Identification and Validation of a Gene Involved in Anchorage-Independent Cell Growth Control Using a Library of Randomized Hairpin Ribozymes

Genomics 66, 274 –283 (2000) doi:10.1006/geno.2000.6230, available online at http://www.idealibrary.com on

Identification and Validation of a Gene Involved in AnchorageIndependent Cell Growth Control Using a Library of Randomized Hairpin Ribozymes Peter J. Welch, Eric G. Marcusson, Qi-Xiang Li, Carmela Beger,* Martin Kru¨ger,* Chen Zhou, Mark Leavitt, Flossie Wong-Staal,* and Jack R. Barber 1 Immusol Incorporated, 3050 Science Park Road, San Diego, California 92121; and *Department of Medicine, University of California, San Diego, La Jolla, California 92093 Received December 2, 1999; accepted April 10, 2000

We have developed a library of hairpin ribozyme genes that can be delivered and expressed in mammalian cells with the purpose of identifying genes involved in a specific phenotype. By applying the appropriate phenotypic selection criteria in tissue culture, we can enrich for ribozymes that knock down expression of an unknown gene or genes in a particular pathway. Once specific ribozymes are selected, their target binding sequence is used to identify and clone the target gene. We have applied this technology to identify a putative tumor suppressor gene that has been activated in HF cells, a nontransformed revertant of HeLa cells. Using soft agar growth as the selection criteria for gain of transformation, we have isolated ribozymes capable of triggering anchorage-independent growth. Isolation of one of these ribozymes, Rz 568, led to the identification and cloning of the human homologue of the Drosophila gene ppan, a gene involved in DNA replication, cell proliferation, and larval development. This novel human gene, PPAN, was verified as the biologically relevant target of Rz 568 by creating five additional “target validation” ribozymes directed against additional sites in the PPAN mRNA. Rz 568 and all of the target validation ribozymes reduced the level of PPAN mRNA in cells and promoted anchorage-independent growth. Exogenous expression of PPAN in HeLa and A549 tumor cells reduced their ability to grow in soft agar, underscoring its role in regulating anchorage-dependent growth. This study describes a novel method for gene discovery where the intracellular application of hairpin ribozyme libraries was used to identify a novel gene based solely on a phenotype. © 2000 Academic Press

INTRODUCTION

Identifying a gene or gene family responsible for a particular phenotype is crucial to the deciphering of 1 To whom correspondence should be addressed. Telephone: (858) 677-0182. Fax: (858) 677-0587. E-mail: [email protected].

0888-7543/00 $35.00 Copyright © 2000 by Academic Press All rights of reproduction in any form reserved.

any biological mechanism and our understanding of disease. Differential expression arrays and other chip technologies have greatly advanced our capabilities in deducing which genes are turned on or off in a particular phenotype. However, since these new technologies identify all differentially regulated genes, they fall short of identifying which gene “causes” the phenotype and which genes are simply downstream “effects.” This report describes the use of a ribozyme gene vector library that is capable of specifically identifying genes that are directly involved in producing a particular phenotype. Ribozymes (Rz) have been used successfully to specifically knock down intracellular expression of a variety of viral and cellular targets (for reviews see James and Gibson 1998; Welch et al., 1998). To create a hairpin ribozyme that binds to and cleaves a particular substrate, the bases of its binding arms (helixes 1 and 2, see Fig. 1) are made complementary to the sequences 5⬘ and 3⬘ of a GUC in the desired substrate RNA. This has been used successfully to create ribozymes for (1) antiviral treatment, (2) reducing specific cellular gene expression, and (3) target validation (for reviews see James and Gibson, 1998; Welch et al., 1998). Now, by randomizing the binding arms in the ribozyme, one can create a “library” of ribozymes capable of cleaving any mRNA that contains a GUC. The gene discovery strategy begins by stably delivering a library of ribozyme genes to cells in culture, with at least one different ribozyme gene per cell. By selecting the library-transduced cells for a particular phenotype of interest, we can identify the ribozyme(s) that is responsible for generation of the phenotype. Due to the duration of typical phenotypic selections, most applications require stable ribozyme expression, which can be achieved by transfection or viral vector transduction. Once the library cells have gone through one round of selection, ribozyme genes can be “rescued” and reintroduced into fresh cells. By repeating this

