The Transgenic RNAi Project at Harvard Medical School: Resources and Validation

  1. Lizabeth A. Perkins*,
  2. Laura Holderbaum*,
  3. Rong Tao*,
  4. Yanhui Hu*,
  5. Richelle Sopko*,
  6. Kim McCall†,
  7. Donghui Yang-Zhou*,
  8. Ian Flockhart*,
  9. Richard Binari*,‡,
  10. Hye-Seok Shim*,
  11. Audrey Miller*,
  12. Amy Housden*,
  13. Marianna Foos*,
  14. Sakara Randkelv*,
  15. Colleen Kelley*,
  16. Pema Namgyal*,
  17. Christians Villalta*,‡,
  18. Lu-Ping Liu*,§,
  19. Xia Jiang§,
  20. Qiao Huan-Huan§,
  21. Xia Wang§,
  22. Asao Fujiyama**,
  23. Atsushi Toyoda**,
  24. Kathleen Ayers††,
  25. Allison Blum‡,‡‡,
  26. Benjamin Czech§§,
  27. Ralph Neumuller*,
  28. Dong Yan*,
  29. Amanda Cavallaro‡,***,
  30. Karen Hibbard‡,***,
  31. Don Hall‡,***,
  32. Lynn Cooley††,
  33. Gregory J. Hannon§§,
  34. Ruth Lehmann‡,‡‡,
  35. Annette Parks†††,
  36. Stephanie E. Mohr*,
  37. Ryu Ueda**,
  38. Shu Kondo*,‡‡‡,
  39. Jian-Quan Ni*,§ and
  40. Norbert Perrimon*,‡,1
  1. *Department of Genetics, Harvard Medical School, Boston, Massachusetts 02115
  2. †Boston University, Boston, Massachusetts 02215
  3. ‡Howard Hughes Medical Institute, Boston, Massachusetts 02115
  4. §TsingHua Fly Center, Beijing, 100084, China
  5. **Comparative Genomics Laboratory, National Institute of Genetics, Shizuoka 411-8540, Japan
  6. ‡‡Skirball Institute, Department of Cell Biology, New York University School of Medicine, New York, New York 10016
  7. §§CRUK Cambridge Institute, University of Cambridge, Cambridge, CB2 1TN, United Kingdom
  8. ***Janelia Farm Research Institute ,Asburn, Virginia, 20147
  9. ††Department of Genetics, Yale University, New Haven, Connecticut 06510
  10. †††Bloomington Drosophila Stock Center Bloomington, Indiana, 47405
  11. ‡‡‡Invertebrate Genetics Laboratory, National Institute of Genetics, Shizuoka 411-8540, Japan
  1. 1Corresponding author: Department of Genetics, Harvard Medical School, 77 Ave. Louis Pasteur, Boston, MA 02115. E-mail: perrimon{at}receptor.med.harvard.edu

Abstract

To facilitate large-scale functional studies in Drosophila, the Drosophila Transgenic RNAi Project (TRiP) at Harvard Medical School (HMS) was established along with several goals: developing efficient vectors for RNAi that work in all tissues, generating a genome-scale collection of RNAi stocks with input from the community, distributing the lines as they are generated through existing stock centers, validating as many lines as possible using RT–qPCR and phenotypic analyses, and developing tools and web resources for identifying RNAi lines and retrieving existing information on their quality. With these goals in mind, here we describe in detail the various tools we developed and the status of the collection, which is currently composed of 11,491 lines and covering 71% of Drosophila genes. Data on the characterization of the lines either by RT–qPCR or phenotype is available on a dedicated website, the RNAi Stock Validation and Phenotypes Project (RSVP, http://www.flyrnai.org/RSVP.html), and stocks are available from three stock centers, the Bloomington Drosophila Stock Center (United States), National Institute of Genetics (Japan), and TsingHua Fly Center (China).

  • RNAi
  • Drosophila
  • screens
  • phenotypes
  • functional genomics

A striking finding from the genomic revolution and whole-genome sequencing is the amount of information missing on gene function. Although Drosophila is arguably the best-understood multicellular organism and a proven model system for human diseases, mutations mapped to specific genes with readily detectable phenotypes have been isolated for ∼15% of the >13919 annotated fly coding genes (http://flybase.org/; FlyBase R6.06). The lack of information on the majority of genes (the “phenotype gap”) suggests that researchers have been unable to either assay their roles experimentally and/or resolve issues of functional redundancy. In addition, some phenotypes may be only detected on specific diets and environments. Further, our understanding of the function of many genes for which we have some information is limited by pleiotropy, whereby an earlier function of the gene prevents analysis of later functions.

The availability of in vivo RNAi has revolutionized the ability of Drosophila researchers to disrupt the activity of single genes with spatial and temporal resolution (Dietzl et al. 2007; see review by Perrimon et al. 2010), and thus address the phenotype gap. Motivated by the power of the approach and the needs of the community, three large-scale efforts, the Vienna Drosophila RNAi Center (VDRC, http://stockcenter.vdrc.at/control/main), the National Institute of Genetics (NIG, http://www.shigen.nig.ac.jp/fly/nigfly/index.jsp), and the Drosophila Transgenic RNAi Project (TRiP) at Harvard Medical School (HMS) (http://www.flyrnai.org/TRiP-HOME.html) have over the years generated large numbers of RNAi lines that aim to cover all Drosophila genes. These resources are proving invaluable to address a myriad of questions in various biological and biomedical fields including but not limited to research in the areas of cell biology, signal transduction, cancer, neurodegeneration, metabolism, and behavior.

