We here present the first evidence of virus induced gene silencing in roots of a monocotyledonous plant species. The transcript levels of HvIPS1, HvPHR1 or HvPHO2 were significantly decreased in the roots of barley plants infected with BSMV-IPS1, BSMV-PHR1 or BSMV-PHO2247 relative to plants infected with the control construct BSMV-GFP250. On the other hand, our attempts at silencing the HvPht1;1 and HvCel1 genes failed. The difference in silencing efficiency for the five genes could be related to differences in experimental conditions or to characteristics of the silencing inducing sequences, the viral constructs, the target genes, or a combination hereof.
Growth conditions, and in particular temperature, are known to be important for the success of VIGS experiments . However all experiments reported here were performed in growth chambers at the same temperature, 20°C. Successful gene silencing in roots was observed under hydroponic growth conditions (HvIPS1) as well as in plants grown in soil (HvPHR1 and HvPHO2), and unsuccessful VIGS results were also obtained under both conditions (HvPht1;1 and HvCel1, respectively). Furthermore, the successful silencing of HvIPS1 was obtained under P-starvation conditions, while the silencing of HvPHR1 and HvPHO2 was achieved under Pi repleteness conditions. Thus the different outcomes with these five target genes seem unrelated to the growth conditions.
Previous studies have indicated that the plant Dicer-like proteins producing siRNAs from dsRNA show a preference for GC rich regions [43, 44], suggesting that GC rich gene fragments would be more likely to induce successful VIGS. An analysis of the silencing inducing fragments used here failed to show any correlation between GC content and silencing, with the successful HvIPS1 fragment having the lowest GC content (41.8%) and the unsuccessful HvPht1;1 fragment the highest (58.1%). See also Additional file 5: GC content in silencing fragments.
The length of the silencing inducing sequence might also influence silencing efficiency. Fragments as short as 33 nt inserted into potato virus X were shown to induce silencing of the PDS gene in Nicotiana benthamiana, but a 368 nt fragment produced stronger silencing . In previous VIGS experiments with BSMV in barley, efficient silencing of the PDS gene has been obtained with fragments ranging from 120 nt to 1215 nt [3, 4, 28]. In the experiments reported here, the successful silencing inducing sequences were 247, 251 or 253 nt in length, respectively, while the unsuccessful sequences were 368 or 401 nt in length or an inverted repeat of 58 nt. Thus, the length of the silencing inducing sequence does not seem directly correlated to the efficiency of silencing.
The length of the sequence inserted into a viral vector may however indirectly impact on the silencing efficiency by influencing the stability of the modified virus. Many viral vectors show a tendency to loose the inserted sequence over time [8, 46, 47]. We have previously shown that the transient nature of BSMV-induced gene silencing is related to instability of the modified virus; the shortest insert tested (128 nt) was most stable, with instability increasing progressively with increasing length up to the longest tested (584 nt) . This prompted us to investigate the stability of the viral constructs used here. The successful construct BSMV-IPS1 proved to be relatively stable, while the unsuccessful BSMV-Pht1;1 was highly unstable. A test of five more constructs, BSMV-Pht1;4, BSMV-Pht1;7, BSMV-PHR1, BSMV-PHO2247 and BSMV-PHO2387, indicated that BSMV-PHR1 and the two BSMV-PHO2 constructs had a stability similar to BSMV-IPS1, and indeed both BSMV-PHR1 and BSMV-PHO2247 were able to induce significant silencing of their target genes in barley roots. On the other hand, three further viral constructs only inducing weak silencing of their target, HvCel1, were all highly unstable. Thus, the stability of the individual viral constructs can account for the differences in silencing efficiency observed.
The reason for the great variation in stability between individual viral constructs is unclear. The length of the insert may be of importance, since three of the most stable constructs, BSMV-IPS1, BSMV-HvPHR1 and BSMV-PHO2247, also carried the shortest of the simple sense inserts, and an inverse correlation between stability and insert length has previously been shown for the PDS gene . However, the BSMV-GFP250 construct was highly unstable, while BSMV-PHO2387 appeared relatively stable, demonstrating that length is not the sole determinant of stability. Possibly, secondary structure of the insert may interfere with steps in the viral infection cycle such as replication or assembly of viral particles. In agreement with this hypothesis, the inverted repeat construct, BSMV-Cel1-IR, was highly unstable despite carrying an insert of only 124 nt in total length. The expected hairpin-loop structure of the inverted repeat may be especially prone to interfere with viral infection steps. The secondary structures of the other inserts in the viral genome context are more difficult to predict, and other features such as local GC content or cryptic motifs might also interfere with the viral infection cycle. A negative effect of the insert on virus fitness would favour the emergence of viral constructs with deletions of the insert. In a previous VIGS study in pea, different silencing efficiencies were obtained when two different Nin gene fragments of equal size were inserted into a Pea early browning virus-based vector. The difference was attributed to differences in viral fitness, since the viral construct inducing the most efficient silencing also accumulated in larger amounts in infected plants .
