GENE impedance: a natural process for control of gene expression and the origin of RNA interference

ARTICLE IN PRESS

Journal of Theoretical Biology 233 (2005) 301–314 www.elsevier.com/locate/yjtbi

GENE impedance: a natural process for control of gene expression and the origin of RNA interference M. Bahman Bahramiana, Helmut Zarblb, a

175 Brushy Plain Road, Suite 3-C6, Branford, CT 06405-2617, USA Fred Hutchinson Cancer Research Center, Mail Stop C1-015, P.O. Box 19024, Seattle, WA 98109-1024, USA

b

Received 8 March 2004; received in revised form 21 September 2004; accepted 7 October 2004

Abstract Gene expression is controlled by coordinated transcriptional and post-transcriptional mechanisms. Normally, expression of a gene switches on and off in response to specific physiological signals that are triggered by cellular demand for the gene products at a given time. Based on our previous studies and the scientific literature, we hypothesize that when a gene promoter switches to transcriptional repression mode, transcription of the gene ceases, and a small amount of double-stranded RNA (dsRNA) is synthesized by the RNA polymerase switching to the opposite DNA strand at the termination region of the gene. These dsRNA structures, which result from normal transcriptional repression, can then be processed into short interfering RNAs (siRNAs) within the nucleus. These molecules subsequently direct specific cleavage of the cognate mRNAs and interfere with their translation through sequence complementarily. We further hypothesize that cellular defense mechanisms invoked by invading genetic elements could be rooted in this fundamental regulatory pathway that we call ‘‘GENE impedance’’, or simply, GENEi. Here, we present a working model that illustrates how transcription-termination and transcription-arrest can contribute to the regulation of gene expression via GENEi. In our model RNAi is only one component of GENEi, which is a more generalized mechanism of gene regulation. r 2004 Elsevier Ltd. All rights reserved. Keywords: GENEi model; RNAi; Small interfering RNAs; Homology-dependent gene silencing; Transcription termination

1. Introduction A phenomenon that was revealed in the 1990s and is still unfolding has invigorated molecular biology research and holds great promises for development of gene-based therapies. We speak of homology-dependent Abbreviations: TTRC, Transcription termination/repression complex; TTU, Transcription termination unit; TRU, Transcription repression unit; 30 UTR, 30 untranslated region; 30 AUT, 30 adjacent untranscribed region; siRNAs, Small interfering RNAs; dsRNA, double-stranded RNA. Horizontal arrows over the genes indicate the direction of transcription, and the thickness of the arrow depicts the strength of transcription. Corresponding author. Tel.: +1 206 667 4107; fax: +1 206 667 5815. E-mail addresses: [email protected] (M.B. Bahramian), [email protected] (H. Zarbl). 0022-5193/$ - see front matter r 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.jtbi.2004.10.007

gene silencing in eukaryotic organisms. One aspect of this intriguing phenomenon that has gathered considerable attention is RNA interference (RNAi). As currently recognized, RNAi is a form of post-transcriptional gene silencing (PTGS), where short, double-stranded (ds) RNAs induce degradation of the homologous endogenous transcripts. The term ‘RNAi’ was originally coined by Fire et al. (1998) to describe occurrence of a specificmRNA interference process in the nematode worm Caenorhabditis elegans mediated by homologous double-stranded RNA that are longer than 25 base pairs (bp). Later, it was shown by others (see below) that the actual ‘‘effector’’ molecules in RNAi are short (o25 bp) dsRNAs. We have asked ourselves two basic questions: (1) What is the origin of the complex and ubiquitous pathway of RNAi? (2) What was the selective advantage for maintaining this highly

ARTICLE IN PRESS 302

M.B. Bahramian, H. Zarbl / Journal of Theoretical Biology 233 (2005) 301–314

conserved pathway throughout the evolution of plants and animals? We believe the answer to both questions is that RNAi is a component of a more fundamental cellular mechanism that allows for the selective elimination of transcripts already present in the cell at the time a gene’s expression is shut off. Conservation of this mechanism, which we call ‘‘gene impedance’’ and abbreviate to ‘‘GENEi’’, among eukaryotes, suggests that the inability to degrade at least some long-lived mRNAs may be deleterious to the cell. Like RNAi and the related phenomena, GENEi addresses the same end result, which is efficient and selective elimination of unwanted mRNA, but also cessation of transcription of the gene and interference with translation of its remaining transcripts. Contrary to the common belief that once transcription from a locus stops, remaining copies of the mRNAs and proteins corresponding to that locus decay at a rate determined simply by the intrinsic stabilities of their sequences and translational control, we hypothesize that in some, if not most, cases transcription arrest in eukaryotic organisms is associated with accelerated decay of the corresponding mRNA and most likely inhibition of translation of the mRNA, involving small double-stranded interfering RNAs. This decay involves a fundamental process of control of gene expression that we call GENEi that comprises the very interesting dsRNA-dependent short interfering RNA (siRNA) mechanism. The basis for our proposal is that gene expression is discontinuous and responsive to cellular demand for a particular gene product at a given time. Therefore, when sufficient amount of a given gene product accumulates, all processes leading to further accumulation of that gene product, including the persistence of transcripts and their translation in the cell, must be stopped. Based on our published results (Bahramian and Zarbl, 1999) and the current literature, we propose that cells have evolved distinct consorted mechanisms that we call gene impedance and we define as, ‘‘the coordinated cellular mechanisms for efficient and specific transcriptionalinhibition, degradation of the mRNA and inhibition of translation of the transcripts of a target gene in a cell or a population of cells in response to physiological signals’’. The basis for our theoretical model is the observation that the effect of the naturally occurring phenomenon of gene impedance can be mimicked in cells by transient introduction of double-stranded polynucleotides homologous to regions of the target gene. Our GENEi model also unifies several unexplained biological phenomena, including the metabolism of stable mRNA, RNA interference by small interfering RNAs, transcription-termination, transcription-arrest, and endogenous synthesis of antisense RNA. The field of homology-dependent gene silencing grew from disparate observations in a variety of experimental

systems and acquired terms such as ‘‘quelling’’, ‘‘cosuppression’’, ‘‘PTGS’’, and ‘‘interference’’, which often describe potentially similar processes. Table 1 outlines various definitions and their contexts. It is not intended in this article to provide a comprehensive review of various forms of gene silencing, as many recent reviews already exist (Dykxhoorn et al., 2003; Hammond et al., 2001; Hutva´gner et al., 2004; Sharp, 2001). We do however want to draw a clear distinction between cosuppression and GENEi, since the two terms are easily confused. We also review the genesis of RNAi, since we posit that this process is a necessary component of GENEi. The details of the mechanisms of gene impedance will be illustrated in the context of normal regulation of gene expression. The purpose of this article is, (a) to clarify the relation between GENEi and RNAi, (b) to put forth a cohesive theory that places the physiological roles for those phenomena within the context of biological processes such as normal termination of transcription and degradation of specific mRNA following transcriptional repression, and (c) to develop testable hypotheses for the proposed mechanisms of GENEi. Originally, we used the term ‘‘transcriptional and post-transcriptional gene silencing’’ to describe the distinct interference phenomenon (Bahramian and Zarbl, 1999). However, ‘‘gene silencing’’ in its broadest interpretation comprises any mechanism that silences a gene, such as, various sequence-homology-dependent silencing terms described in Table 1, gene knockout by homologous recombination, imprinting, DNA methylation, and transvection. Therefore, to avoid confusion, we now rename this distinct phenomenon ‘‘gene impedance’’. The post-transcriptional feature of GENEi (PTGi) mechanistically is identifiable with post-transcriptional gene silencing (PTGS) observed in some of the homology-dependent terms shown in Table 1. In this article, for practical purposes, we use the terms ‘‘PTGi’’ and ‘‘PTGS’’ synonymously.

2. Distinction of GENEi from cosuppression Cosuppression and the related mechanisms have been observed in plants, Drosophila, fungi and C. elegans, but not in higher animals, and particularly not in mammals. Characteristically, following stable integration of one or more copies of homologous transgene(s) into the genome, reduction in expression of the endogenous gene and the transgenes occur in reciprocal fashion (thus, cosuppression). This trans-suppression requires transcription of the transgenes but is independent of the specific promoter sequence used to drive the transcription (Cogoni et al., 1996; Ratcliff et al., 1999; Sharp, 2001).