The TRiP was initiated in 2008 with the goals of developing efficient vectors for RNAi, generating and distributing a genome-scale collection, and providing validation of the fly stocks. Below, we describe details on our TRiP production pipeline (Figure 1), reagents generated, state of the collection, and validation efforts as of May 2015. Note that the project is still ongoing. Regular updates are provided on the TRiP website as well as at FlyBase and stock center websites.

Figure 1
Figure 1

The TRiP Platform: production pipeline flowchart. Details are included in the text.

Materials and Methods

Vector construction

Knockdown vectors:

Construction of the first-generation TRiP knockdown vectors, pVALIUM1 and pVALIUM10, are described in Ni et al. (2008) and Ni et al. (2009), respectively. The second-generation TRiP knockdown vectors, pVALIUM20 and pVALIUM22, are described in Ni et al. (2011). To construct pWALIUM10, pVALIUM10 was first cut by HindIII to remove the vermilion gene and then annealed DNA oligos carrying an AscI cutting site (forward primer, 5′-AGCTTCACGACCTGAGGCGCGCCA-3′, reverse primer, 5′-AGCTTGGCGCGCCTCAGGTCGTGA-3′) were inserted into the linearized vector. The mini-white gene was cut from pUAST using AscI and was then cloned into the AscI site. The orientation of the mini-white gene was confirmed by restriction enzyme digestion and DNA sequencing, and the resulting vector with the correct orientation was named pWALIUM10. pWALIUM20 and pWALIUM22 were constructed in a similar way based on pVALIUM20 and pVALIUM22, respectively, using the same DNA oligos.

Overexpression vectors:

To construct pVALIUM10-roe, pVALIUM10 was cut by XbaI to remove a fragment containing the ftz intron and attR. The resulting vector was named as pVALIUM10-roe. pVALIUM10-roe was further cut by EcoRI and XbaI, and then an annealed DNA product (forward primer, 5′-AATTCGCAGATCTCCATATGAGCTAGCTACTAGTGTC-3′, reverse primer, 5′-CTAGACACTAGTAGCTAGCTCATATGGAGATCTGCG-3′) was inserted to generate pVALIUM10-moe. pWALIUM10-roe and pWALIUM-moe were constructed in the same way as described above, using pWALIUM10 and the same DNA oligos.

Construction of a QUAS vector for RNAi:

From the TRiP RNAi vector, WALIUM20, the two copies of loxp-5xUAS sequences were deleted and replaced with the 94-bp QUAS sequence (five copies of the QF binding site). Briefly, the QUAS fragment was obtained by PCR from the pQUAST vector (Potter et al. 2010). The linearized WALIUM20 backbone was generated by PCR from the WALIUM20 vector to exclude the two copies of loxp-5xUAS. Then the QUAS fragment was inserted into the linearized WALIUM20 backbone using an in-fusion reaction (Clontech).

All of the TRiP vectors described above are available from the PlasmID repository of the Dana Farber/Harvard Cancer Center (DF/HCC) DNA Resource Core at Harvard Medical School (http://plasmid.med.harvard.edu/PLASMID/).

Generation of Transgenic RNAi lines

dsRNA lines:

Long dsRNA hairpins were cloned into VALIUM series vectors (VALIUM1 or VALIUM10) and injected into embryos for targeted phiC31-mediated integration at genomic attP landing sites on the second (attP40) or third (attP2) chromosomes as described by Ni et al. (2008) and Ni et al. (2009). All transgenic lines were sequenced to confirm the identity of the dsRNA hairpin.

shRNA lines:

shRNAs (21 bp) were cloned into VALIUM series vectors (VALIUM20, VALIUM21, or VALIUM22) and injected into embryos for targeted phiC31-mediated integration as described by Ni et al. (2011). All transgenic lines were sequenced to confirm the identity of the shRNA and miR-1 scaffold.

RNA Isolation, Reverse Transcription, and Real-Time qPCR

Detailed protocols are provided in the Extended Materials and Methods of Sopko et al. (2014). Briefly, RNA was isolated by guanidinium thiocyanate-phenol-chloroform extraction using TRIzol (Life Technologies) and glass-bead-based cell disruption. Genomic DNA was eliminated by incubation with DNase (QIAGEN), and samples were processed for cleanup with an RNeasy MinElute Cleanup Kit (QIAGEN). One microgram of purified RNA was incubated with a mix of oligo(dT) and random hexamer primers and with iScript RT (iScript cDNA Synthesis Kit, Bio-Rad) for complementary DNA (cDNA) synthesis. cDNA was used as the template for amplification, using validated primers in iQ SYBR Green Supermix with a CFX96 real-time PCR detection system (Bio-Rad). Query gene expression was relative to a control sample, normalized to the expression of three reference genes: ribosomal protein L32, α-tubulin, and either nuclear fallout or Gapdh1, using the ΔΔCT analysis method.

Rescue constructs

Fosmid rescue constructs:

D. persimilis clones were obtained from the DGRC and then retrofitted for site-specific insertion into D. melanogaster as described in detail in Kondo et al. (2009). Fosmids appropriate for testing rescue of specific genes were identified using the RNAi Rescue online tool (http://www.flyrnai.org/cgi-bin/RNAi_find_rescue_compl.pl) (Kondo, et al. 2009).

