Our project aimed to engineer yeast strains that produce optimal small interfering RNAs (siRNAs) against the Deformed wing virus (DWV), a bee pathogen. The engineering decisions were based on published results, computational modeling, and empirical findings derived from our laboratory investigations. Within this segment of Wiki, we elucidate the rationale behind our decisions while engineering yeast strains to test suppression of viral sequences by RNA interference (RNAi).

The aims and uncertainties

The primary goal of our project is to create genetically modified yeast strains expressing small hairpin RNAs (shRNAs) that are then processed into siRNAs. The siRNAs can induce RNAi response through the short interfering RNA (siRNA) pathway. These siRNAs can selectively target mRNA of DWV, suppressing replication and translation of viral proteins. When consumed by bees as a food supplement, our engineered yeast and its siRNAs could protect the bee populations from DWV.

We have mapped several uncertainties affecting the success of this project:

  1. Can highly efficient anti-DWV siRNAs be designed? Can we distinguish efficient siRNA from less efficient ones at the design stage? What factors are important to consider while designing?
  2. How can we test the designed siRNAs? The success of this approach depends on finding and validating efficient siRNAs. As testing the siRNAs in bees with the DWV is time-consuming, expensive and complicated, as it would require working with viruses, we designed a reporter system to evaluate the siRNAs in yeast.
  3. Do the anti-DWV siRNAs have off-target effects? siRNAs have been found to unintentionally affect genes other than their intended targets. Therefore, a thorough assessment of potential off-target effects, especially those that could impact vital genes, is necessary for the developed siRNAs.
  4. Do the siRNAs retain necessary activity while passing from yeast to the bee cell? RNA molecules are prone to degradation, raising the question of siRNA stability after the siRNA has been ingested by bees. Further investigation is needed to determine whether these siRNA molecules can pass through the bee's digestive system without degradation. Previous studies have indicated that orally delivering siRNA to bees reduces DWV-related mortality (Desai et al., 2012), indicating that the small RNAs are capable of passing through the digestive system while retaining activity. Also, depending on genetic modifications, the yeast can be engineered to produce either shRNAs or siRNAs. If further studies reveal differences in the uptake and stability of shRNAs compared to siRNAs, this knowledge would be integrated into the design of the yeast strain.

Design: From bread and beer to siRNA. How to make S. cerevisiae produce siRNAs

While RNAi is a widespread mechanism in eukaryotes, our literature search identified that yeast Saccharomyces cerevisiae does not have the RNAi machinery and thus is unable to generate siRNAs. However, a closely related yeast, Saccharomyces castellii, does have the genes required for siRNA generation and RNAi. Previous studies have constituted siRNA production in S. cerevisiae through the introduction of Argonaute1 (Ago1) and Dicer1 (Dcr1) from S. castellii (Drinnenberg et al., 2009) (Fig. 1). We used computational modeling to evaluate the potential of processing shRNAs to siRNAs in yeast using non-native Dicer. The simulation suggests that siRNA processing in yeast is not a limiting step in siRNA production in yeast. Therefore, set out to express Dicer in yeast to produce siRNAs.

Figure 1. AGO1 and DCR1 expression cassettes.

Following the reconstitution of siRNA production in S. cerevisiae, the necessity arose to incorporate a template for the generation of specific siRNAs. We used the DWV genome as a reference in pursuit of suitable siRNA candidates, and leveraged the functionality of the GenScript siRNA Target Finder tool (SiRNA Target Finder, Genscript) to identify optimal target sequences within the virus genome. Following the analysis of the DWV genome with the Genscript’s SiRNA Target Finder, ten variants of potential siRNA were obtained. Subsequently, drawing inspiration from the noted high potency and sustainable effects of shRNA, as established by Rao et al. (2009), we aimed to synthesize shRNAs from the provided siRNA sense strands, the design of shRNAs was done in accordance with the guidelines suggested by ThermoFisher (SiRNA Design Guidelines | Technical Bulletin #506). All shRNA variants can be found in Table 1.

Based on the design guidelines, we followed these recommendations:

  1. The whole DWV genome was used as the reference sequence to search for potential siRNA sequences.
  2. All our designed siRNAs started with AA dinucleotide at their 5’-end since it was reported by Elbashir et al., 2001 as the most efficient option.
  3. Since some siRNA may target highly structured or protein-associated regions of DWV mRNA, we designed ten different siRNA variants whose target sequences are located throughout the DWV genome.
  4. In our shRNA construct the UUCAAGAGA loop sequence was used.

