A rapid genetic screening system for identifying gene-specific suppression constructs for use in human cells

Abstract

We describe a rapid cell-based genetic screen using fission yeast for identifying efficient gene suppression constructs (GSCs) from large libraries (10⁵) for any target sequence for use in human cells. In this system, target sequences are fused to the 5′ end of the lacZ reporter gene and expressed in yeast. Random fragment expression libraries derived from the target sequence are screened in the fusion gene-expressing strain using the lacZ gene-encoded colony color phenotype. We demonstrate the utility of this screening assay by identifying a range of different GSCs for the fission yeast ura4 gene and human c-myc and Chk1 sequences, including rare efficient suppressors. GSCs specific for c-myc were shown to regulate expression of both a c-myc–lacZ fusion gene and the endogenous c-myc gene in human cells.

INTRODUCTION

Recent advances in large-scale DNA sequencing have produced partial or complete genomic sequences for a large number of unicellular (1) and multicellular organisms (2), including humans. A majority of these putative genes have no assigned cellular function. Using DNA microarrays and gene expression profiling, numerous genes have been found that can be implicated in the onset or progression of a specific human disease or trait (3–5). For those genes with a known cellular function, the challenge of determining the precise biological role of this gene or gene product in the specific disease still remains. Furthermore, a large proportion of the identified genes has no known homologs and, as such, their biological role in the disease is unknown.

A general approach to gene function determination in human disease has involved studying cells in which the expression of a candidate gene has been over-expressed or suppressed. In this context, suppression of gene expression is commonly attempted using gene constructs expressing antisense RNAs, ribozymes or dominant negative proteins (6,7). Unfortunately, the full potential of these trans-acting RNAs and proteins has not been realized due, in part, to the lack of a rapid method for selecting the most effective construct for each specific gene. To some extent, this limitation has been addressed by the development of cell-based systems in which random gene fragments are expressed in mammalian cells, followed by clonal screening for those that modify a target gene-specific phenotype (8,9). The main advantage of using this approach is the isolation of the best gene-encoded suppression molecule in vivo from a pool of different plasmid constructs. However, at present this approach in mammalian cells is time-consuming, has low throughput and is limited to target genes associated with an easily assayable phenotype. There is a real need in the genomics era for a rapid and simple in vivo cell-based genetic screen for finding a wide range of functional gene suppression constructs (GSCs) for a large number of different gene targets, of both known and unknown function, for use in human cells.

We have developed an expression library screening technology using yeast cells that permits the identification of gene-specific suppressors. This cell-based assay has high throughput with the capacity to rapidly identify and rank order hundreds of GSCs from pools of >10⁵ different plasmids. The types of suppressors found using this assay include antisense RNAs, sense RNAs, trans-dominant polypeptides and promoter decoys. The GSCs confer a wide range of levels of suppression and provide the potential to quantitatively modulate the expression of a specific gene. In addition, these constructs can then be rapidly tested in a transient human cell assay for effectiveness against the endogenous target gene. In this paper, we describe this screening assay and demonstrate the usefulness of yeast-derived c-myc gene-specific GSCs for regulating expression of the c-myc gene in human cells.

MATERIALS AND METHODS

Yeast media and methods

All yeast strains were maintained on YES or EMM media (10). Yeast transformation, plasmid isolation and β-galactosidase enzyme assays were performed as previously described (11). PCR amplification of inserts from plasmid DNA within whole yeast cells was essentially as described in (12) with pre-treatment of cells in 50 mM Tris–HCl (pH 8.0) containing 10 mg/ml zymolyase-20T (ICN Biomedicals, Aurora, OH) for 30 min at 30°C. To these cells was added directly the PCR mix containing 50 mM KCl, 10 mM Tris–HCl (pH 8.3), 200 µM dNTPs, 100 ng of primers and 1.2 U of AmpliTaq DNA polymerase (Perkin-Elmer).

Yeast strain construction

All DNA used to construct specific yeast strains was integrated at the ura4 locus using ura4 flanking sequences contained on pNEB195 (11) and homologous recombination (13). For strain RB3-2, the adh1 promoter and ura4 3′ untranslated region (UTR) were derived as a PCR product by amplification from pURAS (14) and cloned as a HindIII fragment into pGT5, a derivative of pNEB195. The lacZ BamHI fragment derived from pNEBβ2 (11) was then subcloned downstream of the adh1 promoter. This lacZ gene expression cassette was used to replace the ura4 gene in strain 972 using 5-fluoroorotic acid (FOA) selection. The leu1-32 allele was introduced by mating as described (11) to produce RB3-2 (h^–, ura4::adh1-lacZ, leu1-32).