ARTICLE IN PRESS M.B. Bahramian, H. Zarbl / Journal of Theoretical Biology 233 (2005) 301–314

303

Table 1 Various homology-dependent gene-silencing phenomena Term and Definition

Context

Post-transcriptional gene silencing (PTGS) This is a general term that applies to all eukaryotic organisms and involves double-stranded (ds)RNA. The transcription of the gene is unaffected; however, gene expression is lost because mRNA molecules become unstable. PTGS provides nucleotide sequence-specific protection against a variety of foreign genetic elements.

In higher plants this system provides protection against viruses, whereas in Caenorhabditis elegans, Drosophila and Chlamydomonas the targeted elements are transposons. Presumably, in PTGS of transgenes, the cell perceives the foreign DNA or the corresponding RNA as if it were a virus or a transposon.

Transcriptional gene silencing (TGS) Gene expression is reduced by a blockade at the transcriptional level. Evidence indicates that transcriptional repression might be caused by chromatin modification or DNA methylation.

This is generally observed in plants but has also been seen in animals. Generally in plants, transcriptional inactivation requires homology between promoters.

Cosuppression or transgene-induced silencing Silencing is caused by the presence of transgenes in the genome. Repression is usually related to copy number. Tandemly arrayed transgenes are more effective inducers of silencing than dispersed transgenes, with inverted repeats being the most effective. Silencing can occur transcriptionally or post-transcriptionally. This transsuppression requires transcription of the transgenes but is independent of the specific promoter sequence used to drive the transcription.

Cosuppression has been recorded in plants, Drosophila, fungi and C. elegans, but not in higher animals. Characteristically, following stable integration of one or more copies of homologous transgene into the genome, reduction in expression of the endogenous gene and the transgenes occur in reciprocal fashion (thus, cosuppression).

Virus-induced silencing Silencing is induced by the presence of viral genomic RNA. Only replication-competent viruses cause silencing, indicating that dsRNA molecules might be the inducing agents.

Restriction of virus growth in plants is mediated by PTGS, which can be initiated by production of dsRNA replicative intermediates. This silencing of gene expression is gene specific.

Quelling This term is specific for transgene-induced PTGS in Neurospora crassa.

During the vegetative phase in Neurospora crassa, the transgene can induce reversible gene silencing.

RNAi This type of PTGS is induced directly by dsRNA. GENEi This is ‘‘the coordinated cellular mechanisms for efficient and specific transcriptional-inhibition, degradation of the mRNA and inhibition of translation of the transcripts of a target gene in a cell or a population of cells in response to physiological signals’’. Gene impedance can also be triggered by transient introduction of double-stranded polynucleotides homologous to regions of the target gene. GENEi functions transcriptionally and post-transcriptionally, and could operate independent of the level of transcription of the transgenes.

It was first defined in C. elegans and seems to be mechanistically related, if not identical, to PTGS in plants. It was first reported in mammalian cells, but applies to various eukaryotic organisms.

Homology-dependent gene silencing grew from disparate observations in a wide variety of organisms and acquired diverse terms that often describe potentially similar processes. This table outlines the definitions of various terms and their context. The terms are generally defined on the basis of the inducing agent and the mechanism of silencing.

By contrast, if one stably transfects any transgene that is homologous to an endogenous gene and not fused to a heterologous promoter into mammalian cells, one would never observe cosuppression. For example, integration of the full-length pro-a1(I) collagen gene construct (pWTC1) into rodent cell genome results in the transgene expression levels indistinguishable from that of the endogenous gene (Barker et al., 1991); therefore, no ‘cosuppression’ occurs. In addition to the fact that integration of the transgene(s) into a mammalian genome does not result in cosuppression, there are other significant differences between cosuppression and

GENEi. Cosuppression requires the transcription of the transgenes and is influenced by the level of this expression. However, our initial studies indicated impedance of the homologous endogene could occur after transient transfection with the same procollagen pWTC1 construct, even though there was no detectable transgene mRNA expression (Bahramian and Zarbl, 1999). We also tested a number of progressive 50 proa1(I) collagen promoter deletion (promoter bashing) constructs. Even though these constructs varied greatly in their ability to express transgenes, each of these constructs suppressed the transcription of the

ARTICLE IN PRESS 304

M.B. Bahramian, H. Zarbl / Journal of Theoretical Biology 233 (2005) 301–314

endogenous pro-a1(I) collagen gene to exactly the same level. Thus, the level of endogenous gene impedance was unrelated to the strength of the transgene promoter. In spite of these differences, cosuppression and GENEi are mechanistically linked through dsRNA-induced PTGS.

nuclease, or to serve as the template for amplification, remained a mystery, because it could not be detected in worms that were undergoing silencing. Thus, research on discovering the mechanism of PTGS continued using only long dsRNA, which was generally accepted to be the actual ‘‘effector’’ molecule.

3. Historical relationship between RNAi and GENEi

3.2. Effectors of DNA-triggered PTGS

3.1. Genesis of RNAi

The first evidence that the actual effector molecules that regulate PTGS are short dsRNA, surprisingly, did not come from studies of dsRNA in invertebrate animals and plants; rather it was revealed by studies on dsDNA-mediated PTGS in mammalian cells (Bahramian and Zarbl, 1999). Transient introduction into mammalian cells of DNA vectors containing homologous sequences to the 30 end of an endogenous gene triggered specific and efficient degradation of the cognate mRNA (Bahramian and Zarbl, 1999). This effect persisted and was undiminished through at least seven generations of cell divisions, consistent with the induction of a catalytic reaction that utilized sequencespecific ‘‘effector’’—or ‘‘guide’’—molecules for targeting the cognate mRNA. For the post-transcriptional part of gene impedance (PTGi, PTGS) to occur, communication must take place between the silenced gene in the nucleus and the cognate mRNA in the cytoplasm. Specific degradation of mRNA guided by protein or DNA has never been a realistic possibility as the need for a custom-tailored ribonuclease for each type of messenger RNA would be both inefficient and unrealistic. DNA is confined to the nucleus by nuclear membrane. Transgenic DNA is either efficiently taken up by cell nucleus or remains strongly associated with it (Bahramian and Zarbl, 1994). In any case, high molecular weight DNA, mechanistically speaking, would be a cumbersome and inefficient effector molecule for specific mRNA degradation. Moreover, DNA fragments homologous to structural parts of endogenous genes have been introduced transiently into animal cells for decades. In mammalian gene expression studies, such DNA could have been partially degraded in the cytoplasm while in transit to the nucleus, but has never been shown to cause degradation of the cognate mRNA. Communications between DNA confined in the nucleus and the metabolic activities in the cytoplasm are always mediated through RNA. However, antisense RNA generated in the nucleus is another unfitting candidate as the effector molecule for degradation of complementary mRNA in the cytoplasm. First, antisense RNAs lack the protective 30 -polyadenylation and the 50 -7mG-cap structures. Moreover, since they are not translated, antisense RNAs do not benefit from protective structures, such as ribosomes, all making antisense RNAs very unstable, if at all transported to the

Fire et al. (1998) discovered that the injection of double-stranded RNA (dsRNA) into nematode worm Caenorhabditis elegans led to an efficient sequencespecific gene silencing at post-transcriptional level, which is referred to as RNA interference (RNAi). The RNAs that triggered specific gene silencing were large, 4300 bp, and were delivered by injection into adult worms (Fire et al., 1998), soaking worms in dsRNA or by feeding worms Escherichia coli that expressed the dsRNA (Tabara et al., 1998). RNAi mediated by the introduction of long dsRNA was soon demonstrated in other small animals, including flies (Kennerdell and Carthew, 1998) and trypanosomes (Ngo et al., 1998), and was identified with the previously observed PTGS phenomena in plants (Jorgensen, 1995; Vaucheret et al., 1998; Waterhouse et al., 1998) and the fungus Neurospora (Cogoni and Macino, 1997). It was initially presumed by most researchers in the field that the RNAi was not applicable to mammals because the introduction of any dsRNA that is longer than 30 bp into the cytoplasm of mammalian cells induces the sequence-independent ‘‘interferon’’ response (Hunter et al., 1975; Stark et al., 1998). Interferon triggers the nonspecific degradation of mRNAs by inducing the synthetase that generates 20 ,50 oligoadenylates, which in turn activate RNAse L (Elbashir et al., 2001; Minks et al., 1979). In addition, interferon activates the protein kinase PKR, which phosphorylates the translation initiation factor eIF2a leading to a global inhibition of mRNA translation (Stark et al., 1998). Fire et al. (1998) speculated that the key to the interference process is the recognition of the target mRNA by the dsRNA, or a derivative thereof. Because the incoming dsRNA did not need be present at stoichiometric levels, they proposed the existence of either a highly efficient catalytic machine or an amplification step. In the former case, they posited that the machinery of RNAi would search out homologous mRNA, which would subsequently be destroyed by a nuclease that was directed to them by base-pairing interactions. In the latter case, they proposed an amplification of the signal, possibly through a replicative expansion of either the dsRNA or its derivative (Fire et al., 1998; Hammond et al., 2001). The nature of the derivative that functioned either to direct the

