Here we describe a protocol for employing high-throughput RNAi screening to uncover host targets that can be manipulated to enhance oncolytic virus therapy, specifically rhabodvirus and vaccinia virus therapy, but it can be readily adapted to other oncolytic virus platforms or for discovering host genes that modulate virus replication generally.
High-throughput genome-wide RNAi (RNA interference) screening technology has been widely used for discovering host factors that impact virus replication. Here we present the application of this technology to uncovering host targets that specifically modulate the replication of Maraba virus, an oncolytic rhabdovirus, and vaccinia virus with the goal of enhancing therapy. While the protocol has been tested for use with oncolytic Maraba virus and oncolytic vaccinia virus, this approach is applicable to other oncolytic viruses and can also be utilized for identifying host targets that modulate virus replication in mammalian cells in general. This protocol describes the development and validation of an assay for high-throughput RNAi screening in mammalian cells, the key considerations and preparation steps important for conducting a primary high-throughput RNAi screen, and a step-by-step guide for conducting a primary high-throughput RNAi screen; in addition, it broadly outlines the methods for conducting secondary screen validation and tertiary validation studies. The benefit of high-throughput RNAi screening is that it allows one to catalogue, in an extensive and unbiased fashion, host factors that modulate any aspect of virus replication for which one can develop an in vitro assay such as infectivity, burst size, and cytotoxicity. It has the power to uncover biotherapeutic targets unforeseen based on current knowledge.
High-throughput genome-wide RNAi screening has proven to be an invaluable tool for revealing biological insights in diverse areas of study including virus biology. Over the past decade or so, numerous groups have undertaken cell-based large-scale RNAi screening to extensively identify host genes that modulate virus replication (Reviewed in1,2,3,4,5,6,7). Interestingly, a large number of these screens have been conducted in cancer cells and several with viruses that are current oncolytic virus candidates including vaccinia virus8,9,10, Myxoma virus11, herpes simplex virus12, vesicular stomatitis virus13,14, and Maraba virus15. While the focus of the majority of these studies was on identifying host factors that influence some aspect of virus replication and not on identifying biotherapeutic targets for enhancing oncolytic virus therapy, valuable data could be mined from these data sets in this regard. It is clear that the powerful potential to identify host targets that can be manipulated to enhance oncolytic virus therapy has not been fully exploited.
A plethora of oncolytic virus platforms are currently being tested in preclinical and clinical studies including herpes simplex virus, reovirus and vaccinia virus, which have had demonstrable success in late-stage clinical trials. For the majority of these platforms, efforts have been directed toward altering the virus genomes to enhance therapy. For instance, to increase tumor specificity, specific virus genes have been deleted or mutated to significantly restrict viral replication in normal cells, but not in tumor cells. In some cases, transgenes have been added like GM-CSF (granulocyte macrophage colony-stimulating factor) to potentiate the immune response or NIS (sodium iodide symporter) to enable in vivo imaging and cellular radioiodide uptake for therapeutic purposes. However, there has been very little work focused on manipulating host/tumor genes and exploring the impact of host/tumor genes on oncolytic virus replication. With the advent of high-throughput screening technology, it is now possible to probe host-virus interactions on a genome scale and to decipher opportunities to manipulate the host genome to enhance oncolytic virus therapy (Reviewed in16,17,18).
We reported the first study demonstrating proof-of-concept for using high-throughput genetic screening to identify host targets that could be manipulated to enhance oncolytic virus therapy. To discover host genes that modulate Maraba virus oncolysis, an oncolytic rhabdovirus currently being tested in phase I and II trials, we conducted genome-wide RNAi screens across three different tumor cell lines in duplicate with an siRNA (short interfering RNA) library targeting 18,120 genes15. Pathway analysis of hits identified in the screens revealed enrichment of members of the ER (endoplasmic reticulum) stress response pathways. We selected 10 hits within these pathways for secondary validation using siRNA with different targeting sequences from those of the primary screen. As part of tertiary validation, we conducted a rescue experiment with IRE1α (inositol-requiring enzyme 1 alpha) to confirm that the hit was on target. In vitro and in vivo testing using a small molecule inhibitor of IRE1α resulted in dramatically enhanced oncolytic efficacy.