274

RIBOZYME LIBRARIES IDENTIFY A HUMAN GROWTH CONTROL GENE

process, we enrich for ribozymes that target genes involved in the phenotype. The identification of the phenotypically relevant gene is based upon the known complementarity of the binding sequence of the selected ribozyme, consisting of the two target binding arms and the GUC. The resulting 16-base ribozyme sequence tag (or RST) is used to identify the gene by any of several conventional methods: (1) BLAST search of the gene databases; (2) as a primer in 3⬘ and 5⬘ RACE, and/or (3) as a probe to screen gene libraries. Once a gene is cloned, ribozymes targeting additional sequences unique to the candidate gene can then be employed to “validate” that the gene is indeed involved in producing the phenotype. As our first test of this technology, we wanted a system with the following criteria: (1) a positive, selectable cell culture phenotype, (2) low background, and (3) a biologically relevant target. Boylan et al. (1996) described an interesting stable transformation revertant, HF, isolated from HeLa cervical cancer cells following exposure to the mutagen EMS. These HF cells have a nontransformed appearance with flat, nonrefractile morphology. HF cells have lost both their ability to form colonies in soft agar and their tumorigenic potential in nude mice. This was not due to loss of HPV-18 E6 or E7 proteins, and in fact ectopic expression of E6 and E7 did not restore the transformed phenotype. Furthermore, fusion of HF with the parental HeLa cells resulted in the nontransformed phenotype, suggesting a dominant phenotype in HF cells that the authors postulated to be due to activation of a tumor suppressor (Boylan et al., 1996). This system fits each criterion: (1) a positive, selectable (soft agar) growth advantage, (2) up to 2000-fold difference from background, and (3) the potential to discover a tumor suppressor capable of reversing transformation of cancer cells. Here we describe the application of a hairpin ribozyme library and the selection of ribozymes that promote soft agar growth of HF cells. One active ribozyme led to the cloning of a novel human gene that we have named PPAN, since it is the human homologue of the Drosophila Peter Pan gene ppan, which is involved in DNA replication, cell proliferation, and cell– cell communication (Migeon et al., 1999). Additional ribozymes targeting other sites within the PPAN mRNA validated its function by promoting anchorageindependent growth. Finally, reintroduction of the PPAN cDNA into transformed HeLa cells blocked their anchorage independence, thereby establishing this technology as a viable method for gene identification based on function. MATERIALS AND METHODS Construction of the ribozyme library. To generate the Rz plasmid library, the parental vector, pLHPM, was digested with restriction enzymes MluI and BstBI, and the ⬃6-kb vector DNA was purified by agarose gel electrophoresis. To create the ribozyme library insert, three oligonucleotides (Integrated DNA Technologies) were an-

275

nealed: Oligo 1 (5⬘pCGCGTACCAGGTAATATACCACGGACCGAAGTCCGTGTGTTTCTCTGGTNNNNTTCTNNNNNNNNGGATCCTGTTTCCGCCCGGTTT3⬘), Oligo 2 (3⬘ATCCTCCATTATATGGTGCCTGp5⬘), and Oligo 3 (3⬘GGACAAAGGCGGGCCAAAGCp5⬘). The oligos are synthesized with a 5⬘ phosphate. Following oligonucleotide annealing, the resulting DNA fragment contained compatible overhanging bases for MluI and BstBI sites necessary for ligation into the vector. Electrocompetent bacteria (Gibco BRL) were transformed with the ligation mixture, and a total of 5 ⫻ 10 7 ribozymecontaining bacterial colonies were obtained. To evaluate statistically the “randomness” of the library, we used the formula for a two-sided approximate binomial confidence interval: E ⫽ 1.96* (P*(1-P)/N), where P is the expected proportion of each nucleotide in a given position (this value for DNA bases equals 25% or P ⫽ 0.25), E is the desired confidence interval around P (i.e., P ⫾ E), and N is the required sample size. Since we wanted to know the proportion of each base within 5% (E ⫽ 0.05), the required sample size is 289. Since each ribozyme molecule contains 12 independent positions, the number of individual ribozyme genes that need to be sequenced out of the library equals 289 divided by 12, or about 25 molecules. Cell culture, transfection, and soft agar assays. Both HeLa and HF cells were cultured at 37°C in DMEM (Gibco BRL) supplemented with 10% FBS, L-glutamine, sodium pyruvate, and antibiotics. For stable library delivery, 1 ⫻ 10 8 HF cells were transfected with the ribozyme plasmid library using standard calcium phosphate methods. Twenty-four hours posttransfection, cells were selected with G418 (500 ␮g/ml) for 2 weeks. Approximately 1 ⫻ 10 7 stable transfectants were generated as determined by colony formation, and all colonies were pooled prior to soft agar selection. Soft agar selection of the library was performed in forty 150-mm 2 plates, prelayered with 12 ml of a 1:1 mixture of 1.2% Select Agar (Gibco BRL): 2⫻ MEM/20% FBS. After the prelayer had solidified, 3 ⫻ 10 5 cells were plated in the “cell layer” consisting of 12 ml of a 1:1 mixture of 0.6% Select Agar: 2⫻ MEM/20% FBS. As a control, 1.2 ⫻ 10 6 HF cells stably transfected with an unrelated Rz, CNR3, were plated into four 150-mm 2 soft agar plates. As comparisons, 3 ⫻ 10 5 HeLa or HF parental (untransfected) cells were plated into one 150-mm 2 plate each. The cell layers were allowed to solidify prior to incubation at 37°C. Soft agar plates were fed once per week by layering 8 ml of a freshly prepared 1:1 mixture of 0.6% Agar Select: 2⫻ MEM/20% FBS. Colonies were visible by 2 weeks and picked for expansion and analysis at 3 weeks. For the second round of soft agar selection, 300 colonies from the library expressing cells, 100 colonies from the CNR3 HF control, or 30 colonies from either HeLa or HF parental cell lines were picked from the first round, pooled, and expanded for 2 weeks in normal medium. Second-round soft agar selection was performed with 3 ⫻ 10 5 cells in one 150 mm 2 plate for each cell type. For transfections requiring hygromycin (Gibco BRL) or puromycin (Sigma) selection, final concentrations were 250 and 1 ␮g/ml, respectively. Ribozyme rescue. Ribozyme genes were rescued by two independent methods, viral rescue and PCR rescue. Viral rescue was performed by transient transfection of the Rz-expressing cells using the lipid transfection reagent LT1 (Mirus, Panvera Corp.), 6.3 ␮g each pGag-pol (Moloney gag and pol), and pVSV-G (vesicular stomatitis virus G glycoprotein to serve as the retroviral envelope) per 100-mm 2 dish, according to the manufacturer’s instructions. Twenty-four to 48 h later, viral supernatant was recovered and filtered (0.2 ␮m) prior to transduction of fresh HF cells in the presence of 4 ␮g/ml polybrene. PCR rescue was performed on genomic DNA, isolated from the selected cells using the QIAamp Blood Kit (Qiagen). PCR primers within the vector amplified a 300-bp fragment containing the ribozyme genes. The PCR product (containing a pool of Rz genes) was digested with BamHI and MluI and ligated into pLHPM digested with the same enzymes. The resulting bacterial clones were pooled, and purified DNA was used for cell transfections.