To construct the fusion strains, two integrating vectors were developed. For the first vector, a 350 bp fragment containing the SV40 early promoter was PCR-amplified from pSVβ (Clontech) and subcloned in place of the adh1 promoter on a pGEM3Zf-based plasmid upstream of the ura4 UTR (14). A HindIII fragment containing the SV40 promoter and ura4 UTR was subcloned into pGT5. The 3.1 kb lacZ DNA was amplified using pSVβ as a template and subcloned as an EcoRI–BamHI fragment downstream of the SV40 promoter to create the integration vector pINTVEC. The second integration plasmid, pSIV, was produced by replacing the SV40 promoter in pINTVEC with the 745 bp adh1 promoter (14).

The ura4 sequence was amplified from pREP4 (15) using the primers 5′-TGGGGATCCAAGCTTAATTCGAGCTCGGTACAGCTTGG-3′ and 5′-AAGGGATCCCTACTTCTGG-AATAGCTCAGAGG-3′ and subcloned into pINTVEC in-frame with the lacZ gene to produce the ura4–lacZ integration vector. This plasmid was used to transform RB3-2 and ura4–lacZ gene integrants were selected on EMM. For the c-myc sequence, the region surrounding exons 2 and 3 of the c-myc cDNA was PCR-amplified from the Quick-Screen human cDNA library (Clontech) using the primers 5′-CGAGGATCCTTGCAGCTGCTTAGA-3′ and 5′-TAGGGATCCCGCAC-AAGAGTTCCGTAG-3′. This 1379 bp fragment was subcloned as a BamHI fragment into pSIV downstream of the adh1 promoter to generate the c-myc–lacZ integration plasmid. Following transformation of this plasmid into strain NCYC1913 (h-, leu1-32; ATCC), c-myc–lacZ integrants were obtained by plating onto EMM containing leucine and FOA. In the case of the Chk1 target, the cDNA was received as a gift from Dr W. Luyten (Janssen Research Foundation, Belgium) and subcloned as a blunt end fragment into the SmaI site downstream of the SV40 promoter in pINTVEC to produce the Chk1–lacZ integration vector. Selection for integrants containing this fusion gene was performed as described for the c-myc–lacZ fusion strain. All of the fusion gene-expressing strains were characterized for integrated plasmid structure, expression of the fusion mRNA, β-galactosidase activity and blue colony color phenotype.

Construction of expression libraries

The yeast plasmid pGT118 was used in constructing all random fragment expression libraries. This vector was created by modifying pREP1 (16) to include a single ATG codon in the nmt1 promoter region, a unique BglII cloning site between this promoter and the nmt1 UTR, and TAA stop codons in three reading frames downstream of the BglII site. The gene-specific fragments used as substrates for constructing the libraries were PCR-amplified using the following primers: ura4, 5′-TGG-AAGCTTAATTCGAGCTCGGTACAGCTTGG-3′ and 5′-TTG-GGATCCAGAATGCTGAGAAAGTCTTTGCTG-3′; c-myc, 5′-TGGCCTTCGCTTTTCTTTAAGCAAGAG-3′ and 5′-TAG-GGATCCCGCACAAGAGTTCCGTAG-3′; Chk1, 5′-TGTA-AGCTTGATATCGAATTCGATTACTCACTATAG-3′ and 5′-CGAAAGCTTGATATCTATCAGATGTGGCAGGAAG-CCAAACCTT-3′. The templates used for each target were the respective integration vectors. These amplified substrates were partially digested with DNase I. Fragments >300 bp were isolated using Sepharose CL-4B columns (Pharmacia Biotech), while those required for the smaller insert libraries were eluted from agarose gels. All fragments were end-filled with T4 DNA polymerase and Klenow Fragment and ligated to either BglII or BamHI linkers (New England Biolabs) prior to cloning into pGT118. Expression library plasmids were transformed into DH10B cells (Bio-Rad Laboratories) and clones expanded as single colonies. Each library was shown to represent the entire target gene substrate.