Workenhe et al.12 also conducted a genome-wide RNAi screen using a pooled lentiviral shRNA (short hairpin RNA) library targeting 16,056 genes to identify host factors restricting the human herpes simplex virus type 1 (HSV-1) mutant KM100-mediated oncolysis of breast cancer cells. In the primary screen, they identified 343 genes the knockdown of which lead to enhanced cytotoxicity of KM100-infected cells over mock-infected cells; they selected 24 of these genes for secondary validation. Out of the 24 genes, 8 genes were confirmed in the secondary screen with one of them being SRSF2 (serine/arginine-rich splicing factor 2) which was subsequently confirmed using a genetic rescue experiment during tertiary validation. The group went on to identify a DNA topoisomerase inhibitor chemotherapeutic that reduced the phosphorylation of SRSF2 and showed that the combination treatment of HSV-1 KM100 with this inhibitor lead to extended survival of TUBO tumor-bearing mice.
In the aforementioned studies, a similar workflow was followed. Each began with a primary genome-wide RNAi screen followed by a secondary screen on a select number of hits identified in the primary screen. This was followed by tertiary validation studies, which included confirming that the hit was on-target by performing genetic rescue experiments in vitro with ultimate validation performed by manipulating the target gene product in vivo. In this article, we first provide a general protocol for the development and validation of a high-throughput siRNA-screening assay, which involves determining the optimal transfection conditions, amount of virus, and length of infection. We also offer a detailed protocol for conducting a primary a high-throughput siRNA screen as well as an overall description of the method for performing secondary screen validation and tertiary validation.
1. Development and validation of a high-throughput siRNA screening assay
NOTE: For all experiments, use Hepes-buffered growth media appropriate for each cell line. See Figure 1 for an overview of the workflow from assay optimization to tertiary validation.
2. Conducting the primary high throughput screen
NOTE: Prior to commencing the RNAi screen, microtiter plates (e.g. 384-well plates) need to be arrayed with the siRNA library and screening controls. The following steps are best conducted by two persons with one person (i.e. Person A) being primarily responsible for dispensing the transfection reagent and cells across the plates, and the second person (i.e. Person B) being responsible for centrifuging the plates immediately prior to transfection and for preparing the cells. Included in parentheses are some specifics for processing one batch of 33 plates (i.e. half of the library) with the 786-O cell line.
3. Secondary screen validation of hits with the TRC (The RNAi Consortium) shRNA library
NOTE: See Discussion for details on selecting hits for secondary validation and possible screening technologies. If siRNA technology is selected for secondary validation, the same assay development and validation protocol can be used as was used for the primary screen (see 1).
4. Tertiary validation
NOTE: The purpose of tertiary validation is to primarily confirm that the hit is on target, tumor specific, and enhances efficacy in vivo. One can also test whether it modulates spread across a spectrum of tumor cell lines
An overview of the workflow for identifying host targets to enhance oncolytic virus therapy is presented in Figure 1.
As illustrated in Figure 1, the critical first step to conducting a high-throughput RNAi screen is assay development. Figure 2A provides a sample plate layout for the transfection optimization assay. Figure 2B consists of representative images of expected results, and Figure 2C is a representative plot of the expected results. Following siRNA knockdown of PLK-1, a gene that induces cell death and can be used as a positive control to monitor siRNA knockdown efficiency, one would expect the cell survival to be between 0-30%, unless the cell line has a long doubling time in which case it will take longer to observe the cell death phenotype. The non-targeting siRNA should also be minimally toxic to the cells with a similar relative survival to that of untreated cells.
Another key element of assay development is to determine the MOI at which to infect the cells and the length of time for infection. Figure 3A provides a sample plate layout for determining the amount of virus and the length of incubation with virus. Figure 3B depicts the results of a live time-course of 786-O cells infected with vvDD-eGFP (oncolytic vaccinia virus expressing enhanced GFP with the Thymidine Kinase and Vaccinia Growth Factor genes deleted) at various MOIs. Representative images of cells infected with vvDD-eGFP at an MOI of 0.05 are shown in Figure 3C. Based on the results plotted in Figure 3B, one might select an MOI of 0.05 and a length of infection of 21 h as this would allow for the detection of increases and decreases in spread as this time-point is within a linear range, and the estimated Z-factor at this time-point is 0.6 for cohorts that increase and for cohorts that decrease spread. As part of validating this assay, one will want to fix the cells at this MOI and time-point, and calculate the Z-factor.