276

WELCH ET AL.

FIG. 1. Schematic illustration of the hairpin ribozyme library. (A) The hairpin ribozyme consists of a 52-nucleotide RNA molecule (shaded, in uppercase letters) that binds and cleaves an RNA substrate (lowercase letters). Two helices, helixes 1 and 2, form between the ribozyme and the substrate via Watson–Crick basepairing. Helix 2 must be 4 bp while helix 1 can be of variable length. Empirically, we have found that a helix 1 of 8 bp is optimal for high intracellular specificity and fast enzymatic turnover (unpublished results). The substrate RNA must contain a GUC for maximal cleavage, and cleavage occurs immediately 5⬘ of the G as indicated. The catalytic, but not substrate binding, activity of the ribozyme can be disabled by mutating the AAA in loop 2 to CGU, as shown. (B) The ribozyme library was cloned into the retroviral vector, pLHPM, as described under Materials and Methods. The vector consists of the 5⬘LTR driving expression of the neomycin resistance gene and the SV40 early promoter driving puromycin resistance, thus allowing dual selection when necessary. Ribozyme gene expression is under the control of the human tRNA val promoter, as shown. RNA and Northern blot analysis. Total cellular RNA was prepared using the RNAgents Kit (Promega), and 20 ␮g total RNA was electrophoresed on formaldehyde gels using standard procedures. RNA was transferred to Zeta-Probe membranes (Bio-Rad) by capillary action, as recommended by the manufacturer. Northern hybridizations were performed with QuikHyb (Stratagene) according to the manufacturer’s instructions, using the full-length PPAN cDNA random-prime-labeled with the High Prime DNA labeling kit (Boehringer Mannheim). Northern signals were quantitated by PhosphorImager (Molecular Dynamics). Gene cloning and validation. For 5⬘ RACE, poly(A) ⫹ mRNA was prepared from HF cells using the Poly(A)Pure kit (Ambion; Austin, TX). The mRNA was used as template for the Marathon cDNA amplification kit (Clontech, Palo Alto, CA). Briefly, a first-strand cDNA was synthesized from the mRNA and used as a template in a second-strand synthesis reaction. The ends of the double-stranded cDNAs were made blunt with Klenow enzyme, and adapters were ligated to the blunt ends. 5⬘ RACE was performed with a primer complementary to the adapters (AP1, 5⬘CCATCCTAATACGACTCACTATAGGGC3⬘) and a primer that matches the target recognition site of Rz 568 (568 sequence is underlined) (5⬘CGATGCTCCTCTAGACTCGAGGGTACCACCTCCCCGACNCCCT3⬘). The PCR was performed with primer concentrations of 200 nM, AmpliTaq Gold polymerase (Perkin–Elmer, Branchburg, NJ), and the following cycle parameters: 94°C for 10 min; 5 cycles of 94°C for 30 s, 68°C for 4 min; 28 cycles of 94°C for 30 s, 59°C for 30 s, 68°C for 4 min; 72°C for 7 min. The reactions products were gel-purified and cloned into a TA cloning vector (Invitrogen, Carlsbad, CA). 3⬘ RACE was performed using HF poly(A)⫹ mRNA in a reverse transcription reaction using an anchored poly(T)-TAG primer (5⬘GGCCACGCGTCGACTAGTACTTTTTTTTTTTTTTTTTV3⬘, where V is G, A, or C) using Superscript reverse transcriptase (Gibco BRL) according to the manufacturer’s instructions. PCR was performed using a genespecific primer for PPAN (5⬘CGGCTCACCGAGATCGGCCC3⬘) and a primer complementary to the poly(T) TAG region (5⬘GGCCACGCGTCGACTAGTACT3⬘) using the following cycle parameters: 94°C for 10

min; 35 cycles of 94°C for 30 s, 55°C for 30 s, 72°C for 4 min. The resulting PCR product was gel-purified and cloned into a TA cloning vector. To obtain the complete PPAN cDNA, the 5⬘ and 3⬘ RACE products were joined together using the common HgaI site at nucleotide position 958. The final ligation product was verified by overlapping sequencing reactions in both directions. To create the frameshift mutant of PPAN, the unique BssHII site at nucleotide position 135 (amino acid 12) was digested, and the overhanging ends were filled in with Klenow polymerase. The resulting blunt ends were religated, thus shifting the coding frame by 1 base. The frameshift was verified by DNA sequencing, and this new reading frame continues for 53 amino acids before a translation stop codon. For PPAN expression in HeLa and A549 cells, the PPAN cDNA was blunt-end-cloned in both the sense and the antisense orientations into the unique HpaI site in pLNCX (Clontech). Cloning of the target validation ribozymes (as well as d568) was performed as previously described (Welch et al., 1996, 1997). Following digestion with BamHI and MluI, the ribozyme genes were ligated into pLHPM-based vectors digested with the same enzymes. To allow for their coselection posttransfection, target validation ribozymes were cloned into identical plasmids except for their selectable marker (TV1, hygromycin R; TV2 and TV4, neomycin R; TV3 and TV5, puromycin R). Ribozyme sequences were verified by DNA sequencing prior to cell transfections.

