All of these miRNAs, except for miR827, were members of 21 families that are conserved in diverse plant species. The abundance of miR NAs varied greatly. MiRNA families highly conserved across plant species, such as miR166, miR167, nothing and miR168, were sequenced more than 10,000 times, whereas previously known stress induced members, such as miR395 and miR399, were detected less than 10 times, indicating that tissue specific expres sion patterns of miRNAs are related to their functions. In contrast, most rice or monocot specific miRNAs were detected with low read numbers, except for miR444 and miR528, which were represented by 3,917 and 6,305 cop ies, respectively. There were significant variations in expression levels for members of the same family. For example, the abun dance of the miR159 family varied from 9 to 7,113 reads.

Similarly, the abundance of members of the miR166 and miR164 families were also highly variable. Twenty previously reported non conserved miRNA families were not detected in our dataset. A major reason for this might be the limited low sequencing depth, at which the ex pression level of this group of miRNAs might have been too low to be detected in our library. Another factor may have been the different subspecies and cultivar used compared with previous work. We found that the loca tions of many miRNA reads varied within a 2 nt range from the 5 or 3 ends of annotated miRNA sequences. Some of these variants even had similar reads compared with those annotated in miRBase. For example, the annotated miR1870 had 11 reads in our libraries, whereas the other 22 nt variants had 14 reads.

Interest ingly, some miRNA s had higher read numbers than the corresponding miRNAs. For example, miR529 and miR2124 had more reads than their respective miRNAs, 135 vs 0 and 117 vs 1, respectively, suggesting that miRNA may play a major role in these cases. Identification of 11 novel miRNAs in developing caryopses To find novel miRNAs, we first mapped all the small RNAs to the sequenced indica cultivar 9311 genome because Baifeng B is an indica landrace. Secondary structures of sequences around the small RNAs were produced using Mfold. These putative miRNA precur sors were then used to find miRNA s, which are consid ered strong evidence for DCL1 derived products. We found 11 regions that satisfied these criteria and considered them to be novel miRNA gene candidates.

Most novel miRNAs showed weak expression levels. The reads for their miRNA s were even lower. All of these newly identified miRNAs appeared to be rice specific and had not been reported in other species. Most novel miRNAs were not detectable by northern blotting, except Can miR 10, but all were confirmed by using more sensitive array analysis. Surprisingly, novel miRNAs discovered in previous deep sequencing Dacomitinib of rice grain small RNAs were rarely present in our dataset.