Screening expression libraries

A total of 2 µg of library DNA was used to transform the specific fusion gene-expressing strain and transformants selected on EMM containing uracil. All transformants were overlayed with agarose–Xgal medium and incubated at 37°C to permit colony color development. The agarose–Xgal medium was composed of 0.5 M sodium phosphate, 0.5% agarose and 2% dimethylformamide, with SDS and X-gal added to varying concentrations dependent on the level of β-galactosidase activity displayed by the strain. The concentrations of the latter components were pre-determined for each fusion gene-expressing strain following quantitation of the β-galactosidase activity using a standard enzyme assay. This was accomplished by titrating the SDS and X-gal concentrations in combination with one another and varying the incubation temperature and time. This strategy provides the flexibility to detect blue colony color for yeast strains expressing both low and high levels of β-galactosidase activity. For example, detection of the blue colony color for the ura4–lacZ fusion gene-expressing strain required 0.01% SDS and 200 µg/ml X-gal and incubation at 37°C for 3 h. In contrast, the strain expressing the c-myc–lacZ gene produced detectable blue colony color using 0.01% SDS and 1000 µg/ml X-gal when incubated at 37°C overnight.

All yeast transformant colonies were screened as above in comparison with the parental strain transformed with pGT118. Light blue colonies were recovered and assayed for β-galactosidase activity. The insert contained on the resident plasmid was PCR-amplified, using pGT118-specific primers, and sequenced.

Plasmids for human cells

The c-myc–lacZ fusion gene contained on the yeast integration vector was subcloned into pcDNA3 (Invitrogen) by first transferring the lacZ sequence as a BamHI–EcoRI fragment, followed by insertion of the c-myc region at the BamHI site between the CMV immediate early promoter and the lacZ insert. Plasmid pIND-GFP was constructed by subcloning the green fluorescent protein (GFP) expression cassette from pGREENLANTERN (Bio-Rad Laboratories) onto pIND (Invitrogen) as a BglII fragment. This plasmid was used as the parental vector into which PCR-amplified c-myc inserts from yeast-derived GSCs were subcloned as HindIII–XbaI fragments in the antisense orientation downstream of the heat shock promoter. To construct plasmid pFL, the c-myc DNA contained in the c-myc–lacZ fusion gene was subcloned as a BamHI fragment under control of the heat shock promoter in pIND-GFP. Plasmid pFFL was created by subcloning the c-myc insert, containing all three exons, from pSP65cmyc (17) into pIND-GFP as a HindIII–XbaI fragment.

Cell culture and methods

EcR293 human embryonic kidney cells (Invitrogen) were maintained in DMEM containing 10% bovine serum. A total of 40 µg of plasmid DNA was transfected into 1 × 10⁷ cells using standard electroporation conditions. For co-transfection, 1 µg of the fusion gene-containing plasmid and 39 µg of the c-myc antisense plasmid were used. β-galactosidase activity was determined using the β-galactosidase Enzyme Assay System (Promega) according to the manufacturer’s instructions. Fluorescence-activated cell (FACS) analysis for GFP expression was performed on the Becton Dickinson FACSORT, while cell sorting for GFP-expressing cells was completed using the MoFlo (Cytomation, Inc., CO). Total RNA was extracted from cells using the Total RNA Isolation Reagent (Advanced Biotechnologies) according to the manufacturer’s recommendations. Northern hybridization was executed essentially as previously described (11). Western blot analysis of c-myc protein was performed as described in (18) with the following modifications. Cell lysates were prepared using PBS containing 1% Triton, 1% sodium deoxycholate, 0.5 M NaCl, 50 mM Tris, 10 mM EDTA, 0.1% SDS, 1 µg/ml aprotinin, 100 µg/ml PMSF and 10 µg/ml leupeptin. The c-myc protein was detected using an Anti-Myc polyclonal antibody (Upstate Biotechnology, Lake Placid, NY), a horseradish peroxidase-linked anti-rabbit Ig secondary antibody and the luminol/enhancer chemiluminescent substrate (Amersham).

