OLIGONUCLEOTIDES 17:35–43 (2007)© Mary Ann Liebert, Inc.DOI: 10.1089/oli.2006.0067

Tolerance of RNA Interference Toward Modifications of the 5� Antisense Phosphate of Small Interfering RNA

SAMIT SHAH and SIMON H. FRIEDMAN

ABSTRACT

Bringing RNA interference (RNAi) under the control of light will allow the spacing, timing, and de-gree of gene expression to be controlled. We have previously shown that RNAi by small interfering(si) RNA can be modulated through randomly incorporated photolabile groups. Our and others in-terest is to find key locations on siRNA that can completely block RNAi until irradiation releasescompletely active siRNA. Some literature suggests that the 5� phosphate of the antisense strand ofsiRNA cannot be modified without completely blocking RNAi. We have examined this site as a po-tential switch for light control of RNAi and present evidence that siRNA modified at the 5� antisensephosphate can still cause RNAi, although not at the level effected by fully native siRNA. This con-trasts with results from the literature, which suggest that modification of the 5� antisense phosphatewill completely abrogate RNAi in siRNA. We have used mass spectrometry to identify and quanti-tate possible impurities that may be responsible for residual RNAi and show that they are present at1% or less. Our results suggest that there is an inherent tolerance of the RNAi machinery towardmodification of the 5� antisense phosphate.

INTRODUCTION

SMALL INTERFERING RNAs (siRNA) are proving to beuseful tools for exploring the biological consequences

of a specific gene’s expression (Elbashir et al., 2001a,b;Tuschl, 2001; Plasterk, 2002; Dykxhoorn et al., 2003). Inthe interest of expanding the utility of siRNA, multiplegroups have examined a range of nucleic acid modifica-tions, the aim being to impart new properties, such asgreater nuclease stability and enhanced cellular uptake(Amarzguioui et al., 2003; Czauderna et al., 2003; Harborthet al., 2003; Elmen et al., 2005; Li et al., 2005; Choung etal., 2006; Dande et al., 2006). Our group has explored theincorporation of light-sensitive modifications into siRNAto impart the ability to toggle RNA interference (RNAi)(Shah et al., 2005). The ultimate aim of light-activatedRNA interference (LARI) is to control the spacing, timing,and degree of gene expression by controlling the spacing,

timing, and amount of light applied to cells containingmodified siRNA. In our recent publication, we showed thatrandom modification of siRNA with the di-methoxy-nitro-phenyl-ethyl (DMNPE) group partially blocks RNAi untilirradiation releases the photolabile group and RNAi is in-creased. In addition, we showed that increasing the numberof photolabile groups on the duplex increased the blockageof RNAi, but simultaneously lead to a greater challenge inreleasing fully active siRNA. Our interest is in finding keypoints of modification that completely abrogate RNAi untilirradiation releases fully native siRNA. We believe that thiswill provide the tool that is most adept at controlling RNAiwith light.

In addition to this use of random modification, we havealso investigated the specific modification of the 5� phos-phate of the antisense strand of an siRNA duplex. Thiswas based on literature precedent suggesting that block-ing this phosphate would completely abrogate RNA in-

Division of Pharmaceutical Sciences, School of Pharmacy, University of Missouri, Kansas City, MO 64110.

35

terference (Chiu and Rana, 2002). Initially, this appearsto be an ideal strategy, because it could potentially allowthe complete blocking of RNAi. Recently, others haveindicated that this approach is effective (Nguyen et al.,2006). In fact, the literature addressing this question isvaried, with some showing that modification of the 5�phosphate completely abrogates RNAi, (Chiu and Rana,2002; Czauderna et al., 2003; Nguyen et al., 2006) andsome literature indicating that modification of the 5�phosphate is tolerated (Schwarz et al., 2002; Harborth etal., 2003; Jakymiw et al., 2005). Because of the impor-tance of this issue for our work, and potentially others,we have attempted to address this literature discrepancy.What we have observed, and what we report here, is thatthe modification of the 5� phosphate with a variety ofgroups only partially blocks RNAi. We have confirmedthis by examining both photolabile and non-photolabile5� antisense phosphate modifications on siRNAs that tar-get two different genes. In addition, we have confirmedthat this partial blocking of RNAi exists at different con-centrations of siRNA, and is therefore not an artifact ofconcentration.

