siRNA and miRNA

Introduction:

siRNAs and miRNAs are recently discovered classes of short RNA molecules that have crucial roles eukaryotic cells. They are also used as novel tools for single and high-throughput gene knockdown. siRNAs were discovered in 1998 in Caenorhabditis elegans while investigating the knockdown effects exhibited by both the sense and antisense strand to the target mRNA (Fire et al. 1998). Andrew Fire and Craig Mello were awarded the 2006 Nobel Prize in Physiology and Medicine for this discovery. One of the first observations that led to this breakthrough was the complete loss of purple pigmentation from Petunia plants when investigators tried to overexpress the purple pigment gene. (Napoli et al. 1990, van der Krol et al. 1990) However, these studies were unable to provide a mechanistic explanation for this phenomenon. Later, Guo and Kemphaus (1995) found that both the sense and the antisense strand of a target mRNA is sufficient for knockdown in C.elegans and finally Fire et al. (1998) discovered that in both cases the most potent silencer of mRNA activity was some double stranded RNA (dsRNA) contamination.

dsRNAs are cleaved into 21-23nt short interfering RNAs (siRNAs) by an enzyme called Dicer and they get incorporated into a ribonucleoprotein complex called RISC, which catalyzes the degradation of the complementary target mRNA ( Zamore et al. 2000, Hamilton and Baulcombe 1999, Hammond et al. 2000). This process is called RNA interference (RNAi) and it is highly conserved among many eukaryotes including plants, nematodes, insects, and mammals. A great animation of the mechanism of RNAi can be found at: http://www.nature.com/focus/rnai/animations/animation/animation.htm

siRNAs provide a very efficient tool for knocking-down the function of any gene of interest. Hence, there is a great potential for developing novel therapies for many diseases. Preliminary studies using RNAi to treat herpes infections, macular degeneration, Alzheimer’s disease and diabetes have been quite promising (Hong et al. 2006, Palliser et al. 2006, Soutschek et al 2004). However, there are major hurdles to be overcome before RNAi based drugs can be clinically utilized. These include the problem of specific delivery and the possibility of knockdown of unintended targets.

In C.elegans the introduction of siRNAs is relatively straightforward. There are three main methods: 1-Microinjection into embryos 2- Feeding dsRNA expressing bacteria 3- Soaking in dsRNA containing media. The last two methods are relatively cheap and not as time consuming as microinjection. The efficacy of the last two methods depend on the propagation of the dsRNA molecule in between the worm cells. The transport of dsRNAs in C.elegans and plants is dependent on a putative transmembrane protein called SID1 which is lacking from mammals and Drosophila (Feinberg EH, Hunter CP 2003, Winston et al. 2002). High-throughput functional genomics approaches utilize feeding methods either in liquid culture or on agar plates. Currently, libraries of dsRNA expressing bacteria for nearly 18000 genes in the C.elegans genome are available (Ahringer library and Vidal library). In Drosophila, in addition to microinjection into the embryo, two other methods are standardly used. One of them involves “bombarding” the embryo with siRNA molecules and the other requires genetically engineering the fly to carry an inverted copy of the target gene.

dsRNA delivery in mammals pose even more difficulties. Unlike C.elegans, long stretches of dsRNA cannot be used in mammals due to the interferon response. The mammalian cells recognize the foreign long double stranded RNA as a virus and activate the interferon response which shuts down all transcription in the cell eventually leading to apoptosis. Hence developing efficient ways for siRNA delivery is crucial in implementing RNAi technology for novel therapies.

Another challenge in developing RNAi based therapies is the recently emerging concern about the unintended knockdown of untargeted genes. RNAi was originally considered to be highly specific , eliminating the function of a single target gene. However, recent studies have suggested that off target effects might be much more prevalent (Jackson et al. 2003, Lin et al. 2005, Birmingham et al. 2006). These off target effects pose a serious complication for effective implementation of RNAi based drugs in mammals since minimization of side effects is a crucial concern for any drug.

RNAi has also become an efficient method for high-throughput screens (Friedman and Perrimon 2004). The new RNAi technology has been used to conduct high-throughput screens in the model organism C.elegans which led to annotation of many previously unannotated genes as well as identification of genome wide patterns (Fraser et al. 2000, Kamath et al 2003). This system is currently being used to conduct multiple gene knockdowns in a high throughput fashion (Ahringer J, personal communication).

Until recently, it was believed that the fundamental process of gene regulation required only the activity of certain protein transcription factors. However, a newly-discovered class of small RNAs known as microRNAs (miRNAs) has added an additional layer to this concept of gene control, and their study is currently of great interest to molecular biologists, chemists, and computational biologists. microRNAs are ~22-nucleotide long, small RNAs that silence their target mRNAs by binding to miRNA target sites in the mRNA 3’ untranslated region (3’UTR) via base-pair complementarity and inducing inhibition of translation or occasionally mRNA degradation. miRNAs make up an estimated 1-2% of mammalian genomes and are predicted to control ~30% of genes.