Table 1. shRNA designed through GenScript’s siRNA tool and ThermoFisher’s siRNA Design Guidelines. Column ‘Start’ notes the position of siRNA target site in DWV genome.

shRNA No. Sequence sense (obtained from GenScript) + LOOP + antisense Start GC% Scores ΔE/Thermodynamic

All our designed siRNA were cloned into high copy number yeast vectors under control of pGAL1 promoter (Fig. 2).

Figure 2. shRNA expression construct.

To simplify the assessment of our designed shRNA efficiency, we engineered a sensor where a specific segment of the DWV genome is placed in 3’-UTR of Green Fluorescent Protein (GFP), between the GFP STOP codon and tCYC1 terminator. As upon RNAi, the target mRNA that directs the protein translation undergoes degradation, the efficiency of RNA silencing can be estimated by measuring GFP fluorescence signal in these cells.

Figure 3. siRNA activity reporter expression construct.

Build: Assembling the expression cassettes and engineering the yeast strains

We employed molecular cloning techniques, including ligation, restriction, bacterial and yeast transformation, to construct our entities.

Our first objective was to create an expression cassette that incorporates both AGO1 and DCR1 genes. To achieve this, we synthesized these genes as artificial DNA fragments. Subsequently, we inserted these fragments into the pUC19 vector (Fig. 4, pUC19 cloning), which later served as the PCR templates. During the PCR amplification process, BsmBI and BsaI enzyme recognition sites, and compatible overhangs required for the Mo-Clo pYTK kit were incorporated into both the AGO1 and DCR1 sequences (Lee et al., 2015). The PCR fragments were cloned into pYTK001 entry vectors using Golden Gate assembly with BsmBI restriction enzyme, resulting in plasmid vectors categorized as type 3 parts. During the assembly, our PCR fragments replaced BsmBI-flanked GFP dropout cassette. After transformation of bacterial cells with the Golden Gate reaction mix, color coding simplified the selection of colonies carrying correctly assembled plasmids: green colonies contained the initial pYTK001 vector, while in white colonies the GFP was replaced with our inserts

Figure 4. Scheme illustrating AGO1 and DCR1 cloning into pUC19 vector followed by the generation of MoClo kit custom type 3 parts.

At the next level assembly with BsaI enzyme, our custom type 3 parts containing AGO1 and DCR1 (pYTK001-Ago1 and pYTK001-Dcr1) were used to create plasmid vectors carrying complete transcriptional units. AGO1 and DCR1 genes were placed under the control of strong constitutive promoters, pPGK1 and pTEF1, respectively. In addition, the destination vector needed for the final multigene assembly was generated. The complete scheme and full list of parts used for the assembly of transcriptional unit plasmids and the destination vector are represented in Fig. 5.

Figure 5. Scheme illustrating Golden Gate assemblies of the AGO1 and DCR1 transcriptional unit plasmids, the LEU2 integration vector and final multigene cassette.

BsmBI restriction enzyme was used for the final assembly of multigene plasmid, carrying both AGO1 and DCR1 transcriptional units along with LEU2 selection marker gene, and 5’- and 3’-leu2 homologous arms for yeast transformation (Fig. 5, BsmBI assembly).

Moving forward, our focus shifted to the creation of the GFP reporter construct. For this we used a pRS304-based vector that harbored EGFP coding sequence under the control of pGAL1 promoter and tCYC1 terminator. Utilizing iPCR, a fusion was established, linking the viral target region to the 3’-UTR of the EGFP gene (Fig. 6).

Figure 6. Cloning scheme for the siRNA reporter plasmid.

The final construct we aimed to generate was the shRNA expression vector. As the first step, the sense-loop-antisense sequences of the shRNAs were ordered as separate oligonucleotides for each strand and were annealed before the Golden Gate assembly (Fig. 7A). Each oligonucleotide included an overhang from its 5'-end to ensure correct assembly into the pYTK001 entry vector using the BsmBI restriction enzyme. The shRNA sequence replaced the GFP expression cassette in pYTK001, and after assembly, it fell into the part 3 category. Upon transformation with the Golden Gate reaction mixture, bacterial colonies carrying correctly assembled pYTK001-shRNAs displayed a white coloration, in contrast to green colonies carrying initial pYTK001 entry vectors. In expression vectors, shRNA were cloned under the control of an inducible pGAL1 promoter in 2μ-based high copy number plasmid. The expression vectors were assembled with the BsaI restriction enzyme. The complete scheme and the list of Mo-Clo plasmid parts used for the assembly of shRNA expression vectors are presented in Fig. 7B.