Other workers in the field have suggested that smallquantities of 5� unmodified impurities may be responsi-ble for the residual RNAi that can be observed with 5�modified samples (Nguyen et al., 2006). Using a quanti-tative mass spectrometry method, we have confirmedthat the modified antisense strands used have only small(�1%) amounts of deprotected side products and there-fore can not be responsible for the residual RNAi ob-served. Thus, it appears that the 5� antisense phosphate ofsiRNA can tolerate modification and still maintain someRNA, albeit at a lower level than in native siRNA. Thisinformation may prove useful for the incorporation of 5�phosphate modifications to modulate important siRNAproperties, such as transport, bioavailability, and conju-gation of reporter molecules.

MATERIALS AND METHODS

siRNA preparation

Single-stranded RNA oligonucleotides of the desiredsequence were chemically synthesized by IDT (Coralville,

SHAH AND FRIEDMAN36

FIG. 1. (A–H) The different siRNA duplexes used in this study. The bottom line in each pair represents the antisense strand ofthe duplex. Also indicated are the specific sequences used to target GFP and RFP.

IA) or Dharmacon (Lafayette, CO). The 5�-modifiedoligonucleotides were obtained by incorporating the de-sired phosphoramidite at the time of the synthesis. Theinternal linker containing siRNA was designed to incor-porate the photocleavable spacer between the 5� end ofthe antisense strand and the 3� end of the sense strand.The oligonucleotides were annealed by mixing equalamounts of the oligonucleotides in tris-Acetate-EDTAbuffer (12.3 mM Tris, 12.3 mM acetate, 0.3 mM EDTApH 8.3 at room temperature), heating at 80°C for 15 min-utes and then allowing them to cool down to room tem-perature over a period of 2–3 hours. All siRNAs werestored at 24.7 �M final concentration in Tris-acetate-EDTA buffer at �20°C.

Culture and transfection of cells

The human cervical carcinoma-derived HeLa cell line(CCL-2) was procured from the American Type CultureCollection (ATCC, Rockville, MD). HeLa cells were cul-tured at 37°C in humidified atmosphere of 5% CO2/95%air in Dulbecco’s-modified Eagle’s medium (DMEM; In-vitrogen) supplemented with 10% fetal bovine serum(FBS), 100 units/mL penicillin, and 100 �g/mL strepto-mycin (Invitrogen). Cells were regularly passaged at sub-confluence and plated at 70% confluency in 96-wellplates (Corning) 18–20 hours prior to transfection. Fortransfection, the medium in each well was replaced witha mixture containing 0.132 �g of pEGFP-C1 plasmid(Clontech), 0.132 �g of pDsRed2-N1(Clontech), 1.62pmol (13.52 nM) siRNA (unless otherwise indicated),and 1.125 �L lipofectamine (Invitrogen) in 120 �L ofserum-reduced OPTI-MEM (Invitrogen). Cells were in-cubated in the transfection mixture for 6–8 hours, and thetransfection mixture was removed and cells were furthercultured in 200 �L of antibiotic free DMEM supple-mented with 10% FBS. For experiments where wellswere exposed to ultraviolet (UV) light, the transfectionmixture in all wells was replaced with 100 �L of serum-reduced OPTI-MEM. The UV-treated cells were exposedto UV light from a UV lamp (Blak-Ray Lamp, ModelXX-15L, 115V, 60Hz, 0.68 Amps, UVP, Upland, CA) ata distance of 10 cm from the lamp for the specified time.Cells were protected from short-length UV by a WG-320longpass filter (Edmunds Industrial Optics). The serum-reduced OPTI-MEM was removed after UV exposureand cells were then cultured in 200 �L of antibiotic-freeDMEM supplemented with 10% FBS. The cells were in-cubated for 42 hours, and 100 �L of phosphate-bufferedsaline (PBS; Invitrogen) was then added to each well af-ter washing the cells with 200 �L of PBS. Green fluores-cent protein (GFP) and red fluorescent protein (RFP) ex-pression were quantitated by fluorescence spectroscopyin a microplate reader (GFP excitation 485 nm, emission535 nm; RFP excitation 546 nm, emission 595 nm). Nor-

