Natural variation in siRNA activity and the influence of chemical and mutational modifications Torgeir Holen A thesis in partial fulfilment of the requirements for the degree of Doctor Philosophiae Επισκεπτεον δε µαλλον και περι του ηγεµονος, ει κεντρον εχει η µη. (But further enquiry is necessary in fact about the leader, as to whether he has a sting or not) Aristotle, Historia Animalium1 1 The Biotechnology Centre of Oslo & Department Group of Clinical Medicine, Faculty of Medicine University of Oslo 2 - Kolofonside- settes opp av Unipub Forlag - ISBN 82-8072-077-4 3 Contents 0.5 Acknowledgements 0.6 A note on this thesis writing 0.7 List of papers 0.8 Abbreviations 1 An introduction to the RNA interference field 1.0 A short overview 1.1 The three central publications in RNAi 1.2 The immune system hypothesis of RNA interference 1.3 Biochemical investigations of dsRNA and siRNA 1.4 The current RNAi model and a challenger 1.5 Non-complimentary short RNA, stRNA, inhibit translation 1.6 Synthetic siRNA versions 1.7 Endogenously expressed short hairpin RNA (shRNA) 1.8 Using siRNA and shRNA against viruses 2 Discussion of results in thesis papers I, II & III 2.0 Results introduction - a summary of ribozyme experiments 2.1 RNA interference in mammalian cells using dsRNA 2.2 Positional effects in a series of siRNA 2.3 On the possibility of finding local targeting rules 2.4 Investigating local area around hTF167i 2.5. Endogenous hTF mRNA targets differ markedly in siRNA accessibility 2.6 The main hypothesis: Natural variation in siRNA activity 2.7 Tissue Factor protein and procoagulant activity also show position effects 2.8 Tolerance of siRNA for mutations 2.9 Tolerance of siRNA for chemical modifications 2.10 Competition from inactive siRNA can block active siRNA 2.11 Cleavage fragment production from siRNA might indicate RNAi exonuclease 2.12 Dose-response experiments 2.13 Propagation of silencing not detected in human cells 2.14 PSKH1 siRNAs had not much activity against endogenous mRNA 2.15 Antisense siRNA compared with double-stranded siRNA 2.16 Blocking 3' OH does block activity in antisense siRNA 4 0.5 Acknowledgements Without the encouragement and stimulating company of Mohammed Amarzguioui and Gaute Brede, this work would never have seen the dusky light of our misty country. Also I would like to acknowledge the work of Merete T. Wiiger for her contributions to our coagulation assays, and Dimitrios Mantzilas for his assistance with the MALDI-TOF. Eshrat Babaie is acknowledged for her running of the nuclic acid synthesis lab, Henning Johansen for his running of the sequencing lab and Liv Bjørland for her running of the chemical solutions lab. And all the other people at the Biotechnology Centre that makes the wheels go around. You are appreciated. A special thanks to Erlend Ragnhildstveit for presenting that paper by Hammond et al at our Journal Club. Finally I would like to thank professor Hans Prydz for hiring me when I came back from America, helping me write a grant application and also financing me for four months after my grant money ran out. Last, but not least, I would like to acknowledge his unique style of leadership, giving us the freedom to fail or succeed, as all of us, so to speak, are the blacksmiths of our own destiny. Torgeir Holen, Oslo, January 2002 0.6 A note on this thesis writing For some reason writing up this thesis seems to me somewhat meaningless. If I had additional scientific insight that could be justified by the data, I should have included it in the papers. Then again, anybody who claims that life has a meaning is selling something, and the University of Oslo is selling doctorates. The organizing principles of this thesis text are firstly chronological, to convey the richness of a scientific field and a PHD thesis work, undergoing, so to speak, gastrulation. Secondly, topical, to discuss the results in the three thesis papers as a whole, and in relation to various interesting aspects and inconsistensies of RNA interference in the published literature. And finally, eclectic, as RNA silencing is already such a wide and rapidly expanding field that a 5 work like this cannot possibly do justice to all the exiting things that are happening. The strain that these three lines of exposition put on the text is necessarily visible. Consequently, the text have been expanded somewhat to make it more readable and endurable for the very few people who will ever read it. I hope it is appreciated. 0.7 List of papers and unpublished data I) Holen, T., Amarzguioui, M., Wiiger, M.T., Babaie, E. and Prydz, H. (2002). Positional effects of short interfering RNAs targeting the human coagulation trigger Tissue Factor. Nucleic Acids Research, 30, 1757-1766. II) Amarzguioui, M., Holen, T., Babaie, E. and Prydz, H. (2003). Tolerance for mutations and chemical modifications in a siRNA. Nucleic Acids Research, 31, 589-595. III) Holen, T., Amarzguioui, M., Babaie, E. and Prydz, H. (2003). Similarity of RNAi antisense and siRNA behaviour indicate a common RNA interference pathway. Nucleic Acids Research (in press). IV) In addition comes some unpublished figures and unpublished data. 6 0.8 Abbreviations 2'MOE 2'-O-(2-methoxy)ethyl 3'deoxy 3' deoxyribose on 3' terminal nucleotide 3' OH 3' hydroxyl Ago-2 Argonaute 2 Arabidopsis Arabidopsis thaliana ATP adenosine triphosphate bp basepair(s) CAF carpel factory CAP 7-methylguanylate on 5' nt C. elegans Caenorhabditis elegans Drosophila Drosophila melanogaster DNA dexoyribonucleic acid dsRNA double-stranded RNA E.coli Escherichia coli esiRNA endoribonuclease prepared siRNA Fig. figure FITC fluorescein isothiocyanate g gravities (multiples of 9.8 m/s2) GC guanosine and cytosine GFP green fluorescent protein h hour(s) H. sapiens Homo sapiens hTF human Tissue Factor HIV human immunodeficiency virus kDa kilo Dalton Luc luciferase MALDI-TOF matrix assisted laser desorption ionization - time of flight microRNA small RNA expressed from a micro gene mRNA messenger RNA mTF murine Tissue Factor nM nanomolar N. crassa Neurospora crassa 7 nt nucleotide(s) ODN oligodeoxynucleotide, oligo oligo oligonucleotide PAGE polyacrylamide gel electrophoresis PCR polymerase chain reaction pGL2 plasmid GL2 (luciferase) PHD doctor (of philosophy) PSKH1 protein serine kinase H1 PTGS post-transcriptional gene silencing PKR protein kinase R rbz ribozyme RdRP RNA-dependent RNA polymerase RISC RNA induced silencing complex RNA ribonucleic acid RNAi RNA interference RNase ribonuclease RSV Rous Sarcoma Virus S2 Schneider cells S100 centrifugal supernatant after 100 000 g shRNA short hairpin RNA siRNA short interfering RNA SNP single nucleotide polymorphism snRNA small nuclear RNA ssRNA single-stranded RNA stRNA short temporal RNA, a microRNA TRE tetracycline response element TGS transcriptional gene silencing TT two successive thymidine nucleotides UTR untranslated region 8 1 An introduction to the RNA interference field 1.0 A short overview Five years after Andrew Fire did his ground-breaking discovery of RNA interference (RNAi), by showing that double-stranded RNA (dsRNA) can efficiently interfere with the corresponding gene,5 a few things have become clear: - RNA interference is an over a billion years old immune system against virus and transposons - the mechanism of action is starting to be uncovered, as a range of proteins have been identified by mutational screens in various species and biochemical in vitro assays have been developed - RNAi can be manipulated by introduction of various forms of RNA to knock down mRNA of interest, something which has implications for both research and clinical applications - related microRNA that are known to regulate development in Caenorhabditis elegans by inhibiting translation of target mRNA, have been found in humans, but their function is unknown - chemical modifications to the triggering RNA agents are tolerated by the biological machinery, thus opening for further improvement in techniques 1.1 The three central publications in RNAi Plant biologists had been seeing the effects of RNA interference since the early 90ies, as plants tend to resist transgenic manipulation by turning off the offending genes by either TGS (Transcriptional Gene Silencing) or PTGS (Post-Transcriptional Gene Silencing).10 PTGS is presumably RNA interference, and plant researchers Andrew J. Hamilton and David C. Baulcombe contributed with the key observation that silenced genes tended to have strange, short (~ 25 nucleotides) RNAs associated with them.6 This observation led to Thomas Tuschl’s breakthrough into the hitherto inaccessible field of human genetics, by demonstrating how these short interfering RNA (siRNA) bypass both other RNAi requirements and the mammalian virus detection system that had blocked RNAi in mammalian cells so far.7 These three publications, by Fire et al, Hamilton & Baulcombe and Elbashir et al are, in my view, key to understanding development of the RNA interference field. 9 1.2 The immune system hypothesis of RNA interference I will initially concentrate on the hypothesis that RNA interference is a billion years old immune system. How do we know this? Primarily we know this due to comparative biology studies done in C. elegans along with studies on plant viruses. In two key publications Ketting et al 8 and Tabara et al 9 identified genes in C. elegans (mut-2, mut-7, rde-2 and rde-3) that when mutated gave rise to stimulation of transposon activity, while having reduced RNAi. Especially mut-7 was interesting, as its inactivation gave rise to a rapidly mutating mutator strain. As this gene has sequence similarity with proteins with 3'-5' exonuclease domains like RNaseD and Werner syndrome protein, Ketting et al proposed a speculative model where an exonuclease degrades mRNA guided by the double-stranded RNA.8 Tabara et al, however, warned against excessive speculation and pointed out that several mutator strains have intact RNAi, among them mut-6. In addition comes mutations that inactivate RNAi, rde-1 and rde-4, that do not mobilize transposons. All this suggest additional complexity.9 Another group of genes (mes-2, mes-3, mes-4 and mes-6) are involved in silencing of transgenes in the germline in C. elegans.11 Testing these mutants, Tabara et al, found normal levels of RNAi.9 However, Dudley et al found that mes-3, mes-4 and mes-6 were in fact needed for RNAi, but that this requirement could be bypassed by using very low levels of dsRNA.12 Obviously, there are still complexities to be unravelled in the genome's immune system.13 Overall, this indicates that RNA interference has a role in transposon suppression, at least in C. elegans. RNAi mutants have also been found in other species such as Drosophila melanogaster, Neurospora crassa and Arabidopsis thaliana. There are interesting sequence similarities between certain RNAi genes, especially in certain domains. For example, mut-7 homologues (in the general sense of the word) were found in screens for RNA interference deficiency in other species, as qde-3 in N. crassa.14 A long series of such conserved RNAi genes have now been found, as shown in a limited summary table (Table 1). 10 C. elegans (putative Arabidopsis Drosophila N. crassa H. sapiens qde-3 14 WRN function) mut-7 8 (RNaseD, wrn, 3'-5' exonucleases) Ego1, Rrf-1, Rrf-2, Rrf-3 18, Sde1/Sgs2 24 - no homologue- qde-1 16 - no homologue - 22 Rde-1 9 (initiator...) 52, 53 AGO1, AGO1, AUB/Sting, qde-2 15, HIWI, EIF2C 20 alg1, alg2 20 + 21 homologs PINHEAD 20 PIWI20 Ago-2 28, 20 20 AGO4 49 Ago-3 29 dcr-1 97 (several RNase-III CAF, SIN1, SUS1 Dicer + several proteins 26) 10 others 26 Upf1/Smg-2 22 (transposon Sde-3 22 (RNA silencing in C. reinhardthii)17 helicase?) 19 (RdRP in tomato 25 ) (RISC components?!) hDicer 26, 27 GB110 (mouse) 22 9 rde-4 (initiator ?!, bind dsRNA,98 bypassed by injected siRNA 53) rde-2, rde-3, mut-2, mut-8-, mut-9 9, 111 mes-2, mes-3, mes-4, mes-6 Enhancer of zeste, 9, 11, 12 extra sex combs 12 (polycomb homologs ) mut-14 (RNA helicase?, needed for antisense) 52 Table 1 A limited summary of sequence similarity in RNAi mutants Another interesting sequence similarity group is the one of ego-1, rrf-1, rrf-2 and rrf3 18, 31, as they are homologues of RNAi genes in other species (Table1), in addition to having sequence similarity in some domains to a known tomato RNA-dependent RNA polymerase.25 As double-stranded RNA stands centrally in RNAi, a lot has been made out of this, something that will be discussed in section 1.4. Furthermore, when rrf-1 is mutated (rrf-1-/-), RNAi activity was abolished.31, 32 However, rrf-2-/- showed no difference, and rrf-3-/- actually showed stimulated RNAi interference, a phenomenon that has recently been put to use testing dsRNA that previously could not induce RNAi phenotypes.32 The most convincing line of evidence, however, for a role for RNAi in virus and transposon defence, is the discovery of viruses in plants that have evolved proteins to specifically knock out RNA interference.10 Such parasite-immune system interactions are well known and a central piece of evidence for RNA interference being a virus defence. Interestingly, Hc-Pro, one such viral protein that can knock out siRNA production, fails to eliminate the still mysterious signal that can spread throughout the plant from an infection site, and make the 11 rest of the plant immune.19 Recently it has been shown that viruses also can inactivate RNA interference in Drosophila, and, impressively, that the inhibitor FHV protein, B2, actually can inhibit RNA silencing in plants too.33 To conclude this part, the mystery of why Andrew Fire’s double-stranded RNA should be able to knock out the corresponding mRNA target seems to have been solved: Doublestranded RNA is a symptom of virus entry or transposon activity, and an immune system has evolved to use this symptom, quite elegantly, to take care of the problem, sometime before the separation of various single-celled, plant and animal eukaryote lines well over a billion years ago.21, 30 1.3 Biochemical investigations of dsRNA and siRNA To solve the problem of mechanism, in vitro reconstitution of RNA interference in Drosophila embryo lysates was developed by Thomas Tuschl, Phillip Zamore and coworkers.34 This in vitro assay was used to show a curious dilution phenomena. From the work of Fire and coworkers it was known that the dsRNA injected into C. elegans would have been diluted to just a few molecules per cell when the 100 descendant embryos of the F1 generation showed sign of target silencing at their 500-cell stage.5 Some silencing could even be seen in the F2 generation, something which gave rise to the hypothesis that this resistance to dilution indicated that an underlying enzymatic system existed, and, possibly, also propagator proteins replenishing the supply of the silencing agent, especially as rde-2 and mut-7 were required for F2 generation interference.70 Tuschl et al showed that although 10 nM dsRNA gave good in vitro silencing, a 10-fold dilution of dsRNA to 1 nM abolished activity. On the other hand, if the dsRNA was diluted by adding fresh lysate every 30 minutes for 3 hours, for a final dilution of 64-fold, the silencing activity was as potent as ever. Something seemed to be happening to the dsRNA. Tuschl and coworkers did not solve the riddle this time around, but they did show that the hypothetical dsRNA-modifying factors could be titrated out, by adding 200 fold higher concentration