Skip Navigation
National Cancer Institute U.S. National Institutes of Health National Cancer Institute
RNAi
  • shRNA Info

shRNA Validation Project

A Cooperative Project Between the CGAP and the ICBP

Integrative Cancer Biology Program

Purpose
Background
Materials and Methods
Results
  • Table 1
  • Figure 1
  • Table 2
    Summary and Conclusions
    References

  • Purpose

    The purpose of this project was to design and test a set of shRNAs for validation as tools to target high priority cancer genes. After validation, the target sequences of the shRNAs were deposited into the NCI's Cancer Genome Anatomy Project (CGAP) open access, public database. All shRNA constructs are commercially available. The validated shRNAs are designed to serve as tools for researchers to target and modulate the expression of well-known cancer-associated genes.

    Background

    RNA interference (RNAi) is an essential tool to study gene function and it is especially powerful in advancing cancer research (1-3). The shRNA Validation Project expands researchers' ability to exploit RNAi to better understand gene expression and cancer. In this project, shRNAs for 136 high priority cancer-related genes were tested and annotated for their effect(s) on gene expression. The testing and annotation of these shRNA constructs in a lentiviral vector system constituted the validation of these shRNAs. For each of the 136 cancer-related genes, three shRNAmir constructs were designed, tested and annotated.

    The GIPZ lentiviral shRNAmir library (Open Biosystems) was employed for this project. The unique advantages of this library are:

    1) The use of a proprietary shRNA design algorithm that incorporates the latest knowledge for identifying the best hairpins (4),

    2) All hairpins are embedded in a micro-RNA context which increases production of mature small RNAs and increases knockdown efficiency (5),

    3) The lentiviral vector backbone includes TurboGFP, a marker for establishing transduction efficiency and tracking gene expression, and a proven CMV promoter to drive shRNA transcription (unpublished data).


    Materials and Methods

    Cell Lines
    An ovarian (OVCAR-8) and breast (MCF-7) cancer cell line, both obtained from ATCC, were used to express the shRNAmir constructs. The following conditions were used for maintenance. OVCAR-8 was grown in Dulbecco's Minimum Essential Medium (10% Fetal Bovine Serum (FBS), 1% pen-strep). MCF-7 was grown in ATCC complete growth medium: Minimum Essential Medium (Eagle) with 2 mM L-glutamine and Earle's BSS adjusted to contain 1.5 g/L sodium bicarbonate, 0.1 mM non-essential amino acids and 1 mM sodium pyruvate and supplemented with 0.01 mg/ml bovine insulin, 90%; 10% FBS. Both cell lines were grown at 37˚C, 5% CO2.

    Construction of shRNAmir Library
    The shRNAs were selected from the shRNAmir library developed by Greg Hannon and Steve Elledge (5-7). The shRNAmir sequences were incorporated into the pGIPZ lentiviral expression vector to produce the GIPZ lentiviral shRNAmir library (Open Biosystems), which was used to test the 136 cancer gene targets.

    Virus Production and Transduction
    The viral stocks were produced using the Trans-Lentiviral™ Packaging System (Open Biosystems) modified for use with a Ca2PO4 protocol in 96-well format. Viral supernatants were stored at -80˚C until used. A sampling of individual wells of virus was titered to determine the average transforming units (TU) in HEK293T cells (see Download Plugin Adobe Acrobat Reader "Protocol III - Transduction and Titering", in the product insert for Trans-Lentiviral™ Packaging System). The average range of the sample titers were 1-5x106 (TU/ml). This functional titer range was determined using the TurboGFP marker in the HEK293 cell line.