RESULTS

Construction of a Ribozyme Gene Expression Library A replication-deficient retroviral vector was constructed for stable delivery and expression of the randomized ribozyme gene library (pLHPM, Fig. 1B). The vector confers both neomycin and puromycin resistance to stably transfected or transduced cells. Use of a

277

RIBOZYME LIBRARIES IDENTIFY A HUMAN GROWTH CONTROL GENE

retroviral vector allowed for application of the library in either plasmid or viral form and allowed simple and efficient ribozyme gene rescue and reapplication during subsequent rounds of selection. Stable expression of functionally active ribozymes in mammalian cells has been described using a variety of eukaryotic promoters (for reviews see James and Gibson, 1998; Welch et al., 1998). We and others have previously found the human tRNA val promoter to be optimal for intracellular ribozyme gene expression (Bertrand et al., 1997; Welch et al., 1996, 1997 and unpublished observations). This promoter is ideal for the ubiquitous, constitutive, high-level expression of small untranslated RNAs. Following the cloning of the randomized hairpin ribozyme genes into pLHPM (see Materials and Methods), we evaluated the “randomness” of the plasmid library by both statistical and functional analyses. A complete ribozyme library of this design, with 12 random positions, would contain 4 12, or 1.67 ⫻ 10 7, different members. For the statistical analysis, 40 individual bacterial transformants were picked and sequenced (see Materials and Methods). This allowed an evaluation of the complexity of the library without having to sequence each library member manually. The frequencies of the four nucleotides, with 95% confidence limits, in the random positions of the ribozymes were as follows: G, 22.3 ⫾ 6.1; A, 31.9 ⫾ 7.0; T, 27.3 ⫾ 7.8; and C, 18.0 ⫾ 5.1. Since the expected frequency for each base is 25%, each base appears to be randomly represented (except for C, which may be lower than expected). These variations most likely result from the unbalanced incorporation of nucleotides during the chemical synthesis of the oligonucleotides and could reduce the complexity of the library. Thus, ribozymes against Grich target sequences may be underrepresented in the library. For a functional evaluation of the library’s complexity, we used in vitro cleavage to determine whether ribozymes that target known RNA substrates were present in the library pool. This involved transcribing the entire ribozyme library in one reaction from a T7 promoter recognition sequence and then testing the pool’s ability to cleave a variety of different RNA substrates (of both cellular and viral origin) in vitro. Six of seven known RNA targets were properly and efficiently cleaved by the in vitro transcribed library (not shown). This qualitative analysis suggested a sufficiently complex library of ribozyme genes, and the lack of cleavage of one target of seven may reflect the slight nonrandomness suggested by the base composition described above. Specific Ribozymes Selected from the Library Promote HF Soft Agar Growth HF cells were transfected with the ribozyme library, and approximately 1 ⫻ 10 7 individual stable transfectants were produced following G418 selection. To de-

TABLE 1 Enrichment of Ribozyme-Expressing Cells That Grow in Soft Agar Cell type

1st round of selection a

2nd round of selection

HeLa HF parental HF control Rz HF Rz library

50,000 10 20 45

50,000 25 48 15,000

a

Number of soft agar colonies per 10 5 cells plated.

termine whether any of these had regained their transformed phenotype, 1.2 ⫻ 10 7 library-transfected cells were plated into 40 soft agar plates. As a control, HF cells were stably transfected with an unrelated ribozyme, CNR3, which targets the negative RNA strand of human hepatitis C virus. Following 3 weeks in soft agar, colonies appeared in both the Rz library and the CNR3 control Rz; however, the library-expressing cells produced 2.5-fold more colonies than the control Rz and 4-fold more than untransfected HF cells (Table 1). To determine whether the cells that grew as colonies in soft agar had a stable phenotype, pools of soft agar colonies from the library and each of the controls were picked from the first round, expanded, and replated for a second round of soft agar selection. Both the HF parental and the HF-control cells showed only modest (2- to 3-fold) enrichment in soft agar growth, indicating that colony growth in the controls was due mostly to unstable, stochastic processes. In contrast, the libraryexpressing cells showed a dramatic 300-fold increase, suggesting that ribozymes from the library stably enhanced soft agar growth (Table 1). One advantage of our experimental approach is that pools of Rz genes can readily be “rescued” from the selected cells and reintroduced into fresh HF cells. This process allows for the further enrichment of ribozymes that truly cause the phenotype as opposed to those that are found coincidentally in any spurious background colonies. Two methods of ribozyme gene rescue (viral rescue and PCR rescue) were performed in parallel on the pool of 300 colonies from the first round of soft agar. The first method, viral rescue, takes advantage of the fact that the Rz expression cassette is located between packagable retroviral LTRs. Selected cells were transiently transfected with the retroviral gag, pol, and VSV-G envelope genes, resulting in the secretion of viral particles that contain the selected Rz gene(s). Fresh HF cells were then transduced with the infectious supernatant, selected with G418, and plated into soft agar. Sequence analysis from the resulting individual soft agar colonies revealed enrichment of one ribozyme, called Rz 568, present in 3 of 10 clones. The second method of Rz gene rescue was by PCR of the genomic DNA from the selected pool of cells, followed by batch recloning of the Rz genes into the pLHPM vector. Fresh HF cells were then stably transfected with this plasmid pool and plated into soft agar. In this

278

WELCH ET AL.