of competitor dsRNA. Furthermore, they showed that the mRNA target was not mechanically blocked by some association with dsRNA itself, but degraded. Finally, they showed that the silencing activity declined with the decreasing length of dsRNA, a dsRNA of only 49 bp having no activity at all. 12 Hamilton and Baulcombe,6 on the other hand, had chosen to search for small RNAs associated with the silenced state in plants. They reasoned that if these hypothetical antisense RNA molecules were of a similar size to typical mRNAs, they would have been detected by routine RNA analysis. They thus started looking for small RNA, finding that four classes of PTGS in plants, three transgene induced and one virus induced, showed the presence of circa 25 nucleotides long sense and antisense RNA molecules on 15 % PAGE gels blotted to Hybond filters. This was a vital clue to further progress. Hamilton and Baulcombe, however, failed to point out that these ~25 nt RNA could be double-stranded. They proposed that the 25-nt antisense RNA species were not degradation products of the target RNA because they have antisense polarity, so that a more likely source of these RNAs is the transcription of an RNA template, thus favouring the RdRP hypothesis that we will come to later. Meanwhile Tuschl and Zamore had used their in vitro assay to try to find the dsRNAmodifying specific factors. Presumably spurred on by Hamilton and Baulcombe’s ~25 nt RNA, they demonstrated breakdown of radioactive dsRNA to radioactive ~21-23 nt sense and antisense RNA in Drosophila lysates.35 Radioactively labelled, 5’-capped, antisense mRNA was not degraded, nor was presence of target mRNA necessary. Furthermore, both the target mRNA and dsRNA was cleaved at 21-23 nucleotide, or basepair, intervals, suggesting that the 21-23 nt RNAs were guiding the cleavage. In Cold Spring Harbour, another group, led by Gregory Hannon, had started to fractionate Drosophila S2 cell lysates with dsRNA-induced mRNA degradation activity.36 They found that a partial purification (using the S100 supernatant fractionated by an anion-exchange column) contained a clear mRNA degradation peak with a ~25 nt RNA species, that presumably conferred the target specificity. They termed this enzyme RISC for RNA-induced silencing complex. Their lack of observation of stable cleavage endproducts, led them to propose that RISC is an exonuclease, possibly linked to an endonuclease. Or, alternatively, that an RNAi endonuclease makes an initial endonucleolytic cut and that non-specific exonucleases complete the process.36 However, Tuschl and Zamore’s demonstration of obvious cleavage 13 products in Drosophila embryo lysates,35 and, incidentally, our own demonstration of in vivo cleavage products in human HaCaT cells,2 suggest that RISC is primarily an endonuclease, although possibly linked to an exonuclease, as discussed in section 2.11. The first demonstration of RNA interference existing in vertebrates came in zebrafish (Danio rerio).37 However, there were unspecific effects, as seen by the use of the control lacZ control dsRNA. Thus the injection procedure was seen as generally impractical and less robust.38, 39 At this point, inspired by the work of Hammond et al,36 we tried to construct a dsRNA hairpin expressed against a gene of interest, PSKH1,40 as discussed in section 2.1. Hannon’s group went on to find the dsRNA modifying enzyme and called it Dicer,26 and, accordingly, also named the not yet characterized RISC nuclease Slicer.41 In this paper,26 they claim that the RISC activity could be cleared from the extract by a centrifugation of 100 000 g for 60 minutes, leaving the Dicer activity in the supernatant. Strangely, in the previous paper,36 such a centrifugation (100 000 g, 60 minutes) removed ribosomes, but left the RISC. This inconsistency is somewhat confusing. However, in the first paper,36 the 100 000 g centrifugation that left RISC in the supernatant, was proceeded by a 200 000 g centrifugation for 3 hours, before RISC was resuspended by use of salt extraction. Presumably the resuspension process broke up the RISC-ribosome complex and recovering only the so called soluble RISC.28 There also have been some other indications of ribosome association of RNAi,42 and Kennerdell et al report that untranslated mRNA remain resistant to RNAi in Drosophila embryos.43 Hannon has pointed out that their RISC complex varies from other complexes isolated, e.g. it has not been possible to demonstrate mRNA cleavage fragments, indicating that an exonuclease exist in this particular complex.128 That RISC complexes consists of several subunits, that breaks down differently under different conditions, has been amply shown by the three different estimates of the size of the RISC complex.28,44,45 The main point here, however, is that Dicer activity and RISC activity could be separated. This point is somewhat disputed, as some has claimed that Dicer itself is the RISC, in a process termed degradative PCR,46 as further discussed in section 1.4. It should be pointed out that Bernstein et al did not isolate Dicer biochemically.26 Rather they searched the Drosophila genome databases for candidates with sequence similarity to the 14 known dsRNA nuclease RNase-III family. Three classes of hits were found and representatives of these three classes were tested. A candidate gene CG4792, having two RNase-III domains and an amino-terminal helicase domain, showed dsRNA cleavage activity after having been expressed, in a T7-tagged version, in Drosophila S2 cells. Testing whether this enzyme, Dicer, really was causing the production of short RNA, an antibody against the C-terminus of Dicer demonstrated that nuclease activity could be precipitated from cell extracts. Dicer was shown to require ATP as a cofactor. However, Zhang and coworkers found that ATP was not required, and suggested that if ATP participates in the Dicer reaction in mammalian cells, it might be involved in product release needed for the multiple turnover of the enzyme/substrate complex.109 To establish the link of Dicer with RNAi, dsRNA with Dicer sequence was used to deplete Dicer mRNA, which resulted in a seven-fold drop in Dicer activity.26 Elegantly, if somewhat confusing and open to interpretation, RNAi was thus used to demonstrate Dicer requirement in RNAi. A similar technique was used by Hutvagner et al, utilizing siRNA, to establish the link of human Dicer to maturation of let-7, an stRNA.27 However, when immunoprecipitating Dicer, Ago-2, a RISC component, follows, suggesting that Dicer indeed is a component of RISC, but, in certain cases, fails to process dsRNA.28 Obviously, the exact relationships between RISC and Dicer is not yet established. In a very thorough work, using a wide variety of dsRNA, Thomas Tuschl’s new group in Göttingen finally conclusively proved that the 21-22 nt double-stranded RNA (coined siRNA) mediated the RNA interference, by demonstrating uncoupling of dsRNA from the siRNA effect.47 Biochemically isolated siRNA were shown to have a characteristic RNaseIII-like 2-nt overhang at the 3’ end, and terminal 2’ and 3’ hydroxyl groups. Cloning these small RNA (a technique later used for microRNA isolation, see section 1.5) and then synthesizing identical copies by organic chemistry, showed that these synthetic siRNAs had mRNA cleavage activity in Drosophila embryo lysates in the absence of dsRNA. Furthermore, if these overhangs were considerably longer or were absent (blunt), the silencing was lost. This in fact, emphasizes the double-stranded nature of the RNA strands, as it could reasonably be claimed that it was the single-stranded RNA from spontaneous unwinding in vivo that was the underlying cause of the RNAi. A complication arising here, however, is the recent demonstration that the antisense strand of 15 an siRNA, can under certain circumstances work on its own, 2, 3, 45, 48, 52 as will be discussed in the section 2.15. Due to responsiveness to alkaline phosphatase treatment, the 5’ end seemed to have a phosphate group,47 but this was not satisfactorily elucidated before the work of Nykänen et al, who showed that blocking the 5' end with a methoxy moiety blocks silencing in Drosophila lysates.44 Tuschl's group also showed that the mRNA cleavage site is situated centrally in the corresponding siRNA, 47 later determined to be precisely 11 nt from the 5’end of the antisense strand. 69 It was predicted that it would be possible to cleave a target RNA at almost any position, and that RNA interference differed from other types of antisense molecules by not having to search for the optimal target sites.48, 50 This hypothesis we later tested in the first thesis paper,2 as will be discussed in section 2.2. Tuschl’s group went on to use these chemically synthesized siRNAs in cells,7 a breakthrough that is dealt with in section 1.6. 1.4 The current RNAi model and a challenger This short review of the central works of the RNA interference field leaves us with the following model (Fig.1): Aberrant dsRNA is recognized and cleaved by Dicer into short 2123 nt siRNA. These siRNAs are incorporated into RISC, which then can specifically recognize the target RNA and destroy it. This process can be manipulated to accept foreign dsRNA or synthetic siRNA, the topic of section 1.6. A challenge to the current model is the mechanism proposed by Lipardi et al,46 and supported by some data presented by Sijen et al,31 a mechanism which has been termed degradative PCR. The idea is that short RNA, complementary to the mRNA, can anneal to mRNA, and then be extended via the accessible 3’ hydroxyl, by a putative RNAi RNA-dependent RNA Polymerase (RdRP). This would create a new dsRNA, consisting of the target mRNA and a newly synthesized strand, which then can again be cleaved by Dicer to produce more siRNA, and the process is repeated in a chain reaction (Fig.1). 16 17 The idea is attractive. It would serve as a propagating system, giving rise to the lack of dilution of the silencing in C. elegans descendants of the injected worm, as transcribed mRNA would continually be converted into siRNA. It would also dispense of the need for the RISC nuclease, as Dicer would be the active ingredient in RISC. Lipardi et al demonstrated some expected effects of such a system in Drosophila, such as the production of secondary siRNA. Hot-labelled siRNA, incubated with S100 fraction of embryo extracts, led to production of longer transcripts being radioactively labelled, but only when a RNA template was present that was complementary to the siRNA. Interestingly, these longer transcripts then are degraded over a period of 180 minutes, something that Lipardi et al interpret as dsRNA being processed back to siRNA. An objection to these experiments are that irrelevant viral RdRPs might be present. Other groups have also failed in trying to reestablish the Lipardi-mechanism in Drosophila. Sijen et al,31 in C. elegans, went a step further, by demonstrating what they called transitive RNAi, the effect of which is that dsRNA of type X, injected into a worm against a target transcript YX, not only can knock out YX, but also the independent transcript containing only sequence Y. In short, silencing seemed to have spread from X sequences, towards the 5’ end, into Y sequences, and from there to the separate transcript Y. Also, the inverse was shown not to be true, as a target transcript with the sequence XY will not give rise to transitive silencing of Y. More specifically, X here stands for unc22 and Y is GFP.31 Spreading of silencing, even through promoter sequences, has also been seen in Drosophila,90 as discussed in section 2.10. Unfortunately this mechanism has been perceived as inadequate for several reasons: Firstly, nobody has yet biochemically demonstrated any RNAi RdRP-activity, any tortuous argument of sequence similarity to tomato RdRP notwithstanding. Also, humans, and more importantly, Drosophila, have in any case no such RdRP homologue, even if C. elegans does.22 Secondly, Dicer needs ATP and cleavage in vitro can proceed without it,44, 66 even if this argument is weakened by a recent study on human Dicer showing no need for ATP.109 Thirdly, Bernstein et al claimed to be able to separate Dicer and RISC activity by centrifugation,26 although what kind of centrifugal steps performed was not stated clearly.26 An allegedly inactive Dicer was later co-immunoprecipitated with Ago-2, as noted above. Fourthly, propagative effects are not seen with siRNA in mammalian cells as RNAi silencing 18 fades out over time, in our case at 3-4 days post-transfection.2 This has been reproduced in other works.54 Finally, an siRNA with a blocked 3’ end, by a FITC-group, that therefore can not be elongated, had as potent silencing activity as the wild-type siRNA.2 This has been done by Zamore's group, who also showed that 3' blocked siRNA also worked in Drosophila. To control for the possible regeneration of 3'-hydroxyl by loss of the terminal nucleotide, a series of progressively shorter antisense siRNA was tested. The activity was then lower, thus showing that loss of the 3' nucleotide could not possibly give as high activity as the wild-type 21 nt siRNA.48 Recemtly, we have also repeated this experiment, obtaining the same strong silencing, with a siRNA blocked at the 3' end by using a 3' deoxy nucleotide, as discussed in section 2.9. Obviously, the alternative mechanism proposed by Sijen et al,31 whereby the dsRNA is produced by self-priming by the cleaved 5' end of the mRNA, is not invalidated by blocked 3' ends on siRNA antisense strand. However, the absence of transitive silencing when knocking down one isoform, and then reintroducing an almost identical isoform,55 in addition to the loss of siRNA silencing over time,2 would indicate that in human cells this propagative mechanism is not active either. Thus, whatever mechanisms that underlie the data of Lipardi et al and Sijen et al, it can not be universal, as silencing in mammalian cells does not seem to need it. 1.5 Non-complimentary short RNA, stRNA, inhibit translation Yet, there is obviously much that is not known. The investigations into microRNA started with Victor Ambros’ description of heterochronic C. elegans embryo development mutants in 1989. He showed that lin-4 was a regulator of the larva-to-adult molting stage switch. 56 Strangely, the loci this gene was mapped to, turned out to be empty of functional open reading frames. Lee et al thus proposed that lin-4 was expressed as a short RNA that interacted with complementary sequences in the 3’ UTR of the downstream gene of lin-14.57 Later, another such microRNA (also called stRNA (short temporal RNA), after the molting-time effects in C. elegans embryos), let-7, was found.58 This gene has also a homologue that is expressed in humans,59 although the function here is unknown. Using Tuschl’s technique for cloning of small RNA,47 in addition to bioinformatic methods, this class of microgenes was shown to contain at least hundreds of different genes in different organisms.60-62 Their structure seem to be a ~70 nt hairpin RNA that is expressed, and then 19 processed by Dicer into 21-22 nt RNA similar to siRNA, but with only the antisense strand being retained.27 However, recent results investigating temperature dependence of RNA silencing, indicate that temperature-labile siRNA and non-affected stRNA are generated by different nuclease complexes.63 The plant system might be different in this respect, although, since it also has two populations of siRNA with different sizes,64 having recently been shown to originate from dsRNA by plant Dicer.65 Furthermore, in a very recent report, an Arabidopsis Dicer-like mutant, carpel factory (CAF), did not compromise PTGS.68 The stRNA, paired with the mRNA target, forms an interesting non-basepairing bulge in the central region,58 as will be discussed in section 2.8. The functions of these microgenes are unknown. The mechanism of action is only known in the case of lin-4 and let-7 in C. elegans, where they seem to be inhibiting the translation of the target mRNA. By adding pre-let-7 to Drosophila lysates and also in HeLa lysates RNA cleavage activity can be generated if stRNA can basepair perfectly to a target in an mRNA target strand. The function of this bizarre spillover from stRNA to siRNA was proposed by the authors to serve no function in human cells.66 Further research is needed. 