C911 constructs:

Fifteen efficient shRNAs were chosen for generation of C911 versions: aurora (aur, FBgn0000147), Tao-1 (FBgn0031030), wee (FBgn0011737), grapes (grp, FBgn0261278), PAK-kinase (Pak, FBgn0014001), Sak kinase (SAK, FBgn0026371), Cyclin-dependent kinase 8 (Cdk8, FBgn0015618), loki (lok, FBgn0019686), hopscotch (hop, FBgn0004864), dropout (dop, FBgn0036511), gilgamesh (gish, FBgn0250823), and slipper (slpr, FBgn0030018). Twenty-one-basepair sequences identical to the original targeting shRNA but with complementary nucleotides at positions 9–11 were synthesized, cloned into pVALIUM20 or pVALIUM22 (Supporting Information, Table S3), and injected into embryos for targeted phiC31-mediated integration as described by Ni et al. (2011). Injection was at the same attP site as the original efficient targeting shRNA. All transgenic lines were sequenced.

Data availability

All TRiP stocks are available at the BDSC (LINK), NIG (LINK) and THFC (LINK). All vectors are available from the DF/HCC DNA Resource Core PlasmID repository at Harvard Medical School. The order page for the TRiP vectors can be found at http://plasmid.med.harvard.edu/PLASMID/GetVectorsByType.do?type=drosophila%20in%20vitro%20and%20in%20vivo%20expression. Maps with attributes, complete sequences, and detailed cloning protocols can be found on the TRiP website (http://www.flyrnai.org/TRiP-HOME.html) under Reagents, Maps, and Protocols.

Results and Discussion

The TRiP Vectors

Vectors for RNAi:

Over the years, the TRiP has generated a series of 22 knockdown vectors (Vermilion-AttB-Loxp-Intron-UAS-MCS; VALIUM), to facilitate the incorporation of RNAi hairpins into attP landing sites (Ni et al. 2008, 2009, 2011; Table 1; Figure S1). In Table 1 and Figure S1 we list the most commonly used vectors as well as those not previously reported. All VALIUM vectors contain a wild-type copy of vermilion as a selectable marker and an attB sequence to allow for phiC31 targeted integration at genomic attP landing sites (Groth et al. 2004). vermilion was chosen rather than mini-white as the proper gene dosage of white has been found to be important in behavioral studies (An et al. 2000). The VALIUM vectors were also designed with two pentamers of upstream activation sequence (UAS) sequences, one of which can be removed using the Cre/loxP system. Thus, based on findings that UAS sites promote transcription in an additive fashion, if Gal4 is not limiting, the modular number of UAS copies allows for generation of a phenotypic series. Moreover, because the attP chromosome is usually homozygous viable, it is possible to generate 5X, 10X, 15X, and 20XUAS combinations (Ni et al. 2008). In addition to manipulating the number of UAS sequences, the level of RNAi knockdown can also be altered by using Gal4 lines of various strengths, rearing flies at different temperatures, or via coexpression of UAS-Dicer2 (Dietzl et al. 2007). Note that Dicer2 (Dcr2) is only effective with VALIUM1 and VALIUM10-series RNAi fly stocks, as Dcr2 processes long double-strand RNAs (dsRNAs) but not the short hairpin RNAs (shRNAs) used with other TRiP vectors (see below).

View this table:
Table 1 Summary of RNAi and overexpression vectors from the TRiP

The first-generation knockdown vectors chosen by the TRiP for RNAi stock production were VALIUM1 and VALIUM10 (Table 1 and Figure S1). Both allow expression of long dsRNA hairpins, usually between 400 and 600 bp. These are very effective for RNAi in somatic tissues but are not as effective in the female germline (Ni et al. 2008, 2009). Subsequently, we showed that shRNAs containing a 21-bp targeting sequence embedded into a micro-RNA (miR-1) backbone are very effective for gene knockdown in both the germline and soma (Ni et al. 2011). For shRNA expression we developed the second-generation knockdown vectors, VALIUM20, VALIUM21, and VALIUM22 (Table 1 and Figure S1)(Ni et al. 2011). All subsequent TRiP lines were generated with shRNAs in VALIUM20 (for knockdown in germline or soma) or VALIUM22 (germline only).

Since some researchers prefer to use mini-white as the selectable marker for transgenesis, we also generated new versions of the VALIUM vectors in which vermilion is replaced with white (WALIUM10, WALIUM20, and WALIUM22; Table 1 and Figure S1). Except for the selectable marker, the WALIUM vectors have the same attributes as their vermilion containing counterparts, VALIUM10, VALIUM20, and VALIUM22, respectively. We, and others, have used these WALIUM vectors and found that they function as well as the VALIUM vectors for transgenic RNAi.

A QUAS vector for RNAi:

The Q system (Potter et al. 2010) provides an alternative to the GAL4/UAS system. The Q system is particularly valuable when the expression of two different genes needs to be targeted to two different cell types in the same fly, which is made possible by combining the Q and Gal4/UAS systems. For example, to express geneX in the germline and geneY in follicle cells, flies of genotype germline-Gal4, UAS-geneX; follicle cell-QF, QUAS-geneY, can be generated. To allow such applications, we built and tested a QUAS vector. The new vector, pW20-QUAST, is effective for knockdown, as shown for the genes white and draper (Figure 2), and should be effective when combined with Gal4/UAS.