In 1993, Victor Ambros’ group reported that mutations in a small RNA known as lin-4 resulted in increased translation of the protein LIN-14 in C. elegans. Surprisingly, the ~70bp hairpin precursor was processed into 21-24 nt dsRNA by Dicer. These shorter RNA products were complimentary to a number of sites in the lin-14 mRNA 3’UTR. In 2000, Gary Ruvkun cloned a second small RNA known as let-7 in C. elegans. Let-7 also has fly and human homologs and has subsequently been used as a standard component in many experiments on miRNA function. In a year, many new miRNAs were cloned: 20 miRNAs Drosophila, 30 in humans, and 60 in C. elegans.

miRNA genes are conserved across species and transcribed by RNA polymerase II (pol II). Some appear in introns, while others exist in intergenic regions or other parts of the genome. Some are clustered (in flies, >50% are this way) and are co-transcribed as one transcript. miRNA genes are often located far from the regulated gene. They are typically abundantly expressed (50K molecules/cell), often in a cell-type specific manner. The maturation of miRNA precursors closely resembles processing of siRNAs, including hairpin cleavage by Dicer into ~22nt dsRNAs, helicase unwinding of these dsRNA, and incorporation into RISC (RNAi-Induced Silencing Complex). These short RNAs then proceed to direct repression of translation of target genes, or in some cases, cleavage of the target mRNA. miRNAs appear to be important for developmental processes and regulation of transcription factors.

COMPUTATIONAL PROBLEMS IN THE FIELD:

siRNA Design:

Many different algorithms have recently been developed to computationally design efficient and specific siRNAs. Arguably, the two most important resources for siRNA design are Ambion© and siRNA resources by Chris Burge and colleagues at the Whitehead Institute (see links below). There are a few methods for obtaining a desired RNA. The two most often used ones are the direct chemical synthesis and design of siRNA hairpins that are encoded by siRNA expression vectors or siRNA expression cassettes. Current benchmark for commercially available siRNAs is 70% decrease in mRNA levels in HeLa cells 48 hours after the administration and 50-60%success rate in generating and purifying the desired siRNA.

Theoretically, any part of a gene can be targeted using siRNAs yet some properties of the designed siRNA increases the efficiency drastically. First of all, siRNAs act in the cytoplasm hence one needs to target mature mRNA and not introns. Furthermore, targeting 3'UTR is generally more effective than targeting 5'UTR. Current algorithms for siRNA design also take into account several chemical properties of the siRNA. The following are some of these properties:

• 21-23nt siRNA including a 1-2nt U(T) overhang are generally used hence one should find a region of the target that involves a region with two adenosine nucleotides.
• Slightly less than fifty percent GC content seems to be ideal for efficiency.
• Recently the strand with a less stable 5' end has been shown to become selectively associated into the RISC complex.

However, an open secondary structure does not guarantee that the other strand will never be incorporated into RISC. Specificity is also a major concern in siRNA design hence one should ideally blast both the sense and the antisense strand to the whole genome and make sure that there are no significant hits.

The specificity concern for siRNAs is relatively new since the original RNAi papers including Fire et al. (1998) seemed to suggest that siRNAs are extremely specific. However, Jackson et al. (2003) have shown that targeting different regions of the same mRNA resulted in different microarray expression profiles and concluded that siRNAs have genome-wide unintentional effects. All of the current siRNA design algorithms rely on Blast similarity to predict potential off targets. However, the following figure from Birmingham et al (2006) proves the inadequacy of this method:

Figure 2 from Birmingham et al. (2006)

New algorithms are needed to take these new findings into consideration in order to design highly specific and efficient siRNAs. In addition to the standard algorithms described here artificial neural network based (Heusken et al. 2005) algorithms have been developed and these have identified new free energy related parameters that are significantly correlated with siRNA silencing potential (Shabalina et al. 2006).

miRNAs

Computational miRNA Prediction

In the past, it had become fairly straightforward to predict genes, but how could this be done with miRNA genes? Chris Burge approached this problem by collecting a number of hairpin loops. Using the Genome Project and bioinformatics, miRNA genes could be predicted based on the miRNAs that had been cloned in other organisms.

The earlier approaches for miRNA prediction relied on experimental findings and involved 1.) a “BLAST” approach to find known miRNA orthologs in other organisms and 2.) a search for hairpin loops near a known miRNA under the assumption that many miRNAs occur in clusters.