1.6 Synthetic siRNA versions A limiting obstacle to use of RNA-induced silencing in mammalian cells was the induction of PKR/interferon/panic response to double-stranded RNA.72 This virus defence shuts down non-specifically gene activity in a cell, even if some mammalian cell types seem to lack this response and thus permit the use of long dsRNA.73, 74 Tuschl and coworkers' breakthrough article demonstrated that synthetic siRNA, as developed in their earlier work in Drosophila,47 could bypass this virus defence and be able to knock down targets in mammalian cells.7 Elbashir et al prepared three siRNA against three different luciferase reporters, pGL2, pGL3 and pRL, and cotransfected these along with two reporter plasmids into various Drosophila, mouse, monkey and human cell lines. In all cases there were clear knock-down of the targeted reporter, as seen by standardization with the control reporter, although the effect was somewhat variable for the human cell lines, something that inspired us to start looking at position effects with siRNA, as discussed in section 2.2. The double-stranded siRNA had 21 nt RNA strands, ending in 2 nt non-basepairing termini. These overhanging 3' ends Elbashir et al replaced with DNA thymidines instead of uridine, 20 with the rationale that it was cheaper and could perhaps give better protection against degradation. A control, using UU instead of TT, showed no significant difference, thus siRNA was shown for the first time to tolerate chemical modifications, as further discussed in section 2.9. It should also be noted that the last T, or U, in the antisense strand of GL2, uGL2 and GL3, did not basepair with the mRNA target, as this is CA in both X65324 and U47296, for pGL2-Control and pGL3-Control, respectively, and thus constituted a kind of mutation, as further discussed in section 2.8. In contrast, longer dsRNA, ~50 bp or ~500 bp, did induce a 10-20 fold and 200 fold reduction, respectively, of all reporters in a non-specific fashion, assumedly by the PKR system. Even so, a specific depletion of the target reporter could be seen superimposed onto the non-specific results. Why this cohabitation of the RNAi system and the PKR system should exist, is still unclear. Elbashir et al then used siRNA to knock out endogenous proteins lamin A/C, lamin B1, NuMA and vimentin.7 Caplen et al presented much the same siRNA results a few months later,75 with the additional claim that 22-23 nt siRNAs were somewhat more efficient, of the 21-27 nt strand range tested, in contrast with Elbashir et al, who found that 21 nt duplexes were the most efficient, of the 20-25 nt strands tested.69 Caplen et al also demonstrated directly that the PKR response was not activated by siRNA, something that have been supported by later studies.76-78 Furthermore, these oligos had no DNA 3' end overhangs, but were no more efficient. Interestingly, siRNA antisense oligos were found to have no activity, as also many others have reported later, something that will be discussed in section 2.15 on antisense siRNA. Boutla et al were the first to publish a report showing siRNA to work in vivo in Drosophila embryos, and that a central mutation was partly tolerated.79 However, Elbashir et al also published an article on the functional anatomy of siRNA late in 2001, reporting that a central mutation abolished silencing activity in their Drosophila embryo lysate,69 in apparent contrast with both Boutla et al 79 and our own results,2 as discussed in section 2.8. Furthermore, Elbashir et al found that DNA nucleotide substitution was tolerated in the 3’ end of the RNA strands (up to 4 DNA nucleotides at 3' end), but that full substitution of DNA or 2’-O-methyl nucleotides abolished activity, as will be discussed in section 2.9. They also found support for their earlier finding that 2 nt 3’ overhang siRNA were the most effective,47, 69 while finding that some overhang sequences in the antisense strand were more effective than others (UG, 21 UU, TdG and TT), possibly due to the penultimate nucleotide contributing to target recognition.69 Nykänen et al showed by fractionation of cell lysate that it is not the siRNA incorporated into the RISC complex that is the active complex, but a RISC complex containing the singlestranded antisense strand of the siRNA, a complex designated RISC*.44 This active complex size was found to be less than 232 kDa (inactive RISC ca. 360 kDa) something that clashes with Hammond et al’s size of 500 kDa,28 and Tuschl’s recent report of 90-160 kDa.45 Furthermore, it was reported that two of the four sequential steps found (Fig.1), needed ATP, both the siRNA unwinding and mRNA cleavage, in addition to there seeming to be a pair of putative gateway kinases and phosphatases that cycled phosphates on the 5’ end of free siRNA. As noted above, when this 5'-hydroxyl is blocked by a methoxy moiety the activity of the siRNA is lost.44 1.7 Endogenously expressed short hairpin RNA (shRNA) A third class of RNA molecules, in addition to siRNA and stRNA, must briefly be mentioned. Although it has been shown that dsRNA as short as 36 basepairs are inactive,47 and that dsRNA longer than 30 bp will induce the PKR/interferon/response,72, 75 many groups commenced to try expressing hairpin RNA shorter than 30 bp. Transcription in vivo, would be an advantage, as delivery of siRNA can be a problem, although use of viral vectors developed for gene therapy have its own problems.80, 131 Use of shRNA gave good results.76, 81-89 The basis of this technology is that RNA polymerase III will terminate transcription of a transcript after 5 successive thymidines in the template and the transcript is cleaved after two uridines.81 Thus a short transcript, in the form of a hairpin, or with two complementary and separate strands, is possible.82, 86 The shRNA seems to be processed by Dicer to siRNA, whether they were longer mimics of let-7 hairpins, or simple, 29 bp hairpins.84 Although Brummelkamp et al found that the 9 nt loop is clearly better than a 5 nt and 7 nt loop, 81 something that also is corroborated by Castanotto et al,89 a wide range of loops connecting the two strands have been used with good results: Paul et al and Paddison et al used 4-loops (UUCG and UUAA, respectively),83, 84 while Yu et al and Yang et al used 3loops (AUG and UCU, respectively).87, 88 Sui et al reports that a 6-loop works well,85 while 22 Jacque et al find no difference between 3-, 5- and 7-loops.76 More research seems to be needed also here. Ohkawa and Taira, significantly, have constructed a RNA polymerase III promoter that is regulatable, by fusing this promoter with a Tetracycline Response Element (TRE).71 This system will obviously find many applications for tissue and time-point specific induction of RNAi silencing. In fact, Taira is in the process of commercialising this plasmid, as requests for it recently have become too exceptionally intense and time consuming to respond to. 1.8 Using siRNA and shRNA against viruses A long series of impressive results against virus targets have been announced the last 12 months. The first report was from John Rossi's group, where targeting HIV-1 rev transcripts, in an cotransfection assay, gave an impressive 10 000-fold reduction of p24, using siRNA expressed from two complementary U6 snRNA promoter constructs, a construct that is now in the process of being reconstructed for delivery by lentivirus.86 Jacque et al, using synthetic siRNA against several regions of the HIV-1 genome such as LTR, vif and nef, found that in a cotransfection assay, virus production was reduced 30-50 fold. More importantly, even if the HIV-1 infection process itself was not interupted, production of viral cDNA intermediates were dramatically inhibited if the cells had been transfected by siRNA against the HIV-1 genome, even if the virus was added up to 4 days after the siRNA transfection.76 A third group, led by Phillip A. Sharp, demonstrated siRNA inhibition against HIV-1 cellular receptor CD4, Gag and GFP-for-Nef targets,91 and recently Martinez et al demonstrated inhibited HIV-1 replication by using siRNA against CXCR4 and CCR5.92 In a very interesting piece of work, Gitlin et al, targeted polio-virus by transfecting synthetic siRNA into HeLa cells before infection with the cytolytic and apoptosis-inducing Mahoney strain.77 Remarkably, siRNA inhibited plaque formation completely. In co-electroporation assays, where target mRNA was introduced along with siRNA, maximum silencing was seen already after 3 hours. However, the silencing was lost over time, and when the emerging strains were sequenced, it was found that they contained mutations in the siRNA-targeted site, as will be discussed in section 2.8. The first use of siRNA against viruses in a full animal, was done by Hu et al, where co-electroporation of RSV proviral DNA and siRNA into the developing chick neural tube, inhibited viral replication.93 Recently, Kapadia et al has reported inhibition of hepatitis C replication in Huh-7 cells.78 23 2 Discussion of results in thesis papers I, II & III 2.0 Results introduction - a short summary of ribozyme experiments This thesis work intended to use ribozymes for functional genomic studies on PSKH1.94 A year of artefact-prone work ensued, where the luciferase-based assay system repeatedly demonstrated good results (ca 60-80 % depletion of target signal), a silencing signal which was partially lost when the assay was tried out under different conditions, e.g. like the shift from COS-1 cells to HeLa cells or the shift from using total protein content to using Renilla Luciferase as an internal control (data not shown, unpublished). Such assay dependence will be discussed in section 2.5. It was finally decided to do inactivation mutations to the ribozyme core. It is known that hammerhead ribozymes have the conserved core (3-9) sequence CUGANGA, where G-5, G-8 and A-9 are vital for cleavage activity.95,96 Interestingly, our mutated ribozymes turned out to be just as active, or even more active, than the wildtype ribozymes against the Tissue Factor reporter construct TF-Luc (Fig.2B, unpublished). Also note that different hTF ribozymes seem to have different depletion capability (Fig.2A, unpublished), a position effect similar to the one seen previously with ribozymes against PSKH1-Luc.103 We went on to investigate position effects with siRNA, as discussed in section 2.2. That mutated ribozymes were just as effective as the non-mutated ribozymes, probably means that the Tissue Factor ribozymes had no significant catalytic effect upon their target mRNA. Some irregularities discovered in ribozyme synthesis and storage, threw some doubt upon these conclusions, as ribozymes had been exposed to unbuffered, acidic conditions. This could possibly account for both the lack of catalysis, the position effect seen, and some of the non-specific effects. Routine testing of the ribozymes for degradation with PAGE gels had disclosed no significant degradation, but there seemed to be a possibility of acidic conditions having led to depurination in certain positions. MALDI-TOF mass spectrometry, however, although no optimalization was performed, and only rough peaks were obtained, showed no sign of peak shift due to loss of purines, and thus seemed intact (data not shown, unpublished). In addition, theoretical judgement expected that such depurination would have led to breakdown and cleavage of the RNA strand, supporting the hypothesis of the ribozymes indeed having no significant enzymatic activity. 24 A 120 100 80 60 40 20 hT F1 78 hT F2 61 hT F2 70 hT F3 34 hT F3 83 hT F4 52 hT F4 89 hT F5 07 hT F5 24 hT F5 73 hT F7 80 co nt ro l 0 B 120 100 80 60 40 20 0 control rbz hTF383 383mut1 383mut2 hTF489 489mut1 489mut2 Fig.2 Tissue Factor ribozymes exhibit position effects (A) and strong activity after having the conserved core mutated (B). Cotransfection of ribozymes and TF-Luc reporter was done essentially as described for siRNA in paper I. Ribozymes 383mut1 and 489mut1 have methylated G-5. Ribozymes 383mut2 and 489mut2 have a G5A mutation in addition to methylation of this A. To explain the sporadic depletion effects seen with ribozymes in certain assays, one should look to the long list of improvements of ribozymes tha have been tried out, one of them using hybridizing DNA arms on a RNA ribozyme core.95 This, in in vitro reactions on short RNA 25 targets, will improve both enzymatic turnover (DNA:RNA duplexes bind less tight, allowing higher turnover) and protect against degradation. In addition, RNaseH, is known to recognize and cleave RNA annealed to DNA. Furthermore, these DNA arms can also be synthesized with phosphorothioate nucleotides, with the intention of improving the degradation resistance even more. Finally, RNaseH is known to be very promiscuous in its cleavage of such thiophosphodiester DNA:RNA hybrids, in some cases only needing 5-7 hybridizing basepairs.39, 99, 100 Thus, our ribozymes were probably in fact depending on RNaseH action. Although there seemed to be some weak activity against endogenous Tissue Factor mRNA (data not shown, unpublished), probably due to RNase H action, control ribozymes actually stimulated Tissue Factor expression 2-3 fold (data not shown, unpublished). Tissue Factor, in its role as the principal initiator of blood coagulation, is known to respond to a wide range of agents causing perturbation of vascular endothelium, among them bacterial lipopolysaccharide, lectins, phorbol ester, endotoxins and inflammatory cytokines,101 thus the response to extensively modified nucleic acids is not surprising, as this might be perceived as foreign material. As will be discussed in section 2.5, siRNA, interestingly did not induce endogenous Tissue Factor. 2.1 An attempt at RNA interference in mammalian cells using long dsRNA Our first ribozyme target was PSKH1,40 and a lot of effort was spent to get PSKH1 ribozymes to work reliably both in our cotransfection assay 103 (and data not published) and against endogenous PSKH1 mRNA in HeLa cells, with only limited success (data not shown, unpublished). Thus, in the wake of Hammond et al’s paper,36 a plasmid expressing a dsRNA of the protein coding region, was attempted constructed. A second copy 1275 bp coding region of PSKH1 was attempted ligated bidirectionally into a BamHI site after the first copy of the PSKH1 coding region, in what should have produced 50 % clones of each variant. However, it turned out that our standard E. coli strain Dh5alpha only produced direct repeats after selection for clones. In all 12 clones with direct repeats were isolated, but no inverted repeats. Likewise, an attempt to construct a similar TF dsRNA construct was aborted. The problems was solved by trying out the RecABC-deficient E.coli strain SURE-R, and a PSKH1 dsRNA hairpin construct was finally obtained, indicating that our standard strain Dh5alpha indeed rejected inverted repeats by recombination (data not shown, unpublished). 