RESULTS

A universal genetic screen to identify GSCs

In developing a universal system for genetically identifying effective GSCs specific for any target sequence, we chose a gene fusion approach using the lacZ reporter (Fig. 1). Using previously described lacZ antisense plasmids, we first showed a correlation between β-galactosidase activity and the degree of blue colony color. Based on this relationship, we then developed a genetic screen for rare colonies expressing at least 20% less β-galactosidase activity than control colonies. To examine the utility of this phenotypic assay for finding novel GSCs from gene-specific expression libraries, we constructed a small-scale library from the adh1 promoter-driven lacZ gene and screened it in strain RB3-2, stably expressing the target lacZ gene. A screen of 1.4 × 10⁴ transformants produced 26 different GSCs encoding single antisense RNAs, chimeric antisense RNAs, sense RNAs, truncated protein domains and promoter decoys (data not shown). This confirmed the usefulness of the lacZ gene-specific colony color assay as an indicator of β-galactosidase activity and demonstrated that this genetic screen had potential to locate functional GSCs encoding novel gene suppressors.

Figure 1.

Experimental strategy for expression library construction, screening plasmids in yeast and testing of GSCs in human cells. A target DNA is randomly fragmented to construct a gene-specific expression library. This same target DNA is fused to the N-terminus of the lacZ gene and the fusion gene expressed in yeast. The expression library is transformed into this strain and independent transformants screened for a reduction in the lacZ-encoded blue colony color phenotype. These transformants are characterized for β-galactosidase activity to quantitate the degree of suppression of fusion gene expression. The insert contained on the GSC is sequenced and subcloned to a mammalian expression vector for testing in human cells. The abbreviations 5′ and pA indicate the promoter and the 3′ UTR regions, respectively.

As an initial check of the fusion concept for identifying GSCs, we used the Schizosaccharomyces pombe ura4 sequence as a test gene. A ura4–lacZ gene-expressing strain was transformed with an expression library derived from the SV40 promoter and ura4 sequences. A total of 47 000 transformants were screened for their colony color phenotype and 291 light blue transformants identified. The range of suppression of β-galactosidase activity among these transformants was 29–77%, and suppression was shown to be dependent on insert transcription. Plasmids were isolated from 17 different primary transformants and re-tested in the fusion gene-expressing strain. For all plasmids, the degree of suppression was reproduced in separate experiments, indicating that the reduction in β-galactosidase activity was specific for the resident GSC. Analysis of the inserts on plasmids from 44 light blue transformants revealed 29 different antisense inserts and two sense inserts. The top 26 antisense insert-containing GSCs reduced β-galactosidase activity to a greater extent than a rationally designed plasmid expressing an antisense RNA complementary to the entire ura4–lacZ fusion mRNA. All of the inserts were more effective than the full-length ura4 sequence expressed in the antisense orientation, which suppressed β-galactosidase activity by 29%. This analysis indicated that using the lacZ gene-specific cellular phenotype was an efficient way of identifying highly effective GSCs capable of suppressing target–lacZ fusion gene expression, and that this genetic screen yielded more effective GSCs than those obtained by empirical design.

GSCs for human genes

To demonstrate the universal nature of the genetic screen and its application to non-microbial target sequences, we performed a series of screens for GSCs using two human genes. We chose as targets the c-myc and Chk1 genes, both of which encode proteins involved in cell cycle regulation (19,20). Each of these sequences was expressed as a fusion mRNA in S.pombe. Expression libraries for each target were screened using the colony color assay in the strain expressing the fusion gene. From 14 000 yeast transformants screened for activity against the c-myc–lacZ fusion, ~2.5% were identified as having a light blue phenotype (Table 1). Of these transformants, 28 were assayed for β-galactosidase activity and all displayed suppression ranging from 20 to 72%. The inserts contained on these GSCs were sequenced and aligned with the target gene as shown in Figure 2. Two classes of inserts were identified including 19 different antisense inserts and three in the sense orientation. The GSCs encoding c-myc antisense RNAs reduced β-galactosidase activity between 43 and 72%. A total of 85% of these inserts were derived from a region of the c-myc sequence containing the ATG codon, suggesting that complementarity to this portion of the fusion mRNA may be an important component of the gene-encoded c-myc antisense RNAs. Four of these RNAs were more effective than an empirically designed antisense RNA complementary to the entire c-myc sub-region of the fusion mRNA, which reduced β-galactosidase activity by 65%.

Table 1. Summary of the expression library screens in fission yeast.