Figure 7. Overview of shRNA expression vector cloning. (A) First, the DNA oligonucleotides for the shRNA are annealed and ligated to pYTK001. (B) Then, BsaI is used to assemble the indicated fragments to form the shRNA expression plasmid.

All constructed plasmid vectors are listed in Table 2.

Table 2. Constructed plasmids

Name Backbone/Plasmids used for GG assembly Content Description
p184 Golden Gate assembly LEU2 integration cassette Plasmid for assembly of Ago1 + Dcr1 multi gene cassette
p196 pRS304 pGAL1_EGFP_tCYC1 Backbone plasmid for shRNA sensor construction
p198 pRS304 pGAL1_EGFP*siRNAv1_tCYC1 Plasmids containing GFP sensor fused to siRNA target
p199 pRS304 pGAL1_EGFP*siRNAv2_tCYC1
p200 pRS304 pGAL1_EGFP*siRNAv3_tCYC1
p201 pRS304 pGAL1_EGFP*siRNAv4_tCYC1
p202 pRS304 pGAL1_EGFP*siRNAv5_tCYC1
p203 pRS304 pGAL1_EGFP*siRNAv6_tCYC1
p204 pRS304 pGAL1_EGFP*siRNAv7_tCYC1
p205 pRS304 pGAL1_EGFP*siRNAv8_tCYC1
p206 pRS304 pGAL1_EGFP*siRNAv9_tCYC1
p207 pRS304 pGAL1_EGFP*siRNAv10_tCYC1
p209 Golden Gate assembly pPGK1-AGO1-tPGK1 Plasmid containing Ago1 transcriptional unit
p213 Golden Gate assembly shRNA_v1 - shRNA expression vectors shRNA expression vectors
p214 Golden Gate assembly shRNA_v2
p215 Golden Gate assembly shRNA_v3
p216 Golden Gate assembly shRNA_v4
p217 Golden Gate assembly shRNA_v5
p218 Golden Gate assembly shRNA_v6
p219 Golden Gate assembly shRNA_v7
p220 Golden Gate assembly shRNA_v8
p221 Golden Gate assembly shRNA_v9
p222 Golden Gate assembly shRNA_v10
p223 Golden Gate assembly pTEF1-DCR1-tPGK1 Plasmid containing Ago1 transcriptional unit
p224 Golden Gate assembly pPGK1-AGO1-tPGK1 + pTEF1-DCR1-tPGK1 Plasmid containing Ago1 and Dcr1 transcriptional units

Yeast strains were engineered by transforming parent strains with constructed plasmid vectors. All yeast strains constructed and used in the study are listed in Table 3.

Table 3. Yeast strains used in the study.

Strain name Genotype Description
DOM90 w303 MATa {leu2-3,112 trp1-1 can1-100 ura3-1 ade2-1 his3-11,15 bar1::hisG} [phi+] Background strain
I63 DOM90 Leu2::Ago1+Dcr1 Strain expressing Ago1 and Dcr1. It was used to transform with vectors expressing shRNA and target sequences
I66 DOM90 Leu2::Ago1+Dcr1 Trp1::GFP-Target_v3 Strain expressing Ago1, Dcr1, and GFP_V3 target
I70 DOM90 Leu2::Ago1+Dcr1 Trp1::GFP-Target_v7 Strain expressing Ago1, Dcr1, and GFP_V7 target
I73 DOM90 Leu2::Ago1+Dcr1 Trp1::GFP-Target_v10 Strain expressing Ago1, Dcr1, and GFP_V10 target
I76 DOM90 Leu2::Ago1+Dcr1 Trp1::GFP-Target_v3 URA3::shRNA V3 Strain expressing shRNA_V3 and its GFP_V3 target
I80 DOM90 Leu2::Ago1+Dcr1 Trp1::GFP-Target_v7 URA3::shRNA V7> Strain expressing shRNA_V7 and its GFP_V7 target
I83 DOM90 Leu2::Ago1+Dcr1 Trp1::GFP-Target_v10 URA3::shRNA V10 Strain expressing of shRNA_V10 and its GFP_V10 target
I88 DOM90 Leu2::Ago1+Dcr1 Trp1::GFP-Target_v7 URA3::shRNA V3 Strain expressing of shRNA_V3 and GFP_V7 target
I89 DOM90 Trp1::GFP URA3::siRNA anti-GFP Strain expressing of anti-GFP siRNA and GFP but lacking Ago1 and Dcr1

TEST the Effect of siRNA on GFP Expression

To assess the efficiency of siRNAs we designed a sensor that consisted of GFP fused with the viral target sequence for the siRNA. If the siRNA is active and efficient, the mRNA will be degraded leading to no or decreased GFP fluorescence signal compared to cells without siRNA treatment.