26 Tested in our cotransfection assay against our PSKH1-LUC reporter plasmid, the dsRNAexpressing construct showed no activity (data not shown, unpublished). Why this didn’t work was unclear, at least to us, until Tuschl’s group published their paper on short interfering siRNA avoiding the PKR-response.7 The methodology, however, was later proved to be sound, as two plasmids expressing two versions of shRNA (hTFsh and hTFsh+6) against Tissue Factor, based on John Rossi’s plasmid construct,86 proved to deplete TF-LUC in our assay by ca. 85 % and 70 %, respectively, in our cotransfection assays (data not shown, unpublished). Sites on mRNA targets can be differentially accessible to ribozymes and oligodeoxynucleotides.103 Therefore, Thomas Tuschl’s breakthrough, showing that RNA interference could be obtained with short dsRNA (siRNA), made us consider a comparison between ribozyme and siRNA position effects, as it seemed to us that certain siRNA species had limited efficiency.7,47 2.2 Positional effects in a series of siRNA A series of siRNA was synthesized to test whether our ribozyme position effects also would occur with siRNA (Fig.1A - paper I). Four siRNA were made against each of our two human genes of interest, PSKH1 and Tissue Factor (hTF), with two of these four targeting the best ribozyme sites and two targeting the least active sites, on each mRNA. Surprisingly, using Tissue Factor siRNA and reporter construct (Fig.1B - paper I), a strong position effect was found (Fig.1C - paper I), but it did not correlate with the ribozyme data. This has some implications for certain ideas of mRNA target accessibility, as discussed in section 2.3. To further investigate the positional effect, we synthesized several series of siRNAs against new target sites that could be thought to have different degrees of secondary mRNA structure or general accessibility. The translation initiation site (hTF77i) and the 3'-end of the coding region (hTF929i) were both essentially inaccessible targets. Two other siRNA targeted heavily basepaired regions as predicted by MFold.104 One (hTF256i) demonstrated intermediate activity while the other (hTF459i), had low activity (Fig. 1C - paper I). From this position effect, supported by further position effect data from other measurement assays (discussed in section 2.4, 2.5 and 2.15), springs the main hypothesis of this thesis, that 27 siRNA exhibits natural variation in activity, as explicitly posed in section 2.6. Further discussion of possible targeting rules are deferred to section 2.4. Support for the existence of siRNA position effects is emerging from the literature, as several examples of inactive or less active siRNAs in mammalian cells have been described,78, 82, 84, 86, 105-107, 119, 124, 134 while a weak positional effect in Drosophila lysates can be inferred from published data.47, 69 In C. elegans, Simmer et al managed to activate dozens of previously inactive dsRNA stretches, using the RNAi-enhancing rrf-3 negative strain, and in some cases created RNAi knockout phenotypes from genes that had hitherto not been responsive to dsRNA.32 Yang et al, working with a wide range of different techniques, reported some inactive chemically synthesized and in vitro transcribed siRNA, as well as inactive shRNA. Interestingly, the authors also established the new technique of esiRNA (endoribonucleaseprepared siRNA), in which an overlapping set of siRNAs are produced in vitro by partial digestion with E. coli RNase III, a technique that can be superior even to long dsRNA.88 Processing of dsRNA by Dicer starts from a fixed end and proceeds in a sequential manner,47,109 and is thus expected to produce a largely non-overlapping set of siRNAs. Cleavage of a dsRNA by E.coli RNase III will thus create a larger and more complete set of siRNAs than the non-overlapping set produced by Dicer from dsRNA. The higher activity of the esiRNA supports the argument that different siRNAs have different activities. Otherwise any set of siRNA from a dsRNA would have the same activity. That esiRNA might not be a wholly superior technique to siRNA, as esiRNA will average out the siRNA performence, can possibly be inferred from our competition studies, as discussed in section 2.10. Recently Taira's group also has used recombinant Dicer to produce siRNA from dsRNA in vitro (calling it diced siRNA), finding that both puromycin and H-ras mRNA could be more depleted using diced siRNA than any other of 10 specific positions using siRNA. Some of these 20 siRNA had also low activity, giving further support of the existence of position effects.110 28 2.3 On the possibility of finding local targeting rules One way to deal with RNA target selection, has been to postulate open areas more vulnerable for attack and then try to find these theoretical local targets by computational methods based on mRNA sequence and measured energies for such structures as hairpins. An example is such secondary structure prediction software as Mfold.104 We found no correlation between MFold predictions and our siRNA data, nor, as noted above, with our ribozyme data. This supports other reports in the literature finding MFold inadequate as a target prediction tool.110 In any case, mRNA is not naked in vivo, but packed with RNA binding proteins. These RNA binding proteins, in addition to, and in cooperation with, secondary RNA structure, is possibly responsible for our observed position effects, as also pointed out previously for certain less than fully active siRNA.7 Furthermore, probably few antisense oligo can work in an exquisitely specific manner without protein effectors, making mRNA target cleavage a complex interaction between mRNA structure, mRNA binding proteins, antisense basepairing and effector protein(s). Such a complicated process, involving many factors, seems less likely to have so simple local rules that they can be inferred from experiments typically featuring a few mRNA targets and a few dozen antisense agents. More to the point, this multifactor process might easily be imagined to be the physical basis for the main hypothesis of this thesis, namely a continual range of siRNA with differing silencing activity, as posed in section 2.6. Even the pharmaceutical company ISIS, which routinely screens hundreds of targets with massive tiling of oligos, has found no rules increasing the success rate to better than one good hit in 50, although later generations of oligos, like 2’MOE, have a somewhat higher success rate. A review of the literature found that at best, in large studies, one in eight antisense oligonucleotides are active. Despite this, 1655 citations reported success with only a single oligo.99 In this perspective, effective targeting rules might still be some way off. 2.4 Investigating local area around hTF167i Going on to investigate the target positions close to our best agent hTF167i, we found a phenomenon illustrating this local targeting rule problem very nicely. A series of siRNA, each sharing 18 out of 21 nucleotides with each neighbour (Fig.2A - paper I) had very different activities (Fig.2B - paper I). E.g. hTF158i and hTF161 differed by only 9 and 6 nucleotides, 29 respectively, from hTF167i, thus sharing 12 and 15 nucleotides, respectively, yet they were practically inactive. Alluringly, hTF167i lies in the middle of a window of very low mRNA GC content, suggesting that target accessibility might be due to this easily determined factor. Furthermore, Tuschl et al, in addition to licensed manufacturers of siRNA, have previously recommended approximately 50 % GC content.124 However, a review of the siRNA in the published literature and their targets, found only a weak bias of uncertain biological significance in favour of low GC content. Recently Taira's group also found no GC bias in their sample of 20 siRNA against a puromycin-resistance gene and H-ras. Interestingly, one of their best siRNA against H-ras had a GC-percentage of 84 %, while one of the worst had 53 % GC.110 2.5. Endogenous hTF mRNA targets differ markedly in siRNA accessibility An effort had been going on for a while to try out our best ribozymes against endogenous mRNA targets, with very little success, as the best Tissue Factor ribozyme, capable of ca. 75 % target depletion in our cotransfection assay (Fig.2A, unpublished), had only limited activity against the endogenous target (data not shown, unpublished). For PSKH1 the results were similar. Such lack of robustness in activity, can also be inferred from the literature, where good initial results seldom seem to be followed up by application in more stringent and complicated assay systems. An interesting example from the Tissue Factor field is Stephens & Rivers obtaining already in 1997 an impressive 80.4 % (+/- 2.2 %) suppression of Tissue Factor activity in primary monocytes with an antisense oligodeoxy nucleotide (ODN).113 This remarkable success has not been followed up by the authors, or anybody else, even if almost 1600 articles with keyword Tissue Factor has been published in the six years hence. Such an assay system dependence, or contextual dependency, is perceived to be chronic in the antisense field.99,100 With this system dependence in mind we turned from reporter constructs to endogenous targets for siRNA. There we demonstrated that the pattern of depletion in our cotransfection assays were mirrored in results on endogenous Tissue Factor mRNA targets, both on a global (Fig.3A - paper I) and on a local (Fig.3B - paper I) mRNA target scale. Reassuringly, endogenous Tissue Factor expression, known to respond to many 30 environmental stimuli,101 was not stimulated by the introduction of siRNA, as seen by the transfection of the irrelevant control siRNAs PSK314i, PSK546i, PSK566i, PSK739i (Fig.3Apaper I). The position effects were in fact more clear-cut with our Northern assay than with the cotransfection assay, as the siRNA with some activity in cotransfections had practically no activity in Northerns. That there might be false positives in the cotransfection assay, something that we have seen indications of, emphasizes a certain weakness of a reporter system dependent on forced expression of fusion constructs from plasmids transiently transfected. The overall similar pattern of depletion shows that the cotransfection assay is of value as a fast and convenient tool for initial investigation. However, the cotransfections results must later be backed up with data on endogenous targets. An illustrating example, among several others in the RNAi field, of trusting cotransfection data perhaps a bit too far, can be seen in a recent paper by Chiu and Rana, where they claim that crosslinking of siRNA does not completely abolish silencing, as 30 % silencing was still retained, in a cotransfection assay.114 The similar pattern in cotransfection assays and with endogenous targets, have implications also for the search for local targeting rules, as the fusion mRNA, hTF-LUC, was targeted much the same way as the hTF mRNA itself. Furthermore, preliminary results from mouse cells indicate that position effects are strong also in mTF, and, interestingly, targets corresponding to our best hTF siRNAs are also the best in mouse cells, even if the conserved sequence similarity between hTF and mTF is only ca. 66% (data not shown, unpublished). This probably means that there are local rules, shaped by the local nucleotide pattern, although they might not be easily elucidated or generalized. 2.6 The main hypothesis: Natural variation in siRNA activity Thus far we have demonstrated inactive siRNA and active siRNA. Furthermore, there probably exists a balance between RNAi mRNA depletion and mRNA production during the long decline of silencing from 24 h till the effect fade completely 2-3 days later, as discussed in section 2.13. But what is the situation for our best candidate hTF167i at the point of maximum silencing at 24 h? A major hypothesis of this thesis is that this siRNA at this point has excess capacity, as in able to cleave more mRNA if it was present. Two good siRNA candidates would generally obtain the same high level of silencing, as 31 transfection efficiency limits in the silencing assay obscure any differences. Thus, the existence of the excess capacity class of siRNA must be demonstrated indirectly by the high tolerance for mutational and chemical modifications, as will be discussed in section 2.8 and 2.9, respectively. Dosage-response experiments discussed in section 2.12 and 2.15, can show more directly that at certain siRNA concentrations silencing capacity is saturated, and could be used to differentiate excess capacity siRNA. Finally, time-course experiments, as discussed in section 2.13, can show that certain agents silence faster and more persistently, making it possible to differentiate candidates that all give complete silencing under standard conditions. More theoretically and abstractly, a continuum of siRNA with differing activities from very high activity to inactive there can be imagined. Practically it might be easier to think of siRNA as falling into three broad categories, and more importantly, this is easier to demonstrate: 1) Inactive siRNA. 2) Intermediate activity siRNA. 3) Excess capacity siRNA. Certain observations in our papers, and in the literature, become explainable in the light of this hypothesis. 2.7 Tissue Factor protein and procoagulant activity also show position effects Going on to measure the downstream protein level, and, more importantly, the physiological process of blood clotting, we transfected HaCaT cells 48 h before harvest, to allow for protein half-life of 8 hours.115 Again there was a pronounced position effect, with Tissue Factor protein being depleted ten-fold by the best candidate hTF167i and five-fold by hTF372i (Fig.3C - paper I), while hTF478i had no activity (data not shown). Finally, using a procoagulant assay,116 we showed that hTF167i reduced activity five-fold, with hTF372i also here having less effect (Fig.3C - paper I), thus demonstrating siRNA position effect in four different assays. This robust effect shows that RNA interference holds great promise for downregulation, not only of gene transcripts, but of gene function. 