2Greenebaum Cancer Center, University of Maryland School of Medicine, Baltimore, MD, USA
3Department of Physiology, University of Maryland School of Medicine, Baltimore, MD, USA
4Department of Pediatrics, University of Maryland School of Medicine, Baltimore, MD, USA
5Sidney Kimmel Comprehensive Cancer Center, Division of Biostatistics and Bioinformatics, Johns Hopkins University School of Medicine, Baltimore, MD, USA
MicroRNAs (miRs) regulate cellular processes by modulating gene expression. Although transcriptomic studies have identified numerous miRs differentially expressed in diseased versus normal cells, expression analysis alone cannot distinguish miRs driving a disease phenotype from those merely associated with the disease. To address this limitation, we developed miR-HTS, a method for unbiased high-throughput screening of the miRNome to identify functionally relevant miRs. Herein, we applied miR-HTS to simultaneously analyze the effects of 578 lentivirally transduced human miRs or miR clusters on growth of the IMR90 human lung fibroblast cell line. Growth-regulatory miRs were identified by quantitating the representation (i.e., relative abundance) of cells overexpressing each miR over a one-month culture of IMR90, using a panel of custom-designed quantitative real-time PCR (qPCR) assays specific for each transduced miR expression cassette. The miR-HTS identified 4 miRs previously reported to inhibit the growth of human lung–derived cell lines and 55 novel growth-inhibitory miR candidates. Nine of 12 (75%) selected candidate miRs were validated and shown to inhibit IMR90 cell growth. Thus, this novel lentiviral library- and qPCR-based miR-HTS technology provides a sensitive platform for functional screening that is straightforward and relatively inexpensive.
MicroRNAs (miRs) are small non-coding RNAs that regulate gene expression at multiple levels (1-3). MiRs appear to regulate most coding genes and thereby participate in essentially all biological processes including cell survival, proliferation and differentiation (4-6). Consequently, alterations in miR expression and function play pathophysiological roles in cancers and other human diseases (6-8). For example, expression of the miR-34 family is down-regulated in various cancers, and has been shown to be directly transactivated by p53, serving as a potent mediator of p53-dependent tumor suppressive mechanisms, including apoptosis, cell cycle arrest, and senecence (reviewed in Reference (9). Delivery of miR-34a mimics inhibited growth of lung cancers with low miR-34a expression in mouse models (10, 11). A striking observation in this and other preclinical and clinical studies is the lack of toxicity of miR-based therapeutics to normal cells and tissues, suggesting a large therapeutic window for miR-based therapies (8, 12).
Substantial transcriptomic studies have identified many miRs differentially expressed between diseased and normal cells, as well as between different cell lineages and differentiation stages throughout developmental processes (e.g., during hematopoiesis) (6, 7, 13). However, differential expression of a particular miR does not guarantee a physiological or pathophysiologic role for that miR. Changes in miR expression may be the result of the diseased state rather than the cause, and the function of a miR can be altered without changes in its expression level [e.g., due to context-dependent availability of auxiliary factors required for miR function (14, 15)] – such miRs might be excellent biomarkers, but would not be operative therapeutic targets. High-throughput functional screens could distinguish “driver” miRs from “passenger” miRs and thus speed the identification of therapeutically relevant miRs (16).
Unfortunately, current approaches for genome-wide functional/phenotypic screening of the human miRNome require equipment or expertise not readily available in many laboratories and are often prohibitively expensive. Approaches and examples of miR functional screens have been comprehensively reviewed in (16). One approach utilizes transfection of synthetic miR mimic/inhibitor libraries, in which synthetic miRs are arrayed individually in multiple 96- or 384-well plates. Such a miR mimic library approach was used successfully to identify miRs involved in G1 arrest of DGCR8 knockout ES cells (17). Nevertheless, in this approach, screening is limited by (i) the need for cells that can be transfected efficiently and (ii) the use of transient assays, which is disadvantageoussince synthetic miRs are diluted with every cell division. These limitations can be overcome by infection of cells with miR overexpression lentivirus (lenti-miR) libraries, with the individual lenti-miRs virus arrayed in 96-well plates. This approach has been used successfully to identify miRs suppressing growth (18) or metastasis (19) of cancer cell lines. The advantage of this format with one miR (mimic, inhibitor, or virus) per well is that the candidate miRs can be attributed very simply to enhance or decrease a specific function. However, arrayed format screens require assays adaptable to microplates and often require robotics to minimize well-to-well variation and other error-prone screening problems. These requirements and associated costs prevent broad application of arrayed miR functional screens.
We introduce miR-HTS, a novel functional screening method for the human microRNA-ome (miRNome) that is high-throughput and sensitive. Using a lentiviral miR library and quantitative PCR, this method provides a sensitive platform for functional screening of the miRNome that is straightforward and cost-effective.