FIG. 2. Expression of catalytically active Rz568 confers anchorage independence to HF cells. HF cells were stably transfected with pLHPM vector only (HF), or pLHPM expressing either Rz 568 (HF-568) or catalytically disabled 568 (HF-d568). Transfected cells were plated into soft agar and the colonies resulting from two rounds of soft agar selection are shown. As a positive control for soft agar growth, untransfected HeLa cells were plated in soft agar in parallel.

round of rescue and reintroduction, Rz 568 was present in 5 of 10 soft agar colonies. The apparent enrichment of Rz 568 compared to other ribozymes present in viral and PCR rescued soft agar colonies suggested that Rz 568 might be particularly effective at conferring a selective growth advantage to HF cells plated in soft agar. To verify this, we stably transfected the 568 ribozyme gene into fresh HF cells. As a control, the catalytically disabled form of 568 (d568, Fig. 1A) was similarly cloned and transfected. After two rounds of selection in soft agar, Rz 568 (but not d568) significantly promoted HF soft agar growth (Fig. 2), verifying that Rz 568 alone can confer anchorage independence. Equally important, since d568 had no effect, we can conclude that the catalytic activity of Rz 568 is required for the phenotype, presumably by cleaving an mRNA involved in maintaining anchoragedependent growth in HF cells. In addition to Rz 568, only one other ribozyme, Rz 619, was repeated in multiple soft agar colonies, suggesting that the remaining ribozymes might be less

active stimulators of anchorage-independent growth. Rz 619, but not its disabled version, was confirmed to increase soft agar colony growth significantly (data not shown). Our initial attempts to clone the target gene for Rz 619 have been unsuccessful. Therefore, we have focused on Rz 568, while efforts are continuing to clone the target gene for Rz 619. Rz 568 Targets a Human Gene Related to the Drosophila peter pan Gene Since ribozymes recognize their cognate targets by sequence complementarity, the sequence of a ribozyme that causes a phenotype through its catalytic activity predicts a sequence tag that can be used to clone the target gene. This RST is 16 bases long, consisting of the two target binding arms (helixes 1 and 2), the requisite GUC in the target, and one degenerate base (see Fig. 1A). The RST can be used to BLAST search the gene and EST databases, and it can be used as a primer for 3⬘ and 5⬘ RACE. BLAST searches of the public data-

RIBOZYME LIBRARIES IDENTIFY A HUMAN GROWTH CONTROL GENE

bases yielded several matches; however, most were genes of unknown function, and none appeared related to anchorage-dependent growth. Since none of the database matches was an obvious target gene, we chose to clone the target gene from HF mRNA using the 568 RST as a primer for 5⬘ RACE (see Materials and Methods). Several PCR products were generated from HF mRNA. To verify the presence of a complete 568 target site in these messages, we created larger gene-specific primers upstream of the potential 568 RST sites. Next, we used these primers to perform 3⬘ RACE. Unfortunately, sequencing data of many of these 3⬘ RACE products revealed that most did not contain the complete 568 RST and were therefore due to mispriming events in PCR (not shown). Only one of the fragments demonstrably contained the 568 RST as determined by sequencing the 3⬘ RACE product. This cDNA had matches to several incomplete cDNAs in the human EST databases. More interestingly, the putative translated amino acid sequence matched a newly described Drosophila gene, peter pan (ppan), that was recently shown to be involved in cell growth, DNA replication, and possibly cell– cell communication during development (Migeon et al., 1999). To clone the rest of this human cDNA, we used a 20-bp gene-specific primer in a 3⬘ RACE and ligated the 5⬘ and 3⬘ halves together to generate the full-length cDNA of 1653 nucleotides (see Materials and Methods). The cDNA contains a Kozak ATG translation start site at nucleotide position 103, which is believed to be the start of the protein reading frame due to the fact that there is no other ATG upstream in the same open reading frame. The region potentially codes for a 473amino-acid protein with a calculated molecular mass of approximately 53 kDa. The binding of Rz 568 to the cloned PPAN substrate target mRNA is shown in Fig. 3A. At this point it is unclear if we have obtained the bona fide 5⬘ end of the message, although it appears we have cloned the entire protein coding region. We have named this human homologue of ppan, PPAN. This gene appears to be highly conserved evolutionarily and includes homologues in mouse, Drosophila, Caenorhabditis elegans, yeast and Arabidopsis (Migeon et al., 1999; see Fig. 3C). It should be noted that the previously postulated PPAN and murine PPAN amino acid sequences, based on compilation of EST fragments (Migeon et al., 1999), differ slightly from our reported sequence. This is most likely due in part to the incompleteness of the available ESTs and their proposed compilation. Additionally, several human ESTs in the public databases suggest several alternatively spliced variants of PPAN. Our own RT-PCR results indicate the full-length mRNA to be the predominant species in HeLa and HF, cells; however, we can detect at least one splice variant at very low levels (not shown). Finally, we wished to determine whether Rz 568 affected the mRNA levels of PPAN in HF cells. Northern analysis using the full-length PPAN cDNA identified a single 1.6-kb mRNA. Cells stably expressing Rz