32 2.8 Tolerance of siRNA for mutations In line with the ribozyme inactivation experiments (Fig.2B, unpublished), we wanted to try mutational inactivation of siRNA. Two versions of hTF167i, with one or two central mutations, were synthesized and named M1 and M2 (later renamed s10 and ds10/11 in paper II: Fig.6a - paper I, Fig.1 - paper II). Later, this central position was shown to be close to the putative active site of the RISC, as the cleavage site was found to be exactly 11 nt from the 3' end of antisense strand.69 In addition, GC pairs were chosen for mutation to disrupt more hydrogen bonds possibly involved in hybridisation. Satisfyingly, these two mutations showed a gradual decline of activity (Fig.6C - paper I). Interestingly, M2 had higher activity in the Northern assay than in the TF-LUC assay, where M2 was almost inactive. In the perspective of a recent prediction by Hutvagner & Zamore, where it was proposed that bulges in the stRNA is causing the translational inhibition rather than mRNA degradation,66 our experiment indicated that the nature of the bulge is also important. If a central bulge was all that is needed, as indeed seems to be the case for the AU-bulge for let-7 targeting two targets in lin-41,58 there would not be much degradation of mRNA, (although there would be some, as there seems to be spill-over into RISC by stRNA),66 while the protein would be silenced. This is actually the opposite of what we see in our experiment with M1 and M2, where endogenous TF mRNA is being degraded, while TF-LUC is not significantly silenced (Fig.6 B,C - paper I). Further support of this is found in a very recent report, where Zeng & Cullen performed a systematic series of mutations on the human microRNA miR-30, showing that the most sensitive area was not the bulge, but the stemstructure at the 5' and 3' end, just outside the processed stRNA.117 Other combinations of modified and wild-type strands, thus with a bulge in the siRNA itself, were also tolerated well, with the combination M1/wt (sense/antisense) and M2/wt siRNA not having much less activity than wild-type, while the mutation in the antisense strand was more deleterious (Fig.6B,C - paper I). This has previously been shown for dsRNA,67 and for 3' end overhangs,69 and has recently been reproduced.112 DNA:RNA, or DNA:DNA oligos, on the other hand, had little activity, although the antisense RNA had some interesting weak activity possibly due to antisense siRNA effects (Fig.6B,C - paper I), as will be discussed in section 2.15. 33 Wanting to explore the tolerance for mutations further, we did study introducing a series of G:C mutations from one end of hTF167i to the other. We found that hTF167i had a general tolerance to mutations, with less tolerance for mutations at 3’ end of the siRNA (Fig.2 - paper II), possibly facilitating design of single-nucleotide polymorphism (SNP) specific siRNA. However, preliminary results from recent targeting of A:T pairs in the 3' end, show that s14 and s17 seem to be well tolerated (data not shown, unpublished). Tuschl and coworkers found the diametrically opposite effect, namely that double mutations in the 3' end of the siRNA were less deleterious to silencing than mutations in the 5' end of the siRNA, in a Drosophila lysate assay.69 Overall, our best agent hTF167i seem to exhibit great tolerance for mutation, supporting the existence of siRNA with excess capacity. Further support for this view can be found in preliminary data showing the intermediate activity siRNA hTF372i and hTF173i being more affected by s7-mutations than hTF167i (data not shown, unpublished). The most recent review of RNAi still claims that one of the unique features of RNAi is its exquisite sequence specificity, where a single mismatch can abolish activity,118 while Tuschl’s latest review conclude that the silencing effect is extraordinarily specific, because one nucleotide mismatch between the target RNA and the central region of the siRNA is frequently sufficient to prevent silencing.54 However, more reports on tolerance of single mutations are now available. In the first published paper on siRNA mutations, other than the 3' end mismatches of Tuschl's siRNA,7, 47 Boutla et al reported that a mutated siRNA with a single centrally located mismatch relative to the mRNA target sequence, retained substantial activity when injected into Drosophila embryos.79 Jacque et al found that a single mismatch in siRNA targeting the HIV LTR resulted in only partial loss of activity, while another siRNA targeting the HIV VIF exhibited almost full activity.76 In a GFP-fluorescence cotransfection assay, Yu et al 's single mutation in the antisense strand was tolerated, while, paradoxically, the corresponding mutation in the sense strand led to reduced silencing, although against the antisense target this mutated siRNA had good activity.87 In contrast, Elbashir et al found that a single mismatch, in the tenth or eleventh basepair (position), counting from the 5' end of the siRNA, was deleterious to activity in a Drosophila embryo lysate assay.69 That there is no mRNA production in such in vitro assays complicates 34 matters, as an siRNA that in vivo had not sufficient capacity to influence the mRNA level, might still be effective against limited amounts of RNA target. However, abolishment of in vivo activity by a single mutation has been reported. Brummelkamp et al,81 using a short hairpin RNA (shRNA), processed to siRNA by Dicer,84 achieved inactivation by single mutations in either the second or ninth position from the putative 5’ end of the shRNA. Gitlin et al, argued the case for single mutation inactivation more strongly by isolating siRNAresistant polio virus strains containing single mutations in the target site on the genomic RNA, corresponding to the sixth or ninth nucleotide of the siRNA, counted from the 5’ end of the sense strand.7 The ninth position was also mutated by Martinez et al, resulting in strong inhibition of a mutant form of p53, while leaving wild-type p53 untouched.108 In the context of the hypothesis of natural variation in siRNA activity, it is possible that these results are explainable by these particular siRNA falling into the intermediate capacity group, and thus having no excess capacity to resist mutations. To support this view, it can be observed that Gitlin et al also tested antisense siRNA, showing this to have no activity. Since antisense siRNA do have activity, although at a lower efficiency, as discussed in section 2.15, it might be argued that if Gitlin et al's siRNA had excess capacity, its antisense siRNA partner should also have had some remaining activity. In stark contrast to single mutations, four mutations abolished activity completely for Jaques et al.76 Complete abolishment of activity has also been reported by Gitlin et al, Klahre et al and Garrus et al, for siRNAs with 5, 6 and 7 mutations, respectively.77, 120, 121 Paddison et al reported that a triple mutation in an shRNA abolished activity in a cotransfection assay,84 as did Elbashir et al for the pair of GL2 and GL3 siRNA.7,124 Thus, a double mutation is perhaps the threshold for toleration for most siRNA. Kisielow et al reported that an siRNA resulting in essentially complete knockdown of the expression of its target gene, was unable to inhibit the expression of an isoform from a transiently transfected plasmid, an isoform having two non-contiguous mutations in the recognition sequence of the siRNA.55 The positions of these mutations correspond to our ds10/13 double mutant, which, along with two other double mutations (Fig.1 - paper II), also exhibited low activity in our assay (Fig.2 - paper II). Yu et al, Lassus et al and Wilda et al also reported abolishment of activity with a double mutation.87,122,123 35 Therefore, our results, and the literature, so far, is consistent with the hypothesis of siRNA falling into three general groups of siRNA ranked on their activity, as posed in section 2.6: Inactive siRNA, intermediate activity siRNA and excess capacity siRNA. This latest group of excess capacity siRNA, seen in conjunction with our results showing a gradual decline of activity when introducing modifications in our best agent hTF167i (section 2.9), might be interesting for siRNA-based drug development. 2.9 Tolerance of siRNA for chemical modifications As noted in the introduction, the siRNAs first used by Elbashir et al in mammalian cells, were not wildtype siRNA, as they were synthesized with two 3’ end overhangs of deoxyribose thymidine nucleotides (TT), one of which did not basepair with the mRNA target.7 Our siRNA, having to target ribozyme sites, and following Tuschl's example, also did not basepair with mRNA in these overhangs in the penultimate nucletide. At the time we worried that this might be the reason for some of our inactive siRNAs (Fig.1 paper I), and the loss of silencing seen in time-course experiments, as discussed in section 2.13. Since a major objective initially was to try to detect such long-term silencing (or propagative effects) that had been seen in C. elegans,5 we decided to test both base-pairing, non-TT deoxyribose overhangs for hTF167i (hTF167i-BO), in addition to resynthesizing pure RNA strand versions of hTF167i, hTF173i, hTF256i, hTF372i, hTF478i and hTF562i with base-pairing overhangs. In all cases, TT-nonbasepairing overhangs, DNA-basepairing overhangs and all-RNA siRNA, we found no significant change of activity between versions measured at 24 h, or in time-course experiments, using either Northerns (Fig.6D, and data not shown - paper I) or reporter plasmids in serial transfection assays (data not shown). Thus our work supported Tuschl's conclusions of some tolerance for modifications in the 3' end of siRNA.7, 47, 69 More extensive tolerance for DNA modifications in the 3' end was also found by Tuschl's group, as even 4 DNA nucleotides, at each 3' end, did not affect activity much, although full DNA or 2-O-methyl substitution abolished activity.69 This tolerance has recently been exploited very elegantly by Martinez et al to isolate RISC complexes by way of photo-cleavable linkers from the 3' end of the RNA strand to a biotin marker.45 More 5' and 3' modifications will be discussed in section 2.16. Wanting to use fluorescent siRNA, we synthesized an antisense oligo version, with a FITC 36 group tagged to the 3' end, targeting position hTF167. In a siRNA duplex, hTF167-FITC, this oligo had no significant loss of activity (Fig.6B,C - paper I), something which has attracted some interest in the RNA interference field, because this is not consistent with the mechanism proposed by Lipardi et al,46 as mentioned in the introduction. Also, since we had been bothered somewhat by the thought of synthesis failures being responsible for the lack of activity in some of our siRNA, even if they were tested for annealing on non-denaturing PAGE gels, these repeated rounds of synthesis gave additional support for the observed position dependence. During the mutational screening (section 2.8), we indeed ran into two cases of documented difficulties with synthesis failure. We were able to detect which synthesized RNA strand was faulty, when they failed to anneal in our routine PAGE controls, by annealing each to different sense and antisense versions of the two strands. In two cases, one of the two RNA strands failed this cross-annealing test, and were then resynthesized (data not shown, unpublished). Our time-course experiments (section 2.13) and the observed tolerance for 3' end modifications, in the context of our earlier experiences with ribozymes, led us to the thought of trying out a series of experiments modifying the ends of the RNA oligos, with the objective of reducing the vulnerability to degradation. However, since RNA oligos in the RNA interference system must be incorporated into a protein complex, we expected chemical modifications to affect the silencing activity in a negative way. This in contrast with the antisense field generally, where the appropriate modifications can improve oligo activity.95, 132 We wanted first to test three different chemical modifications that earlier have been used in the ribozyme field for improvement of nuclease resistance, namely phosphothioates, 2-Omethylation and 2-O-allylation.95, 103 With the possible loss of activity in mind, oligos were chosen with double substitution in the 3' end only. We expected the 5' end to be vulnerable, since this is where putative RNAi kinases and RNAi phosphatases cycle a phosphate group.44 A series with a single modification in the 5' end and the 3' end each was also synthesized. In all, 6 siRNA were synthesized, P0+2, P1+1, M0+2, M1+1, A0+2, A1+1. In this pilot experiment the primary objective was to estimate the activity loss, as such minor modifications can not be expected to stabilize oligos against degradation. Interestingly, five of these siRNAs showed no significant loss of activity (Fig.3 - paper II), 37 while the siRNA with an allyl group on the 5' nucleotide had markedly lower activity than wildtype siRNA and the other modified siRNA. We proposed that this might be due to disturbance of 5' phosphate regulating enzymes,44 although it should be noted that A1+1 still had good silencing capability. This general tolerance to these and other modifications, in our best agent hTF167i, can be interpreted as supporting the existence of the excess capacity siRNA. Abandoning allylation, another series of siRNA were synthesized with more extensive 2-Omethylation or phosphorothioate backbones: M2+2, M2+4, P2+2 and P2+4. At this point oligo degradation should start to be inhibited,95 but the question was whether the silencing activity loss was acceptable. Satisfyingly, these siRNAs still showed strong activity (Fig.3 paper II), although somewhat lower than wildtype, and more persistent silencing in our timecourse assay (Fig.4 - paper II). Unfortunately, the phosphorothioate-modified siRNAs now exhibited cytotoxicity, in particular P2+4, which was removed from further analysis. Interestingly, oligos with more extensive phosphorothioation have been used with good effect.95,99,132 Indeed, the pharmaceutical company ISIS have drugs in clinical trials with such modifications.132 Why our oligos had strong toxicity when others use them without major side effects even in preclinical and initial clinical tests, is unclear, although it is possible that this is an example of the assay system dependence discussed in section 2.5, or, perhaps, our modified oligos interfered with the RNAi components themselves. A final series of modification was tried out, as we had in mind that Tuschl's fully 2-Omethylated siRNA had no activity at al.69 A series of siRNA were prepared: M4+4, M4+6, M6+6, M6+8. A gradual decline of activity would support the hypothesis of the existence of a continual range of siRNA activities rather than abrupt thresholds. Indeed, this last series of modifications finally seemed to exhaust the activity of our best candidate hTF167i, as M6+8 (having only 7 pure RNA basepairs left in the centre of the siRNA) had almost no silencing effect (figure 3A - paper III). In one representative experiment, the activities of these siRNA exhibited 28%, 42%, 68% and 75% residual expression, respectively (data not shown - paper II). The last chemically modified oligo produced so far, a 3' deoxy version of the antisense strand of hTF167i, also had activity indistinguishable from the wildtype, or, indeed, hTF167-FITC, when incorporated in an siRNA duplex hTF167-3d (Fig.3C - paper III). Working on its own, 38 as an antisense siRNA oligo, it had severely limited activity compared with wildtype antisense siRNA, as will be discussed in section 2.16. 