Figure 2
Figure 2

The pW20-QUAST vector for RNAi. (A) Drosophila eyes from flies expressing QUAST-white-RNAi (QUAST-white-i) (top) or QUAST-draper-i (bottom) with the tubulin-QF driver or TM6B controls. Eye pigmentation is reduced in flies expressing white-i but not draper-i. The difference in eye pigmentation between QUAST-draper-i/+; tubulin-QF/+ and QUAST-draper-i/+; TM6B/+ flies is due to the number of transgenes. (B) Stage 14 egg chambers from flies expressing QUAST-white-i (left) or QUAST-draper-i (right) with the tubulin-QF driver. Nurse cells have been cleared normally in the white-i egg chamber, but nurse cells persist (arrow) in the draper-i egg chamber. (C) Map of the QUAS vector.

Vectors for overexpression:

For many genes, overexpression phenotypes can provide valuable information on gene function and provide tools for epistasis experiments and genetic screens. After using TRiP RNAi lines for gene knockdown, members of the fly community asked the TRiP to develop a vector that allowed them to overexpress genes using a comparable strategy. Thus, we generated vermilion and mini-white versions of VALIUM10 for overexpression experiments (Table 1 and Figure S1). Specifically, we generated pVALIUM10-roe, pVALIUM10-moe, pWALIUM10-roe, and pWALIUM10-moe. These vectors allow cloning for overexpression by recombinational cloning ("roe” versions for recombination overexpression) or into a multicloning site ("moe” for multicloning overexpression). Note: that all of the TRiP vectors described above are available from the DF/HCC Plasmid Resource Core in Boston (http://plasmid.med.harvard.edu/PLASMID/).

The TRiP collection

Generation of the collection:

All fly stocks generated at the TRiP are inserted into one of two attP sites, attP40 on the left arm of the second chromosome at 25C6 or attP2 on the left arm of the third chromosome at 68A4. These sites were selected for their abilities to provide high levels of induced expression of the transgenes, yet maintain low basal expression when the transgenes are not induced (Markstein et al. 2008). The landing site chosen by the TRiP for hairpin insertion is guided first by the preference of the community member nominating the gene and second by the TRiP. If a TRiP stock for a particular gene is available in one location a second TRiP stock for the same gene will be generated in the second location. In May 2015, of the 11,491 total TRiP stocks, 5232 (45%) are in attP40 (II) and 6259 (54%) are in attP2 (III).

We used both single-construct cloning and pooled library approaches to generate the TRiP plasmid constructs. Either dsRNA or shRNA constructs were generated individually in 96-well plates, or, in the case of shRNAs, they were selected from the shRNA libraries generated in VALIUM20 and VALIUM22 starting from a pool of 83,256 unique shRNA oligonucleotides synthesized on glass slide microarrays (Ni et al. 2011). For the TRiP constructs that were constructed individually in 96-well plates, we used two different approaches to generate transformed fly stocks. Early in the project, constructs were injected individually into attP40- or attP2-bearing lines and transformants were recovered. As only one attB insert can integrate by phiC31-mediated recombination into an attP site (Groth et al. 2004), we later injected pools of constructs, established transgenic lines, and then subsequently characterized the inserted DNA by sequencing. This approach proved to be extremely efficient and quickly became the method of choice (Figure 3). Typically 20–25 constructs per pool were injected, although in some cases more complex pools from the VALIUM20 and VALIUM22 libraries were injected (see above).

Figure 3
Figure 3

Pooled injection results. Examples of the frequencies of independent shRNA transgenic lines recovered from six different pools of individual constructs. Details from single G0’s from Pools 3 and 5 are shown. Note that seven different shRNA transformants were recovered from a single G0 from pool 3 that contains 15 different constructs.

From the shRNA libraries, we initially injected pools derived from the VALIUM20 library, established transgenic lines, and then determined the sequence of the inserted vector for each line. However, we found that approximately one-third of the lines carried either an empty vector or a vector with an incorrect shRNA sequence, which likely reflected the error frequency of the original library. Thus, we decided to first sequence individual clones and pool only those with a correct sequence. We retransformed the library into bacteria and sequenced 31,953 clones. Of those, 19,239 had a correct shRNA sequence, representing 8694 unique clones targeting 4856 different genes. To generate fly stocks from these “quality selected” libraries, we pooled and then injected 3903 clones targeting genes for which no VALIUM20 stocks had been made.

State of the collection:

To date the TRiP has generated 11,491 stocks and has ∼3386 in final production stages, and ∼53 additional genes have been nominated by the Drosophila community or are part of the Gene Groups and Hu-Dis Projects described below (Figure 4 and Figure 5). Altogether, the collection covers 9803 unique FBgns or 71% of the genes in the fly genome (Flybase Release 6.05) with 81% of highly conserved genes represented as determined using the DRSC Integrative Ortholog Prediction Tool (DIOPT), (Hu et al. 2011).

Figure 4
Figure 4

Summary of available TRiP fly stocks. (A) The percentage of RNAi fly stocks available in each vector is indicated. (B) Main features of the vectors and details on the stocks available: (*) Still in production, accepting nominations; (1) JF: generated at Janelia Farm; (2, 3, and 7) HMS: generated at the TRiP at Harvard Medical School; (4) HMC: generated at the TRiP (HMS) in collaboration with the TsingHua Fly Center (THFC), China; (5) HMJ: generated at the TRiP in collaboration with the National Institute of Genetics (NIG), Japan; (6) Generated at the TRiP, the TRiP and THFC, or the TRiP and NIG.