    The two cell lines, OVCAR-8 and MCF-7, were seeded into 24-well plates. OVCAR-8 cells were seeded at 5 x 104 cells/well and MCF-7 cells were seeded at 1 x 105 cells/well and transduced the following day. OVCAR-8 and MCF-7 cells were transduced with unconcentrated virus at a multiplicity of infection (MOI) of 0.4-2.0, producing on average one functional insertion per cell (i.e. low to single copy number; data not shown). Forty-eight hours post-transduction, cell populations were subjected to puromycin selection. OVCAR-8 transduced populations underwent 3 days of puromycin selection at 30 µg/ml. MCF-7 transduced populations were subjected to 4 days of puromycin selection at 5 µg/ml. The differences in selection time reflected the differing growth rates of MCF-7 and OVCAR-8. MCF-7 required one extra day of growth to obtain the cell mass necessary to isolate sufficient amount of high-quality RNA.

    cDNA Generation
    The cells were lysed and RNA was purified (96 well RNeasy kits, Qiagen). Qiagen's protocol for lysis and purification was followed with the following caveats. The lysis buffer was supplemented with 10 μl/ml of β-mercaptoethanol. OVCAR-8 was lysed in 350 μl and MCF-7 was lysed in 150 μl; different amounts of lysis buffer were required in order to isolate high-quality RNA from the two cell lines (data not shown). The RNA was eluted from the column in 110 μl of RNase-Free Water (supplied with the kit). The RNA was reverse transcribed into cDNA following the protocol included with the High Capacity cDNA Archiving Kit(Applied Biosystems), the only exception being that cDNA was generated in a 20 μl reaction using 80 ng to 150 ng of total RNA.

    Quantitative Real-Time PCR
    Gene expression was measured by quantitative real time PCR Taq-man® Gene Expression Assays (Applied Biosystems). Using robotics, the primers and probes were mixed with the cDNA then 5 μl of the mix was distributed to individual wells of 384-well plates. These plates were frozen and shipped on dry ice to the Vanderbilt Microarray Shared Resource (VMSR). The VMSR added 5μl of master mix TaqMan® Gene Expression Master Mix (Applied Biosystems) to the reaction and amplified the samples on the 7900HT Fast Real-Time PCR System (Applied Biosystems).

    The eukaryotic 18S rRNA (FAM MGB Probe, Non-Primer Limited)(Applied Biosystems) was used as the endogenous control. Each sample's target gene and endogenous control were measured in triplicate. All probes were minor groove binding (MGB) and FAM labeled.

    The remaining gene expression was measured (as a percentage) compared to a non-silencing control (Download Plugin Adobe Acrobat Reader Non-silencing-GIPZ lentiviral shRNAmir control, Open Biosystems). This non-silencing control is identical to the vector used to knockdown target genes with the exception that the shRNA itself has no homology to any sequence in the human, mouse or rat genomes. Therefore, the non-silencing control replicates the process of transduction and stimulation of the RNAi pathway without inducing direct knockdown of any gene.


    Results
    The following figure and tables summarize the data generated to produce a collection of cancer-relevant shRNAs.

    Effect of shRNAmirs: Most of the 136 Cancer-Related Genes have at Least 1 shRNAmir that Represses Gene Expression ≥50%.

    Table 1: Percentage of Genes with Either: (A) ≤50% or (B) ≤30% Remaining Gene Expression.
    Table 1A: Remaining Gene Expression: 50% or less
    (50% remaining gene expression equates to 50% knockdown)
    • OVCAR-8: For 87% of the cancer-related genes targeted, a 50% knockdown in gene expression was seen with at least one of the shRNAmirs.
    • MCF-7: For 60% of the cancer-related genes targeted, a 50% knockdown in gene expression was seen with at least one of the shRNAmirs.

    Table 1B: Remaining Gene Expression: 30% or less
    (30% remaining gene expression equates to 70% knockdown)

    • OVCAR-8: For 70% of the cancer-related genes targeted, a 70% knockdown in gene expression was seen with at least one of the shRNAmirs.
    • MCF-7: For 32% of the cancer-related genes targeted, a 70% knockdown in gene expression was seen with at least one of the shRNAmirs.

    Average Knockdown Efficiency is Greater in OVCAR-8 Cell Populations Versus MCF-7 Cell Populations.