2.10 Competition from inactive siRNA can block active siRNA That the observed limit (at ca. 10 % residual mRNA) to depletion is probably caused by untransfected cells is now well established, but at the time we wished for a stronger effect. It was reasoned that the use of single siRNA is not a biologically relevant situation, as Dicer would produce a range of different siRNA. It was possible that there existed cooperative effects between pairs of siRNA, as also suggested by others.136 One scenario was that multiple RISC complexes could be stimulating each other to give faster mRNA cleavage when several of them were situated near each other on the same mRNA. Testing this idea by performing a range of experiments using several siRNA simultaneously, we obtained some initial success that did not hold up to scrutiny. It seemed possible, however, that the excess capacity of hTF167i in the transfected cells masked the cooperative effect. One way to differentiate siRNA that have excess capacity, would be to measure the persistence of the silencing, therefore the idea of RISC cooperation was considered falsified when timecourse experiments showed that a triplet of hTF167i, hTF372i and hTF562i did not give a significant increased persistence (data not shown, unpublished data). During these experiments it was noticed, however, that inactive siRNA could affect active siRNA in a negative fashion. Exploring this effect, in a cotransfection assay, we found that with a ratio of 2:1 (20 nM inactive: 10 nM active siRNA) a weak inhibition could be seen, even if 10 nM active siRNA was well into the saturation area. Furthermore, when the ratio was increased to 9:1 (27 nM inactive : 3 nM active siRNA), the silencing was practically abolished even if 3 nM hTF167i normally should have full silencing capacity (Fig.4A - paper I). The competition effect could also be demonstrated with different active siRNA and blocked by different inactive siRNAs (Fig.4B, C - paper I). In addition siRNA competition was demonstrated in the three other assays, endogenous mRNA, protein and procoagulation (Fig.4D - paper I). Where along the RNAi pathway the competition occurred (Fig.1), we were not able to determine but the order of siRNA transfection did not affect hTF167i activity in competition experiments. This suggests that siRNA incorporation into RISC is not irreversible (data not shown). Two other studies using siRNA competition have now been published. McManus et 39 al report that inactive siRNA can compete with active siRNA when electroporated into E10 cells, finding an inability to silence both CD4 and CD8alpha targets simultaneously,134 while Martinez et al found that siRNA incorporation into RISC was irreversible, in a HeLa S100 extract, and thus impregnable to competition.45 Biologcially, this competition effect poses some interesting questions. How can RNA interference cope with more than 9 targets if a 9:1 ratio (inactive:active siRNA) can block siRNA silencing? Virus and transposons also can produce large amounts of dsRNA, so how will a prodigious production of newly Dicer-cleaved siRNA affect suppression of other targets? How can RNAi silence a target that has very few open target positions, such as Tissue Factor, and PSKH1 (section 2.14), seemingly have? Would not Dicer then produce many inactive siRNA competing with the active siRNA produced? Furthermore, as siRNA has been proposed a role in clinical applications, will not a flood of siRNA against a drug target led to transposon (and/or virus) release and the mutational effects seen in C. elegans mutator strains? Finally, will not the above considerations, if valid, make necessary a link from RNA interference system to a more permanent silencing, such as chromosomal silencing, with RNA interference being an analogue to the innate immune system compared with immune memory? Data for such a link between RNAi and chromosomal silencing are scarce and scattered. Phillip A. Sharp has proposed a link to the Polycomb-system,135 based on the data, among others, from James Birchler and coworkers' series of experiments in Drosophila,90,129,130 where silencing (cosuppression) seems to flow from homologous sequences in the expressed mRNA region, and through promoter sequences to silence a gene with only sequence similarity in non-expressed regions. More specifically, the transgene construct w-Adh, with promotor sequence from w and transcribed region from Adh, silenced not only endogenous Adh, but also affected Adh-w. The latter, consisting of promotor sequence from Adh and transcribed region from w, had no sequence similarity with w-Adh in the transcribed region, and only promoter sequences in common with Adh. Furthermore, Adh-w is only silenced when endogenous Adh is present.90 Birchler’s group also demonstrated a partial loss of silencing when using Polycombgroup (PcG) mutants, a group of genes implicated in chromosomal silencing in Drosophila. Also, certain RNAi mutants in C. elegans have sequence similarity with Polycombgroup genes,12 40 something that is especially interesting as RNAi homologues have now been implicated in centromere silencing in S. pombe.141 Furthermore, methylation of DNA induced by dsRNA in plants is well established.142 Still, the mechanism of RNA induced DNA silencing is yet very unclear. A last note on competition must be made, as the esiRNA technique have in two distinct cases proven to be superior to siRNA.88, 110 However, our competition data indicate that esiRNA will average out the siRNA performance, as the postulated few fully superior siRNA would be outnumbered by siRNA of a more average activity. On the other hand, there might be mechanisms for selection of the best siRNA, as an analogue to the classical immune system's antibody selection and affinity maturation. Verification or falsification of this idea must await a study testing a full tiling of an mRNA target, to see whether the best siRNA still are better than esiRNA, as yet a very expensive exercise. 2.11 Cleavage fragment production from siRNA suggests the existence of an RNAi exonuclease There seems to be a relation between the phenomena of siRNA excess capacity and siRNA cleavage fragment production. The appearances of the latter (Fig.3A - paper I) are puzzling. The 3’ end cleavage fragments that we would be isolating with our poly-dT beads, thus an mRNA not protected by a 5’ CAP, should not be stable more than a few minutes in vivo.138, 139 Likewise, single-stranded siRNA oligos are rapidly degraded in lysates.27,48 So why do we see a signal that appears to be as strong as the main band at maximum depletion (10 % +/- 1.2 % residual signal) there at all? We have been forced to consider that there must be some kind of stabilization after endonucleolytic cleavage, perhaps due to a rate-limiting factor at a later stage in the RNAi pathway. More specifically, the appearance of the cleavage fragment only occurs when the silencing effect is the strongest. It disappeared when the concentration of siRNA is decreased (Fig.6E paper I), when the siRNA is made less efficient by mutations (Fig.6C - paper I), or when the siRNAs are targeting less accessible positions like hTF173 (data not shown). Interestingly, this cleavage fragment appearance thus seems to correlate with the speed with which the silencing is established (Fig.5A - paper I). If one can assume that our best agents hTF167i, hTF167-M1 and hTF372i have a higher depletion speed, that is, higher mRNA 41 cleavage rate than hTF173i and hTF167-M2, it would seem reasonable that the cleavage product is a classical rate-limiting result due to a saturated, and mRNA-stabilizing, downstream factor. We proposed that this limiting factor might be the RNAi exonuclease postulated, as the RISC, by Hammond et al, in Drosophila.36 One candidate would be the human mut-7 homologue, Werner syndrome protein (WRN), as this has an 3'-5' exonuclease domain.8 Some effort, over a dozen experiments, have been spent to try develop an in vitro assay in HaCaT cells to investigate this cleavage fragment phenomena, so far with little success. However, when mRNA transcription was turned off using Actinomycin D, it was found that natural turnover of Tissue Factor was faster than RNA interference, as could have been expected by the halflife of Tissue Factor and the depletion speed of RNAi (Fig.5A - paper I). More interestingly, the two effects are additive (Fig.5A - paper III). This can be seen to support the existence of an independent RNAi exonuclease, which possibly can be saturated. Whereas if RNAi tagged mRNA for degradation by non-specific exonuclease, as proposed by Hammond et at,8 such saturation would not seem possible. A system, like the degradosome, which handles the continual mRNA degradation of thousands of different genes, would not be saturated this easily. Incidentally, the requirement for DEAD-box RNA helicase in the degradosome for 3'5' degradation,140 also suggests other roles for the RNAi RNA helicases than making initial mRNA cleavage possible or unwinding siRNA in RISC. 2.12 Dose-response experiments The excess capacity siRNA hypothesis has been supported by dose-response experiments, demonstrating a saturation level of circa 5 nM in the cotransfection assay (Fig.4A, lower curve - paper I), and circa 30 nM in the Northern assay, above which more siRNA will not result in more silencing, and thus can be said to have excess silencing capacity. Saturation of silencing above a certain threshold has been seen also by others.7, 48, 124, 134 However, our dose-response experiments have been criticized by an anonymous referee to paper I. The referee's main objection was that at lower concentration, the complexation between nucleic acids and the lipid transfection agents might be less efficient, but this argument founders on the fact that all siRNA doses were complexed together in one volume, at standard complexation concentration. A range of transfection concentrations was obtained when this complexation solution was diluted into different volumes laid upon the cells. 42 Complexation creates small lipid micelles, containing siRNA, as can be visualized using fluorescent tagged siRNA. These, presumably, will fuse with cells, when hitting the cell membrane, creating a speckled appearance, with many speckles per cell (Fig.5B - paper I). It is still possible that there might be threshold effects, but the main point here is that the doseresponse curves actually seem continuous, from full activity to no activity (Fig.4A - paper I). Lowering the transfection concentration will probably only led to fewer hits per cell, and not any abrupt threshold effects, although this experiment has not yet been done. 2.13 Propagation of silencing not detected in human cells Silencing induced by siRNA is slow to reach a maximum (Fig.5A - paper I) and fades out 3-4 days post-transfection (Fig.5C - paper I). Similar results have recently been obtained by Phillip Sharp's group.134 Other groups have measured later recovery of expression. 54, 88 The relatively rapid recovery of expression is perhaps due to the high transcription rate and rapid turnover of Tissue Factor mRNA, which has a half-life of only 0.5-1.5 hours.101 Support of this notion can be found in the fact that challenge with greater amounts of TF-Luc reporter plasmids, using the serial transfection assay, showed less silencing when using more plasmid (data not shown - paper I), as already noted in section 2.5. In addition, the HaCaT cells typically expanded 10-fold during the 6 day period of these experiments, leading to lowering of siRNA concentration in the cells. The significance of this loss of silencing is that it constitutes an argument against a system of propagation (as in the Lipardi-mechanism or transitive RNAi) in human cells, as noted in section 1.4. There had been a report that RNAi silencing could be mediated by the growth medium.133 Furthermore, RNAi silencing does cross cellular boundaries in C. elegans.5 Therefore we tried incubating cells in the presence of siRNA. No unaided uptake was found, however, as the cells showed no silencing when transfected with reporter plasmid the following day. Recently, functional, unaided uptake has been observed with rat cholangiocytes.119 Furthermore, when co-culturing cells transfected with siRNA alongside cells transfected with reporter plasmids, no flow of silencing from the siRNA transfected cells were detected (data not shown). Another apparent ambiguity in siRNA uptake is the already noted inability of siRNA to silence the last 10 % of the mRNA in mammalian cells. The most plausible reason is that Northerns include the cells that resist transfection. This is supported by the work of Tuschl’s 43 group when it was observed that cells are being silenced in a binary manner, single cells appearing totally unsilenced in a field of silenced cells.7 Experiments with a fluorescent siRNA, hTF167-FITC, showed a speckled distribution on all cells examined. The fluorescent signal was detectable from 2 h post-transfection until and beyond 22 h post-transfection (Fig.5B - paper I). This probably reflects small lipid micelles fused to the cells, as noted in section 2.12. That all cells showed speckles, could mean that the cells immune to silencing have micelles fused to them, but that the siRNA are not released from the lipid micelles for some reason, perhaps due to lack of lipid rearrangment dependent upon cell cycle. Also, slow or absent release could partly explain the slow build-up of silencing (Fig.5A - paper I). Some support for this view can be found in experiments done by Gitlin et al where electroporation of the siL siRNA into HeLa cells, led to almost full silencing of both luciferase target already after 2-3 hours.77 The mRNA target was co-electroporated into the cells, although, and might lack the protein coverage, or at least the same kind of protein coverage resulting from RNA polymerase II transcription and processing by the splicing machinery. However, McManus et al also used electroporation in E10 cells and observed maximum silencing at 36 h.134 A series of experiments testing a non-liposome transfection agents, calcium phosphate precipitation, to see whether this would affect speed of silencing, showed in prelimenary experiments strong silencing in cotransfection experiments (data not shown, unpublished). However, calcium levels needed for efficient transfection were somewhat toxic to HeLa cells. Furthermore, HaCaT cells, which express our endogenous target TF mRNA, is not compatible with high calcium levels, as this will disturb the TF expression.101 We have also performed a series of pilot experiments on primary monocytes, which can be stimulated to express hTF,101 but these have proved much harder to transfect than cell lines (unpublished, data not shown). However, there have been many reports of natural uptake of nucleic acid oligos in vivo,132 in addition to reported results with siRNA in whole mice,51,102 thus we remain optimistic that using siRNA against hTF, and the corresponding mTF in whole mice, should still be possible. 2.14 PSKH1 siRNA had not much activity against endogenous mRNA One siRNA in the first series of four siRNA against PSKH1 had good activity (60-80 %) 44 against PSKH1-Luc in our cotransfection assay (data not shown, unpublished). Several series of new siRNA against PSKH1, in total 12, all ended up showing no substantial activity against endogenously expressed PSKH1 in HeLa cells. Interestingly, PSKH1 mRNA is expressed at a low level, even in HeLa cells, possibly indicating that RNAi might have a lower threshold for mRNA. Alternatively, PSKH1 mRNA might be particularly packed with RNA binding proteins. A third possibility is that the high GC percentage of PSKH1 is influencing the silencing. However, testing this hypothesis of low GC content being more accessible, by synthesizing a series of six siRNA against low GC sites in PSKH1 had no more success (data not shown, unpublished). 