An efficient RNA interference screening strategy for gene functional analysis

Abstract

Background: RNA interference (RNAi) is commonly applied in genome-scale gene functional screens. However, a one-on-one RNAi analysis that targets each gene is cost-ineffective and laborious. Previous studies have indicated that siRNAs can also affect RNAs that are near-perfectly complementary, and this phenomenon has been termed an off-target effect. This phenomenon implies that it is possible to silence several genes simultaneously with a carefully designed siRNA.

Results: We propose a strategy that is combined with a heuristic algorithm to design suitable siRNAs that can target multiple genes and a group testing method that would reduce the number of required RNAi experiments in a large-scale RNAi analysis. To verify the efficacy of our strategy, we used the Orchid expressed sequence tag data as a case study to screen the putative transcription factors that are involved in plant disease responses. According to our computation, 94 qualified siRNAs were sufficient to examine all of the predicated 229 transcription factors. In addition, among the 94 computer-designed siRNAs, an siRNA that targets both TF15 (a previously identified transcription factor that is involved in the plant disease-response pathway) and TF21 was introduced into orchids. The experimental results showed that this siRNA can simultaneously silence TF15 and TF21, and application of our strategy successfully confirmed that TF15 is involved in plant defense responses. Interestingly, our second-round analysis, which used an siRNA specific to TF21, indicated that TF21 is a previously unidentified transcription factor that is related to plant defense responses.

Conclusions: Our computational results showed that it is possible to screen all genes with fewer experiments than would be required for the traditional one-on-one RNAi screening. We also verified that our strategy is capable of identifying genes that are involved in a specific phenotype.

Figure 1

Schematic diagram of each step of our heuristic method. (a) The process of enumerating the subsequences from the candidate-genes. The green sequence represents a candidate-gene, and the blue sequences represent the subsequences that were derived from this candidate-gene by the sliding scan. (b) Diagram of the relationship between the subsequences. The blue lines C₁, C₂,…, C₈ represent the candidate-genes, and S₁, S₂,…, S₁₁ are the subsequences that were enumerated from these candidate-genes. A solid line between two subsequences indicates that the two subsequences are neighbors, and the dotted line indicates that the two subsequences are far_neighbors. Each subsequence is marked as a candidate for a qualified sequence. (c) Diagram of the relationship between the marked subsequences after the far_neighbor examination. The subsequences S₅, S₉ and S₁₀ are unmarked because they all contained a far_neighbor. However, S₂ is still marked because one of its neighbors was located in C₄, and its far_neighborS₅ was also located in C₄. In this situation, we are not concerned about whether the siRNA that was designed based on S₂ will recognize S₅ because C₄ is already the target gene of this siRNA. Therefore, S₂ is still marked as a candidate for the qualified sequence. (d) Diagram of the relationship between the marked subsequences after the powerful subsequence examination. S₁, S₄ and S₆ are unmarked because they are not powerful subsequences and are all dominated by S₂. (e) Diagram of the relationship between the marked subsequences after the excluded-gene hit examination. E_i is one of the excluded-genes, and the dot-dashed line indicates that the Hamming distance between a marked subsequence and a substring that is located in an excluded-gene is less than d_N; this scenario also indicates that this marked subsequence contains an excluded-gene hit. Because any subsequences that contain an excluded-gene hit will be unmarked, S₁₁ is unmarked.

Figure 2

Flowchart of the design of the qualified siRNAs that were used in the first-round analysis of our strategy. The first three stages demonstrate the flow of the heuristic algorithm that we proposed. The qualified sequences can be derived by this heuristic algorithm, and the complement sequences of these qualified sequences are the qualified siRNAs. Based on the availability of the qualified siRNAs, we can then select the siRNAs for the first-round RNAi analysis from the qualified siRNAs in stage 4.

Figure 3

Schematic diagram of the hierarchical RNAi analysis for identifying genes that are involved in a specific phenotype. R = {r₁, r₂, …, r_k} is the set of qualified siRNAs that were selected by our modified algorithm. The candidate-genes in each ellipse are the target genes of each available siRNA r_i. In the first-round RNAi analysis, we can quickly determine which candidate-genes could be related to the target phenotype based on the RNAi experiment results. The green ellipse represents the target phenotype that was affected in the RNAi experiment with r_i, whereas the blue ellipse indicates that the target phenotype was not affected. In the second-round RNAi analysis, we further examined the candidate-genes in the green ellipses to precisely identify the related genes. However, the complementary strand of the selected siRNA, r^′_i, can also be loaded into the RISC, and it could facilitate RNAi. A gene that contains a subsequence that contains no more than d_T mismatches to r^′_i would be regarded as the possible target gene of r^′_i, which is defined as X_i. To precisely identify the associated gene from the genes that are affected by r_i and r^′_i, we would silence those genes one-on-one to identify which genes were definitively associated with the target phenotype.

Figure 4

Using siRNA-G41 to simultaneously silence three target genes. The effects of silencing three putative transcription factors in plants that were transfected with a hairpin RNA-G41 are shown. The RNA levels of Pha106, Pha157 and PhaTF187 were analyzed in healthy (H) plants, in plants that were infiltrated with Agrobacterium (agro), in plants that were infiltrated with Agrobacterium that carried an empty pB7GWIW2 vector (agro-v) or in plants that were infiltrated with Agrobacterium that carried pB7GWIW2 to deliver hairpin RNAs designed from siRNA-G41 (hpRNA-G41). Phalaenopsis Ubiquitin 10 was used as an internal control.

Figure 5

Using siRNA-G55 in first-round analysis. The effects of silencing the marker genes that are involved in the salicylic acid (SA)-related plant defense response pathway in plants that were transfected with a hairpin RNA-G55. The RNA levels of PhaPR1, PhaNPR1, PhaTF15, PhaTF21 and PhaTF60 were analyzed in healthy (H) plants, plants that were infiltrated with Agrobacterium that carried an empty pB7GWIW2 vector (agro-v), or Agrobacterium that carried pB7GWIW2 to deliver different hairpin RNAs that were designed from siRNA G55 (hpRNA-G55) or were specific to PhaTF60 (PhaTF60-hpRNA). The Agrobacterium strain that carried a partial PhaTF15 cDNA was used as a positive control (P), and PhaTF60 was used as a negative control. The plants that were pretreated with SA to induce the SA-related plant defense response are indicated. Phalaenopsis Ubiquitin 10 was used as an internal control.

Figure 6

Identifying the associated gene more precisely in the second-round analysis. The knockdown effects of the marker genes involved in the salicylic acid (SA)-related plant defense response pathway in plants that were transfected with hairpin RNA and that are specific to PhaTF15 and PhaTF21. The RNA levels of PhaPR1, PhaNPR1, PhaTF15 and PhaTF60 were analyzed in healthy (H) plants, plants that were infiltrated with Agrobacterium that carried an empty pB7GWIW2 vector (agro-v), or Agrobacterium that carried pB7GWIW2 to deliver different hairpin RNAs that were specific to PhaTF15 (PhaTF15 hpRNA), PhaTF21 (PhaTF21 hpRNA), PhaRNA (PhaRNA), or PhaTF60 (PhaTF60 hpRNA). The Agrobacterium strain that carried the partial PhaTF15 cDNA was used as a positive control (P), and PhaTF60 was used as a negative control. The plants that were pretreated with salicylic acid (SA) to induce the SA-related plant defense response are indicated. Phalaenopsis Ubiquitin 10 was used as an internal control.