ARTICLE IN PRESS M.B. Bahramian, H. Zarbl / Journal of Theoretical Biology 233 (2005) 301–314

cytoplasm. More importantly, antisense RNA is a minor component of nuclear RNA, and is indistinguishable from other single-stranded RNA present in the nucleus from the view of a putative ribonuclease that could bind to it and use it as a guide to selectively degrade a specific mRNA in the cytoplasm. Therefore, finding a specific antisense RNA in the nucleus for a ribonuclease is like looking for a needle in a field of haystacks. Taken together, available data, prior knowledge plus studies on dsDNA-mediated PTGS in mammalian cells (Bahramian and Zarbl, 1999), implicitly implicated siRNAs as the only possible sequence-specific effector molecules in PTGS in mammalian cells. The existence of naturally occurring, large and heterogeneous-length antisense RNA transcripts from the 30 end of the endogenous mammalian genes, including collagen genes was demonstrated by previous studies (Farrell and Lukens, 1995). Such antisense RNA can hybridize with the corresponding sense nuclear transcripts to form dsRNA. Since long (430 bp) dsRNA in the cytoplasm of mammalian cells provokes interferon response, it could not be considered a potential effector molecule in mammals. It was also known that introduction of moderate quantities of short (o30 bp or so) dsRNA into mammalian cells does not provoke interferon response (Hunter et al., 1975; Stark et al., 1998). Based on this prior knowledge, i.e., by logical elimination of all other candidates for carrying the nucleotide-sequence code from DNA to mRNA, it was obvious that PTGS triggered by transient homologous dsDNA could only be mediated by short dsRNA homologous to coding or regulatory regions within the endogenous-gene transcript. Further, it was predictable from the same logic that any long dsRNA generated in the nucleus by mRNA-antisense RNA hybridization, as a consequence of nuclear introduction of homologous dsDNA, must be processed to short dsRNA (o30 bp or so) before transport to the cytoplasm in order to avoid induction of the interferon response. The resulting siRNA molecules could then serve as effector molecules in specific mRNA degradation. The nexus between the originally discovered phenomenon of RNAi (Fire et al., 1998) and siRNA that was implicitly implicated in the dsDNA-mediated PTGS in mammalian cells (Bahramian and Zarbl, 1999), which is a component of the mechanism we now call GENEi, was recognized by several investigators (Elbashir et al., 2001; Fire, 1999; Gitlin et al., 2002; Marathe et al., 2000; Pal-Bhadra et al., 2002; Tuschl et al., 1999; Yang et al., 2001; Yi et al., 2003), and led to a flurry of investigations in other systems. Subsequently, 21–25 bp dsRNAs were detected in plants that were undergoing PTGS (Hamilton and Baulcombe, 1999); but in this case, it was impossible to distinguish whether the short dsRNAs were the mediators or the products of PTGS. In mouse oocytes (Svoboda et al., 2000) and in early mouse

305

embryos (Wianny and Zernicka-Goetz, 2000), which are incompetent in the ‘‘interferon response’’ pathway, sequence specific interference with gene function by long dsRNA was demonstrated. In Drosophila embryo extracts, it was demonstrated that long dsRNA could be cleaved into short interfering dsRNA species (siRNAs) of 22 bp (Zamore et al., 2000). Subsequently, the mediators of sequence-specific mRNA degradation were demonstrated to be 21- and 22-nucleotide small interfering RNAs (siRNAs) generated by ribonuclease III (Dicer) cleavage from longer dsRNAs (Bernstein et al., 2001; Elbashir et al., 2001; Hammond et al., 2000; Provost et al., 2002; Zhang et al., 2002). A detail of the mechanism of action of RNAi is recently reviewed (Dykxhoorn et al., 2003).

4. Proposed physiological roles for GENEi and RNAi In this section we introduce the hypothesis that GENEi is essentially an ancient cellular mechanism for precise control of gene expression following repression of transcription in eukaryotic organisms. We hypothesize that the PTGi aspect of GENEi was subsequently adopted for protection of organisms against foreign genetic elements. 4.1. Control of gene expression and transcription termination Gene expression at the transcriptional level is controlled by a series of cis-acting DNA sequence elements, including promoters, enhancers, silencers and locus-control elements, and cognate transcription factors that bind to these elements. Together these DNA sequences and proteins regulate the formation of the RNA polymerase II (Pol II) complex to control the production of pre-mRNA molecules. The pre-mRNA molecules then undergo several steps of processing before becoming functional mRNA. Introns are removed, a 7-methyl-guanylate (m7G) cap structure is added at the 50 end of the transcript, and after sequencespecific endonucleolytic cleavage of the primary transcript, a stretch of 100–250 adenine residues, poly(A) tail, is added to the 30 end. Modulation, or fine-tuning, of gene expression is achieved by post-transcriptional events in the cytoplasm that regulate mRNA translation, mRNA turnover, and subcellular localization. The latter processes are mediated through cis-acting regulatory features including stem-loop structures, upstream initiation codons and open reading frames, internal ribosome entry sites and various cis-acting elements that are bound by RNA-binding proteins. For a review, see Mignone et al. (2002). The messenger RNA processing reactions of capping, splicing and polyadenylation occur cotranscriptionally.

ARTICLE IN PRESS 306

M.B. Bahramian, H. Zarbl / Journal of Theoretical Biology 233 (2005) 301–314

They not only influence one another’s efficiency and specificity, but are also coordinated by transcription. The phosphorylated carboxyl-terminal domain (CTD) of Pol II provides key molecular contacts with these mRNA processing reactions throughout transcriptional elongation and termination (Proudfoot et al., 2002, for review). Prior to the addition of poly(A), the pre-mRNA must be cleaved. The site of cleavage in most pre-mRNAs lies between the highly conserved poly(A) addition signal (AAUAAA) and a downstream sequence element (DSE), which is a U- or GU-rich motif. Cleavage itself occurs predominantly at a CA dinucleotide. A U-rich upstream sequence located 50 of the poly(A) site (AAUAAA) has also been identified. Two multisubunit protein complexes recognize these cis elements. The 160 kDa subunit of the cleavage and polyadenylation specificity factor (CPSF) has been shown to interact with the poly(A) site, but the other three subunits (CPSF-100, CPSF-73, and CPSF-30) most likely contribute to the specificity and strength of binding. The DSE represents a platform for the interaction with the cleavage stimulatory factor (CstF) via its 64 kDa subunit. The binding of CstF and CPSF appears to be cooperative in that CstF binding to the DSE greatly enhances the affinity of CPSF to the poly(A) site and vice versa. Two additional factors are essential to direct cleavage of the pre-mRNA: cleavage factor I (CF I) and cleavage factor II (CF II). Poly(A) polymerase (PAP) is usually required for the cleavage reaction and together with CPSF directs poly(A) addition. Poly(A) binding protein PABP II binds the emerging poly(A) tail and in turn enhances the processivity of the PAP. Proper transcriptional termination is critical for successful gene expression. First, it allows release of transcripts. Second, it allows release of Pol II from the DNA template, facilitating the recycling of Pol II back to the promoter for further rounds of transcription. Third, it will ensure that downstream promoters are not perturbed by read-through polymerases that have failed to terminate at upstream genes. Contrary to logic, transcription does not stop at the poly(A) site. Rather, the polymerase continues for another 200–2000 bases beyond this site before transcription is aborted. Pol II proceeds up to the site of termination before transcript cleavage takes place, indicating that 30 end processing is the trigger for Pol II termination. Transcript cleavage at the poly(A) site leaves two products which have quite different fates; the upstream product, destined to become mRNA, is stabilized by poly(A) addition, whereas the downstream RNA is destabilized by cleavage, separating it from the body of the transcript and is rapidly degraded. Since it is clear that RNA processing signals affect transcription, it follows that the reverse effect on

transcriptional events influencing RNA processing occurs. Transcriptional pause sites located downstream of the poly(A) site cause a transient pause to Pol II progression and so enhance poly(A) site recognition, increase the efficiency of cleavage and polyadenylation and thereby promote termination. Detailed studies of nascent transcription in the 30 flanking regions of a number of genes have indicated that sequences located downstream of the poly(A) site are important players in the termination process. Transcription termination takes place in two distinct steps. In the first step, the nascent RNA is cleaved cotranscriptionally at specific subsections of the ‘‘termination region’’, possibly leading to its release from Pol II. (This is distinct from cleavage of the nascent RNA at the poly(A) site.) In a subsequent step, Pol II disengages from the DNA. Hence Pol II continues to transcribe the DNA, at least for a short distance, after cotranscription cleavage of the nascent RNA. No consensus sequences have been found between different ‘‘termination regions’’ studied. There is expanding evidence that dedicated signals and factors operate downstream of the poly(A) site and are also required for termination to occur.