Figure 5
Figure 5

TRiP fly stocks corresponding to specific gene categories. (A) TRiP coverage of specific gene groups. (B) TRiP coverage of human disease orthologs. List of genes in each category is available at http://www.flyrnai.org/glad as well as the TRiP website at http://www.flyrnai.org/TRiP-HOME.html. The number of genes in each category is indicated next to the name category. Note that the lists on GLAD are updated periodically so the numbers may not match perfectly.

Approximately 2445 TRiP stocks generated with long dsRNA hairpins were inserted into first-generation VALIUM vectors, VALIUM1 and VALIUM10 (Figure 4). Primers for the TRiP double-strand hairpins were designed using the amplicon design tool SnapDragon at the Drosophila RNAi Screening Center (DRSC; http://www.flyrnai.org/DRSC-HOME.html) (Flockhart et al. 2012). When possible, sequences common to all splice forms of the gene of interest were chosen as long as they did not include 19-bp matches to other sequences in the genome. Specific and detailed descriptions of the generation of long dsRNA hairpins can be found in Ni et al. (2008, 2009).

All remaining TRiP stocks (>9000) were generated with shRNAs inserted into the second-generation VALIUM vectors, VALIUM20, VALIUM21, and VALIUM22 (Figure 4). Our experience and elsewhere has shown that, in general, shRNAs are more effective than long dsRNAs (Ni et al. 2011). For design of the shRNAs, sequences common to all splice forms of a gene are identified, and any subsequences 16 bp or longer that match other genes are not considered in the selection of the 21-bp targeting sequence. Each subsequence is given a score (calculated as described by Vert et al. 2006) and the highest scoring sequences are selected. Top- and bottom-strand oligos are designed using an automated Perl program developed internally, and the hairpins were cloned into the vector (see Ni et al. 2011).

Approximately 4595 genes have been nominated by the Drosophila community and over the years >260 different investigators nominated lists of genes. In addition, we prioritized specific categories of genes to help researchers perform targeted screens. We defined lists of 23 major gene categories (kinases, transcription factors, secreted proteins, etc.; 14 categories are shown in Figure 5A) and generated shRNA lines to provide comprehensive sets of RNAi stocks. We have generated a web resource, named GLAD (for Gene List Annotation at the DRSC; http://www.flyrnai.org/glad) where lists of major gene categories, as well as subcategories, can be found and downloaded (Hu et al. 2015). Of special interest, we assembled a TRiP collection representing Drosophila orthologs of genes associated with human diseases (Human Disease TRiP Project, HuDis-TRiP, http://www.flyrnai.org/HuDis). There are currently TRiP fly stocks in the HuDis-TRiP collection for 1575 Drosophila orthologs of human disease-associated genes (Figure 5B). These include 85% coverage for 670 high-confidence Drosophila orthologs of high-confidence disease-associated human genes.

The “TRiP Toolbox”:

In addition to the TRiP RNAi lines, the TRiP provides a set of TRiP Toolbox stocks, including injection stocks for labs wishing to generate their own RNAi lines (e.g., isoform-specific lines or other custom lines that would not be appropriate for the TRiP community nomination and production pipeline), as well as commonly used Gal4 lines combined in most cases with UAS-Dcr2 to enhance knockdown using long dsRNAs (i.e., in VALIUM1 or VALIUM10). A list of available lines is available at http://www.flyrnai.org/TRiP-TBX.html and in Table S1, as well as at stock center websites (e.g., http://flystocks.bio.indiana.edu/Browse/TRiPtb.htm).

Distribution of TRiP fly stocks

All completed stocks are annotated on the TRiP website and on FlyBase and made immediately available to the research community through the Bloomington Drosophila Stock Center (BDSC), the NIG (http://www.shigen.nig.ac.jp/fly/nigfly/), and the TsingHua Fly Center (http://center.biomed.tsinghua.edu.cn/public/eq-category/all/modelanimalfacility). Currently, the BDSC is distributing 10,434 RNAi stocks, the NIG has 5961, and the THFC has 8591. As of May 2015, 286,658 TRiP RNAi stocks have been distributed from the BDSC to 1271 different user groups at 640 different organizations in 39 countries. A total of 1347 TRiP stocks have been distributed from the NIG, Japan, and 34,000 from the THFC, China.

Validation of the TRiP lines

RNAi Stock Validation and Phenotypes Project:

Challenges with RNAi include evaluation of the efficiency (level of knockdown) and specificity (potential off-target effects, OTEs). To facilitate the selection of the best available RNAi fly stocks, we initiated the RSVP to evaluate the performance of existing TRiP fly stocks. We performed validation experiments (phenotype analyses and RT–qPCR) for a number of lines, collected available published information, and established a web resource (http://www.flyrnai.org/RSVP.html) to provide the community with all available information relevant to performance of the lines. Importantly, on the RSVP web pages, community members have the opportunity to share their own data relevant to TRiP stocks (Figure 6).

Figure 6
Figure 6

The validation of TRiP lines. (A) Knockdown efficiency of shRNAs analyzed from 0- to 4-hr embryos derived from maternal-Gal4, shRNA females. (B) Percentage of shRNA lines generating >50% knockdown and their targeting regions (CDS, coding sequence; 5′ and 3′ UTR). (C) Snapshot of RSVP webpage. Example RSVP “details page” layout for a TRiP line targeting Notch (Line ID: HMS00009). “Validation Test Results” are curated from publications, TRiP data, online community input, or personal communications to the TRiP. Bottom, text boxes for further input by the community. At RSVP there is a details page for each TRiP line generated, as well as for RNAi fly stocks in other major public collections.