279

568 consistently showed a 30 –35% reduction in PPAN message relative to a G3PDH internal control (Fig. 3B) while neither d568 nor the unrelated Rz CNR3 had any significant effect on PPAN mRNA levels. Knowing that ppan is an essential gene in Drosophila (Migeon et al., 1999), it was not surprising that the selected ribozyme, Rz 568, exhibited only a partial knock-down in PPAN expression. Presumably, a more complete reduction in PPAN mRNA would not be tolerated in a cell and therefore would be selected against. Interestingly, we also consistently saw a 15–20% difference in PPAN levels in HeLa vs HF cells. It is unclear how significant this difference is between cell types; however, it does indicate that small differences in the level of PPAN expression may contribute to the phenotypic differences observed between HeLa and HF cells. Validation of the Role of PPAN in AnchorageDependent Growth To validate whether the Rz 568-mediated knockdown of PPAN mRNA in HF cells was truly promoting soft agar growth, we designed several other ribozymes against additional, unique GUC target sites within the PPAN mRNA. Five “target validation” ribozyme sites (Table 2, TV1 through TV5) were chosen within PPAN. TV1, 2, and 3 were all located within 150 bases of the 568 Rz site, where we hoped the RNA secondary structure was sufficiently open and available for cleavage. TV4 and 5 were chosen near the 5⬘ end of the mRNA, at or before the ATG translation start site, which has been shown to be often accessible and vulnerable to ribozyme-mediated cleavage in vivo (for reviews see James and Gibson, 1998; Welch et al., 1998). To maximize the degree of PPAN knock-down, one or more TV Rz genes were stably transfected into HF cells using different selectable markers (see Materials and Methods), followed by soft agar selection. All TV transfections yielded prominent soft agar growth while transfection of a control Rz had no effect (Table 2), strongly suggesting that PPAN was indeed the phenotypically relevant target of the 568 Rz. As further confirmation, we also created three Rz against each of two different (not PPAN) ESTs of unknown function that were identified from a BLAST database search with the 568 RST. None of those six Rz, alone or in combinations of three in the same cell, showed any soft agar growth above background (not shown). These data further implicate PPAN’s involvement in the soft agar phenotype. Finally, as expected, each of the TV-transfected cell populations, but not the control, showed a reduction in PPAN mRNA following soft agar selection (Table 2), thus correlating the soft agar phenotype with a ribozyme-mediated gene knock-down. Exogenous Expression of PPAN in HeLa Cells Confers Anchorage Dependence If PPAN was indeed involved in preventing HF growth in soft agar, we hypothesized that increased

FIG. 3. Ribozyme 568 targets a human gene with a known Drosophila homologue. (A) Schematic illustration of the 568 ribozyme binding its target in the PPAN mRNA, with the cleavage site indicated. PPAN was cloned using sequential 5⬘ and 3⬘ RACE using the 568 Rz sequence as the initial primer (see Materials and Methods). (B) Reduction of PPAN mRNA in HeLa cells and in HF cells expressing the 568 ribozyme. Northern analysis was performed using HF parental cells, HeLa cells, and HF cells expressing ribozyme CNR3 (control), 568, or its catalytically disabled version, d568. PPAN mRNA levels were normalized to internal G3PDH mRNA, and values are reported as a percentage, where HF is set to 100%. Quantification was performed by PhosphorImager analysis (Molecular Dynamics), and data averaged from three to four independent experiments were plotted. (C) Alignment of PPAN amino acid sequence with putative homologues from Drosophila and mouse. Hs, Homo sapiens; Mm, Mus musculus (compilation of ESTs AI325663, AA756790, and AA575760); and Dm, Drosophila melanogaster (Migeon et al., 1999).

280 WELCH ET AL.

RIBOZYME LIBRARIES IDENTIFY A HUMAN GROWTH CONTROL GENE

281

TABLE 2

TABLE 3

Validation of PPAN in Anchorage Dependence and as the Functional Target of Rz 568

High-Level Overexpression of PPAN Can Induce Toxicity in HeLa, but Not HF Cells

Colony growth in soft agar a Ribozyme

1st round

2nd round

PPAN mRNA b

Control TV1 TV2 ⫹ TV3 TV4 ⫹ TV5

6 215 75 103

15 ⬎50,000 ⬎50,000 ⬎50,000

101 ⫾ 2 82 ⫾ 9 76 ⫾ 13 74 ⫾ 17

Number of soft agar colonies per 10 5 cells. PPAN Northern analysis from three independent experiments after the first round of soft agar selection, relative to G3PDH mRNA, ⫾ standard error.

Plasmid transfected a

HeLa colonies b

HF colonies

Vector control WT FS

1720 80 1440

760 840 680

a pIRESHyg (Clontech) containing the wildtype PPAN (WT) or a frameshift mutation (FS). b Number of colonies, per 150-mm 2 dish, following transfection and 2 weeks of hygromycin selection.

a b

expression of PPAN in transformed HeLa cells should block their ability to grow in soft agar. To this end, we stably delivered the PPAN cDNA to HeLa cells, under the control of the CMV promoter, using a retroviral vector. As a control, either the empty viral vector (LNCX) or the antisense of PPAN was delivered to HeLa cells. Following 2 weeks of G418 selection, no discernible differences were observed in the growth rate or morphology of the three different cell types and a sixfold overexpression of the transgene was verified by Northern analysis (not shown). Cells were then plated in soft agar and monitored for 3 weeks. Expression of PPAN, but not its antisense or the vector control, significantly reduced the ability of HeLa cells to grow in soft agar (Fig. 4). Interestingly, HeLa cells appear to be sensitive to the levels of PPAN expression. Under conditions where the PPAN cDNA was introduced by transfection, and transient overexpression reached nearly 20-fold over the endogenous PPAN mRNA level, PPAN expression induced toxicity (Table 3). Expression of the wildtype protein appeared to be responsible for this toxicity since expression of a frameshifted mutation had no

effect (Table 3). Interestingly, no toxic effect was observed in HF cells, suggesting that overexpression could lead to tumor-specific cell death. These results were obtained with multiple plasmid preparations using several different expression cassettes (not shown), so it is unlikely to be an experimental artifact. Furthermore, this observation may be significant considering that Migeon et al. (1999) reported a similar cell death following overexpression of wildtype ppan in actively dividing Drosophila wing imaginal discs. Finally, to determine whether PPAN-dependent suppression of anchorage-independent growth was restricted to the HeLa/HF system, we stably introduced the PPAN cDNA, via retroviral vector, into the lung carcinoma cell line A549. Similar to the effect seen in HeLa cells, A549 cells expressing PPAN (but not the vector control) exhibited a 50% reduction in their ability to grow in soft agar (not shown). Together, these results indicate that moderate expression of PPAN can promote anchorage dependence, in at least two transformed cell lines, and that down-regulation of the gene can lead to anchorage-independent growth. DISCUSSION