4.2. GENEi and transcription arrest connection When physiological signals are relayed to an endogenous gene to stop transcription, it seems logical that the cytoplasmic transcripts of the repressed gene should also be destroyed coordinately and efficiently, particularly if the mRNA and/or the proteins encoded by the gene are long lived. Continued expression of a particular gene after transcriptional arrest is not only wasteful, but also could have adverse physiological consequences. Thus, it seems unlikely that the cell depends exclusively on generalized posttranscriptional modulatory processes to gradually degrade a specific mRNA or inhibit its translation. In fast-growing bacterial cells, the requirement for timely removal of specific mRNA is generally addressed by evolution of a dynamic system of synthesis and degradation of all mRNA species. By contrast mRNA molecules in more complex, slower growing eukaryotic organisms are remarkably stable, and individual mRNA species show tremendous variation in halflives. Overriding the mechanisms that allow for differential stability of eukaryotic mRNAs is thus prerequisite for precise control of gene expression. We hypothesize that eukaryotic cells have evolved a more refined mechanism that comprises RNAi and allows for selective removal of the stable mRNA molecules. We refer to this archaic cellular process, which appears to be conserved in all eukaryotes from Neurospora to man, as gene impedance.

ARTICLE IN PRESS M.B. Bahramian, H. Zarbl / Journal of Theoretical Biology 233 (2005) 301–314

307

4.3. GENEi as the likely source of ubiquitous antiviral immunity

5.1. Studies that significantly contributed to the hypothesis

It has been stated that RNAi, the oldest and most ubiquitous antiviral system, appeared before the divergence of plants and animals (Sharp, 2001). In animals, RNAi may represent an ancient antiviral response, just as post-transcriptional gene silencing appears to protect plants from viral infection (Baulcombe, 1999; Grant, 1999; Hamilton and Baulcombe, 1999; Ratcliff et al., 1999). Although it may seem logical that RNAi first evolved for protection against foreign genetic elements, there are several reasons to suggest that it could represent an opportunistic use of a pre-existed geneexpression regulatory mechanism, i.e., gene impedance. RNAi is a double-stranded RNA-dependent mechanism that operates in two distinct steps. In step one, the dsRNA introduced to the cell by invading viruses, or any other mean, is degraded to fragments of 22 bp by a dsRNA-endonuclease (ribonuclease III, Dicer) that is sequence non-specific. This step of RNAi is sufficient for inactivation of the foreign genetic elements that involve dsRNA in their life cycle. The second step is a more complex mechanism involving multiple factors and enzymes, and the use of the small dsRNA as ‘‘guide molecules’’ to inactivate the nucleotide sequence related mRNAs. The second step appears to be directed toward limiting expression of DNA viruses. While the RNAi processes would enhance survival of virus-infected cell, precise regulation of gene expression, especially for the regulatory proteins, is also necessary for cell survival. Sequence-specific control of the transcripts of eukaryotic genes, the complexity of the RNAi mechanism, and dependence of viral life cycle on biology of the host cell, are among reasons that the RNAi/PTGS pathway is just as likely to have evolved as a basic mechanism to control gene expression. The cellular mechanism could subsequently have been co-opted to defend against viruses, rather than the other way around. Alternatively, the two functions could have evolved in parallel.

Combining the sensitive technique of ribonuclease protection with the highly quantitative technique of phosphorimaging, while simultaneously monitoring the efficiency of transient transfections into the nuclei of cells, enabled accurate quantification of transient expression of the transgenic procollagen genes in presence of the endogenous gene expression (Bahramian and Zarbl, 1999). The transgene present in the cosmid (pWTC1) carried a 21-bp insert in the 50 untranslated region that made it possible to differentiate its transcripts from the transcripts of the endogenous gene (Barker et al., 1991). Surprisingly we discovered that not only were the transgenic procollagen genes totally silent soon after transfection, but their presence in the cell nucleus also caused a rapid and dramatic reduction in the steady-state level of the endogenous procollagen gene transcripts. Moreover, the normally long half-life of the pro-a1(I) collagen mRNA indicated to us that the endogenous procollagen transcripts were degraded at a highly accelerated rate in the transiently transfected cells. This gene specific effect on mRNA stability effect persisted for days after transfection, indicating the induction of an active process of specific mRNA degradation. A number of obvious reservations were readily dismissed: We demonstrated that transient transgeneinduced gene impedance was not the result of stress from, for example, the shock of electroporation or chemicals used in transfection. The silencing was specific and mediated by homologous DNA; vector sequences alone were ineffective in suppressing the expression of the endogenous gene. The GENEi phenomenon was also highly specific for a particular gene and did not affect expression of other genes. For example, an endogenous control gene, glyceraldehyde-3-phosphate dehydrogenase (GAPDH), expressed uniformly in all of the cell lines tested under all conditions. We also showed that GENEi was not limited to a specific cell line or species. Rodent pro-a1(I) collagen is a structural gene of 16 kbp in length, comprising 51 short exons. Studies have demonstrated that the promoter of the pro-a1(I) collagen contains numerous cis-acting control elements and is subject to a complex pattern of transcription regulation. The promoter region and the first intron have been implicated in transcription control during development (Chan et al., 1991; Nehls et al., 1992), tissue-specific expression (Kratochwil et al., 1989; Ravazzolo et al., 1991), carcinogenesis (Slack et al., 1992) and in response to a variety of physiological signals (Ramirez and di Liberto, 1990; Ritzenthaler et al., 1991). In order to study the role of these two regulatory regions in the GENEi phenomenon, we used

5. A proposed model for GENEi mechanisms In this section we present a model that illustrates possible mechanisms of transcriptional and post-transcriptional GENEi in the general context of gene expression, as a gene goes through cycles of transcription, termination of transcription and periodic transcription arrest. Such a generic model needs to be consistent with the current literature and our published results (Bahramian and Zarbl, 1999). To facilitate this communication, we begin with reviewing the key experiments and the conclusions of our studies that significantly contributed to the hypothesis, before proceeding with description of the generic model.

ARTICLE IN PRESS 308

M.B. Bahramian, H. Zarbl / Journal of Theoretical Biology 233 (2005) 301–314

a series of progressive deletion constructs that were transiently transfected into various cell lines (Bahramian and Zarbl, 1999). These experiments demonstrated that the promoter deletion constructs containing DNA sequences between the basal promoter, starting at bp 220 with respect to transcription start, and the first quarter of intron 1 are involved in transcriptional gene impedance (TGi). When constructs that did not include any translated sequences of pro-a1(I) collagen mRNA (up to the middle of first exon) were introduced into cells, there was 50% reduction in the steady state level of transcripts from the endogenous gene. Adding further sequences up to first quarter of intron 1 reduced the level of transcripts by another 20%. However, inclusion of DNA sequences downstream from the first quarter of intron 1 to the end of exon 5, or upstream promoter sequences between bp 3500 and 220, did not further contribute to the endogenous gene impedance. The extent of endogenous procollagen gene impedance was independent of transgene expression levels. This suggests that TGi is independent of transcription of the transgenes and is dependent on the presence on the transgenes of genespecific elements located in the 50 region of the target gene. Crucial insights into the mechanisms of GENEi came from our observation in non-tumorigenic revertants isolated from v-fos transformed cells (EMS-1-19) (Zarbl et al., 1987). When the same 50 promoter deletion constructs described above were transiently transfected into these revertants, no impedance of the endogenous procollagen gene was detected. The revertant cells harbor mutations in downstream effectors of v-fos induced transformation. These mutations allow for transcription of the pro-a1(I) collagen gene, despite the continued expression of a functional v-fos oncogene (Zarbl et al., 1987). As stated above, the DNA sequences between the basal promoter, starting at bp 220 with respect to transcription start, and the first quarter of intron 1 are involved in TGi. Absence of GENEi in EMS-1-19 revertant cells transfected by transgenes carrying such sequences indicated that in these cells pro-a1(I) collagen gene has become independent of TGi and independent of v-fos induced transcriptional suppression, possibly as a result of mutations in a general transcription factor. This mutation must be qualitative rather than quantitative, because a quantitative mutation in the general transcription factor cannot explain the absence of TGi in the revertants. The mutation in the general transcription factor renders the activation of RNA polymerase complex independent of a genespecific transcriptional cofactor, which is required in the normal and v-fos transformed cells. However, another component of gene muting was also shown to involve other regions of the gene, particularly the 30 end. The component of GENEi