To date we have performed RT–qPCR validation analysis for >500 TRiP fly stocks. Most RT–qPCR analyses were performed in early embryos from MTD-Gal4; UAS-shRNA animals as described in Sopko et al. (2014). On average, 65% of TRiP stocks display knockdown efficiencies of >50% (Figure 6A). To facilitate searching for primers appropriate for RT–qPCR analysis, we assembled FlyPrimerBank (http://www.flyrnai.org/FlyPrimerBank) (Hu et al. 2013b). Relevant to long dsRNA reagents, the tool indicates if a given primer pair should be avoided as the primers are predicted to amplify the dsRNA itself.

RSVP also collects information relevant to the phenotypes observed with specific Gal4 lines. In particular, results from various large-scale screens have been included. Among these are germline and maternal effect screens using the MTD–Gal4 driver (Staller et al. 2013; Sopko et al. 2014; Yan et al. 2014), a muscle screen using the muscle Dmef2–Gal4 line (Perrimon and Randkelv, unpublished results), a screen in gut stem cells using the esg-Gal4 (Zeng et al. 2015), and a screen of maternally expressed genes involved in embryonic patterning (Liu and Lasko 2015). As of May 2015, and specific for the TRiP stocks, the RSVP contains 8334 data entries for 5202 TRiP stocks representing 3735 fly genes. This information is particularly helpful for selecting lines that have strong knockdown phenotypes; moreover, if multiple RNAi lines targeting the same gene have the same phenotypes with similar Gal4 lines, then it suggests that these RNAi reagents are likely on target. Finally, this information will help cull the lines that are not optimal for knockdown efficiencies or that are associated with OTEs.

Interestingly, based on RT–qPCR and phenotypic analyses, not all regions of a gene appear optimal for hairpin design. Specifically, shRNAs designed to target 5′ or 3′ UTRs of genes produce effective knockdown in ∼60 or ∼30% of fly stocks, respectively. By contrast, up to 85% of TRiP lines with shRNAs designed to the coding region of a gene produce effective RNA knockdown as determined by RT–qPCR and phenotypic analyses (Figure 6B).

Most recently, the RSVP has incorporated from FlyBase data entries for RNAi lines generated by the NIG and the VDRC making it possible for researchers to choose stocks from among the various RNAi collections. In total the RSVP now also contains 23,451 data entries for 17,782 RNAi lines representing 11,346 genes.

Rescuing with fosmids:

A critical issue when a phenotype is observed with a specific RNAi line is to evaluate whether the phenotype reflects knockdown of the intended target gene or results from an OTE. The OTE issue and various approaches to addressing the problem have been discussed at length (e.g., see review by Perrimon et al. 2010). To test for OTEs, the gold standard method, although relatively cumbersome, is a rescue approach, such as rescue using genomic DNA from a different species (Kondo et al. 2009; Langer et al. 2010). An alternative is to rescue using an ORF from a UAS construct, although interpretation of the result using this approach could be complicated if overexpression of the ORF alone in a wild-type background generates a phenotype. To date we have generated 38 fosmid lines with DNA from D. persimilis (Table S2). We selected D. persimilis as we have at the DRSC access to a fosmid library of 70,000 clones, and this species is closely related to D. pseudoobscura, which we have shown previously to be ideal for such rescue experiments (Kondo et al. 2009).

By combining shRNA and fosmid-bearing chromosomes we tested whether providing a D. persimilis ortholog could rescue the presumed on-target gene and were able to rescue shRNA-induced phenotypes about two-thirds of the time. This moderate failure rate may be attributed to the DNA excision method used for fosmid clone library generation, which may inadvertently truncate gene promoter and termination sequences. Additionally, the large blocks of cloned D. persimilis DNA may no longer be regulated like native chromatin and consequently this might affect transcription factor accessibility and general transcription.

Addressing OTEs with C911 constructs:

An alternative method for assessment of OTEs is to examine C911 versions of the targeting shRNA (Buehler et al. 2012). C911 versions are near identical shRNAs but with mismatches at positions 9–11. The mismatches are predicted to disrupt on-target binding but preserve seed sequence-mediated OTEs, since the anti-sense and sense seed sequences remain intact. We generated C911 versions of 15 shRNAs for which expression in the germline results in defective hatching (see Table S3). As expected for on-target shRNA constructs, C911 mutations restored hatching to wild-type rates for nearly all shRNAs. The one exception was for a construct targeting drop out (dop), which only partially rescued the phenotype (Figure 7A). We further verified by RT–qPCR that mutation of these three residues eliminates knockdown of on-target transcript for 13 targets. For the exceptions, dop and loki, the C911 construct only partially restored mRNA to wild-type levels (Figure 7B). These data suggest that C911 constructs may serve as more suitable controls as compared with common negative control constructs, e.g., constructs expressing eGFP or shRNAs targeting white.

Figure 7
Figure 7

Addressing OTEs with C911 shRNAs. (A) Hatch rate examination of efficient shRNAs and their C911 counterparts. (B) Target expression levels for original shRNA and their C911 counterparts, relative to an EGFP-targeting shRNA.