We have developed a library of hairpin ribozyme genes that can be delivered and expressed in mamma-

FIG. 4. Exogenous expression of PPAN in HeLa cells blocks their anchorage independence. Full-length PPAN cDNA was cloned into the retroviral vector pLNCX (Clontech). As a control, the PPAN cDNA was cloned in antisense orientation (see Materials and Methods). HeLa cells were stably transduced with retrovirus LNCX (vector control), LNC-PPAN sense, or PPAN antisense. Following G418 selection, cells were plated in soft agar and photographed after 20 days.

282

WELCH ET AL.

lian cells with the purpose of identifying genes that are necessary for a specific phenotype. This approach is in contrast to conventional “functional genomics” strategies that start with a gene sequence and attempt to assign a corresponding function to it. Our experimental approach allows us to select first for a desired phenotype and then to use the resulting RST to clone the gene responsible. The function of the putative gene is then validated by creating additional ribozymes against other unique sequences within its mRNA. This strategy has been applied to a nontransformed revertant of HeLa cells in an attempt to identify a dominant tumor suppressor gene. The result was the cloning of a novel human gene, PPAN, that is highly conserved evolutionarily and is involved in the establishment of anchorage-dependent growth. Successful application of this strategy requires a highly complex library. Since we have no sequence information about the target gene, the library must be complex enough to contain Rz against any mRNA. Theoretically our library contains 1.67 ⫻ 10 7 different members, capable of binding any mRNA sequence surrounding the requisite GUC. In practice, this number is most likely somewhat lower due to a number of factors. Our sequencing data indicated a bias for underincorporation of guanosine residues during synthesis of the oligonucleotides, leading to an underrepresentation of cytidine residues in the binding arms of the ribozymes. In our more recent libraries, we have nearly eliminated this bias by compensating for the concentrations of each nucleotide during synthesis of the random regions. Complexity is also lost at each step of production and application of the library, for instance, following transfection where we only obtained 1 ⫻ 10 7 stably transfected cells. Fortunately, hairpin ribozyme sites occur quite frequently, with GUC triplets occurring on average every 64 bases. If this is coupled with the fact that the human genome is estimated to code for roughly 1 to 2 ⫻ 10 5 genes, we believe that even a library that is not completely random will still contain one or more Rz against any potential mRNA target. This may explain why we isolated the 568 Rz but did not pull out any of the equally effective target validation Rz from the initial library screen. Naturally, there will be Rz that target essential genes that are required for cell survival. If these Rz are sufficiently active intracellularly, they will be naturally selected against. The result is a tendency toward identifying genes that are, perhaps, cell type-specific or genes that are unique to pathways not essential for general cell viability. Alternatively, Rz that do target critical cellular genes could be isolated if the efficacy of the Rz is less than optimal. Many factors contribute to Rz activity in vivo, including Rz binding sequence, integration site of the Rz expression cassette, target mRNA secondary structure, and intracellular localization of target and Rz. In fact, in a population of cells stably expressing the same ribozyme gene, we have

observed cell-to-cell variations in the levels of Rz expression and biological activity (unpublished observations). If the selected phenotype can be achieved by a partial knock-down of a particular gene, use of the library increases the likelihood of isolating the desired Rz-expressing cells. Therefore Rz can act as gene knock-downs instead of gene knock-outs and is illustrated in the identification of PPAN. According to the data generated in Drosophila and yeast, PPAN homologues are required cellular genes. As such, we most likely killed any cells where the anti-PPAN Rz were achieving 100% knockout. Indeed, Rz 568 and the target validation Rz in soft agar-selected cells rarely achieved more than a 40% knockdown of PPAN mRNA. Interestingly, the selected Rz 568 did not contain a complete helix 1 of 8 bases binding to the substrate. Instead, a helix 1 with only 7 bases was selected for, which may alter the overall efficiency of that Rz. That level of knock-down was clearly sufficient to promote anchorage-independent growth and for the selection of that phenotype. In fact, this selection step increases the power of our system. We were able to identify a gene with marginally increased expression (⬃20%) in HF cells when compared to HeLa cells. Many systems that look for genes that are differentially expressed between two different cell types (e.g., RNA arrays) would have missed this gene due to the noise in their assays. This is important because the best drug targets are probably gene products that can change a cellular phenotype with only moderate modulation of their activity or expression level. Much of the data generated in Drosophila and yeast suggest that ppan, or its yeast homologues SSF1 and SSF2, is part of a mechanism that controls cell– cell communication and actin cytoskeletal reorganization and how those signals are translated by the cell to promote (or prevent) cell proliferation (Kim and Hirsch, 1998; Migeon et al., 1999; Yu and Hirsch, 1995). This may reflect our observations in HF cells placed in soft agar. Under these unfavorable conditions, mechanisms active in HF cells sense their lack of substrate contact and prevent their proliferation. Indeed, these single cells appear to undergo apoptosis, similar to observations made in other soft agar systems (Orford et al., 1999). When Rz 568 reduces the level of PPAN in these cells, soft agar growth resumes. This indicates that PPAN is part of a pathway that provides a cell with information about its substrate contact and may be involved in the metastatic potential of transformed cells. Current studies are addressing the effect of Rz 568 in other nontransformed cells. Similar to observations made for ppan in Drosophila, we have found that overexpression of PPAN in certain cells can be toxic. Of particular interest is the fact that HeLa cells, but not HF, are killed by the overexpression of PPAN. Endogenous PPAN in HeLa does not appear to signal when the cell is on an inappropriate substrate, perhaps due to additional regulators down-