specified by the 30 sequences revealed different features. Thus transient transfection with a construct that carried both the 50 regulatory regions and the 30 end sequences of the gene (pWTC1) resulted in rapid decline in the steady-state level of the endogenous gene transcripts by greater than 90% in both normal and v-fos transformed cells. Significantly, in the revertant EMS-1-19 cells, the endogenous procollagen transcripts also decreased by 70% when transfected transiently with pWTC1. These results indicated that other regulatory elements present in the remainder of the gene, past the exon 5 to the end of the gene, were involved in post-transcriptional suppression of the endogenous gene. Thus, GENEi is comprised of two distinct mechanisms, transcriptional and post-transcriptional, and these mechanisms can be uncoupled experimentally by using various constructs in different cell lines. 5.2. Description of a generic model for GENEi Fig. 1 depicts a simplified generic model for the operation of the GENEi mechanisms. First, it is conceivable that under normal physiological conditions the transcriptional and post-transcriptional mechanisms of GENEi operate coordinately. Second, we postulate that a set of sequence-specific DNA-binding factors and respective DNA sequence elements are involved in each of the dual mechanisms of GENEi. These cis-acting elements and their cognate binding factors must be specific for each gene or a set of coordinately regulated genes. The requisite sequence element(s) at the 50 end of the gene, and the cognate binding factor(s), designated as a co-activator(s), facilitate the activation of RNA polymerase complex in transcription initiation or elongation. In this model, the transcription termination/repression complex (TTRC) mediates termination of transcription. The TTRC consists of two cooperative components: The first is the transcription termination unit (TTU) that would be common to all genes, or a class of genes. When associated with the second component, TRU (for transcription repression unit), TTU functions in cleavage of nascent transcript from RNA polymerase and in releasing RNA polymerase from DNA template. TRU can also trigger the post-transcriptional feature of GENEi. When binding of the TRU component to TTU is destabilized for any reason, RNA polymerase, instead of disengaging from DNA, turns around and synthesizes RNA from the opposite strand of DNA, resulting in generation of dsRNA either in the form of hairpin or sense–antisense RNA. Double-stranded RNA is subsequently cleaved to 21- and 22-nucleotide siRNAs by ribonuclease III-type enzyme. The siRNAs direct the cleavage of specific mRNA through sequence complementarity by RNAi pathway.

ARTICLE IN PRESS M.B. Bahramian, H. Zarbl / Journal of Theoretical Biology 233 (2005) 301–314

309

Fig. 1. Depiction of transcription state and impedance of a generic eukaryotic gene. (A) The structure of a generic eukaryotic gene combined with an activated RNA polymerase complex after interaction of the polymerase with a transcription factor and a gene-specific co-activator that facilitates initiation or elongation of transcription. The transcription results in generation of appropriately synthesized, released and processed transcripts owing to the presence of TTRC on the gene. This condition applies to the normal cell lines and v-fos transformed cells, although in the latter, transcription rate is greatly reduced due to a decrease in the transcription factor copy number (Zarbl et al., 1987). The transcription state is also applicable to the v-fos revertant EMS-1-19 cells with the exception that the transcription of the latter has become independent of the gene-specific coactivator, owing to a mutation either in the transcription factor or the RNA polymerase. Because in EMS-1-19 and the v-fos transformed cells there are comparable numbers of the transcription factor (only the RNA polymerase activating property has changed), an EMS-1-19 cell expectedly expresses pro-a1(I) collagen at a level intermediate between the wildtype and the v-fos transformed cells. (B) Reduced gene expression results from TGi triggered by transient transfection with various 50 promoter deletion constructs that compete for the limited gene-specific co-activators. This condition affects only the transcription of normal and v-fos-transformed fibroblasts, but not the revertant EMS-1-19 cells that are independent of the gene-specific co-activator. (C) ‘Intrinsic GENEi’ in all cell lines, where the transcription factor is removed by physiological signals, causing transcriptional repression, and TRU is altered by a physiological repressor, allowing formation of dsRNA that leads to PTGi. (D) ‘Alternative GENEi’ induced by homologous transgenes that comprise both the 50 - and the 30 - end sequences. The multiple copies of the transgenes in the nucleus destabilize the gene-specific co-activator and the TRU factors. Limited transcription occurs from both directions, resulting in the formation of dsRNA and subsequent PTGi. This condition applies to all cell lines. However, while the transcription of the procollagen gene is greatly reduced in normal and v-fos transformed cells, transcription of the gene in the revertant EMS-1-19 cells is unaltered owing to the co-activator-independent RNA polymerase complex.

ARTICLE IN PRESS 310

M.B. Bahramian, H. Zarbl / Journal of Theoretical Biology 233 (2005) 301–314

The TRU protein can be destabilized in two ways: First, during normal, ‘intrinsic GENEi’ in response to physiological signals, transcription of the gene becomes repressed by removal of a transcription factor, while mRNA destabilization is mediated by a repressor molecule that interacts with and destabilizes the association of the TRU factor with TTU (Fig. 1C). Second, the ‘alternative GENEi’ could be triggered in response to transient introduction of multiple copies of the targeted endogenous gene into the cell nucleus. Assuming that the concentrations of the gene-specific factors (co-activator(s) and TRU) are limiting in the cell, the transgenes could compete with the endogenous gene for binding these factors (Fig. 1D). This would result in repression of the transcription (but not its total shut off, since the transcription factor is still present) and specific degradation of the endogenous gene transcripts mediated by siRNAs. This model suggests how the two mechanisms can be uncoupled, as was the case in the revertant EMS-1-19 cells. In these cells, the transcription of the procollagen gene was unaffected by GENEi, because transcription was no longer subject to normal regulatory mechanisms. However, post-translational GENEi (PTGi) occurred since the siRNA-mediated pathway remained intact. The proposed model also accommodates both the ‘intrinsic GENEi’ and the ‘alternative GENEi’ induced by various pro-a1(I) collagen constructs, in all of the fibroblast cell lines we studied, despite the complexity of regulation of this gene. Of course, the exact relative positions and sequences of the cis-acting elements described above, as well as molecular identities of the relevant binding factors, transcription activator, and repressor involved are undetermined at this time.

6. Naturally occurring siRNAs and miRNAs Naturally occurring siRNAs that are involved in sequence-specific degradation of cognate mRNA are derived from processing of long dsRNAs by the RNAse III-type enzyme called Dicer (Bernstein et al., 2001). Dicer cleaves long dsRNA into duplexes of 21- to 28nucleotide siRNA that contain 2-nucleotide 30 overhangs with 50 phosphate and 30 hydroxyl termini (Hamilton and Baulcombe, 1999; Hammond et al., 2000; Zamore et al., 2000). Components of the RNAi pathway specifically recognize the siRNA duplex and incorporate a single siRNA strand into a multiprotein RNA-inducing silencing complex (RISC) that cleaves mRNAs containing perfectly complementary sequences, ten nucleotides from the 50 end of the incorporated siRNA strand (Dorsett and Tuschl, 2004; Hammond et al., 2000; Martinez et al., 2002). Dicer produces two types of small regulatory RNAs that regulate gene expression: siRNAs and micro RNAs

(miRNAs). A miRNA is a single-stranded molecule of 21- to 22-nucleotides that is excised from a 70-nucleotide hairpin precursor (Bartel, 2004). The mature miRNAs are incorporated into a protein complex (miRNP) that associates with ribosomes and inhibits translation of mRNAs containing sequences partially complementary to the miRNA in their 30 untranslated region (Brennecke et al., 2003; Xu et al., 2003). Hundreds of endogenous siRNAs and miRNAs have been identified from both plants and animals, yet little is known about their biochemical modes of action or biological functions (Hutva´gner et al., 2004). Recent data suggest that both siRNAs and miRNAs incorporate into similar, perhaps even identical, protein complexes and that a critical determinant of mRNA destruction versus translation regulation is the degree of sequence complementarity between the small RNA and its mRNA target (Zeng et al., 2003; Hutva´gner et al., 2004). In accordance with our hypothesis, these naturally occurring siRNAs or miRNAs could have originated from intrinsic GENEi.