Following the protocol described, we conducted genome-wide RNAi screens across three different tumor cell lines (e.g. OVCAR-8, U373, NCI-H226) in duplicate with an siRNA library targeting 18,120 genes15. Using the MAD method, we identified 1008 hits common to at least 2 out of the 3 cancer cell lines (Figure 4A). Pathway analysis of hits identified in the screens revealed enrichment of members of the UPR (unfolded protein response) and ERAD (Endoplasmic-reticulum-associated protein degradation) pathways (Figure 4A). We selected 10 hits within these pathways for secondary validation using siRNA with different targeting sequences from those of the primary screen (Figure 4B). As part of tertiary validation, we conducted a genetic rescue experiment with IRE1α and ATF6α (activating transcription factor 6 alpha) to confirm that these hits were on target (Figure 4C).
Figure 1: Schematic of workflow for identifying host targets to enhance oncolytic virus therapy. The first step is to develop and validate an assay for high-throughput RNAi screening, which includes optimizing transfection conditions for introducing siRNA into cells, determining the optimal amount of virus (i.e. MOI) to infect the cells with and the length of incubation with virus. The second step is to conduct a high-throughput genome-wide screen. An siRNA library targeting genes across the entire genome is arrayed on microtiter plates. Cells are then reverse transfected with the siRNA library. Following a 48-72 h incubation period to allow for gene silencing, cells are infected with the optimal amount of virus. After the optimal length of incubation with virus, cells are fixed and stained, and subsequently imaged with a high-content microscope. Subsequent image and data analysis will help to identify genes that significantly modulate virus replication, which are referred to as "hits." A secondary validation screen is performed on select hits from the primary screen generally using siRNA or shRNA with different seed regions. Tertiary validation experiments are performed to confirm that the hits are on-target, tumor specific and enhance efficacy in vivo. Please click here to view a larger version of this figure.
Figure 2: Optimization of transfection conditions for high-throughput RNAi screening assay. (A) A sample plate layout for optimizing transfection conditions with two different cell lines. (B) Representative images of 786-O cells (1500 cells/well) stained with Hoechst 33342 following 72 h incubation with transfection reagent diluent alone (i.e. untreated), with transfection reagent and diluent (i.e. mock transfected), with non-targeting siRNA and with siRNA targeting PLK-1. (C) Representative results from transfection optimization experiment showing 22% survival of cells transfected with PLK-1 siRNA (n=24) and 94% survival of cells transfected with non-targeting siRNA (n=8). Error bars = ± SD, standard deviation; scale bars = 200 μm. Please click here to view a larger version of this figure.
Figure 3: Determination of the amount of virus and length of incubation with virus for high-throughput RNAi screening assay. (A) A sample plate layout for optimizing the amount of virus and length of incubation with virus. Columns 1 and 24 contain transfection controls and are left uninfected. Columns 2 to 23 contain untreated cells, mock transfected cells and cells transfected with positive and negative virus controls all infected with a range of MOIs starting with no virus in columns 2 and 3. (B) 786-O cells were reverse transfected with non-targeting siRNA and incubated for 48 h. They were subsequently infected with vvDD-GFP at the MOIs indicated and imaged at 8 h intervals starting at 5 hours post-infection (hpi). The mean GFP area per well (± SD) was calculated for each time-point (n=12) and plotted on the graph. The Z-factor calculated between points A and B is 0.6 and the Z-factor calculated between points B and C is 0.6. (C) Representative live images of a well infected with an MOI of 0.05 at the time-points indicated. Magnification, 10x; 9 fields/well; scale bars = 200 μm. Please click here to view a larger version of this figure.