    Figure 1 :Global Gene Expression Effects of shRNAmirs in MCF-7 and OVCAR-8 Cell Lines.
    -The knockdown effects are grouped to highlight differences in expression modulation between the two cell lines.

    shRNAmirs that result in at least 50% reduction in gene expression:

    • 68% of shRNAmirs tested in OVCAR-8 led to at least a 50% reduction in gene expression.
    • 35% of shRNAmirs tested in MCF-7 led to at least a 50% reduction in gene expression.

    shRNAmirs that result in at most 25% reduction in gene expression:

    • 8.8% of shRNAmirs tested in OVCAR-8 led to 0-25% reduction in gene expression.
    • 20.4% of shRNAmirs tested in MCF-7 led to 0-25% reduction in gene expression.

    Some shRNAmirs Lead to Induction of Gene Expression (also Figure 1)

    shRNAmirs that result in activation of gene expression:
    • 8.5% of shRNAmirs tested in OVCAR-8 led to activation of gene expression.
    • 13.3% of shRNAmirs tested in MCF-7 led to activation of gene expression.
    Expression Annotation and Summary for all Tested shRNAmirs Targeted to Cancer-Related Genes.

    Table 2: Gene List with Corresponding shRNA clones and % remaining gene expression.
    Table 2 Fields:

    Gene: Each gene from the high-priority cancer-related list is identified by its HUGO gene symbol. The gene symbol is hyperlinked.
    The link will direct users to:

    • A description of the gene.
    • A link to the corresponding gene accession number.
    • A list of all of the shRNAmirs designed for that gene.
    • "RNAi View" that shows a graphical representation of where the shRNAmirs are located on the gene transcript.

    Oligo ID: Specific and unique identifier for each shRNAmir. The Oligo ID is hyperlinked. The link will direct users to the distributor's website for more information about the shRNAmir construct, including purchasing options.

    Percent: Percent remaining gene expression, which is the inverse of the percent knockdown. Gene expression measured after transduction with gene-specific shRNAmir divided by the gene expression measured in same cell line transduced with a non-silencing shRNAmir. Gene expression was measured by quantitative real-time PCR and expressed as a percentage.

    Replicates: The number of independent tissue culture wells that received virus expressing a given shRNAmir. Each of these replicate samples was then measured in triplicate by qRT-PCR.

    No Data (ND): Data not available due to undetectable basal expression level of gene target, low nucleic acid yields or other reasons that prohibited accurate quantitative measurement of gene expression.

    Table 2 Summary:

    • 132 genes were tested.
      • Data was generated for a total of 126 genes.
      • OVCAR-8: 116 genes with shRNAmir results.
      • MCF-7: 118 genes with shRNAmir results.
      • 96 of the genes tested resulted in data for 3 shRNAmirs in both cell lines.
    • 3 shRNAmir targets were tested against almost every gene.
      • 393 shRNAmirs were tested in both cell lines.
        • 130 genes were tested with 3 shRNAmirs
        • 1 gene was tested with only 2 shRNAmirs.
        • 1 gene was tested with only 1 shRNAmir.
      • OVCAR-8: 331 shRNAmirs resulted in interpretable data.
      • MCF-7: 341 shRNAmirs resulted in interpretable data.
    • Of the 786 possible data points (393 shRNAmirs in two cell lines), interpretable data was obtained for 672 (86%) of the data points.
      • The evidence suggests that over three-fourths of the uninterpretable data was due to low or undetectable gene expression levels.
    Summary and Conclusions

    A set of 393 GIPZ lentiviral shRNAmirs, targeted to 132 high priority cancer-related genes and validated in MCF-7 and OVCAR-8 cells are now available as a research resource.

    The majority of the hairpins led to a reduction in gene expression. In the OVCAR-8 cell line, 224 shRNAmirs (or 68% of the shRNAmirs) resulted in at least a 50% reduction in gene expression. In MCF-7 cells, 119 of the 337 hairpins resulted in at least a 50% reduction in gene expression. Overall, 52% of the shRNAmirs tested were able to knockdown gene expression by at least 50%. Additionally, for 74% of the genes tested in OVCAR-8 cells, there are at least two shRNAmirs that reduced gene expression by at least 50%; in 44% of genes tested in OVCAR-8, there were two hairpins that knocked down gene expression by 70% or more. (Table 1-A and Table 1-B).