7. Experimental methods for achieving specific GENEi Experimentally, there are two general methods for achieving gene muting, either ‘‘independent’’—or ‘‘dependent’’—on cellular transcription of the GENEi nucleic acid. One example of the first category is direct application of the effector molecules, siRNAs (Elbashir et al., 2001; Svoboda et al., 2000), or some chemically modified forms (analogs), in PTGi. Another example in the same general category involves transient transfection of cells with either the full or partial copies of the target gene (Bahramian and Zarbl, 1999). The transfected sequences comprise selected regulatory region(s), in order to achieve transcriptional and/or post-transcriptional GENEi independent of the expression of the transgenes. For TGi and PTGi, the relevant regulatory elements appear to be polarized to the 50 and 30 termini of a gene, respectively. This is a general rule in accordance with our model. We have shown GENEi operates in our experiments involving the procollagen gene, a prototypical [housekeeping] gene expressed in many cells. Since genespecific DNA elements and respective cognate proteins are likely involved, one should expect these elements to be different in sequence from one gene to another; also, the location of the elements from a constant feature, like transcription start or transcription termination site, could vary from gene to gene. Some analogs of DNA potentially could act like DNA in this category of muting. In PTGi induced by transgenes that do not require transcription, DNA sequences corresponding to the 30 end of the target gene need to be present on the transgenes to induce synthesis of dsRNA from the 30 end

ARTICLE IN PRESS M.B. Bahramian, H. Zarbl / Journal of Theoretical Biology 233 (2005) 301–314

of the target gene, for nuclear processing in the nucleus. The resulting siRNAs then serve as the effector molecules in PTGi. Transient transfection with transgene sequences homologous to the 30 end regulatory region of the targeted endogene are predicted to trigger PTGi in one of two ways. The transgene could trigger production of a small number of antisense RNA transcripts from the 30 end of the target gene. The latter sequences could then combine with the sense transcripts in nucleus to form dsRNA. Alternatively, RNA polymerase, instead of disengaging from DNA, could turn around and synthesize RNA from the opposite strand of DNA, resulting in generation of dsRNA either in the form of hairpin or unlinked dsRNA. In either case the resulting dsRNA molecules would be degraded to siRNAs by Dicer-type activities (Provost et al., 2002; Zhang et al., 2002) in the nucleus, combine with components of the RNAi machinery that shuttle between nucleus and cytoplasm, and used as templates by special ribonucleases to initiate specific degradation of the sequence-related mRNA in the cytoplasm. Double-stranded DNA oligonucleotides have been used, sometime successfully (Gambarotta et al., 1996), as a ‘‘general transcription factor decoy’’ to inhibit gene transcription. Such inhibitions, unlike gene muting and generally speaking, have proven to be inefficient, lack specificity, and large quantities of the oligonucleotides were required over time to sustain a durable suppression. The second category applies to transgenes that must be transcribed inside the cell (transcription-dependent). Within this category, there is a sub-category that pertains exclusively to PTGi and requires somehow providing the specific siRNAs to the cell through transcription. This could be achieved by a variety of methods as follows. One method entails providing siRNAs to the cells by introducing a DNA vector that can transcribe complementary sense and antisense RNA corresponding to a translated or an untranslated region of a target gene mRNA. The resulting dsRNA would be processed to siRNAs in the nucleus. In support of this hypothesis, Yi et al. (2003) employed an approach in which dsRNA is generated by cellular transcription from plasmids containing long (1 kilobase) inverted DNA repeats of the target gene, rather than using dsRNA synthesized in vitro. They showed that gene silencing by this method is effective for endogenously expressed genes as well as for exogenous reporter genes. An analysis of the expression of several endogenous genes and exogenous reporters demonstrated that the silencing effect was specific for the target gene containing sequences within the inverted repeat. Their method eliminates the need to chemically synthesize dsRNA and is not accompanied by global repression of gene expression. Similarly, Yu et al. (2003) reported simulta-

311

neous inhibition of GSK3alpha and GSK3beta using hairpin siRNA expression vectors. Here, RNA is transcribed in one orientation and folds into a ‘‘hairpin structure’’, because it has self-complementary regions that form a dsRNA structure. Other researchers have reported the use of viral promoters for expressing long dsRNA inside the cell. An example is gene silencing by adenovirus-derived siRNA (Shen et al., 2003) that resulted in efficient and specific knockdown of p53 in breast cancer and lung carcinoma cells. RNA-polymerase III dependent promoters expressing hairpin RNAs have been inserted into retroviral vectors and used to achieve effective, specific and stable gene silencing in diverse biological systems (AbbasTerki et al., 2002; Stewart et al., 2003). Antisense RNA transcription is therefore necessary for the generation of the intermediary molecule of communication (siRNA) between the DNA in the nucleus and the mRNA in the cytoplasm in the PTGi mechanism. However, as discussed above, antisense RNA is not the actual effector molecule, or the intermediate molecule of communication between DNA and mRNA degradation. Antisense RNA technologies have been in use for two decades. Historically, chemically modified antisense oligodeoxyribonucleic acids were designed to operate by interaction of antisense RNA with DNA or mRNA to cause inhibition of transcription or translation of an endogenous gene, respectively, thus reducing the level of a target gene product (Scherer and Rossi, 2003; Kurreck, 2003). Despite certain limited successes with antisense RNA technologies, there have been general problems with delivery, stability, dose requirement, specificity and toxicity associated with such techniques. With the understanding of PTGS, GENEi and RNAi, it would be possible to design more effective antisense RNAs, for example, homologous to the 30 translated and untranslated region, which are expressed inside the cell nucleus from heterologous promoters incorporated in transgenic vectors. Such antisense RNA would have the ability to hybridize with the nuclear mRNA and specifically degrade the related cytoplasmic mRNA by activating the RNAi pathway. Fortuitous design of molecules with these characteristics may explain the fact that in some cases, antisense RNAs performed almost as well as double-stranded siRNAs. In this regard, it is significant that similar efficacy of single- and doublestrand siRNAs was demonstrated in human cells (Holen et al., 2003). In a different sub-category, if a component of the ‘‘gene-specific factors’’ according to the above model turned out to be short RNA complementary to the cognate regulatory DNA element, then nuclear overexpression of such ‘‘sense’’—or ‘‘antisense’’—RNA fragments from DNA vectors could possibly destabilize the complex and decrease the target-gene expression.

ARTICLE IN PRESS 312

M.B. Bahramian, H. Zarbl / Journal of Theoretical Biology 233 (2005) 301–314

8. Discussion Fundamental regulatory mechanisms required for maintenance of cellular metabolism and survivals of organisms have been conserved throughout the evolution of plant and animal kingdoms. We believe gene impedance is an example of such fundamental processes. Compared to transcripts in bacteria, eukaryotic mRNAs are generally stable, with some like pro-a1(I) collagen, having half-lives greater than 12 h. Thus, in the absence of an efficient mechanism for selective mRNA degradation of individual mRNA species, translation of the transcripts of a gene to protein products in cytoplasm could continue for many hours after gene transcription had stopped. Thus, an efficient response to signals for elimination of unwanted transcripts seems necessary to maintain a fine balance of cellular proteins. It is for this reason that we maintain that GENEi is a fundamental gene expression regulatory mechanism in animals, plant and lower eukaryotes. This mechanism may have evolved to control over-expression of eukaryotic genes, which could be wasteful and potentially harmful to the cell and the organism. For a major structural protein, such as collagen, the precise control of its expression may not be as critical as for a gene encoding a regulatory molecule that must be rapidly eliminated. Therefore, variation in efficiency of such regulatory mechanism between different genes of the same organism may not be unexpected. Our hypothesis, supported by our previous results and the scientific literature, for the first time, links GENEi and, therefore, RNAi pathway to the actual mechanisms that control gene transcription, including transcription arrest and termination of transcription. The model we have presented, explains in detail the dynamics of interactions of various mechanisms involved during ‘intrinsic’ and ‘alternative’ GENEi. This model is compatible with our current understanding of the mechanisms of transcriptional termination (see Section 4.1). For example, the TTU could be represented by CPSF that interacts with the poly(A) site, while the TRU could be identified with gene specific factors that bind to the downstream ‘‘termination region’’. An important step in our gene impedance hypothesis is that the double-stranded RNA generated in the nucleus during either ‘‘intrinsic’’—or ‘‘alternative’’— GENEi, described in the above model of gene muting, in order to prevent occurrence of interferon response, must be processed in the nucleus to siRNA fragments of less than 30 bp long before being transported to the cytoplasm and being used as the effector molecules in RNAi pathway. Theoretically, it is possible that the long dsRNAs produced by GENEi become associated with nuclear proteins and transported intact to the cytoplasm, where Dicer would process them into siRNAs.

