Bph3-carrying backcross rice lines infested by BPH are sterile
Ms55 is a high-yielding elite indica cultivar in China, but it is susceptible to BPH. To develop BPH-resistant indica cultivars, Ms55 was used as the recurrent parent and backcrossed with TZ21 harboring the resistance gene Bph3 to produce an initial backcross (BC1F) population (Fig. S1). Two Bph3-carrying indica lines, Rby1 and Rby2, were selected from the BC1F population.
To test whether the introgression of Bph3 can improve resistance to BPH, rice plants from the Rby1 and Rby2 lines grew under greenhouse conditions, and each line at the seedling stage was infested with BPH of mixed biotypes (with biotype 2 being the dominant one; the BPH biotypes refer to specific populations of BPH classified according to their virulence on different BPH resistance genes) [24] collected from rice fields in Hangzhou. Ms55 was used as negative control, and TZ21 was regarded as positive control. The Bph3-carrying Rby1, Rby2 and TZ21 lines were not visibly damaged, while BPH infestation resulted in 100% mortality in the susceptible recurrent parent Ms55 at 19 days. When Rby1, Rby2, TZ21 and Ms55 were infested with a fixed number of BPHs, the BPH population on Ms55 increased steadily over time. In contrast, the BPH population on Rby1, Rby2, and TZ21 decreased dramatically within a few hours of infestation. These observations indicated that Rby1 and Rby2 displayed strong resistance to BPH. Although the behavior of Rby1 and Rby2 infested by BPH was similarly to that of Ms55 or TZ21 in terms of their vegetative growth, both rice lines showed panicle enclosure, and only a few seeds were produced on each rice plant, which are typical characteristics of sterile rice plants (Fig. 1c-e); however, the agronomic traits of the Rby1 or Rby2 rice plants not infested by BPH were similar to those of Ms55 or TZ21, including earing and seed formation (Fig. 1a-b).
Multiple RNA viruses were identified in sterile rice plants by deep metatranscriptomic sequencing
Rice plants of lines Rby1 or Rby2 not infested by BPH were as healthy as those of Ms55 or TZ21 (Fig. 1a-b). However, those of lines Rby1 and Rby2 became sterile when they were infested by BPH, and these sterile rice plants did not exhibit any disease symptoms indicative of fungal or bacterial infections or typical characteristics of viral infection, such as pronounced stunting and dark green leaves. The only difference between the healthy and sterile rice plants was BPH infestation; it is known that BPH can transmit viruses as vectors to rice plants, but we did not isolate any viruses from the sterile rice plants. We speculated whether viral covert infection caused rice sterility. To test this hypothesis, we performed deep metatranscriptomic sequencing of rice plants, including Rby1-21 from line Rby1 and Rby2-45 from line Rby2. RNA sequencing of rRNA-depleted libraries yielded 12.69 Gb of data for Rby1-21 and 10.64 Gb of data for Rby2-45, which resulted in 70,923,042 and 84,627,146 reads, respectively. We de novo assembled these reads into 328,146 contigs for Rby1-21 and 586,111 contigs for Rby2-45 (Table 1).
We first examined the assembled contigs that matched previously characterized RNA viruses. We identified five known viruses in Rby1-21 and Rby2-45. Rice tombus-like virus 1 (RTV1) and rice ragged stunt virus (RRSV) were present in both Rby1-21 and Rby2-45, while rice picorna-like virus 1 (RPiV1), rice toti-like virus (RToV) and a brown planthopper virus, Nilaparvata lugens reovirus (NLRV), were present only in Rby2-45 (Table 1).
Next, we identified virus-like sequences by BLASTing against reference sequences from viral genomes available in GenBank. We discovered eight novel viruses in both rice lines (Fig. 2), including three negative-sense RNA viruses and five positive-sense RNA viruses (Table S1). Of the eight RNA viruses, rice mononega-like virus (RMV) and rice peribunya-like virus (RPeV) were present in both Rby1-21 and Rby2-45 (Table S1). Rice picorna-like virus 2 (RPiV2) and rice noda-like virus (RNV) were present only in Rby1-21 (Table 1 and Table S1). Rice phasma-like virus (RPhV), rice tombus-like virus 2 (RTV2), rice tombus-like virus 3 (RTV3) and Rice picorna-like virus 3 (RPiV3) were present only in Rby2-45 (Table 1 and Table S1). Thus, we identified thirteen RNA viruses from Rby1-21 and Rby2-45 in total, including five known RNA viruses and eight novel RNA viruses (Table 1). We detected six RNA viruses in Rby1-21 and eleven RNA viruses in Rby2-45 (Table 1). Of these viruses, four RNA viruses were present in both Rby1-21 and Rby2-45, two RNA viruses were present in only Rby1-21, and seven RNA viruses were present in only Rby2-45 (Fig. S2). However, we did not detect any viruses in Rby1-N65 from line Rby1 or in Rby2-N32 from line Rby2, which were not infested by BPH.