2.15 Antisense siRNA compared with double-stranded siRNA Our best siRNA candidate, hTF167i, has a corresponding antisense siRNA (as-167) that has strong silencing activity at a standard transfection concentration of 200 nM (RNA concentration) (Fig.5 - paper II). This is interesting in the context of a series of papers reporting no activity for antisense siRNA, including Caplen et al, Gitlin et al, Lee et al, Castanotto et al, Novina et al, Hu et al, Chiu et al and Elbashir et al.75, 77, 86, 89, 91, 93, 114, 124 We found that as-167 has a 5-6 fold higher IC50 than its partner hTF167i, and is already approaching a saturation of silencing at 100 nM (Fig.1A - paper III), in a manner similar to saturation effects seen previously (Fig.4A - paper I). The 5' phosphorylation status, crucial for activity in an earlier report,45 did not seem to matter (Fig.5 - paper II). Furthermore, we find that the pattern of position effects are very similar for antisense siRNA targeting same sites as double-stranded siRNA (Fig.1B,C - paper III). Thus, siRNA with intermediate activity, being used as antisense siRNA, could possibly lose all its activity, yielding the results seen in the above cited publications. 75, 77, 86, 89, 91, 93, 114, 124 Excess capacity siRNA, on the other hand, would have spare activity to tolerate the 5-6 fold lower efficiency of antisense siRNA, in a manner seen by our best candidate pair hTF167i / as-167, thus being consistent with our main thesis of natural siRNA variation, as posed in section 2.6. However, the prevailing results regarding the relative efficiencies of siRNA and RNAi antisense are quite contradictory in the literature. Martinez et al states that RISC activity could not be induced by RNAi antisense in Drosophila lysates,45 while Schwarz et al found 45 that it could.48 Furthermore, Martinez et al claim as good antisense effects against the protein A/C lamin in vivo in HeLa cells, as with double-stranded siRNA.45 On the other hand, Schwarz et al report that they need eight-fold more antisense to approach the potency of siRNA.48 A long series of RISC-complexes and RISC-like complexes have been described. 28, 44, 45, 66, 125-127 The hypothesis of a shared RNAi pathway, despite all the cited possible effector complexes, we reasoned, would be falsified if major differences in antisense and doublestranded siRNA were found. We therefore compared a series of antisense siRNA and doublestranded siRNA, and found a close similarity with regards to position effects (Fig.1B,C paper III), appearance of mRNA cleavage fragments (Fig.1C - paper III) and tolerance for mutational and chemical backbone modifications (Fig.2, Fig.3A - paper III). These experiments therefore did not falsify the hypothesis of a shared RNAi pathway. Furthermore, we found that in competition experiments, double-stranded siRNA could block antisense siRNA (Fig.4 - paper III), something which support the notion of a shared RNAi pathway. Finally, when comparing the speed with which silencing is obtained, we found that antisense siRNA worked significantly faster than double-stranded siRNA (Fig.5C - paper III), suggesting that antisense siRNA enters the RNAi pathway at a later stage, as schematically illustrated in Figure 1. 2.16 Blocking 3' OH blocks activity in antisense siRNA Constructing a fluorescent version of hTF167i for use in uptake studies (section 2.17), hTF167-FITC, we found that even an siRNA tagged at the 3’ end of the antisense strand with a bulky FITC group, somewhat surprisingly, had only marginally lower activity than the wildtype (Fig. 6B, 6C). These data were challenged by Zamore's group, on the grounds that we had not proved that all siRNA had been FITC-tagged.48 However, the synthesis of our FITC-tagged oligo was done using nucleotides which already had the FITC group on the 3' hydroxyl. Furthermore, our dose-response experiments show clearly that even if a small fraction did escape FITC-tagging, the concentration of this untagged fraction would be too small to account for the robust silencing both at 24 h and in the time-course studies, as seen in the dose-response experiment (Fig.6E - paper I). However, if the terminal nucleotide was completely removed, in vivo, and thus regenerated 46 the hydroxyl group, the silencing observed could still be due to a free 3' hydroxyl. Zamore and coworkers controlled for this by showing that if the hydroxyl was regenerated by loss of the 3' terminal nucleotide, the siRNA would have lost activity.48 Additional support of the 3' modification tolerance, as noted earlier, comes from Tuschl's group, which used biotin-linkers at the 3' end to isolate RISC complexes, without loss of activity, something that even worked on the 5' end on the sense strand.45 Some closing notes on antisense siRNA must be added, as several experiments indicate that the mechanisms of antisense siRNA and double-stranded siRNA are not entirely identical. We found, as noted in the introduction, that double-stranded siRNA could very well have the 3' end (of the antisense strand) blocked by a FITC-group (Fig.6B,C - paper I). Furthermore, in time-course studies, hTF167-FITC had the same silencing decay found with hTF167i in serial transfection assays at least up to 72 h (data not shown, unpublished data). However, when testing the corresponding antisense siRNA (as-FITC), we found that it had severely limited activity (Fig.3C - paper III). Reasoning that this could be due to inefficient incorporation into RISC*, because of the bulky fluorescent group, we synthesized a version having a hydrogen group (3'deoxy, termed as-3d) replacing the 3' hydroxyl on the 3' terminal nucleotide. This too had practically full activity as a double-stranded siRNA (hTF167-3d), but low activity as an antisense siRNA (Fig.3C - paper III). In apparent contrast with this, Schwarz et al found that in HeLa lysates there was no difference between the cleavage activity of 3' hydroxyl and 3' blocked antisense siRNA.48 The cleavage rate was much weaker than with duplex siRNA, however, but this was accredited to degradation, and they concluded that the 3' hydroxyl had no role in RNAi in flies or humans. Interestingly, they tested 5' blockage of antisense siRNA, but not their 3' blocked antisense siRNA, in vivo in HeLa cells, thus there is no direct inconsistency between our results. Degradation should not be a direct factor, unless a 3' deoxy and 3' FITC are, for some strange reason, particularly vulnerable compared to wild-type as-167. Other results in the literature also indicate a role for the 3' hydroxyl, as Tijsterman et al report that a 3' hydroxyl on antisense siRNA are needed in C. elegans.52 Taira's group report even of inactivated double-stranded siRNA due to 3' modifications.112 So there still might be a role for the 3' hydroxyl. Furthermore, the published results on the 5' end modifications were not completely satisfactory. Schwarz et al's 5' amino-modifier was less destructive of silencing 47 than the 5' methoxy, although these results were obtained in cotransfection assays that might be vulnerable to non-specific effects.48 These paradoxes will possibly yield to further research. But further enquiry is necessary. 48 References 1. Aristotle, Historia Animalium, VIII, XLII, 25, ~335 BCE. (translated and edited by Balme, Peck and Gotthelf, The Loeb Classical Library, Harvard University Press, London, 1991). 2. Holen T, Amarzguioui M, Wiiger MT, Babaie E, Prydz H. Positional effects of short interfering RNAs targeting the human coagulation trigger Tissue Factor. Nucleic Acids Res. 2002 Apr 15;30(8):1757-66. 3. Amarzguioui M, Holen T, Babaie E, Prydz H. Tolerance for mutations and chemical modifications in a siRNA. Nucleic Acids Res. 2003 Jan 15;31(2):589-95. 4. Holen T, Amarzguioui M, Babaie E, Prydz H. Similarity of single-stranded and doublestranded siRNA behaviour suggests they act through a common RNAi pathway (submitted). 5. Fire A, Xu S, Montgomery MK, Kostas SA, Driver SE, Mello CC. Potent and specific genetic interference by double-stranded RNA in Caenorhabditis elegans. Nature. 1998 Feb 19;391(6669):806-11. 6. Hamilton AJ, Baulcombe DC. A species of small antisense RNA in posttranscriptional gene silencing in plants. Science. 1999 Oct 29;286(5441):950-2. 7. Elbashir SM, Harborth J, Lendeckel W, Yalcin A, Weber K, Tuschl T. Duplexes of 21nucleotide RNAs mediate RNA interference in cultured mammalian cells. Nature. 2001 May 24;411(6836):494-8. 8. Ketting RF, Haverkamp TH, van Luenen HG, Plasterk RH. Mut-7 of C. elegans, required for transposon silencing and RNA interference, is a homolog of Werner syndrome helicase and RNaseD. Cell 1999 Oct 15;99(2):133-41 9. Tabara H, Sarkissian M, Kelly WG, Fleenor J, Grishok A, Timmons L, Fire A, Mello CC. The rde-1 gene, RNA interference, and transposon silencing in C. elegans. Cell. 1999 Oct 15;99(2):123-32. 10. Vance V, Vaucheret H. RNA silencing in plants - defense and counterdefense. Science. 2001 Jun 22;292(5525):2277-80. 11. Kelly WG, Fire A. Chromatin silencing and the maintenance of a functional germline in Caenorhabditis elegans. Development. 1998 Jul;125(13):2451-6. 12. Dudley NR, Labbe JC, Goldstein B. Using RNA interference to identify genes required for RNA interference. Proc Natl Acad Sci U S A. 2002 Apr 2;99(7):4191-6. 13. Plasterk RH. RNA silencing: the genome's immune system. Science. 2002 May 17;296(5571):1263-5. 14. Cogoni C, Macino G. Posttranscriptional gene silencing in Neurospora by a RecQ DNA helicase. Science. 1999 Dec 17;286(5448):2342-4. 49 15. Catalanotto C, Azzalin G, Macino G, Cogoni C. Gene silencing in worms and fungi. Nature. 2000 Mar 16;404(6775):245. 16. Cogoni C, Macino G. Gene silencing in Neurospora crassa requires a protein homologous to RNA-dependent RNA polymerase. Nature. 1999 May 13;399(6732):166-9. 17. Wu-Scharf D, Jeong B, Zhang C, Cerutti H. Transgene and transposon silencing in Chlamydomonas reinhardtii by a DEAH-box RNA helicase. Science. 2000 Nov 10;290(5494):1159-62. 18. Smardon A, Spoerke JM, Stacey SC, Klein ME, Mackin N, Maine EM. EGO-1 is related to RNA-directed RNA polymerase and functions in germ-line development and RNA interference in C. elegans. Curr Biol. 2000 Feb 24;10(4):169-78. 19. Mlotshwa S, Voinnet O, Mette MF, Matzke M, Vaucheret H, Ding SW, Pruss G, Vance VB. RNA silencing and the Mobile Silencing Signal. The Plant Cell. 2002. 14: S289-301. 20. Grishok A, Pasquinelli AE, Conte D, Li N, Parrish S, Ha I, Baillie DL, Fire A, Ruvkun G, Mello CC. Genes and mechanisms related to RNA interference regulate expression of the small temporal RNAs that control C. elegans developmental timing. Cell. 2001 Jul 13;106(1):23-34. 21. Javaux EJ, Knoll AH, Walter MR. Morphological and ecological complexity in early eukaryotic ecosystems.Nature. 2001 Jul 5;412(6842):66-9. 22. Dalmay T, Horsefield R, Braunstein TH, Baulcombe DC. SDE3 encodes an RNA helicase required for post-transcriptional gene silencing in Arabidopsis. EMBO J. 2001 Apr 17;20(8):2069-78. 23. Bernstein E, Denli AM, Hannon GJ. The rest is silence. RNA 2001 Nov;7(11):1509-21. 24. Dalmay T, Hamilton A, Rudd S, Angell S, Baulcombe DC. An RNA-dependent RNA polymerase gene in Arabidopsis is required for posttranscriptional gene silencing mediated by a transgene but not by a virus. Cell. 2000 May 26;101(5):543-53. 25. Schiebel W, Pelissier T, Reidel L, Thalmeir S, Schiebel R, Kempe D, Lottspeich F, Sanger HL, Wassenegger M (1998). Isolation of an RNA-directed RNA polymerase-specific cDNA clone from tomato. Plant Cell 10, 2087–2102. 26. Bernstein E, Caudy AA, Hammond SM, Hannon GJ. Role for a bidentate ribonuclease in the initiation step of RNA interference. Nature. 2001 Jan 18;409(6818):363-6. 27. Hutvagner G, McLachlan J, Pasquinelli AE, Balint E, Tuschl T, Zamore PD. A cellular function for the RNA-interference enzyme Dicer in the maturation of the let-7 small temporal RNA. Science. 2001 Aug 3;293(5531):834-8. 28. Hammond SM, Boettcher S, Caudy AA, Kobayashi R, Hannon GJ. Argonaute2, a link between genetic and biochemical analyses of RNAi. Science. 2001 Aug 10;293(5532):114650. 50 29. Williams RW, Rubin GM. ARGONAUTE1 is required for efficient RNA interference in Drosophila embryos. Proc Natl Acad Sci U S A. 2002 May 14;99(10):6889-94. 30. Wang DY, Kumar S, Hedges SB. Divergence time estimates for the early history of animal phyla and the origin of plants, animals and fungi. Proc R Soc Lond B Biol Sci. 1999 Jan 22;266(1415):163-71. 31. Sijen T, Fleenor J, Simmer F, Thijssen KL, Parrish S, Timmons L, Plasterk RH, Fire A. On the role of RNA amplification in dsRNA-triggered gene silencing. Cell. 2001 Nov 16;107(4):465-76. 32. Simmer F, Tijsterman M, Parrish S, Koushika SP, Nonet ML, Fire A, Ahringer J, Plasterk RH. Loss of the putative RNA-directed RNA polymerase RRF-3 makes C. elegans hypersensitive to RNAi. Curr Biol. 2002 Aug 6;12(15):1317-9. 33. Li H, Li WX, Ding SW. Induction and suppression of RNA silencing by an animal virus. Science. 2002 May 17 296(5571):1319-21. 34. Tuschl T, Zamore PD, Lehmann R, Bartel DP, Sharp PA. Targeted mRNA degradation by double-stranded RNA in vitro.Genes Dev. 1999 Dec 15;13(24):3191-7. 35. Zamore PD, Tuschl T, Sharp PA, Bartel DP. RNAi: double-stranded RNA directs the ATP-dependent cleavage of mRNA at 21 to 23 nucleotide intervals.Cell. 2000 Mar 31;101(1):25-33. 36. Hammond SM, Bernstein E, Beach D, Hannon GJ. An RNA-directed nuclease mediates post-transcriptional gene silencing in Drosophila cells. Nature. 2000 Mar 16;404(6775):2936. 37. Wargelius A, Ellingsen S, Fjose A. Double-stranded RNA induces specific developmental defects in zebrafish embryos. Biochem Biophys Res Commun. 1999 Sep 16;263(1):156-61. 38. Carthew RW. Gene silencing by double-stranded RNA. Curr Opin Cell Biol. 2001 Apr;13(2):244-8. Review. 39. Editorial. Targeting zebrafish.Nat Genet. 2000 Oct;26(2):129-30. 40. Brede G, Solheim J, Troen G, Prydz H. Characterization of PSKH1, a novel human protein serine kinase with centrosomal, golgi, and nuclear localization. Genomics. 2000 Nov 15;70(1):82-92. 41. ESF Workshop, St. Goar, april 2002, http://www.functionalgenomics.org.uk/sections/ activitites/Reports/report_germany_2002.htm. 42. Ambros V. microRNAs: tiny regulators with great potential. Cell. 2001 Dec 28;107(7):823-6. Review. 43. Kennerdell JR, Yamaguchi S, Carthew RW. RNAi is activated during Drosophila oocyte maturation in a manner dependent on aubergine and spindle-E. Genes Dev. 2002 Aug 1;16(15):1884-9. 51 44. Nykänen A, Haley B, Zamore PD. ATP requirements and small interfering RNA structure in the RNA interference pathway. Cell. 2001 Nov 2;107(3):309-21. 45. Martinez J, Patkaniowska A, Urlaub H, Luhrmann R, Tuschl T. Single-stranded antisense siRNAs guide target RNA cleavage in RNAi. Cell. 2002 Sep 6;110(5):563-74. 46. Lipardi C, Wei Q, Paterson BM. RNAi as random degradative PCR: siRNA primers convert mRNA into dsRNAs that are degraded to generate new siRNAs. 47. Elbashir SM, Lendeckel W, Tuschl T. RNA interference is mediated by 21- and 22nucleotide RNAs. Genes Dev. 2001 Jan 15;15(2):188-200. 48. Schwarz DS, Hutvagner G, Haley B, Zamore PD. Evidence that siRNAs function as guides, not primers, in the Drosophila and human RNAi pathways. Mol Cell. 2002 Sep;10(3):537-48. 49. Zilberman D, Cao X, Jacobsen SE. ARGONAUTE4 control of locus-specific siRNA accumulation and DNA and histone methylation. Science. 2003 Jan 31;299(5607):716-9. 50. Stein CA. Antisense that comes naturally. Nat Biotechnol. 