To our knowledge, no such mechanism has yet been identified. Hutva´gner et al. (2004) recently have shown that 20 -Omethyl oligonucleotides can act as irreversible, stoichiometric inhibitors of small RNA-directed RNA silencing in vivo and in vitro by blocking the siRNA-programmed RISC. In principle, it would be possible to use 20 -Omethyl oligonucleotides to provide further support for our GENEi hypothesis. But in practice, it would prove challenging, since the dsRNAs generated in the nucleus are heterogeneous in length and could produce a set of diverse-sequence siRNAs for the same target gene. It may be possible to synthesize several long and overlapping sequences of 20 -O-methyl oligonucleotides, and cotransfect these with GENEi DNA and hope that they trap significant number of activated RISC molecules to reduce silencing. Exploitation of the post-transcriptional feature of GENEi that is mediated by short interfering dsRNA has become a routine tool for transient knockdown of gene expression in a wide range of organisms. It has impacted the study of gene function, including the development of transgenic animals, gene silencing in somatic tissues and functional genomic studies. It also holds the promise of therapeutic gene silencing. Understanding the overall function and mechanisms of GENEi indeed has more potential use than RNAi, which is only a component of the GENEi mechanism. By using the appropriate regulatory DNA sequences as the ‘GENEi nucleic acid’, for example, one could both inhibit transcription and induce specific degradation of the transcripts of a target gene efficiently, independent of transgene expression. This would offer an enormous advantage over transgenes that need to be expressed and siRNAs, since therapy is often challenged by the double uncertainty associated with transgene delivery and expression. Also, as recent studies show (Bridge et al., 2003; Sledz et al., 2003), RNAi has the potential to activate the ‘interferon response’ pathway. GENEi also offers the choice of transcriptional and/or post-transcriptional gene impedance. Blocking transcription of a gene and removing its transcripts at the same time provides a more effective, durable and economic method of gene silencing. It is known that siRNAs that target different regions of the same gene vary considerably in their effectiveness, and the rules governing efficient siRNA-directed gene silencing remain undefined (Dorsett and Tuschl, 2004; Dykxhoorn et al., 2003; Holen et al., 2002). The base composition of the siRNA sequence is probably not the only determinant of how efficiently it will silence a gene. Other factors that are likely to have a role include the secondary structure of the mRNA target and the binding of RNA-binding proteins (Dorsett and Tuschl, 2004; Dykxhoorn et al., 2003; Kretschmer-Kazemi Far and Sczakiel, 2003). Since DNA-induced GENEi according to our hypothesis simultaneously releases

ARTICLE IN PRESS M.B. Bahramian, H. Zarbl / Journal of Theoretical Biology 233 (2005) 301–314

diverse sequences of naturally selected siRNAs corresponding to the same gene target, thus method might prove to be more efficacious and cost-effective over individually designed siRNAs. References Abbas-Terki, T., Blanco-Bose, W., Deglon, N., Pralong, W., Aebischer, P., 2002. Lentiviral-mediated RNA interference. Hum. Gene Ther. 13, 2197–2201. Bahramian, M.B., Zarbl, H., 1994. Direct gene quantitation by PCR reveals differential accumulation of ectopic enzyme in Rat-1 cells, v-fos transformants, and revertants. PCR Methods Appl. 4, 145–153. Bahramian, M.B., Zarbl, H., 1999. Transcriptional and posttranscriptional silencing of rodent a1 (I) collagen by a homologous transcriptionally self-silenced transgene. Mol. Cell. Biol. 19, 274–283. Barker, D.D., Wu, H., Hartung, S., Breindl, M., Jaenisch, R., 1991. Retrovirus-induced insertional mutagenesis: mechanism of collagen mutation in Mov 13 mice. Mol. Cell. Biol. 11, 5154–5163. Bartel, D.P., 2004. MicroRNAs: genomics, biogenesis, mechanism, and function. Cell 116, 281–297. Baulcombe, D.C., 1999. Fast forward genetics based on virus-induced gene silencing. Curr. Opin. Plant. Biol. 2, 109–113. Bernstein, E., Caudy, A.A., Hammond, S.M., Hannon, G.J., 2001. Role for a bidentate ribonuclease in the initiation step of RNA interference. Nature 409, 363–366. Brennecke, J., Hipfner, D.R., Stark, A., Russell, R.B., Cohen, S.M., 2003. bantam encodes a developmentally regulated microRNA that controls cell proliferation and regulates the proapoptotic gene hid in Drosophila. Cell 113, 25–36. Bridge, A.J., Pebernard, S., Ducraux, A., Nicoulaz, A.L., Iggo, R., 2003. Induction of an interferon response by RNAi. Nature Genet. 34, 263–264. Chan, H., Hartung, S., Breindl, M., 1991. Retrovirus-induced interference with collagen I gene expression in Mov 13 fibroblasts is maintained in the absence of DNA methylation. Mol. Cell. Biol. 11, 47–54. Cogoni, C., Macino, G., 1997. Isolation of quelling-defective (qde) mutants impaired in posttranscriptional transgene-induced gene silencing in Neurospora crassa. Proc. Natl Acad. Sci. USA 94, 10233–10238. Cogoni, C., Irelan, J.T., Schumacher, M., Schmidhauser, T.J., Selker, E.U., Macino, G., 1996. Transgene silencing of the al-1 gene in vegetative cells of Neurospora is mediated by a cytoplasmic effector and does not depend on DNA–DNA interactions or DNA methylation. EMBO J. 15, 3153–3163. Dorsett, Y., Tuschl, T., 2004. siRNAs: applications in functional genomics and potential as therapeutics. Nature Rev. Drug Discovery 3, 318–329. Dykxhoorn, D.M., Novina, C.D., Sharp, P.A., 2003. Killing the messenger: short RNAs that silence gene expression. Nature Rev. Mol. Cell. Biol. 4, 457–467. Elbashir, S.M., Harborth, J., Lendeckel, W., Yalcin, A., Weber, K., Tuschl, T., 2001. Duplexes of 21-nucleotide RNAs mediate RNA interference in cultured mammalian cells. Nature 411, 494–498. Farrell, C.M., Lukens, L.N., 1995. Naturally occurring antisense transcripts are present in chick embryo chondrocytes simultaneously with the down regulation of alpha 1 (I) collagen gene. J. Biol. Chem. 279, 3400–3408. Fire, A., 1999. RNA-triggered gene silencing. Trends Genet. 15, 358–363. Fire, A., Xu, S., Montgomery, M.K., Kostas, S.A., Driver, S.E., Mello, C.C., 1998. Potent and specific genetic interference by