Target gene	Library complexitya	No. of yeast screened	No. of light blue coloniesb	Range of suppressionc	No. of GSCsd
lacZ	2.5 × 104	14 000	237 (1.7%)	20–54% (27–72%)	26 (0.6%)
ura4	2.0 × 104	47 000	291 (0.6%)	29–77% (39–103%)	31 (0.1%)
c-myce	1.8 × 107	30 000	735 (2.5%)	20–81% (27–108%)	57 (1.7%)
Chk1e	1.1 × 106	40 000	913 (2.3%)	21–80% (28–107%)	50 (1.2%)

^aTotal number of independent recombinant clones as determined by sequencing random inserts.

^bNumber in brackets indicates the percentage of yeast transformants displaying a light blue colour phenotype.

^cNumbers in brackets represent suppression within a yeast population (colony) corrected for GSC-containing cells.

^dNumber in brackets indicates the percentage of plasmids screened that suppress target gene expression in yeast.

^eThe numbers indicated for each column include large and small insert library screens.

Figure 2.

Inserts on GSCs isolated from a screen of the c-myc expression library. At the top is shown the region of the c-myc–lacZ fusion gene used in constructing the library. The bent arrow indicates the site of transcription. The exon 2 and exon 3 boundaries are marked. The arrows in brackets show the orientation of the insert in the yeast plasmid. To the right of each insert is indicated the degree of suppression of β-galactosidase activity mediated by the GSC in comparison to vector control. The asterisks indicate those inserts tested in human cells.

In the screen for Chk1, a similar percentage of light blue colonies were identified as in the c-myc screen (Table 1). A total of 41 light blue transformants were assayed for β-galactosidase activity and all were found to suppress fusion gene expression, with the most effective mediating an 80% reduction in β-galactosidase activity. Further characterization of the resident plasmids indicated the presence of 25 different GSCs, all of which contained antisense oriented inserts derived from the 3′ half of the Chk1 cDNA sequence (data not shown). This pattern of alignment of the inserts from Chk1 GSCs was completely different from that observed for c-myc. To determine whether the lack of antisense inserts from the 5′ end of Chk1 was due to under-representation in the expression library, we sequenced inserts from 30 random clones which would have been present in the initial screen. This analysis indicated that all regions of Chk1 were represented. Another possible explanation for the absence of 5′ antisense inserts identified in yeast was a potential lethal effect of these expression plasmids. To examine this possibility, we expressed three different antisense RNAs, complementary to the 5′ half of the Chk1 part of the fusion mRNA, in the Chk1–lacZ gene-expressing strain. Transformants obtained for each plasmid displayed neither an adverse growth phenotype nor a reduction in β-galactosidase activity. We conclude that the yeast genetic screen can be used to find GSCs for human genes and that the antisense inserts display a target gene-specific alignment pattern.

To further examine the potential of the genetic screen to define sub-regions of a target RNA sensitive to regulation by trans-acting RNAs, we constructed c-myc- and Chk1-specific libraries containing shorter inserts ranging in size from 100 to 400 bp, and screened these in their respective fusion strains. β-galactosidase assays on 56 light blue transformants for each gene showed that the resident GSCs suppressed c-myc–lacZ and Chk1–lacZ gene expression by 20–81% and 21–64%, respectively. The inserts identified on the GSCs were found to be composed of either single fragments or greater than one fragment (Fig. 3). For the c-myc target, a total of 12 different single antisense inserts were identified and all of these originated from two sub-regions of exon 2, consistent with the results obtained with the larger insert library. In the case of Chk1 GSCs, six of the 25 inserts characterized contained unique single antisense fragments and these aligned with a 408 bp region at the 3′ end of the Chk1 sequence. From these results, it is clear that the genetic screen can be used to define potential domains to target for complementary RNA-dependent suppression (Fig. 3). An unexpected result was the identification of GSCs with chimeric inserts as the most effective constructs for both target genes. The high frequency of these chimeric GSCs was not due to over-representation in the expression libraries and, instead, demonstrates the power of the genetic screen for identifying novel GSCs.

Figure 3.

Single and chimeric inserts on GSCs identified from the c-myc expression library containing smaller inserts. The inserts are aligned with the c-myc target sequence and categorized into three sub-groups. Abbreviations and symbols are as indicated in Figure 2. For chimeric inserts, the number within the square brackets defines the position of the fragment within the GSC insert. The asterisks indicate those inserts tested in human cells.