malized signals were generated by taking the GFP/RFPratio for an experimental point, and then normalizing it tothe same ratio for unirradiated, plasmid-only transfectedcells. All fluorescent signals were corrected for the samesignal in cells that were treated only with the transfectionagent lipofectamine. The siRNAs used do not show signsof cell toxicity. This is based on the observation that theuntargeted gene expression (e.g., RFP in the case of GFPtargeting) remains the same, in absolute terms, in siRNA-treated cells and plasmid-only cells.

Electrospray ionization mass spectrometry

Electrospray mass spectrometry was guided by thework of Potier et al. (1994). siRNA (50 �L, 24.7 �M)was treated with 25 �L of 10 M ammonium acetate for10 minutes at room temperature to enhance the displace-ment of sodium ions. The siRNA were then precipitatedby adding 3 volumes of absolute ethanol and 2.5 �L (20mg/mL) of glycogen and cooling to �80°C for 2–3hours. Centrifugation was carried out for 30 minutes at17,400 � g and 0°C to obtain the siRNA pellets. Pelletscontaining varying amounts of a mock impurity oligonu-cleotide in the presence of a set quantity of a standard oli-gonucleotide were obtained in an identical manner. Thepellets were then dissolved in water:acetonitrile (50:50,50 �L) containing 1% triethylamine to make a final con-centration of 24.7 pmol/�L. The siRNA was then intro-duced into the ion source at a flow rate of 7 �L/min. Theelectrospray source was operated in the negative ion

MODIFICATIONS OF siRNA 5� ANTISENSE PHOSPHATE 37

FIG. 2. Reduction in normalized GFP expression with vary-ing concentrations of siRNA. Solid circles represent results us-ing GFP targeting duplex C (siRNA with an N6 modified 5� an-tisense phosphate.) Open circles represent unmodified GFPtargeting siRNA (duplex B). Five experimental points are aver-aged for each value and the standard error indicated.

RFP plasmids allows us to normalize GFP expression toRFP expression when using GFP targeting siRNAs and inaddition to normalize RFP expression to GFP expressionwhile using RFP targeting siRNAs.

To confirm the ability of modification of the 5� anti-sense phosphate to block RNAi completely, we tested aGFP targeting siRNA both with and without a nonphoto-cleavable N6 group modifying its 5� antisense phosphategroup (Fig. 1, duplex C and duplex B, respectively). Al-though we found that the presence of the modification re-duced RNAi, it did not completely abrogate it (Fig. 2).We considered that this difference in effect compared toother authors may be due to variation in siRNA concen-trations used. However, even at different siRNA concen-trations, the difference in GFP expression between modi-fied and nonmodified phosphate remained approximatelyconstant, although the absolute amount of RNA interfer-ence increased with increasing siRNA concentration.

Because the issue of siRNA purity has been suggestedto be key, we further investigated the possible presence

mode, with a needle voltage of �4.2 kV and the sourcetemperature at 150°C. The data were obtained for a m/zof 500–1200 amu and then deconvoluted using bayesianprotein reconstruct software (Bioanalyst) to find allpeaks from 4000 to 9000 amu. After identifying the mostintense point (amu) for each peak, the integration foreach peak was calculated using the trapezoidal method.The intensities of points within five mass units on eitherside of the principal point were integrated.