Online search of TRiP stocks and information

We have assembled a number of resources for online search of existing RNAi lines and to view corresponding information, e.g., on efficiency (see TRiP website http://www.flyrnai.org/TRiP-HOME.html and Table 2). We provide two online tools for direct search and view of RNAi fly stock information (UP-TORR and RSVP), we provide links between these tools and from other results pages to these tools, and we provide organized groups of TRiP stocks, e.g., based on gene function. UP-TORR (http://www.flyrnai.org/up-torr/) allows the search of TRiP and other RNAi fly stocks based on up-to-date gene annotation information (Hu et al. 2013a) and links to RSVP. As described above, RSVP (http://www.flyrnai.org/RSVP.html) allows users to search and view information about knockdown efficiency (RT–qPCR data) and phenotypes (in text format and when available, supplemented with images) for specific RNAi fly stock/Gal4 driver combinations. Moreover, RSVP includes results curated by FlyBase for other major stock collections, such as phenotypes associated with VDRC fly stocks.

View this table:
Table 2 Online resources for search and view of TRiP stocks and corresponding information

To start an RNAi fly stock search with a list of orthologs from other species or with a disease term, results pages from the ortholog search tool DIOPT and the disease-gene ortholog search tool DIOPT-DIST (Hu et al. 2011) include a link that allows the user to carry results to UP-TORR. In addition, a list of TRiP stocks corresponding to ∼670 high-confidence fly orthologs of high-confidence human disease-associated genes (i.e., the HuDis-TRiP stocks; see above) is available for view as a table including BDSC stock IDs and links to RSVP, or as a download (URL at Table 2). Moreover, the TRiP home page provides links that prepopulate the UP-TORR search page with gene groups (e.g., kinases) to facilitate identification of stocks for focused studies (URL at Table 2).

Concluding remarks

Transgenic RNAi fly stocks have become essential tools in the Drosophila molecular genetic toolbox, as illustrated by the number of requests for RNAi fly stocks from stock centers, and the observation that a very large proportion of current Drosophila publications include studies using one or more RNAi fly stocks. On the production side, our own and other efforts are approaching coverage of all Drosophila genes (in particular, coverage of all community-nominated genes). Thus, production efforts might shift to specific needs such as production of replacements for ineffective shRNAs and production of fly stocks targeting specific gene isoforms. Moreover, now that such a large number of RNAi fly stocks are available and being tested, there is an increasing need for curation of the information regarding performance of the fly stocks and inclusion of information in a centralized database such as RSVP. Because researchers often know that a given RNAi fly stock does or does not perform in a manner consistent with expectation (e.g., based on mutant phenotype data) well in advance of publication, the most timely approach is for researchers to provide community feedback on RNAi efficiency independent of publication of research results. More than ever, community feedback is needed to annotate the various RNAi collections.

Acknowledgments

G.J.H., R.L., and N.P. are investigators of the HHMI. The Transgenic RNAi Project (TRiP) at Harvard Medical School (HMS) thanks Aram Comjean for website oversight and Linda Jiang for stock maintenance in the early stages of the TRiP. We thank Kevin Moses and Chonnettia Jones and the HHMI Visitor’s Research Program for support of the work performed at Janelia Farm. We also thank Monti Mercer, Megan Hong, Grace Zheng, Dona Fetter and Jessica Ketting at Janelia Farm for helping to process the transformant lines. R.U. and S.K. thanks Satoru Yoshida and Fumiwo Ejima for their help with DNA sequencing at NIG. We also thank Kathy Matthews and Kevin Cook for their advice, support, and enthusiasm while making the TRiP stocks available to the fly community. Finally, the TRiP at HMS thanks the Drosophila community for their nominations, encouragement, and collegiality. Work at the TRiP and Drosophila RNAi Screening Center is supported by National Institutes of Health (NIH) R01GM084947 (N.P.), R01GM067761 (N.P.), and R24RR032668 (N.P.). Work in Japan is supported by the National Institute of Genetics. Work at the Tsinghua Fly Center is supported by the National Key Technology Research and Development Program of the Ministry of Science and Technology of the People’s Republic of China (no. 2015BAI09B03), the National Basic Research Program (973 Program) (no. 2013CB35102), and the National Natural Science Foundation of China (no. 31371496). Work at Boston University is supported by NIH R01GM060574 (K.M.) and RO1GM094452 (K.M.) and we thank Jon Iker Etchegaray who suggested the idea of the QUAS-RNAi vector. Work at Yale University is supported by NIH RO1GM043301 (L.C.). Work at CSHL was supported by NIH R37GM062534 (G.J.H.). Work at New York University School of Medicine Medical was supported by HHMI and an administrative supplement (American Recovery and Reinvestment Act) to R01HD14900-09 (R.L.). Work at the Bloomington Drosophila Stock Center is supported by the Office of The Director, National Institutes of Health under Award no. P40OD018537.

Footnotes

  • Received July 8, 2015.
  • Accepted August 24, 2015.