Flow cytometry reveals suppression of EGFP expression by siRNA induction in yeast

Flow cytometry offers a means for efficient and precise evaluation of GFP expression at the individual cell level. In our experimental setup, we cultivated genetically modified yeast strains under tightly regulated environmental conditions. Subsequently, upon initiation of siRNA and GFP production, we subjected each yeast cell culture to flow cytometry. This method enables accurate measurement of any changes in GFP signal. A reduction in GFP fluorescence signifies the efficacy of the siRNA.

We used flow cytometry to measure the GFP fluorescence intensities in yeast cultures 24h after inducing the expression of shRNA and the GFP-target sequence reporter. In the absence of shRNA, the GFP-reporter-expressing cultures showed much higher GFP signal in comparison to background fluorescence of non-induced cultures, confirming sufficient expression of the reporter protein (Fig. 8). It is noteworthy that fluorescence intensities among different GFP-target constructs significantly varied, suggesting that the viral sequence introduced into the 3'-UTR of the transcript may influence mRNA stability (not shown). Reduced mRNA stability, in turn, leads to impaired translation and decreased GFP fluorescence. For this reason, to compare the impact of shRNAs on GFP reporter expression, we normalized the GFP fluorescence for each strain with shRNA to the data obtained for its parent strain without shRNA expression. Out of seven tested shRNAs, four showed statistically significant decrease in fluorescence levels, as determined by Two-sample Two-tailed Student’s T-test with unequal variances (p-value at least less than 0.05) (Fig. 8).

As a control to verify the functionality and specificity of our system, we employed strains expressing the GFP_V7 target alone with shRNA_V3. As anticipated, inducing the expression of GFP and shRNA in this strain did not result in any substantial change in GFP fluorescence. Furthermore, we employed a strain lacking Ago1 and Dcr1 but containing the shRNA with the corresponding GFP reporter. No suppression of GFP was observed in this strain, verifying that Ago1 and Dcr1 from S. castelli are essential for RNA interference in S. cerevisiae and that the suppression of GFP expression arises from RNAi.

Figure 8. Expressing shRNAs in the engineered RNAi-capable yeast enables testing of siRNA activities. Plot showing the mean GFP fluorescence intensities of a population of cells expressing the indicated shRNA and the GFP reporters, measured by flow cytometry 24h after induction. The GFP fluorescence data presented was normalized to cells expressing the GFP reporter, but not shRNAs. The mean with standard deviation from 3 biological replicates for each shRNA variant is shown. Error bars indicate standard deviation. Statistical analysis (Two-sample Two-tailed Student’s T-test with unequal variances) was performed. Not Significant (NS) - p-value > 0.05; * - p-value ≤ 0.05; ** - p-value ≤ 0.01; *** - p-value ≤ 0.001.


These experiments lead to three important conclusions. Firstly, the presented work confirms the previously published results, demonstrating that introducing Dicer and Argonaute proteins to S. cerevisiae is sufficient to reconstitute RNAi response. Secondly, the experiments show that when viral sequences are introduced into the 3’-UTR of a GFP-coding transcript, they can be targeted by shRNAs in yeast, offering a straightforward means to assess the efficacy of various shRNAs. Thirdly, we observed partial suppression of GFP-reporter expression with tested anti-DWV siRNAs, making them strong candidates for further testing in bees to investigate possible off-target effects.

Additionally, we also found that the presence of the viral target sequence can influence GFP reporter expression, even in the absence of shRNAs. Although all tested reporters showed sufficiently high GFP expression to allow measuring of RNAi effects, further optimization of the length of the viral sequence and the exact positioning of that sequence should be investigated.

In the current experimental setup, both shRNA and the GFP reporter are expressed from pGAL1 promoter, as was done in Drinnenberg et al. 2009. However, it would be more beneficial to use different inducible promoters for shRNA and GFP to enable the flexibility of expressing either one within the same strain. The current single-promoter design necessitates the use of different strains to measure the reporter fluorescence signal in the absence and presence of shRNA. Utilizing distinct promoters would enable the use of a single strain in which reporter expression can be independently induced from shRNA expression.

In our strains shRNAs are expressed from a high copy number 2μ-plasmid, ensuring extremely high expression levels. In future experiments, we could consider expressing shRNAs from a CEN plasmid, which exists in lower numbers within yeast cells, thereby resulting in reduced shRNA expression. Such an adjustment could potentially enhance the method's sensitivity to quantitative differences in shRNA efficiency.