The efficiency of silencing can also be influenced by characteristics of the target genes, such as accessibility of target sequences for siRNAs, possibly related to target RNA secondary structure [48, 49], or feed-back regulation leading to increased transcription when mRNA levels are reduced by RNA silencing [50, 51]. Genes with high transcription rates and short mRNA half-lives will also appear more refractory to RNAi than genes with slower natural turn-over . We cannot rule out that such factors may also contribute to the different silencing efficiencies we observe.
The HvIPS1 and HvPHO2 genes are expressed in both leaf and root tissues, and BSMV-IPS1 and BSMV-PHO2247 induced similar levels of silencing in both organs. The leaf samples analyzed were taken from leaf no. III where silencing is expected to be most strongly manifested . The root samples were taken after crushing the entire root and thus represent an average of all parts of the root. The significant silencing seen in root tissue with the more stable BSMV constructs suggests that BSMV can induce gene silencing in roots at least as strongly as in leaves. However, the sampling difficulties in roots put greater demands on the silencing efficiency/stability of the virus.
We measured the free Pi contents in the root samples of plants grown under hydroponic conditions. This confirmed that the 0 mM Pi growth conditions indeed led to extremely low levels of free Pi in the plant roots (Additional file 1: Pi content in hydroponics: HvPht1;1 experiment). Silencing of the HvIPS1 gene might have been expected to lead to increased Pi levels, if this gene plays a similar role in Pi regulation in barley as previously shown for the homologous gene in Arabidopsis . However the differences observed were not statistically significant due to the extremely low Pi content of plants grown without addition of Pi (Additional file 6: Pi content in hydroponics: HvIPS1 experiment). Further studies including more sensitive (e.g. P-33 autoradiography) Pi measurements and/or more sophisticated Pi addition regimes will be necessary to establish the impact of HvIPS1 silencing on the Pi uptake in barley. On the other hand, silencing of HvPHO2 under Pi repleteness conditions led to a significant increase in leaf Pi and a similar decrease in root Pi content (Figure 3E). This is in agreement with previous studies in Arabidopsis, where pho2 mutants show increased root-to-shoot translocation of Pi and excessive accumulation of Pi in leaf tissues [31–33]. Our data indicate that the barley PHO2 homologue has similar roles in Pi regulation in barley as previously shown for PHO2 genes in Arabidopsis, and that BSMV-mediated VIGS can be used to explore the roles of barley genes in nutrient acquisition and distribution.
Our VIGS experiments in B. distachyon were successful, in that two out of three plants infected with BSMV-BdPDS developed photobleaching accompagnied by significant reductions in BdPDS mRNA levels (Table 1 Figure 5). These data suggest that BSMV could be a very useful tool for functional genomics in this new model species. We have not attempted gene silencing in B. distachyon roots, but since BSMV was found to infect the roots efficiently (Additional file 3) it is likely that this vector will also be useful for VIGS in B. distachyon roots. The experiments reported here were conducted under greenhouse conditions and the silencing frequency was found to increase between experiment 1 (performed in mid-winter) and experiment 3 (performed in spring), suggesting that optimisation of growth conditions could further increase silencing efficiencies. In barley, efficiency of BSMV-mediated VIGS is highly cultivar dependent, and screening of different B. distachyon accessions may also further improve the VIGS efficiency in this species. For the experiments reported here, line Bd21-3 was chosen since this line is also the preferred line for transformation  and for mutagenesis and TILLING projects by international consortia (e.g., http://www.renewall.eu/). In a recent publication , successful VIGS was demonstrated in the B. distachyon ecotype ABR-1.
Our attempts at using BSMV for VIGS in oat were less promising. Although BSMVCV42 and pseudorecombinants including BSMVCV42-α accumulated to similar levels in infected oat plants as in barley (Figure 6), we found it more difficult to achieve high infection frequencies with in vitro transcripts in oat compared to barley and B. distachyon, and the infected plants showed no or very limited signs of silencing. Although we screened a range of both hexaploid and diploid oat accessions for BSMV susceptibility before choosing two lines for the silencing experiments, it is possible that other accessions may be more suited for VIGS. Furthermore, it is known that growth conditions can significantly affect VIGS efficiency . The experiments reported here were performed in growth chambers under conditions that were previously found to be appropriate for BSMV-mediated VIGS in barley , but further optimisation of temperature and light regimes might lead to improved silencing in oat.
Although the BSMV vectors are generally easy to manipulate during cloning work, occasional cloning difficulties prompted us to explore two previously reported systems for ligation-free cloning. Both the USER cloning strategy  and the T4 DNA polymerase strategy [41, 53] rely on the generation of relatively short sticky ends for inserting fragments into the vectors; in the BSMV vectors reported here, only 8 nt are used for annealing. This means that in contrast to e.g. the GATEWAY cloning strategy, only very few extra nucleotides are introduced into the final virus construct relative to traditional restriction enzyme based cloning strategies. This is particularly important for vectors such as BSMV, where instability of the virus construct increases significantly with increasing length of the foreign sequence . We found the USER strategy to be extremely efficient, fast and reliable for inserting both simple and more complex, inverted repeat fragments into the BSMV vector. The T4 DNA polymerase strategy as implemented here was slightly less efficient, with 6/12 clones containing the correct insert (8/12 in a subsequent experiment (data not shown)). However, due to the general availability of T4 DNA polymerase in most molecular biology laboratories some laboratories may find the T4 vector system preferable.