Yeast-derived GSCs function to regulate gene expression in human cells

The ultimate test of the GSCs found in the yeast screen for the human target genes was their activity in human cells. For this test, we selected nine c-myc-specific GSCs representing the various classes identified (Figs 2 and 3). In addition, we also chose four c-myc expression plasmids, containing antisense inserts, which were shown to be ineffective in the yeast screen. Each of the inserts was subcloned into a mammalian expression vector under control of the ecdysone-inducible heat shock promoter. Initially, these constructs were examined for their effectiveness in regulating expression of the c-myc–lacZ fusion gene in a transient co-transfection assay using 293 cells. Eight out of the nine c-myc GSCs tested reduced β-galactosidase activity in human cells to the same degree or greater than in yeast cells (Fig. 4A). Of the four negative control constructs, two were found to reduce β-galactosidase activity by 50% in the co-transfection assay, suggesting that a subset of GSCs effective in human cells were not functional in yeast (Fig. 4B). These results indicate that at least 90% of the c-myc-specific GSCs from the yeast screen were capable of regulating expression of the fusion target in human cells and that the fusion gene concept was valid.

Figure 4.

β-galactosidase activity in human cells transiently co-transfected with the c-myc–lacZ fusion gene and (A) c-myc GSCs identified using the yeast screen or (B) c-myc plasmids containing inserts that did not suppress fusion gene expression in yeast. A total of 1 × 10⁷ 293 cells were co-electroporated with the c-myc–lacZ target plasmid and independent c-myc library plasmids at a ratio of 1:40. At 24 h post-transfection, ponasterone A was added to a final concentration of 30 µM to induce transcription of the inserts on the c-myc library plasmids. Cells were incubated for an additional 48 h, harvested and analyzed for β-galactosidase activity and GFP expression. Each value represents three replicates of three independent electroporations. The β-galactosidase activity is expressed relative to the control (C) vector.

As a second test of the constructs, their efficacy in controlling expression of the endogenous c-myc gene was investigated. Following transient transfection of 293 cells and induction of expression of plasmid inserts, plasmid-containing cells were isolated and examined for c-myc protein levels. The two negative control plasmids and constructs encoding an antisense RNA complementary to two (pFL) or three (pFFL) exons of c-myc mRNA were unable to affect the level of c-myc protein. In contrast, among the yeast-derived GSCs, two of the nine constructs were shown to reduce c-myc protein by 45% (Fig. 5). One of these plasmids (CM-4) encoded for an antisense RNA complementary to 961 bases of the 5′ end of exon 2 of c-myc mRNA, while the other GSC (CM-40) produced a chimeric antisense RNA. For all of the plasmids tested, we showed that the respective RNAs were expressed at similar steady-state levels and that cells containing these plasmids did not display significant changes in the level of the c-myc target mRNA (data not shown). This analysis showed that 25% of the GSCs found using yeast were effective at controlling c-myc gene expression in human cells, and that these GSCs were more effective than a rationally designed antisense gene (pFFL in Fig. 5) previously shown to reduce c-myc protein (21).

Figure 5.

c-myc protein levels in human cells conditionally expressing c-myc antisense RNA. A total of 3 × 10⁷ cells were electroporated using 40 µg of the c-myc plasmids. Ponasterone A was added as in Figure 4. At 72 h post-transfection, a total of 2 × 10⁶ GFP-expressing cells were isolated by FACS and analyzed for c-myc protein levels. Each histogram represents four to eight replicates from two independent transfections. The c-myc protein level is expressed relative to the control (C) vector.

DISCUSSION

A key requirement for assigning cellular function to a gene sequence is the capacity to control the expression of the gene in vivo. At present, a common way to achieve this in mammalian cells is by using gene constructs encoding trans-acting RNAs or proteins. However, despite the conceptual simplicity of this approach, the success of these rationally designed constructs has been extremely limited due to the many variables that can impact on these forms of gene regulation. To overcome these limitations, we have developed an in vivo genetic assay for rapidly screening >10⁵ different gene-specific plasmids for the rare GSCs effective in human cells. The assay is universal and particularly useful for those target genes without an easily assayable phenotype or for which there is no known function.