RESULTS

The analysis of modified siRNAs was performed using ahigh-throughput gene expression system we have previ-ously described (Shah et al., 2005). Briefly, we co-transfectsiRNAs with GFP- and RFP-expressing plasmids intoHeLa cells that are plated in 96-well plates. This allows usto acquire five replicates of each experimental condition us-ing a fluorescence multiplate reader. The use of GFP and

SHAH AND FRIEDMAN38

FIG. 3. Mass spectrum of mock impurity. Magnified mass spectrum of varying concentrations of a mock impurity (duplex B, an-tisense strand) in the presence of a set concentration of a standard (duplex C, antisense strand). The integration of this data wasused to form the curve in Fig. 4.

of an unmodifed antisense strand, formed either throughan N-1 deletion during synthesis (i.e., a 5� OH, Fig. 1, du-plex A), or through other hydrolytic processes to formthe 5� phosphate (duplex B), by using mass spectrometryto quantitate these potential impurities. To validate thismethod, we generated a standard curve of mass spec-trometer signal integration for varying amounts of amock impurity oligonucleotide in the presence of a setquantity of a standard oligonucleotide. This standard wasthe 5� N6 modified antisense strand for GFP targetingsiRNA (antisense strand of duplex C, calculated molecu-lar weight 6865, observed 6869). The mock impurity wasthe 5� phosphate version of this oligonucleotide, i.e.,missing the N6 group (antisense strand of duplex B, cal-culated molecular weight 6766, observed 6772). Themole percentage of this mock impurity varied between0.5% and 20%. The observed signal of the impurity wasdetermined by using trapezoidal integration of the massspectrometer signal, as normalized to the signal due tothe standard oligonucleotide. We integrate the signal in-cluding points within 5 mass units on either side of thepeak (Fig. 3). Typically, the observed mass is within 10mass units of the expected, and the observed differencein masses between two species within one mass unit.

The resultant measured mole percentages were plottedrelative to the actual mole percent, as determined by UVabsorbance of the individual strands before mixing (Fig.4). The observed near-unit slope and near-zero interceptindicated that this integrated signal is a good representa-tion of the quantity of impurity found in the presence of amain species, particularly at levels of 1% or greater. Us-ing this same approach, we analyzed the N6-modifiedsiRNA targeting GFP described above, examining it forsigns of possible impurities. These included n � 1 syn-

MODIFICATIONS OF siRNA 5� ANTISENSE PHOSPHATE 39

FIG. 4. Relationship between measured and actual concentra-tion of oligonucleotide impurity. Integrated mass spectrum sig-nals for varying concentrations of mock impurity were plottedversus their actual concentrations.

thesis errors (in the case of 5� N6-modified oligonu-cleotides, a simple 5� OH), and possible 5� phosphate-modified antisense strand that could form from hydroly-sis of the 5� N6-modified strand. These species arecalculated to have masses of 6690 and 6772, respec-tively, and were observed to have masses of 6686 and6766. Their observed quantities, taken as a mole percent-age to the normal 5� N6 strand were 0.1% and 1.1%, re-spectively. This further shows that the samples tested hadlittle or no unmodified 5� N6 antisense strand, and furthersupported the result that this 5� antisense modification isnot sufficient to block RNAi completely.

To confirm further the inability of 5� phosphate modi-fication to block RNAi substantially, we examined theeffect of the same N6 modification on a differentsiRNA/target pair, namely an siRNA that targets RFP ex-pression. The results are essentially identical to thosewith GFP (Fig. 5). The unmodified RFP targeting siRNAreduces RFP expression and the 5� antisense phosphatemodification only partially recovers this. As with GFP,this difference in expression with and without 5� modifi-cation varied little with concentration.

In addition, we examined the ability of siRNA with anN3 phosphate modification (duplex D) to affect RNAi. Incontrast with the previously reported result (Chiu andRana, 2002), and in agreement with the results describedabove, this modification also was unable to block RNAicompletely, independent of concentration (Fig. 6).