Figure 4: Genome-wide Screen Identifies ER Stress Response Blockade as a Potent Sensitizer of Rhabdovirus-Mediated Oncolysis. (A) Venn diagram outlining the number of overlapping hits of high-throughput RNAi screens in 3 tumor cell lines, and a table (+, synthetic lethal; -, no interaction) and schematic diagram (hits outlined in red) illustrating key hits within the UPR and ERAD pathways. (B) EC50 shifts (black bars, left y axis) were determined for U373 cells treated with Maraba virus (48 h) following treatment with siRNA (72 h) targeting a series of UPR/ERAD hits from the screen. Relative mRNA expression for each gene following siRNA knockdown (72 h) is depicted in white (right y axis). (C) Cell viability assays were conducted 48 h after Maraba virus infection, in U373 cells ectopically expressing mouse ATF6α (or control GFP) ± siRNA targeting human ATF6α (or non-targeting [NT] control; left panel) or human XBP1(s) (or control) ± siRNA targeting human IRE1α (or control; right panel). Western blots demonstrating gene silencing and ectopic gene expression are shown. EC50 shifts = the shift in the dose of virus required to kill 50% of the cells; Error bars = ±SD; XBP1(s) = X-box binding protein 1 (spliced); In (B), one-way ANOVAs were performed followed by a Bonferroni multiple comparison's post hoc test to derive p values; In (C), student's t-tests were performed to derive p values; *p < 0.05; #p < 0.05. (This figure has been modified from Mahoney et al., 201115). Please click here to view a larger version of this figure.
Here we present a protocol for employing high-throughput RNAi screening to identify host targets that can be manipulated to enhance oncolytic virus therapy. This has been tested successfully with oncolytic Maraba virus and oncolytic vaccinia virus, but, as noted, it can be adapted for use with other oncolytic viruses or other viruses generally to identify host genes that modulate virus replication. The screening protocol is also designed to identify genes that increase or decrease spread, but it can be readily modified for other readouts. Our screens with Maraba virus, for instance, were designed to identify genes modulating oncolysis utilizing a simple resazurin-based vital dye assay to score cell viability (see Figure 4)15. In these screens, following incubation with virus, instead of fixing cells with formaldehyde and staining with Hoechst, we dispensed resazurin dye into each well (final concentration of 20 μg/mL) and, after a 6 h incubation, we measured absorbance (573 nm) with a plate reader. We could have also used cell number based on Hoechst staining as a readout. While using a surrogate reporter to monitor virus replication was unnecessary in this screen, a virus with a reporter protein, particular a fluorescent protein does facilitate screen optimization as one can, for example, monitor virus replication live; furthermore, a virus with a reporter protein is less cumbersome and costly than using antibodies to stain for viral antigens, but it is possible to screen without one. In addition, one could make even further modifications to the assay to investigate host factors that impact on infectivity, burst size, and even more detailed sub-phenotypes like immunological cell death. In each case, one would follow a similar assay development protocol, screening strategy and secondary and tertiary validation methods.
In the protocol, we suggest optimizing transfection conditions ideally with multiple tumor cell lines. There are at least two reasons for optimizing with multiple cell lines. One reason is that not every cell line is amenable to high-throughput screening as some may be difficult to transfect or may not adhere well, but a key reason is to enable screening with multiple cell lines in order to avoid cell intrinsic bias. The choice of cell lines is largely dependent on one's objectives. One may elect to focus on a particular cancer type or select disparate cancer types to cover a broader spectrum. Alternatively, one may wish to focus on tumor cell lines that are resistant to identify host factors that can be manipulated to make the tumor cell lines less resistant to oncolytic virus infection.
In addition to the selection of cells lines, the selection of positive and negative virus controls is an important consideration for any screen. In some instances, virus genes that are required for or restrict replication may not be known or, if they are known, they may be cell-specific. Consequently, one can still determine the approximate robustness and dynamic range of the assay by using surrogate controls. For example, one could use uninfected cells or cells infected with a low MOI as a surrogate positive control for identifying genes that decrease spread upon knockdown. For identifying genes that enhance spread upon knockdown, one could infect cells with a dilution series of virus, and use wells with the maximum signal as a surrogate positive control.
Selection of hits for secondary screen validation and the type of technology to employ is another matter that requires careful deliberation. Candidates are generally selected for secondary screen validation based on the magnitude of the hit, on whether they significantly modulate spread in multiple cancer cell lines (if primary screens were conducted in multiple cell lines) and on biological interest. Bioinformatics tools such as DAVID22,23, PANTHER24, STRING25 and PINdb26 can help to isolate pathways and hits of biological interest. Mercer et al.8 describe the use of open-source image analysis software and have also made available an algorithm that they developed for the visualization and analysis of hits. One factor to take into account when interpreting results is that gene silencing may have different impacts depending on the stage in the virus life cycle that is being investigated. For example, it is possible that knockdown of a gene early in the life cycle may inhibit virus replication, while knockdown at a later stage may increase replication. Often, secondary screens are conducted with siRNA from a different vendor with different seed regions with multiple siRNAs per gene. However, complementary technologies targeting candidate genes can be employed such as CRISPR-Cas9 or lentivirus vectors expressing shRNA.