Literature Cited

    1. An X.,
    2. Armstrong J. D.,
    3. Kaiser K.,
    4. O’Dell K. M.
    , 2000 The effects of ectopic white and transformer expression on Drosophila courtship behavior. J. Neurogenet. 14: 227–243, 271.
    1. Buehler E.,
    2. Chen Y. C.,
    3. Martin S.
    , 2012 C911: a bench-level control for sequence specific siRNA off-target effects. PLoS One 7: e51942.
    1. Dietzl G.,
    2. Chen D.,
    3. Schnorrer F.,
    4. Su K. C.,
    5. Barinova Y.,
    6. et al.
    , 2007 A genome-wide transgenic RNAi library for conditional gene inactivation in Drosophila. Nature 448: 151–156.
    1. Flockhart I. T.,
    2. Booker M.,
    3. Hu Y.,
    4. McElvany B.,
    5. Gilly Q.,
    6. et al.
    , 2012 FlyRNAi.org: the database of the Drosophila RNAi screening center: 2012 update. Nucleic Acids Res. 40: D715–D719.
    1. Groth A. C.,
    2. Fish M.,
    3. Nusse R.,
    4. Calos M. P.
    , 2004 Construction of transgenic Drosophila by using the site-specific integrase from phage phiC31. Genetics 166: 1775–1782.
    1. Hu Y.,
    2. Flockhart I.,
    3. Vinayagam A.,
    4. Bergwitz C.,
    5. Berger B.,
    6. et al.
    , 2011 An integrative approach to ortholog prediction for disease-focused and other functional studies. BMC Bioinformatics 12: 357.
    1. Hu Y.,
    2. Roesel C.,
    3. Flockhart I.,
    4. Perkins L.,
    5. Perrimon N.,
    6. Mohr S.E.
    2013a UP-TORR:online tool for accurate and up-to-date annotation of RNAi reagents. Genetics 195: 37–45.
    1. Hu Y.,
    2. Sopko R.,
    3. Foos M.,
    4. Kelley C.,
    5. Flockhart I.,
    6. et al.
    , 2013b FlyPrimerBank: an online database for Drosophila melanogaster gene expression analysis and knockdown evaluation of RNAi reagents. G3 Genes Genomes Genet. 3: 1607–1616.
    1. Hu Y.,
    2. Comjean A.,
    3. Perkins L. A.,
    4. Perrimon N.,
    5. Mohr S. E.
    , 2015 GLAD: an Online Database of Gene List Annotation for Drosophila. J Genomics 3: 75–81.
    1. Kondo S.,
    2. Booker M.,
    3. Perrimon N.
    , 2009 Cross-species RNAi rescue platform in Drosophila melanogaster. Genetics 183: 1165–1173.
    1. Langer C. C.,
    2. Ejsmont R. K.,
    3. Schonbauer C.,
    4. Schnorrer F.,
    5. Tomancak P.
    , 2010 In vivo RNAi rescue in Drosophila melanogaster with genomic transgenes from Drosophila pseudoobscura. PLoS One 5: e8928.
    1. Liu N.,
    2. Lasko P.
    , 2015 Analysis of RNA interference lines identifies new functions of maternally-expressed genes involved in embryonic patterning in Drosophila melanogaster. G3 (Bethesda). 5: 1025–34.
    1. Markstein M.,
    2. Pitsouli C.,
    3. Villalta C.,
    4. Celniker S. E.,
    5. Perrimon N.
    , 2008 Exploiting position effects and the gypsy retrovirus insulator to engineer precisely expressed transgenes. Nat. Genet. 40: 476–483.
    1. Ni J. Q.,
    2. Markstein M.,
    3. Binari R.,
    4. Pfeiffer B.,
    5. Liu L. P.,
    6. et al.
    , 2008 Vector and parameters for targeted transgenic RNA interference in Drosophila melanogaster. Nat. Methods 5: 49–51.
    1. Ni J. Q.,
    2. Liu L. P.,
    3. Binari R.,
    4. Hardy R.,
    5. Shim H. S.,
    6. et al.
    , 2009 A Drosophila resource of transgenic RNAi lines for neurogenetics. Genetics 182: 1089–1100.
    1. Ni J. Q.,
    2. Zhou R.,
    3. Czech B.,
    4. Liu L. P.,
    5. Holderbaum L.,
    6. et al.
    , 2011 A genome-scale shRNA resource for transgenic RNAi in Drosophila. Nat. Methods. 8(5): 405–7.
    1. Perrimon N.,
    2. Ni J. Q.,
    3. Perkins L.
    , 2010 In vivo RNAi: today and tomorrow. Cold Spring Harb. Perspect. Biol. 2: a003640.
    1. Potter C. J.,
    2. Tasic B.,
    3. Russler E. V.,
    4. Liang L.,
    5. Luo L.
    , 2010 The Q system: a repressible binary system for transgene expression, lineage tracing, and mosaic analysis. Cell 141: 536–548.
    1. Sopko R.,
    2. Foos M.,
    3. Vinayagam A.,
    4. Zhai B.,
    5. Binari R.,
    6. et al.
    , 2014 Combining genetic perturbations and proteomics to examine kinase-phosphatase networks in Drosophila embryos. Dev. Cell 31: 114–127.
    1. Staller M. V.,
    2. Yan D.,
    3. Randklev S.,
    4. Bragdon M. D.,
    5. Wunderlich Z. B.,
    6. et al.
    , 2013 Depleting gene activities in early Drosophila embryos with the “maternal-Gal4-shRNA” system. Genetics 193: 51–61.
    1. Vert J. P.,
    2. Foveau N.,
    3. Lajaunie C.,
    4. Vandenbrouck Y.
    , 2006 An accurate and interpretable model for siRNA efficacy prediction. BMC Bioinformatics 7: 520.
    1. Yan D.,
    2. Neumuller R. A.,
    3. Buckner M.,
    4. Ayers K.,
    5. Li H.,
    6. et al.
    , 2014 A regulatory network of Drosophila germline stem cell self-renewal. Dev. Cell 28: 459–473.
    1. Zeng X.,
    2. Han L.,
    3. Singh S. R.,
    4. Liu H.,
    5. Neumuller R. A.,
    6. et al.
    , 2015 Genome-wide RNAi screen identifies networks involved in intestinal stem cell regulation in Drosophila. Cell Reports 10: 1226–1238.