As our original purpose in this work was to allow thecomplete toggling of RNAi with light, we also examined

FIG. 5. Reduction in normalized RFP expression with vary-ing concentrations of siRNA. Solid circles represent results us-ing RFP targeting duplex C (siRNA with an N6 modified 5� an-tisense phosphate.) Open circles represent unmodified RFPtargeting siRNA (duplex B).

the photocleavable versions of the 5� phosphate modifi-cations described above. This is the photocleavable bi-otin derivative, duplex E, shown in Fig. 1 (Olejnik et al.,1996, 1998). This modification also blocks the 5� phos-phate, but leaves behind a free phosphate group uponphoto-deprotection. In this experiment, we used the GFPtargeting siRNA and examined GFP expression normal-ized to RFP expression. In addition, we examined the 5�photocleavable biotin on both the antisense and the sensestrands. The former showed similar results to those foundwith the non-photocleavable versions, namely that RNAiwas only partially blocked prior to irradiation, and thenachieved near-native siRNA function upon irradiation(Fig. 7). Sense strand modifications showed no blockingof RNAi effect. We also tested the effect of avidin in thepresence of the biotinylated modification, and it showedno improved blocking of RNAi relative to biotin alone(data not shown). We analyzed these photoprotected siRNAs for deprotection using the quantitative massspectrometry method described above. We observed10.7% deprotection by quantitating the 5� phosphate anti-sense strand that forms upon irradiation.

In addition, we also tested the photocleavable aminoderivative (Fig. 1, duplex F), which also modifies the 5�antisense phosphate until photo-deprotection releases thenative siRNA. As with the biotinylated version, thisshowed the inability to block fully RNAi prior to irradia-tion (Fig. 8).

Finally, we also examined a system where the 5� anti-sense phosphate modification is linked to the 3� OH ofthe sense strand using a photocleavable linker (Fig. 1, du-plex G). The rationale behind this was to provide themaximum modification of the native structure prior to ir-radiation to block RNAi completely. We observe that al-though this modification is better able to block RNAi,nontoxic levels of irradiation are unable to effect theRNAi of native siRNA (Fig. 9).

DISCUSSION

The aim of our work is to bring RNAi under the controlof light. Our recent publication describes one approach tothis, using random modification of siRNA with photola-bile groups (Shah et al., 2005). That work was preceded inour laboratory by work examining the blocking of the 5�antisense phosphate of an siRNA with photolabile moi-eties. The rationale behind this approach is that an unmod-ified 5� antisense phosphate is, in theory, required for ansiRNA to be used by the RNA-induced silencing complex(RISC) during RNAi. This has been suggested by a varietyof authors (Chiu and Rana, 2002; Czauderna et al., 2003;Nguyen et al., 2006), but disproven by others (Schwarz etal., 2002; Harborth et al., 2003; Jakymiw et al., 2005). Thetemporary blocking of this phosphate by a photolabilegroup could allow complete blocking of RNAi until pho-tolysis releases fully native and active siRNA.

SHAH AND FRIEDMAN40

FIG. 6. Reduction in normalized GFP expression with vary-ing concentrations of siRNA. Solid circles represent results us-ing GFP targeting duplex D (siRNA with an N3 modified 5� an-tisense phosphate). Open circles represent unmodified GFPtargeting siRNA (duplex B). Five experimental points are aver-aged for each value and the standard error indicated.

FIG. 7. Effect of 5� phosphate modification by photo-cleav-able (PC) biotin group on GFP expression. HeLa cells weretransfected with GFP-expressing plasmids and GFP-targetingsiRNAs, modified with photocleavable 5� phosphate protection.After transfection, cells were irradiated or not, and then allowedto culture for 40 hours. Dark bars represent normalized GFPsignal in cells that were not irradiated. Light bars represent nor-malized GFP signal in cells irradiated for 10 minutes.