The method of analysis is also a critical consideration. We suggest here using the robust Z-score or MAD20,27 with suggested cut-offs for hit calling of >2 or <-2. The cut-offs can be increased or decreased depending on whether the focus is on eliminating more false positives or capturing more false negatives. For instance, in a recent high-throughput screen with vaccinia virus9, the lower cut-off was set at -1.5 (no upper cut-offs were described as the focus was on identifying genes that decreased spread upon knockdown). There are also other methods that can be employed including the z-score, SSMD (strictly standardized mean difference), and the B score20,28,29,30. Barrows et al.31 compared hit lists generated by sum rank, MAD, z-score and SSMD and found each method of analysis resulted in different hit lists. In toto, there is no perfect method or cut-off. Ultimately, secondary screen validation and tertiary validation studies will confirm true hits.
The method of filtering out cytotoxic hits must also be considered. One way is to conduct primary screens in parallel plus or minus virus. This permits the detection of siRNAs that are cytotoxic on their own. This was our approach in our Maraba virus screens15; however, this is not often practical. Perhaps a more practical method, in addition to being robust, is to ascertain hits that are cytotoxic as part of secondary screen validation, especially if one's focus is to identify hits that decrease replication upon knockdown. For instance, in a whole genome primary screen with Myxoma virus, Teferi et al.11 identified 1,588 siRNAs that decreased virus replication upon knockdown. To filter out cytotoxic siRNAs, they conducted a secondary cytotoxicity screen with these siRNAs; they measured cell viability 72 h post-transfection using a resazurin-based cell viability assay. Cytotoxic siRNAs were identified based on a Z-score of ≤ -1.96. Another approach is to simply filter out cytotoxic siRNAs from the primary screening data based on a reduction of cell number by a certain percentage, for example, by a reduction > 50% as in Sivan et al.9, or based on a particular score as in Lee et al.14; in their study of host factors required for vesicular stomatitis virus replication, they excluded siRNA hits that reduced cell viability by > 3.0 standard deviations from the mean. The mean number of cells was determined based on the Hoechst staining of cell nuclei and, although not explicitly stated, the mean was seemingly calculated on a per plate basis as it is clear that the mean GFP signal was calculated on a per plate basis.
A limitation of any high-throughput RNAi screen is the frequent high number of false positives and negatives, which may come as a consequence of the assay design, the technology used or through systematic errors. For example, a protein may significantly affect the replication of the virus, but the time chosen for gene silencing may be too short given the protein's half-life to detect its effect leading to a false negative by virtue of the assay design. It is well established that incomplete knockdown and the off-target effects of siRNA technology can also introduce false positives and negatives. Systematic errors, such as the edge effect32, can generate misleading results as well. Here the signal in the outer wells of the plate may be systematically higher or lower than the rest of the plate due to evaporation. There are strategies to address some of these issues including, for instance: decreasing the concentration of siRNA to reduce off-target effects33, reconfiguring plate layouts to fill the outer wells with media to mitigate the edge effect, or employing computational methods that have been developed to detect and counter systematic errors such as the edge effect32,34,35. A general method to deal with false positives is to conduct a secondary screen following the primary screen. As a way of dealing with false negatives, one may relax the cut-off for hit calling, and include potential false negatives for secondary screening. This, of course, will not capture false negatives that are well below the cut-off. Primary screens should also be conducted in duplicate or triplicate to limit spurious results. Ultimately, no matter what strategies are employed, no high-throughput screen will exhaustively identify all true hits, but it can provide a treasure trove of data from which one is likely to mine some valuable hits that can be capitalized upon.