The key steps in the yeast genetic screen on which the approach is based have all been demonstrated in the present study. The fission yeast S.pombe was selected as the host based on its similarity with human cells in terms of gene regulation and the fact that this yeast maintains one plasmid type per cell and is easily transformed and grown. The establishment of a link between intracellular β-galactosidase activity and the lacZ gene-encoded blue colony color was essential to allow phenotypic screening for rare GSCs from large pools of plasmids. The screen was made universal by using target–lacZ gene fusions and screening target gene-specific libraries for GSCs using the reporter phenotype. This overcame the need to be dependent on target gene-associated phenotypes for screening. The utilization of random fragment libraries eliminated bias in plasmid construction and provided a wide range of different constructs from which to select. Characterization of the GSCs provided information about the origin of the inserts and revealed gene-specific patterns of alignment of these sequences, suggesting that these regions of the target gene were important for post-transcriptional gene regulation. The yeast-derived GSCs were easily tested in a transient assay in human cells to identify those most effective against the endogenous target gene. In the case of the c-myc gene, a screen of 3 × 10⁴ different plasmids in yeast produced two GSCs effective at strongly reducing c-myc protein levels in human 293 cells. The rare nature of these GSCs in the expression library, and the fact that these constructs could not be rationally designed, indicates that the genetic screen is effective and essential for finding functional GSCs for use in human cells. In addition, these GSCs were more effective against c-myc than empirically designed plasmids (21).

Artificially designed antisense RNAs are commonly used tools for suppressing the expression of specific genes or controlling viral replication. The efficacy of this approach is dependent on a number of parameters, one of which is the accessibility of the antisense RNA to the target mRNA in vivo. This accessibility is determined by both the intracellular localization of the two interacting RNAs and the ability of these RNAs to hybridize. A number of in vitro strategies have been developed to identify regions of a target mRNA accessible to hybridization with antisense sequences (22,23). A major limitation of these predictions is that the identified antisense RNAs or oligonucleotides do not always function in vivo. This is not surprising given that the complexity of RNA structures in vivo is determined by long-range secondary and tertiary interactions and RNA binding proteins. The fission yeast screen offers the distinct advantage of identifying effective antisense RNAs within the cellular milieu. Using this screen, we found a variety of GSCs for both c-myc and Chk1 which encoded antisense RNA sequences. Alignment of the inserts from these GSCs with the target sequence produced target gene-specific patterns for the encoded antisense RNAs, suggesting that the identified regions within the target mRNA may be accessible to antisense RNA-mediated gene suppression. Furthermore, by restricting the size of the library inserts, it was possible to further delineate sub-regions of the target mRNA sensitive to antisense RNA-mediated regulation in vivo. This was particularly obvious with the c-myc target in which two potential accessible regions were identifed within exon 2.

Using the random fragment expression libraries it was possible to identify novel trans-acting gene suppression molecules. For both the ura4 and c-myc target genes, we identified GSCs encoding sense RNAs. Sense RNA-mediated suppression has been demonstrated to be an effective way of controlling gene expression (24) and this is the first report of similar activity in yeast. The genetic screen using expression libraries with smaller fragments produced GSCs with chimeric inserts specific for both c-myc and Chk1. Chimeric antisense RNAs have not been commonly used in empirical approaches to designing effective intracellular inhibitors. The advantages of such antisense RNAs are the ability to hybridize with more than one region of the target mRNA and the potential for one of the antisense sequences to act as a facilitator of other antisense sequences contained on the same chimeric RNA (25). A biological approach such as the one we have described is the only way to select such effective and novel chimeric gene constructs.

The present cell-based screens for GSCs using mammalian cells are limited by the number of target genes that can be tested, the requirement for gene-specific phenotypes and the labor-intensive nature of construct delivery (8,9). With the advent of genomics, the number of novel genes being identified is rapidly increasing, while the functions for these sequences remain undetermined. The yeast genetic screen overcomes all of the limitations associated with mammalian cell screens and, most importantly, provides the high throughput capacity required for linking effective GSCs with target gene sequences for use in functional studies. Furthermore, each step in the screen is amenable to robotic automation.

This genetic screen can have several applications. In the first instance, its simplicity allows for the identification of GSCs for a large number of target genes and rapid testing of these plasmids in human cells. At present, we have isolated GSCs for a number of human target genes and are screening these in cell culture. This level of throughput is essential for target validation programs where multiple candidate genes are being evaluated as potential novel drug targets or for their role in the process of a human disease. A second application of the screen involves the identification of GSCs for use in functional genomics to assign cellular function to orphan genes, particularly in organisms where genetic study is difficult. Finally, the yeast-derived GSCs can be used to genetically mimic proposed drug–target interactions and provide information regarding drug toxicity and specificity for the target. These are just three important application areas of this approach for the isolation of rare effective GSCs from large random expression libraries.