RIBOZYME LIBRARIES IDENTIFY A HUMAN GROWTH CONTROL GENE

stream. Overexpression of PPAN may override this control, sending a constitutive signal that the cell is inappropriately anchored. This hypothesis leads us to ask whether the PPAN-induced death occurs via an apoptotic pathway or perhaps some type of cell cycle arrest. Interestingly, Migeon et al. (1999) describe how ppan-induced cell death in Drosophila imaginal disc cells can be blocked/rescued by coexpression of the baculovirus P35 protein, which blocks cell death by inhibition of caspases, suggesting an apoptotic mechanism (Hay et al., 1994). We are currently attempting to address these questions using inducible expression systems. Now that PPAN has been functionally isolated, is this the activated dominant tumor suppressor proposed by Boylan et al. (1996)? Several pieces of evidence suggest that it may be, the most compelling of which is that exogenous expression of PPAN in both HeLa and A549 tumor lines induces anchorage dependence. Ongoing tumor growth studies in athymic mice will also help address this issue. It is curious, however, that expression levels of endogenous PPAN are not radically different between HeLa and HF cells. Also, it is important to note that HF cells expressing Rz 568 do not appear morphologically “transformed” like their parental HeLa cells. Even after two rounds of soft agar selection, they still retain their flat, nonrefractile phenotype (not shown). One possible explanation is that we have partially dissected an anchorage-dependent growth pathway, perhaps downstream of the true tumor suppressor activated in the transition from HeLa to HF. Indeed, there were other ribozymes present in soft agar colonies that could target other genes in the same growth pathway. Only one other ribozyme, Rz 619, was present in multiple soft agar colonies. Rz 619 was also confirmed to have a stronger phenotype than Rz 568 (i.e., produced a larger number of soft agar colonies after transfection of HF cells). Unlike Rz 568, expression of Rz 619 clearly altered the morphology of HF cells to a transformed, highly refractile appearance. Rz 619 does not target the PPAN mRNA nor does it have any obvious database matches. We are currently attempting to clone the cellular target of Rz 619, which, it is hoped, will further illuminate the transformation mechanism in this cell system. It will be interesting to determine whether Rz 619 reduces the ex-

283

pression of another gene product in the same pathway as PPAN. Clearly, however, PPAN is involved in regulating anchorage dependence and thus may play a role in cell– cell communication, tumor progression, and metastasis. ACKNOWLEDGMENTS We thank Dr. Helmut Zarbl for supplying the HF cell line used in this study. We also thank Andrew Craven, Brian Winter, and Xiuying Sun for technical assistance. This work was supported, in part, by a grant from the Department of Energy (DE-FG03-98ER62624).

REFERENCES Bertrand, E., Castanotto, D., Zhou, C., Carbonnelle, C., Lee, N. S., Good, P., Chatterjee, S., Grange, T., Pictet, R., Kohn, D., Engelke, D., and Rossi, J. J. (1997). The expression cassette determines the functional activity of ribozymes in mammalian cells by controlling their intracellular localization. RNA 3: 75– 88. Boylan, M. O., Athanassiou, M., Houle, B., Wang, Y., and Zarbl, H. (1996). Activation of tumor suppressor genes in nontumorigenic revertants of the HeLa cervical carcinoma cell line. Cell Growth Differ. 7: 725–735. Hay, B. A., Wolff, T., and Rubin, G. M. (1994). Expression of baculovirus P35 prevents cell death in Drosophila. Development 120: 2121–2129. James, H. A., and Gibson, I. (1998). The therapeutic potential of ribozymes. Blood 91: 371–382. Kim, J., and Hirsch, J. P. (1998). A nuclear protein that affects mating efficiency in Saccharomyces cerevisiae by altering the morphological response to pheromone. Genetics 149: 795– 805. Migeon, J. C., Garfinkel, M. S., and Edgar, B. A. (1999). Cloning and characterization of peter pan, a novel Drosophila gene required for larval growth. Mol. Biol. Cell 10: 1733–1744. Orford, K., Orford, C. C., and Byers, S. W. (1999). Exogenous expression of beta-catenin regulates contact inhibition, anchorage-independent growth, anoikis and radiation-induced cell cycle arrest. J. Cell Biol. 146: 855– 868. Welch, P. J., Barber, J. R., and Wong-Staal, F. (1998). Expression of ribozymes in gene transfer systems to modulate target RNA levels. Curr. Opin. Biotechnol. 9: 486 – 496. Welch, P. J., Tritz, R., Yei, S., Barber, J., and Yu, M. (1997). Intracellular application of hairpin ribozyme genes against hepatitis B virus. Gene Ther. 4: 736 –743. Welch, P. J., Tritz, R., Yei, S., Leavitt, M., Yu, M., and Barber, J. (1996). A potential therapeutic application of hairpin ribozymes: in vitro and in vivo studies of gene therapy for hepatitis C virus infection. Gene Ther. 3: 994 –1001. Yu, Y., and Hirsch, J. P. (1995). An essential gene parin in Saccharomyces cerevisiae with a potential role in mating. DNA Cell Biol. 14: 411– 418.