An alternative screening strategy is to employ pooled lenti-miR or retroviral miR libraries, in which each virus encodes a unique miR or miR cluster (20-25). This screening strategy requires an assay for a phenotype that results from over- or under-representation of cells overexpressing particular miRs. Because these screens simultaneously assay all the miRs in a pool, they depend on downstream methods to distinguish the candidate miRs. Taking advantage of viral integration into the host genomic DNA, the unique miR transgene sequence of each miR virus can serve as the “barcode” for quantitating the number of cells overexpressing a particular miR. Thus, the under- or overrepresented barcode enables identification of each miR whose overexpression results in a decrease or increase in the number of cells infected with that particular miR in the functional screen. For example, the barcode of a miR whose overexpression inhibits cell growth would decrease over time; Izumiya et al. identified candidate pancreatic cancer suppressive miRs on this basis (22). Conversely, by determining barcodes enriched in migrated cells versus the starting cell population, candidate miRs that promoted invasiveness of cancer cell lines were identified (23, 26). Although these studies successfully employed pooled viral miR libraries in functional screens, their viral miR barcode quantification methods required custom-made microarrays (22, 24), bead-based detection systems (20, 21), or high-throughput sequencing (23, 26). Those approaches require access to costly high-throughput infrastructure and sophisticated informatics analyses, thereby restricting the accessibility of these assays.
We report here a simple barcode quantification strategy based on real-time quantitative PCR (qPCR) as an effective alternative to the aforementioned identification methods, allowing any laboratory with access to a real-time PCR machine to conduct functional screening of pooled miR libraries. Using a pooled human lenti-miR library and custom-designed qPCR assays to quantify lenti-miR barcodes, we established miR-HTS, a novel method that enables depletion- and enrichment-based functional screening of the human miRNome. As proof-of-principle, we used miR-HTS to screen the IMR90 human lung fibroblast cell line for miRs capable of regulating cell growth. We identified four known growth-inhibitory miRs, plus 55 miRs not previously known to regulate growth of IMR90 cells. Nine of 12 (75%) selected candidate miRs were independently validated and confirmed to inhibit IMR90 cell growth, demonstrating the miR-HTS as a straightforward and sensitive alternative to reported approaches. Materials and methods Cell culture
The IMR90 cell line was purchased from ATCC (Manassas, VA, USA), and cultured in Minimum Essential Medium (Mediatech, Manassas, VA, USA) containing 10% fetal bovine serum (Gemini, Sacramento, CA, USA), 1 mM sodium pyruvate (Lonza, Walkersville, MD, USA), 1× GlutaMAX (Life Technologies, Grand Island, NY, USA), 1x penicillin-streptomycin solution (Mediatech). Design of genomic DNA representation qPCR (GRE-qPCR) assays and the GFP loading control assay
To design genomic DNA representation qPCR (GRE-qPCR) assays, the miR (plus flanking) insert sequences of each lenti-miR precursor construct were obtained from System Biosciences (SBI, Mountain View, CA, USA) under a mutual nondisclosure agreement between SBI and the University of Maryland Baltimore. GRE-qPCR assays were designed using Primer Express software from Applied Biosystems (ABI, Foster City, CA, USA) or were custom-designed by ABI and purchased. The internal miR-specific forward primers for each GRE-qPCR assay were custom-plated by ABI at 0.018 nmole per well in 384-well plates. The internal common reverse primer, GFP primers, and external common forward and reverse primers were synthesized by Integrated DNA Technologies (Coralville, IA, USA). The GFP probe and internal common probe were synthesized by ABI. Sequences of all primers and TaqMan probes used in this studyare listed in Supplementary Table S1. During the design of GRE-qPCR assays, care was taken to minimize potential nonspecific binding of internal miR-specific forward primers to unintended targets [due to sequence similarities between lenti-miRs (e.g., paralogous miRs)], by designing primers within the more variable miR flanking sequences. Nevertheless, six of the 578 internal miR-specific primers may fail to distinguish paralogous miRs (see comments in Supplementary Table S1). miR-HTS procedure and data analysis
In each miR-HTS, 2.7 million IMR90 cells were infected at a multiplicity of infection (MOI) of 0.2 with the human lenti-miR pooled virus library (Cat. no. PMIRHPLVA-1, SBI) to achieve ~20% transduced cells. Eight μg/mL polybrene (Sigma-Aldrich, St. Louis, MO, USA) was used as the infection vehicle (27). On days 3 (t0), 11 (t1), 19 (t2) and 27 (t3) after infection, a fraction of the infected culture (≥2 million cells) was harvested and genomic DNA isolated using the DNeasy Blood & Tissue Kit (Qiagen, Valencia, CA, USA). The external common forward and reverse primers (Figure 1A) were used to amplify lentiviral integrants; these amplifications used 1.5 μg genomic DNA (per 100 μL PCR reaction) from each time point, 3% DMSO, 0.4 μM external common forward and reverse primers, 0.2 mM dNTPs, 1× Phusion HF buffer and 2 units Phusion High-Fidelity DNA Polymerase (New England BioLabs, Ipswich, MA, USA). PCR was performed with the following program: 98°C for 2 min, 25 cycles of 98°C for 10 s, 56.4°C for 30 s, 72°C for 90 s, and a final extension at 72°C for 5 min. PCR products were purified using the QIAquick PCR purification kit (Qiagen). 0.2 ng purified PCR products were used as templates for all GRE-qPCR assays; 0.002 ng template was used for the GFP loading control qPCR assay. In addition to the templates, each 20-μL qPCR reaction contained 1× TaqMan Universal Master Mix II (ABI), 900 nM forward and 900 nM reverse qPCR primers, and 250 nM TaqMan probes. qPCR was performed on the ABI 7900HT Fast Real-Time PCR System with the following program: 95°C for 10 min, 40 cycles of 95°C for 15 s, and 60°C for 1 min. The qPCR results were analyzed using Sequence Detection System software (ABI); see Supplementary Methods for a detailed miR-HTS data-analysis protocol.