We identified three novel negative-sense RNA viruses (Table S1). Of these viruses, RPeV is closely related to Penicillium roseopurpureum negative ssRNA virus 1, RPhV is related to Anopheles triannulatus orthophasmavirus, and RMV is similar to Tacheng tick virus 5 (Table S1). In the RdRp phylogeny, RPeV was clustered within the family Peribunyaviridae, and RPhV belonged to the family Phasmaviridae, both of the order Bunyavirales (Fig. 3). RMV clustered within a currently unclassified family of the order Mononegavirales (Fig. 3 and Fig. S3).
We also discovered five undescribed positive-sense RNA viruses in this study (Table S1). In the RdRp phylogeny, these viruses fell within the Tombusviridae, Picornaviridae and Nodaviridae families (Fig. 3). Two positive-sense RNA viruses, RTV2 and RTV3, belonged to the family Tombusviridae (Fig. 3). RTV2 was closely related to soybean leaf-associated ssRNA virus 1, and RTV3 was similar to Setosphaeria turcica ambiguivirus 1 (Table S1). Two positive-sense RNA viruses, RPiV2 and RPiV3, fell within the family Picornaviridae (Fig. 3); RPiV2 was closely related to Picornavirales sp., and RPiV3 was related to Hubei picorna-like virus 35 (Table S1). The remaining RNV clustered within the family Nodaviridae (Fig. 3), and RNV was closely related to Hubei orthoptera virus 4 (Table S1).
Although phylogenetic analysis showed that these sequences are viral in origin, they may represent endogenous viral elements (EVEs) integrated into the rice plant genome [11] or may be derived from surface contamination rather than active infection. To exclude the possibility that these contigs represent EVEs segregating in rice plants, we mapped the raw reads from the Rby1-21 and Rby2-45 genomes to our set of candidate viruses revealed by BLAST and confirmed that no genome mapped at a rate high enough to be consistent with a genome copy of any virus in a particular individual.
Small RNA analysis indicated active virus infections in sterile rice plants
Rice plants utilize small RNA pathways for viral defense [26, 27]. If a virus is active in rice plants, this is often indicated by the presence of an antiviral immune response [28]. Thus, we determined the presence of virus-derived small interfering RNAs (vsiRNAs) in rice plants as being indicative of active viral infection. We constructed small RNA libraries from Rby1-21 and Rby2-45 samples. Both libraries were subjected to 50 bp single-end sequencing, resulting in 24,822,021 and 23,341,707 reads for Rby1-21 and Rby2-45, respectively.
To map the resulting small RNA reads to the putative viruses described above, we first removed the small RNA reads that are specific to rice plants and then aligned the reads to the viral genomes described in Table 1. The results showed that differentially abundant small RNAs were mapped to putative viruses (Table S2), and an abundance of small RNAs were identified for RTV1 (134,715 reads) and RRSV (7,108 reads) from Rby1-21 as well as RTV1 (109,437 reads), RRSV (23,558 reads) and RPeV (3,639 reads) from Rby2-45. Small RNAs with fewer than 200 reads were excluded because of possible random degradation.
To test whether these small RNA sequences were derived from active virus infection, we analyzed all 18-30 nt small RNAs from identified viruses in Rby1-21 and Rby2-45. VsiRNAs detected in plants infected with RNA viruses are predominantly 21 nucleotides in length produced by Dicer-like 4 (DCL4), and 22-nucleotide vsiRNAs are produced by DCL 2 [29, 30]. Our small RNA reads have a size distribution of 21 to 22 nt for RTV1 and RRSV in Rby1-21 as well as RPeV, RTV1, and RRSV in Rby2-45. These small RNAs occur mainly in the sense and antisense orientations (Fig. 4), indicating that the vsiRNAs were produced from double-stranded RNA replicative intermediates and that Dicer-like enzymes cleaved double-stranded RNAs into vsiRNA. The cleaved double-stranded vsiRNAs are bound and sorted by AGOs depending on the 5’-terminal base [31]. To assess the base preference at the 5’ terminus in the rice samples, we calculated the base percentage for the vsiRNAs. As shown for the three viruses (RTV1, RRSV and RPeV) in Fig. 4, base G was the least favored at the 5’-terminal nucleotide for all viruses, and the viruses showed a distinct preference for the remaining three bases at the 5’ terminus. Base U was enriched in 21-nt vsiRNAs for RPeV and RRSV, while base C was enriched in those for RTV1. These results were consistent with those of a previous study [32]. Twenty-one-nucleotide vsiRNAs occurred in regions of identified genomic RNAs (Fig. 5), and those derived from antisense strands showed a high proportion for each virus (Table S3). Collectively, our data demonstrated that RTV1 and RRSV in Rby1-21 as well as RPeV, RTV1 and RRSV in Rby2-45 were active infections.