2001 Aug;19(8):737-8. 51. McCaffrey AP, Meuse L, Pham TT, Conklin DS, Hannon GJ, Kay MA. RNA interference in adult mice. Nature. 2002 Jul 4;418(6893):38-9. 52. Tijsterman M, Ketting RF, Okihara KL, Sijen T, Plasterk RH. RNA helicase MUT-14dependent gene silencing triggered in C. elegans by short antisense RNAs. Science. 2002 Jan 25;295(5555):694-7. 53. Parrish S, Fire A. Distinct roles for RDE-1 and RDE-4 during RNA interference in Caenorhabditis elegans. RNA. 2001 Oct;7(10):1397-402. 54. T. Tuschl and A. Borkhardt. Small Interfering RNAs: A Revolutionary Tool for the Analysis of Gene Function and Gene Therapy. Mol. Interv., June 1, 2002; 2(3): 158 - 167. 55. Kisielow M, Kleiner S, Nagasawa M, Faisal A, Nagamine Y. Isoform-specific knockdown and expression of adaptor protein ShcA using small interfering RNA. Biochem J. 2002 Apr 1;363(Pt 1):1-5. 56. Ambros V. A hierarchy of regulatory genes controls a larva-to-adult developmental switch in C. elegans. Cell. 1989 Apr 7;57(1):49-57. 57. Lee RC, Feinbaum RL, Ambros V. The C. elegans heterochronic gene lin-4 encodes small RNAs with antisense complementarity to lin-14. Cell. 1993 Dec 3;75(5):843-54. 58. Reinhart BJ, Slack FJ, Basson M, Pasquinelli AE, Bettinger JC, Rougvie AE, Horvitz HR, Ruvkun G. The 21-nucleotide let-7 RNA regulates developmental timing in Caenorhabditis elegans. Nature. 2000 Feb 24;403(6772):901-6. 52 59. Pasquinelli AE, Reinhart BJ, Slack F, Martindale MQ, Kuroda MI, Maller B, Hayward DC, Ball EE, Degnan B, Muller P, Spring J, Srinivasan A, Fishman M, Finnerty J, Corbo J, Levine M, Leahy P, Davidson E, Ruvkun G. Conservation of the sequence and temporal expression of let-7 heterochronic regulatory RNA.Nature. 2000 Nov 2;408(6808):86-9. 60. Lee RC, Ambros V. An extensive class of small RNAs in Caenorhabditis elegans. Science. 2001 Oct 26;294(5543):862-4. 61. Lau NC, Lim LP, Weinstein EG, Bartel DP. An abundant class of tiny RNAs with probable regulatory roles in Caenorhabditis elegans. Science. 2001 Oct 26;294(5543):858-62. 62. Lagos-Quintana M, Rauhut R, Lendeckel W, Tuschl T. Identification of novel genes coding for small expressed RNAs. Science. 2001 Oct 26;294(5543):853-8. 63. Szittya G, Silhavy D, Molnar A, Havelda Z, Lovas A, Lakatos L, Banfalvi Z, Burgyan J. Low temperature inhibits RNA silencing-mediated defence by the control of siRNA generation. EMBO J. 2003 Feb 3;22(3):633-40. 64. Hamilton A, Voinnet O, Chappell L, Baulcombe D. Two classes of short interfering RNA in RNA silencing. EMBO J. 2002 Sep 2;21(17):4671-9. 65. Tang G, Reinhart BJ, Bartel DP, Zamore PD. A biochemical framework for RNA silencing in plants. Genes Dev. 2003 Jan 1;17(1):49-63. 66. Hutvagner G, Zamore PD. A microRNA in a multiple-turnover RNAi enzyme complex. Science. 2002 Sep 20;297(5589):2056-60. 67. Parrish S, Fleenor J, Xu S, Mello C, Fire A. Functional anatomy of a dsRNA trigger: differential requirement for the two trigger strands in RNA interference. Mol Cell. 2000 Nov;6(5):1077-87. 68. Finnegan EJ, Margis R, Waterhouse PM. Posttranscriptional Gene Silencing Is Not Compromised in the Arabidopsis CARPEL FACTORY (DICER-LIKE1) Mutant, a Homolog of Dicer-1 from Drosophila. Curr Biol. 2003 Feb 4;13(3):236-40. 69. Elbashir SM, Martinez J, Patkaniowska A, Lendeckel W, Tuschl T. Functional anatomy of siRNAs for mediating efficient RNAi in Drosophila melanogaster embryo lysate. EMBO J. 2001 Dec 3;20(23):6877-88. 70. Grishok A, Tabara H, Mello CC. Genetic requirements for inheritance of RNAi in C. elegans.Science. 2000 Mar 31;287(5462):2494-7. 71. Ohkawa J, Taira K. Control of the functional activity of an antisense RNA by a tetracycline-responsive derivative of the human U6 snRNA promoter. Hum Gene Ther. 2000 Mar 1;11(4):577-85. 72. Williams BR. PKR; a sentinel kinase for cellular stress. Oncogene. 1999 Nov 1;18(45):6112-20. Review 53 73. Wianny F, Zernicka-Goetz M. Specific interference with gene function by doublestranded RNA in early mouse development. Nat Cell Biol. 2000 Feb;2(2):70-5. 74. Svoboda P, Stein P, Hayashi H, Schultz RM. Selective reduction of dormant maternal mRNAs in mouse oocytes by RNA interference. Development. 2000 Oct;127(19):4147-56. 75. Caplen NJ, Parrish S, Imani F, Fire A, Morgan RA. Specific inhibition of gene expression by small double-stranded RNAs in invertebrate and vertebrate systems. Proc Natl Acad Sci U S A. 2001 Aug 14;98(17):9742-7. 76. Jacque JM, Triques K, Stevenson M. Modulation of HIV-1 replication by RNA interference. Nature. 2002 Jul 25;418(6896):435-8. 77. Gitlin L, Karelsky S, Andino R. Short interfering RNA confers intracellular antiviral immunity in human cells. Nature. 2002 Jul 25;418(6896):430-4. 78. Kapadia SB, Brideau-Andersen A, Chisari FV. Interference of hepatitis C virus RNA replication by short interfering RNAs. Proc Natl Acad Sci U S A. 2003 Feb 3 79. Boutla A, Delidakis C, Livadaras I, Tsagris M, Tabler M. Short 5'-phosphorylated doublestranded RNAs induce RNA interference in Drosophila. Curr Biol. 2001 Nov 13;11(22):1776-80. 80. Balter M. Gene therapy on trial. Science. 2000 May 12;288(5468):951-7. 81. Brummelkamp TR, Bernards R, Agami R. A system for stable expression of short interfering RNAs in mammalian cells. Science. 2002 Apr 19;296(5567):550-3. 82. Miyagishi M, Taira K. U6 promoter-driven siRNAs with four uridine 3' overhangs efficiently suppress targeted gene expression in mammalian cells. Nat Biotechnol. 2002 May;20(5):497-500. 83. Paul CP, Good PD, Winer I, Engelke DR. Effective expression of small interfering RNA in human cells. Nat Biotechnol. 2002 May;20(5):505-8. 84. Paddison PJ, Caudy AA, Bernstein E, Hannon GJ, Conklin DS. Short hairpin RNAs (shRNAs) induce sequence-specific silencing in mammalian cells. Genes Dev. 2002 Apr 15;16(8):948-58. 85. Sui G, Soohoo C, Affar el B, Gay F, Shi Y, Forrester WC, Shi Y. A DNA vector-based RNAi technology to suppress gene expression in mammalian cells. Proc Natl Acad Sci U S A. 2002 Apr 16;99(8):5515-20. 86. Lee NS, Dohjima T, Bauer G, Li H, Li MJ, Ehsani A, Salvaterra P, Rossi J. Expression of small interfering RNAs targeted against HIV-1 rev transcripts in human cells. Nat Biotechnol. 2002 May;20(5):500-5. 87. Yu JY, DeRuiter SL, Turner DL. RNA interference by expression of short-interfering RNAs and hairpin RNAs in mammalian cells. Proc Natl Acad Sci U S A. 2002 Apr 54 30;99(9):6047-52. 88. Yang D, Buchholz F, Huang Z, Goga A, Chen CY, Brodsky FM, Bishop JM. Short RNA duplexes produced by hydrolysis with Escherichia coli RNase III mediate effective RNA interference in mammalian cells. Proc Natl Acad Sci U S A. 2002 Jul 23;99(15):9942-7. 89. Castanotto D, Li H, Rossi JJ. Functional siRNA expression from transfected PCR products. RNA. 2002 Nov;8(11):1454-60. 90. Pal-Bhadra M, Bhadra U, Birchler JA. Cosuppression of nonhomologous transgenes in Drosophila involves mutually related endogenous sequences. Cell. 1999 Oct 1;99(1):35-46. 91. Novina CD, Murray MF, Dykxhoorn DM, Beresford PJ, Riess J, Lee SK, Collman RG, Lieberman J, Shankar P, Sharp PA. siRNA-directed inhibition of HIV-1 infection. Nat Med. 2002 Jul;8(7):681-6. 92. Martinez MA, Gutierrez A, Armand-Ugon M, Blanco J, Parera M, Gomez J, Clotet B, Este JA. Suppression of chemokine receptor expression by RNA interference allows for inhibition of HIV-1 replication. AIDS. 2002 Dec 6;16(18):2385-90. 93. Hu WY, Myers CP, Kilzer JM, Pfaff SL, Bushman FD. Inhibition of retroviral pathogenesis by RNA interference. Curr Biol. 2002 Aug 6;12(15):1301-11. 94. The PHD project plan: http://www.nsd.uib.no/nfi/rapport/?keys=22140&language=no 95. Amarzguioui M, Prydz H. Hammerhead ribozyme design and application. Cell Mol Life Sci. 1998 Nov;54(11):1175-202. Review. 96. Usman N, Beigelman L, McSwiggen JA. Hammerhead ribozyme engineering. Curr Opin Struct Biol. 1996 Aug;6(4):527-33. Review. 97. Knight SW, Bass BL.A role for the RNase III enzyme DCR-1 in RNA interference and germ line development in Caenorhabditis elegans. Science. 2001 Sep 21;293(5538):2269-71. 98. Tabara H, Yigit E, Siomi H, Mello CC. The dsRNA binding protein RDE-4 interacts with RDE-1, DCR-1, and a DExH-box helicase to direct RNAi in C. elegans. Cell. 2002 Jun 28;109(7):861-71. 99. Stein CA. Keeping the biotechnology of antisense in context. Nat Biotechnol. 1999 Mar;17(3):209. 100. Branch AD. A good antisense molecule is hard to find. Trends Biochem Sci. 1998 Feb;23(2):45-50. Review. 101. Camerer E, Kolsto AB, Prydz H. Cell biology of tissue factor, the principal initiator of blood coagulation. Thromb Res. 1996 Jan 1;81(1):1-41. 102. Lewis DL, Hagstrom JE, Loomis AG, Wolff JA, Herweijer H. Efficient delivery of siRNA for inhibition of gene expression in postnatal mice. Nat Genet. 2002 Sep;32(1):107-8. 55 103. Amarzguioui M, Brede G, Babaie E, Grotli M, Sproat B, Prydz H. Secondary structure prediction and in vitro accessibility of mRNA as tools in the selection of target sites for ribozymes. Nucleic Acids Res. 2000 Nov 1;28(21):4113-24. 104. Zuker M, Mathews DH, Turner DH. Algorithms and Thermodynamics for RNA Secondary Structure Prediction: A Practical Guide. In Barciszewski, J. and Clark, B.F.C (eds.), RNA Biochemistry and Biotechnology. Kluwer Academic Publishers, Dordrecht, 1999, pp.11-43. 105. Harborth J, Elbashir SM, Bechert K, Tuschl T, Weber K. Identification of essential genes in cultured mammalian cells using small interfering RNAs. J Cell Sci. 2001 Dec;114(Pt 24):4557-65. 106. Krichevsky AM, Kosik KS. RNAi functions in cultured mammalian neurons. Proc Natl Acad Sci U S A. 2002 Sep 3;99(18):11926-9. 107. Scherr M, Battmer K, Winkler T, Heidenreich O, Ganser A, Eder M. Specific inhibition of bcr-abl gene expression by small interfering RNA. Blood. 2003 Feb 15;101(4):1566-9. 108. Martinez LA, Naguibneva I, Lehrmann H, Vervisch A, Tchenio T, Lozano G, HarelBellan A. Synthetic small inhibiting RNAs: efficient tools to inactivate oncogenic mutations and restore p53 pathways. Proc Natl Acad Sci U S A. 2002 Nov 12;99(23):14849-54. 109. Zhang H, Kolb FA, Brondani V, Billy E, Filipowicz W. Human Dicer preferentially cleaves dsRNAs at their termini without a requirement for ATP. EMBO J. 2002 Nov 1;21(21):5875-85. 110. Kawasaki H, Suyama E, Iyo M, Taira K. siRNAs generated by recombinant human Dicer induce specific and significant but target site-independent gene silencing in human cells. Nucleic Acids Res. 2003 Feb 1;31(3):981-7. 111. Ketting RF, Plasterk RH. A genetic link between co-suppression and RNA interference in C. elegans. Nature. 2000 Mar 16;404(6775):296-8. 112. Hamada M, Ohtsuka T, Kawaida R, Koizumi M, Morita K, Furukawa H, Imanishi T, Miyagishi M, Taira K. Effects on RNA interference in gene expression (RNAi) in cultured mammalian cells of mismatches and the introduction of chemical modifications at the 3'-ends of siRNAs. Antisense Nucleic Acid Drug Dev. 2002 Oct;12(5):301-9. 113. Stephens AC, Rivers RP. Suppression of human monocyte tissue factor synthesis by antisense oligodeoxynucleotide. Thromb Res. 1997 Mar 1;85(5):387-98. 114. Chiu YL, Rana TM. RNAi in human cells: basic structural and functional features of small interfering RNA. Mol Cell. 2002 Sep;10(3):549-61. 115. Wiiger MT, Pringle S, Pettersen KS, Narahara N, Prydz H. Effects of binding of ligand (FVIIa) to induced tissue factor in human endothelial cells. Thromb Res. 2000 May 56 15;98(4):311-21. 116. Hvatum M, Prydz H. Studies on tissue thromboplastin. Biochim. Biophys. Acta. 1966. 130: 92-101. 117. Zeng Y, Cullen BR. Sequence requirements for micro RNA processing and function in human cells. RNA. 2003 Jan;9(1):112-123. 118. Shi Y. Mammalian RNAi for the masses. Trends Genet. 2003 Jan;19(1):9-12. 119. Splinter PL, Masyuk AI, LaRusso NF. Specific inhibition of AQP1 water channels in isolated rat intrhepatic bile duct unist by small interfering RNAs. J Biol Chem, 2003 Feb 21; 278(8): 6268 -6274. 120. Klahre U, Crete P, Leuenberger SA, Iglesias VA, Meins F Jr. High molecular weight RNAs and small interfering RNAs induce systemic posttranscriptional gene silencing in plants. Proc Natl Acad Sci U S A. 2002 Sep 3;99(18):11981-6. 121. Garrus JE, von Schwedler UK, Pornillos OW, Morham SG, Zavitz KH, Wang HE, Wettstein DA, Stray KM, Cote M, Rich RL, Myszka DG, Sundquist WI. Tsg101 and the vacuolar protein sorting pathway are essential for HIV-1 budding. Cell. 2001 Oct 5;107(1):55-65. 122. Lassus P, Rodriguez J, Lazebnik Y. Confirming specificity of RNAi in mammalian cells. Sci STKE. 2002 Aug 27;2002(147):PL13. 123. Wilda M, Fuchs U, Wossmann W, Borkhardt A. Killing of leukemic cells with a BCR/ABL fusion gene by RNA interference (RNAi). Oncogene. 2002 Aug 22;21(37):571624. 124. Elbashir SM, Harborth J, Weber K, Tuschl T. Analysis of gene function in somatic mammalian cells using small interfering RNAs. Methods. 2002 Feb;26(2):199-213. 125. Mourelatos Z, Dostie J, Paushkin S, Sharma A, Charroux B, Abel L, Rappsilber J, Mann M, Dreyfuss G. miRNPs: a novel class of ribonucleoproteins containing numerous microRNAs. Genes Dev. 2002 Mar 15;16(6):720-8. 126. Caudy AA, Myers M, Hannon GJ, Hammond SM. Fragile X-related protein and VIG associate with the RNA interference machinery. Genes Dev. 2002 Oct 1;16(19):2491-6. 127. Ishizuka A, Siomi MC, Siomi H. A Drosophila fragile X protein interacts with components of RNAi and ribosomal proteins. Genes Dev. 2002 Oct 1;16(19):2497-508. 128. Hannon GJ. RNA interference. Nature. 2002 Jul 11;418(6894):244-51. Review. 129. Pal-Bhadra M, Bhadra U, Birchler JA. Cosuppression in Drosophila: gene silencing of Alcohol dehydrogenase by white-Adh transgenes is Polycomb dependent. Cell. 1997 Aug 8;90(3):479-90. 57 130. Pal-Bhadra M, Bhadra U, Birchler JA. RNAi related mechanisms affect both transcriptional and posttranscriptional transgene silencing in Drosophila. Mol Cell. 2002 Feb;9(2):315-27. 131. Check E. A tragic setback. Nature. 2002 Nov 14;420(6912):116-8. 132. Jansen B, Zangemeister-Wittke U. Antisense therapy for cancer--the time of truth. Lancet Oncol. 2002 Nov;3(11):672-83. Review. 133. Caplen NJ, Fleenor J, Fire A, Morgan RA. dsRNA-mediated gene silencing in cultured Drosophila cells: a tissue culture model for the analysis of RNA interference. Gene. 2000 Jul 11;252(1-2):95-105. 134. McManus MT, Haines BB, Dillon CP, Whitehurst CE, van Parijs L, Chen J, Sharp PA. Small interfering RNA-mediated gene silencing in T lymphocytes. J Immunol. 2002 Nov 15;169(10):5754-60. 135. Sharp PA. RNA interference--2001. Genes Dev. 2001 Mar 1;15(5):485-90. Review. 136. Hope IA. RNAi surges on: application to cultured mammalian cells. Trends Genet. 2001 Aug;17(8):440. 137. Volpe TA, Kidner C, Hall IM, Teng G, Grewal SI, Martienssen RA. Regulation of heterochromatic silencing and histone H3 lysine-9 methylation by RNAi. Science. 2002 Sep 13;297(5588):1833-7. 138. Gao M, Wilusz CJ, Peltz SW, Wilusz J. A novel mRNA-decapping activity in HeLa cytoplasmic extracts is regulated by AU-rich elements. EMBO J. 2001 Mar 1;20(5):1134-43. 139. Evdokimova V, Ruzanov P, Imataka H, Raught B, Svitkin Y, Ovchinnikov LP, Sonenberg N. The major mRNA-associated protein YB-1 is a potent 5' cap-dependent mRNA stabilizer. EMBO J. 2001 Oct 1;20(19):5491-502. 140. Carpousis AJ, Vanzo NF, Raynal LC. mRNA degradation. A tale of poly(A) and multiprotein machines. Trends Genet. 1999 Jan;15(1):24-8. Review. 141. Dernburg AF, Karpen GH. A chromosome RNAissance. Cell. 2002 Oct 18;111(2):15962. 142. Wang MB, Waterhouse PM. Application of gene silencing in plants. Curr Opin Plant Biol. 2002 Apr;5(2):146-50. Review. 58