In this protocol, we described the use of arrayed siRNA and shRNA technology, though another alternative is the use of pooled shRNA and CRISPR-Cas9 libraries. A pooled shRNA library was employed in the Workenhe et al.12 study outlined in the introduction. Pooled CRISPR-Cas9 technology has been used to conduct genome-scale screens to identify host factors required for replication of viruses such as Zika virus36, West Nile virus37,38, dengue virus39 and hepatitis C virus39. The advantage of using pooled libraries is that they are less expensive to purchase, do not require specialized infrastructure (e.g. liquid handling devices, high-content microscopes, and robotic equipment), nor do they have the high costs of operating and maintaining this infrastructure, and they require fewer consumables. The disadvantage is that the types of readouts are more limited. In most if not all pooled library host-virus interaction screens to date, the readout is cell viability or lack thereof; one cannot readily probe for more complex phenotypes. The choice of format depends on the biological question being asked.
The advancements in genetic screening technology over the past several decades that first enabled screens in bacteria and yeast to now present-day genome-scale screens in human cells have opened the door wide for discovery in diverse fields of study. These genetic screens have not only allowed us to answer fundamental biology questions, but have given us keys to unlocking insights into developing and improving biotherapeutics. While there are still lessons to be learned and challenges to be overcome with this technology, the results to date have been exciting. The discoveries will likely be even greater as novel screening technology including CRISPR-Cas9 libraries, microfluidics and in vivo screening technologies continue to mature.
The authors have nothing to disclose.
This work was supported by grants from the Ontario Institute for Cancer Research, the Canada Foundation for Innovation, the Ottawa Regional Cancer Foundation, and the Terry Fox Research Institute. K.J.A. was supported by a Vanier Canada Graduate Scholarship, a Canadian Institute for Health Research-Master's Award, and an Ontario Graduate Scholarship.
Consumables | |||
siRNA library | GE Dharmacon | G-005005-02 | |
Corning 384-well plate | Corning | 3985 | |
Falcon 384-well plate | Becton Dickinson | 353962 | |
Water | Sigma | W4502 | |
siGenome SMARTpool PLK1 | GE Dharmacon | M-003290-01 | |
AllStars Negative Control siRNA | Qiagen | SI03650318 | |
siGenome Non-targeting siRNA Pool #2 | GE Dharmacon | D-001206-14-20 | |
DMEM/HIGH Glucose | GE Healthcare Life Sciences | SH30022.01 | |
Fetal Bovine Serum | Sigma | F15051-500ML | |
HEPES solution (1M) | Sigma | H3535-100ML | |
Penicillin-Streptomycin 100X solution | GE Healthcare Life Sciences | SV30010 | |
Trypsin-EDTA (0.25%) | Thermo Fisher Scientific | 2500056 | |
DPBS/Modified | GE Healthcare Life Sciences | SH30028.02 | |
Lipofectamine RNAiMAX | Thermo Fisher Scientific | 13778100 | |
Oligofectamine | Thermo Fisher Scientific | 1225011 | |
Opti-MEM | Thermo Fisher Scientific | 22600-050 | |
Resazurin sodium salt | Sigma | R7017-5G | |
Hoechst 33342 | Thermo Fisher Scientific | H21492 | |
Formaldehyde (37% by weight) | Fisher Scientific | F79-4 | |
500 ml bottles | Corning | 430282 | |
Adhesive Sealing Film For Microplates | Excel Scientific | 361006007 | |
Peelable Heat Sealing Foil | Thermo Fisher Scientific | AB-3720 | |
BioTek MicroFlo Select Dispenser cassette | Fisher Scientific | 11-120-625 | |
Name | Company | Catalog Number | Comments |
Equipment | |||
MicroFlo Select (liquid dispenser) | Fisher Scientific | 11120621 | |
BioTek Synergy HT plate reader | Biotek | N/A | |
Opera Imaging and Analysis Instrument | PerkinElmer | HH10000115 | |
KiNEDx Robotic Arm | Peak Analysis and Automation (formerly Peak Robotics) | KX-300-660 | |
Cytomat 24C Series Automated Incubator | Thermo Fisher Scientific | 50080227 | |
JANUS Automated Workstation | PerkinElmer | AJI4M01 | |
Plate Carousel | Peak Analysis and Automation (formerly Peak Robotics) | N/A | |
Alps 3000 plate sealer | Thermo Fisher Scientific | AB-3000 |