    While these results are promising, the effectiveness of the shRNAmir constructs can likely be increased further. Knockdown efficiency of shRNAmirs is highly dependent on the number of functional integrations within the cell. For these experiments, cell populations were transduced at low MOIs resulting in low to single copy integrations. Therefore, increasing the transducing MOIs will increase functional integrations and thus potentially increase the knockdown efficiencies of the hairpins. Additionally, transductions outlined herein were done in the presence of serum. Serum inhibits transduction efficiency and thus knockdown efficiency (unpublished data). Transducing cell populations at higher MOIs in the absence of serum would likely increase the knockdown efficiency for the shRNAmirs.

    A comprehensive view of the hairpins' effects shows a range of gene expression modulation (see Figure 1). In most cases there were cell line specific differences. Generally, greater knockdown effects were observed in the OVCAR-8 cell line than in the MCF-7 cell line. For example, 68% of the hairpins tested in OVCAR-8 resulted in 50% or greater knockdown, while just over 35% did so in MCF-7. These differences were in part due to the cell lines' ability to be transduced. MCF-7 was found to have a lower transduction efficiency than OVCAR-8 (data not shown). These data suggest that the selection of cell lines to study gene expression modulation is important and should be considered in the interpretation of results.

    The data also indicate that the level of basal gene expression should also be considered. Table 2 includes data for 85% of the 393 possible data points that would be generated for each of the 132 genes targeted in the two cell lines. However, data was not generated for 15% of total possible data points. In the majority of the samples that did not generate data, the 18S internal control was detected within the expected range suggesting much of the 15% attrition was likely due to undetectable levels of target gene expression or primer and probe failure. In some instances there is strong evidence suggesting that losses were more likely due to lack of gene expression. For example, the receptor tyrosine kinase AXL was measured in OVCAR-8 demonstrating that the primer and probe set functions. However, all of the MCF-7 AXL samples failed detection. This is consistent with expression levels too low to detect in one cell line but not the other. These data indicate that basal levels of gene expression between different cell lines are an important consideration in the selection, design and interpretation of gene expression analysis.

    Additionally, 73 hairpins (10% of all shRNAmirs) resulted in increased gene expression, i.e. expression levels above that of the non-silencing control. More gene expression activation was seen in MCF-7 cell populations than in OVCAR-8, i.e. 13.3% versus 8.5%, respectively. There are examples in the literature of duplexed RNA inducing transcriptional activation of genes (8-10). It is possible that the results seen here are related to this mechanism, however current data are inconclusive. In summation, there is now a validated shRNAmir clone set for genes directly relevant to cancer research. The availability of this set and the accompanying knockdown efficiency data will enable investigators to apply RNAi technologies immediately to their research efforts with greater confidence.


    References

    1. M. Allen et al., Nat. Genet. 35, 258 (2003).

    2. J. Downward, Oncogene 23, 8334 (2004).

    3. J. Silva, et al., Oncogene 23, 8401 (2004).

    4. Algorithm design, Greg Hannon at Cold Spring Harbor. http://www.cshl.edu/public/SCIENCE/hannon.html 5. J. M. Silva et al., Nat. Genet. 37, 1281 (2005).

    6. P. J. Paddison et al., Nat. Methods 1, 163 (2004).

    7. F. Stegmeier, et al., Proc. Natl. Acad. Sci. 102, 13212 (2005).

    8. J. J. Rossi, Nat. Chem. Biol. 3, 136 (2007).

    9. L. C. Li et al., Proc. Natl. Acad. Sci. 103, 17337 (2006).

    10. B. A. Janowski et al., Nat. Chem. Biol. 3, 166 (2007).