A number of programs exist for systematic miRNA prediction. miRScan was developed in 2003 by Lim et al. for microRNA predictions in human and C. elegans. miRSeeker was developed by Lai et al. in 2003 and uses a similar algorithm. Both programs search for a large number of candidates in non-coding sequences. They then predict secondary structures based on energy and compare between C.elegans/C. briggsae and mouse/human to check for conserved regions. About 40,000 regions are identified in C. elegans that may form loops.

miRScan then scores these candidates based on a number of features:

For each individual feature, the following formula is used to score each miRNA:

s_i(x_i)= log2 [f_i(x_i)/g_i(x_i)]

where f_i(x_i) = frequency of feature in known miRNAs and g_i(x_i) = frequency in all 40K candidates

Certain features are enriched in known miRNAs. If the frequencies are not very different, they can probably not be considered.

The score of each hairpin is then calculated as S = summation s_i(x_i) from i=1,..,7

Each particular feature contributes the following to hairpin scores:

With these scores, one can determine a cutoff that will include most miRNAs. In the following figure, red indicates known miRNAs, and yellow indicates that a predicted miRNA is indeed expressed.

Differences between miRNAs and endogenous siRNAs

• microRNAs are encoded at their own loci, whereas endogenous siRNAs are generally derived from reverse transcription and incorporation of mRNA or viral RNA
• as a result, endogenous siRNAs tend to be less conserved
• miRNAs are derived from hairpin precursors, whereas endogenous siRNAs are derived from long, co-transcribed dsRNAs or long hairpins
• generally, 1 miRNA precursor : 1 miRNA; 1 siRNA precursor : many siRNAs
• miRNAs target the 3’UTR, whereas siRNAs target anywhere in the mRNA
• miRNAs have mismatches with their targets and direct translation repression, whereas siRNAs base pair perfectly and direct target mRNA cleavage

miRNA Target Prediction

Because miRNA targeting tends to be governed by certain rules, mRNA targets can be predicted computationally. TargetScan (Chris Burge) is one such program, which finds human mRNA targets in the following manner:

1. the miRNA is aligned with a 3’UTR

2. a 7-mer perfect match in the 5’ end of the miRNA is required (although the program will show 6-mer sites in targets)

3. the miRNA must extend with good pairing based on the secondary structure

4. the target region must be conserved between humans, mouse, and rat

In the end, the program narrowed candidates down to 451 mammalian miRNA:target interactions based on conservation.

In plants, target prediction uncovers a high % of transcription factors; these mRNAs are degraded upon a perfect match.

Often, there are multiple target sites in a 3’UTR. This is likely to ensure downregulation of gene expression even if there is an alternative polyadenylation event or a deletion somewhere in the UTR that includes one or more of the miRNA target sites. These multiple target sites appear at a certain distance from each other, and such a phenomenon is conserved in many organisms. This is very similar to transcription factor motif-finding, as many miRNAs can regulate the same mRNA.

A registry is available through the Sanger Institute that contains a database of all known miRNAs. (See link below)

miRNA Effect on mRNAs

Until only last year, it was believed that miRNAs only directed translation repression. It now appears that miRNAs may also control mRNA cleavage. It was shown that siRNAs have off-target effects in downregulating other genes. Many of these 3’UTR sequences that were targeted contain miRNA sites with seed regions similar to the siRNA. Thus, the siRNA is acting as an miRNA at off-target mRNAs.

Indeed, when miRNAs were introduced into HeLa cells, the mRNA profile shifted to that of a new cell type. For instance, when brain-specific miR-124 was added, the HeLa cells’ mRNA profile appeared to resemble that of brain cells. A similar phenomenon occurred when muscle-specific miR-1 was added.

Thus, miRNAs can also cleave mRNAs, although this appears to be weaker and is present in regulating tissue-specific gene expression.

Whether there is translation repression or mRNA cleavage of the target gene appears to depend on the strength of base-pairing outside the seed region. All 7 bp of the seed region at the 5’ end of the miRNA must base pair to its target with perfect complimentarity. If the remainder does not match well, there is translation repression; otherwise, the mRNA is cleaved with a strength that is approximately proportional to the strength of the match.

As a result of this new finding, siRNA use may be problematic in causing off-target effects since each miRNA has hundreds of targets.

The effect of miRNA on evolution has been considered. For instance, certain 3’UTRs may evolve to contain “anti-targets” of co-expressed miRNAs to avoid their genes being silenced when they are needed. In this way, genes that must be expressed can still co-exist in the same tissues where a particular miRNA is present by evolving to avoid developing a 7-mer seed sequence.

