC bias, however, is not generic in monocot miRNA productions. From wild Dactylis glomerata naturally infected by Cocksfoot streak potyvirus (CSV), CSV-siRNAs had a greater GC% than the virus genome, indicating that GC rich regions were also preferred by the grass DCLs. This supports the notion that GC preference is an ancient feature for plant DCLs, and suggests that monocot miRNA genes might have evolved to a high GC% resulting in GC bias being not detectable during mature miRNA production.