313

double-stranded RNA in Caenorhabditis elegans. Nature 391, 806–811. Gambarotta, G., Boccaccio, C., Giordano, S., Ando, M., Stella, M.C., Comoglio, P.M., 1996. Ets up-regulates MET transcription. Oncogene 13, 1911–1917. Gitlin, L., Karelsky, S., Andino, P., 2002. Short interfering RNA confers intracellular antiviral immunity in human cells. Nature 418, 430–434. Grant, S.R., 1999. Dissecting the mechanisms of posttranscriptional gene silencing: divide and conquer. Cell 96, 303–306. Hamilton, A.J., Baulcombe, D.C., 1999. A species of small antisense RNA in posttranscriptional gene silencing in plants. Science 286, 950–952. Hammond, S.M., Bernstein, E., Beach, D., Hannon, G.J., 2000. An RNA-directed nuclease mediates post-transcriptional gene silencing in Drosophila cells. Nature 404, 293–296. Hammond, S.M., Caudy, A.A., Hannon, G.J., 2001. Post-transcriptional gene silencing by double-stranded RNA. Nature Rev. Genet. 2, 110–119. Holen, T., Amarzguioui, M., Wiiger, M.T., Babaie, E., Prydz, H., 2002. Positional effects of short interfering RNAs targeting the human coagulation trigger tissue factor. Nucleic Acids Res. 30, 1757–1766. Holen, T., Amarzguioui, M., Babaie, E., Prydz, H., 2003. Similar behaviour of single-strand and double-strand siRNAs suggests they act through a common RNAi pathway. Nucleic Acids Res. 31, 2401–2407. Hunter, T., Hunt, T., Jackson, R.J., Robertson, H.D., 1975. The characteristics of inhibition of protein synthesis by double-stranded ribonucleic acid in reticulocyte lysates. J. Biol. Chem. 250, 409–417. Hutva´gner, G., Simard, M.J., Mello, C.C., Zamore, P.D., 2004. Sequence-specific inhibition of small RNA function. PLoS Biol 2 (4), E98. Jorgensen, R.A., 1995. Cosuppression, flower color patterns and metastable gene expression states. Science 268, 686–690. Kennerdell, J.R., Carthew, R.W., 1998. Use of dsRNA-mediated genetic interference to demonstrate that frizzled and frizzled 2 act in the wingless pathway. Cell 95, 1017–1026. Kratochwil, K., Mark, K.V.D., Kollar, E.J., Jaenisch, R., Mooslehner, M., Schwarz, K., Haase, I., Gmachl, I., Harbers, K., 1989. Retrovirus-induced insertional mutation in Mov13 mice affects collagen I expression in tissue-specific manner. Cell 57, 807–816. Kretschmer-Kazemi Far, R., Sczakiel, G., 2003. The activity of siRNA in mammalian cells is related to structural target accessibility: a comparison with antisense oligonucleotides. Nucleic Acids Res. 31, 4417–4424. Kurreck, J., 2003. Antisense technologies. Improvement through novel chemical modifications. Eur. J. Biochem. 270, 1628–1644. Marathe, R., Anandalakshmi, R., Smith, T.H., Pruss, G.J., Vance, V.B., 2000. RNA viruses as inducers, suppressors and targets of post-transcriptional gene silencing. Plant Mol. Biol. 43, 295–306. Martinez, J., Patkaniowska, A., Urlaub, H., Luhrmann, R., Tuschl, T., 2002. Single-stranded antisense siRNAs guide target RNA cleavage in RNAi. Cell 110, 563–574. Mignone, F., Gissi, C., Liuni, S., Pesole, G., 2002. Untranslated regions of mRNAs. Genome Biol. 3, 1–10. Minks, M.A., West, D.K., Benvin, S., Baglioni, C., 1979. Structural requirements of double-stranded RNA for the activation of 20 , 50 oligo(A) polymerase and protein kinase of interferon-treated HeLa cells. J. Biol. Chem. 254, 10180–10183. Nehls, M.C., Grapilon, M.L., Brenner, D.A., 1992. NF-I/SP1 switch elements regulate collagen a1 (I) gene expression. DNA Cell Biol. 11, 443–452. Ngo, H., Tschudi, C., Gull, K., Ullu, E., 1998. Double-stranded RNA induces mRNA degradation in Trypanasoma brucei. Proc. Natl Acad. Sci. USA 95, 14687–14692.

ARTICLE IN PRESS 314

M.B. Bahramian, H. Zarbl / Journal of Theoretical Biology 233 (2005) 301–314

Pal-Bhadra, M., Bhadra, U., Birchler, J.A., 2002. RNAi related mechanisms affect both transcriptional and posttranscriptional transgene silencing in Drosophila. Mol. Cell 9, 315–327. Proudfoot, N.J., Furger, A., Dye, M.J., 2002. Integrating mRNA processing with transcription. Cell 108, 501–512. Provost, P., Dishart, D., Doucet, J., Frendewey, D., Samuelsson, B., Radmark, O., 2002. Ribonuclease activity and RNA binding of recombinant human Dicer. EMBO J. 21, 5864–5874. Ramirez, F., di Liberto, M., 1990. Complex and diversified regulatory programs control the expression of vertebrate collagen genes. FASEB J. 4, 1616–1623. Ratcliff, F.G., MacFarlane, S.A., Baulcombe, D.C., 1999. Gene silencing without DNA. RNA-mediated cross-protection between viruses. Plant Cell 11, 1207–1216. Ravazzolo, R., Karsenty, G., de Crombrugghe, B., 1991. A fibroblastspecific factor binds to an upstream negative control element in the promoter of the mouse a1 (I) collagen gene. J. Biol. Chem. 266, 7382–7387. Ritzenthaler, J.D., Goldstein, R.H., Fine, A., Lichtler, A., 1991. Transforming-growth-factor-b activation elements in the distal promoter regions of the rat a1 type I collagen gene. Biochem. J. 280, 157–162. Scherer, L.J., Rossi, J.J., 2003. Approaches for the sequence-specific knockdown of mRNA. Nature Biotechnol. 21, 1457–1465. Sharp, P.A., 2001. RNA interference—2001. Genes Dev. 15, 485–490. Shen, C., Buck, A.K., Liu, X., Winkler, M., Reske, S.N., 2003. Gene silencing by adenovirus-delivered siRNA. FEBS Lett. 539, 111–114. Slack, J.L., Parker, M.I., Robinson, V.R., Bornstein, P., 1992. Regulation of collagen I gene expression by ras. Mol. Cell. Biol. 12, 4714–4723. Sledz, C.A., Holko, M., de Veer, M.J., Silverman, R.H., Williams, B.R., 2003. Activation of the interferon system by short-interfering RNAs. Nat. Cell Biol. 5, 834–839. Stark, G.R., Kerr, I.M., Williams, B.R., Silverman, R.H., Schreiber, R.D., 1998. How cells respond to interferons. Annu. Rev. Biochem. 67, 227–264. Stewart, S.A., Dykxhoorn, D.M., Palliser, D., Mizuno, H., Yu, E.Y., An, D.S., Sabatini, D.M., Chen, I.S., Hahn, W.C., Sharp, P.A., Weinberg, R.A., Novina, C.D., 2003. Lentivirus-delivered stable gene silencing by RNAi in primary cells. RNA 9, 493–501.

Svoboda, P., Stein, P., Hayashi, H., Schultz, R.M., 2000. Selective reduction of dormant maternal mRNAs in mouse oocytes by RNA interference. Development 127, 4147–4156. Tabara, H., Grishock, A., Mello, C., 1998. Soaking in the genome sequence. Science 282, 430–431. Tuschl, T., Zamore, P.D., Lehmann, R., Bartel, D.P., Sharp, P.A., 1999. Targeted mRNA degradation by double-stranded RNA in vitro. Genes Dev. 13, 3191–3197. Vaucheret, H., Beclin, C., Elmayan, T., Feuerbach, F., Godon, C., Morel, J.B., Mourrain, P., Palauqui, J.C., Vernhettes, S., 1998. Transgene-induced gene silencing in plants. Plant J. 16, 651–659. Waterhouse, P.M., Graham, M.W., Wang, M.B., 1998. Virus resistance and gene silencing in plants can be induced by simultaneous expression of sense and antisense RNA. Proc. Natl Acad. Sci. USA 95, 13959–13964. Wianny, F., Zernicka-Goetz, M., 2000. Specific interference with gene function by double-stranded RNA in early mouse development. Nature Cell Biol. 2, 70–75. Xu, P., Vernooy, S.Y., Guo, M., Hay, B.A., 2003. The Drosophila microRNA Mir-14 suppresses cell death and is required for normal fat metabolism. Curr. Biol. 13, 790–795. Yang, S., Tutton, S., Pierce, E., Yoon, K., 2001. Specific doublestranded RNA interference in undifferentiated mouse embryonic stem cells. Mol. Cell. Biol. 21, 7807–7816. Yi, C.E., Bekker, J.M., Miller, G., Hill, K.L., Crosbie, R.H., 2003. Specific and potent RNA interference in terminally differentiated myotubes. J. Biol. Chem. 278, 934–939. Yu, J.Y., Taylor, J., De Ruiter, S.L., Vojtek, A.B., Turner, D.L., 2003. Simultaneous inhibition of GSK3 alpha and GSK3 beta using hairpin siRNA expression vectors. Mol. Ther. 7, 228–236. Zamore, P.D., Tuschl, T., Sharp, P.A., Bartel, D.P., 2000. RNAi: double-stranded RNA directs the ATP-dependent cleavage of mRNA at 21-23 nucleotide intervals. Cell 101, 25–33. Zarbl, H., Latreille, J., Jolicoeur, P., 1987. Revertants of v-fostransformed fibroblasts have mutations in cellular genes essential for transformation by other oncogenes. Cell 51, 357–369. Zeng, Y., Yi, R., Cullen, B.R., 2003. MicroRNAs and small interfering RNAs can inhibit mRNA expression by similar mechanisms. Proc. Natl Acad. Sci. USA 100, 9779–9784. Zhang, H., Kolb, F.A., Brondani, V., Billy, E., Filipowicz, W., 2002. Human Dicer preferentially cleaves dsRNAs at their termini without a requirement for ATP. EMBO J. 21, 5875–5885.