References:

1- Birmingham A, Anderson EM, Reynolds A, Ilsley-Tyree D, Leake D, Fedorov Y, Baskerville S, Maksimova E, Robinson K, Karpilow J, Marshall WS, Khvorova A. 2006. 3 ' UTR seed matches, but not overall identity, are associated with RNAi off-targets. NATURE METHODS 3 (3): 199-204 Abstract

2- Fraser AG, Kamath RS, Zipperlen P, Martinez-Campos M, Sohrmann M, Ahringer J. 2000. Functional genomic analysis of C. elegans chromosome I by systematic RNA interference. Nature. 408(6810):325-30. Abstract

3- Hong CS, Goins WF, Goss JR, Burton EA, Glorioso JC. 2006. Herpes simplex virus RNAi and neprilysin gene transfer vectors reduce accumulation of Alzheimer's disease-related amyloid-beta peptide in vivo. Gene Ther. 13(14):1068-1079. Abstract

4- Jackson AL, Bartz SR, Schelter J, Kobayashi SV, Burchard J, Mao M, Li B, Cavet G, Linsley PS. 2003. Expression profiling reveals off-target gene regulation by RNAi. Nature Biotechnology 21 (6): 635-637. Abstract

5- Kamath RS, Fraser AG, Dong Y, Poulin G, Durbin R, Gotta M, Kanapin A, Le Bot N, Moreno S, Sohrmann M, Welchman DP, Zipperlen P, Ahringer J. 2003. Systematic functional analysis of the Caenorhabditis elegans genome using RNAi. Nature. 421(6920):231-7. Abstract

6- Lin X, Ruan X, Anderson MG, McDowell JA, Kroeger PE, Fesik SW, Shen Y. 2005. siRNA-mediated off-target gene silencing triggered by a 7 nt complementation. Nucleic Acids Research 33 (14): 4527-4535. Abstract

7- Palliser, D., Chowdhury, D., Q-YWang, SJ. Lee, RT. Bronso, DM. Knip, J. Lieberman. 2006 An siRNA-based microbicide protects mice from lethal herpes simplex virus 2 infection. Nature 439, 89-94. Abstract

8- Soutschek J, Akinc A, Bramlage B, Charisse K, Constien R, Donoghue M, Elbashir S, Geick A, Hadwiger P, Harborth J, John M, Kesavan V, Lavine G, Pandey RK, Racie T, Rajeev KG, Rohl I, Toudjarska I, Wang G, Wuschko S, Bumcrot D, Koteliansky V, Limmer S, Manoharan M, Vornlocher HP. 2004. Therapeutic silencing of an endogenous gene by systemic administration of modified siRNAs. Nature. 432(7014):173-8. Abstract

9. Hamilton AJ, Baulcombe DC. 1999. A species of small antisense RNA in posttranscriptional gene silencing in plants. Science 286, 950. Abstract

10. Hammond SM, Bernstein E, Beach D, Hannon GJ. 2000. An RNA-directed nuclease mediates post-transcriptional gene silencing in Drosophila cells. Nature 404, 293. Abstract

11. Guo S, Kemphues KJ. 1995. par-1, a gene required for establishing polarity in C. elegans embryos, encodes a putative Ser/Thr kinase that is asymmetrically distributed. Cell 81, 611. Abstract

12. Zamore PD, Tuschl T, Sharp PA, Bartel DP. 2000. RNAi: double-stranded RNA directs the ATP-dependent cleavage of mRNA at 21 to 23 nucleotide intervals. Cell 101, 25 Abstract

13. Feinberg EH, Hunter CP. 2003. Transport of dsRNA into cells by the transmembrane protein SID-1, Science 301(5639), 1545-7. Abstract

14. van der Krol AR, Mur LA, Beld M, Mol JNM, Stuitji AR. 1990. Free in PMC Flavonoid genes in petunia: addition of a limited number of gene copies may lead to a suppression of gene expression. Plant Cell 2, 291. Abstract

15. Bartel BP, Chen CZ (2004) Micromanagers of gene expression: the potentially widespread influence of metazoan microRNAs. Nature Rev Genetics (5): 396-400.

16. Bartel DP (2004). MicroRNAs: genomics, biogenesis, mechanism, and function. Cell (116): 281-297.

17. Cummins JM, Velculescu VE (2006). Implications of microRNA-profiling for cancer diagnosis. Oncogene (46): 6220-6227.

18. Griffiths-Jones S (2006) miRBase: the microRNA sequence database. Methods Mol Bio (342): 129-138.

19. Zhang B, Pan X, Anderson TA (2006) MicroRNA: a new player in stem cells. J Cell Physiol (209): 266-269.

20. Winston WM, Molodowitch C, Hunter CP. 2002. Sytemic RNAi in C.elegans requires putative transmembrane protein SID-1, Science 295(5564), 2456-9. Abstract

Useful Software/Web-based applications and Other Resources

- STATS 115 Lecture

- Ambion RNAi

- Sanger miRBase

- MiRscan

- TargetScan 3.1

- siRNA Selection at WI