The work that we describe in this paper shows that thisblocking of the 5� phosphate is not sufficient to blockRNAi completely. During the preparation of this manu-script, another group, McMaster and co-workers, de-scribed in the literature the modification of the 5� anti-sense phosphate with the identical photolabile moietyand suggested its use in controlling RNAi using light(Nguyen et al., 2006). Our results suggest that thismethod, while allowing modulation of RNAi, does notpermit complete switching from 100% gene expressionto full RNAi. McMaster and co-workers also noticed thisincomplete block of RNAi by 5� antisense phosphatemodification and ascribe it to impure oligonucleotidesthat are missing this terminal modification. These impu-rities can form either during synthesis or due to inadver-tent photo-deprotection of the oligionucleotides prior tothe experiment. Although we think that these possibilitiesare important and ultimately affect the ability of the 5�antisense phosphate-modified oligionucleotides to blockRNAi completely, we think there is, in addition, an inher-ent inability of a modified 5� phosphate to block RNAicompletely.

The strongest evidence that we have of this inherent in-ability of 5� antisense phosphate modification to blockRNAi completely is seen in Fig. 2. The siRNA used inthis experiment has its 5� antisense phosphate modified

with a non-photocleavable group. This eliminates thepossibility of early photo deprotection. Furthermore, thisis an oligonucleotide that has 1% or less of expected im-purities as determined by quantitative mass spectrometry.McMaster and co-workers suggest that a 1% impurity ofnative siRNA can reduce gene expression by 5–10%, butwe have observed a much greater reduction that variesbetween 30 and 60% with the concentration of thesiRNA. This level is consistent with the results that wehave for the photocleavable biotin shown in Fig. 7. Thissample, as described in the Results section, had about10% deprotection at the time of analysis. This is a resultthat is not inconsistent with the results of McMaster andco-workers, who suggested that impurities such as thesemask the true ability of 5� phosphate modification toblock RNAi. In light of our results showing the inabilityof a noncleavable 5� antisense phosphate modification toblock RNAi completely, we believe that there is an inher-ent tolerance to this modification by the RNAi machinerythat prevents a complete block of activity. The results wehave shown for the photocleavable 5� antisense phos-phate modification are actually representative of numer-ous repetitions of this experiment, using oligonucleotidesprovided from two different suppliers, and using increas-

MODIFICATIONS OF siRNA 5� ANTISENSE PHOSPHATE 41

FIG. 8. Effect of 5� phosphate modification by photo-cleav-able (PC) amine group on GFP expression. HeLa cells weretransfected with GFP-expressing plasmids and GFP-targetingsiRNAs, modified with photocleavable 5� phosphate protection.After transfection, cells were irradiated or not, and then allowedto culture for 40 hours. Dark bars represent normalized GFPsignal in cells that were not irradiated. Light bars represent nor-malized GFP signal in cells irradiated for 10 minutes.

FIG. 9. Effect of linking 5� phosphate of anti-sense strandwith 3� OH of sense strand with a photo-labile linker. HeLacells were transfected with GFP-expressing plasmids and GFP-targeting siRNAs, modified with photocleavable 5� phosphateprotection. After transfection, cells were irradiated or not, andthen allowed to culture for 40 hours. Dark bars represent nor-malized GFP signal in cells that were not irradiated. Light barsrepresent normalized GFP signal in cells irradiated for 10 minutes.

ing levels of care in experiments to exclude light. Wehave not had a substantially different result from thatshown in Fig. 6, indicating incomplete block of RNAithrough the modification of the 5� antisense phosphate.

We have further support of this from our previouslypublished results, in which we modify siRNA with theclosely related DMNPE group (Shah et al., 2005). In thisprevious study, we showed that we had approximately1.4 DMNPE groups incorporated on average per siRNA.Upon examination of the mass spectrum of the modifiedsiRNA used in this previous study, we see that primarilythe antisense strand is modified. We see little signs of un-modified antisense siRNA. We assume this is the casebecause in this study, only the antisense strand of thesiRNA had a terminal phosphate group. We have subse-quently shown that terminal phosphates are much easierto protect with DMNPE than internal sites and that if the5� terminal antisense phosphate is not present to reactwith DMNPE, blockage of RNAi is severely limited(manuscript in preparation). With a single blockinggroup incorporated freshly into siRNA (thereby creatingduplex H) we get a similar incomplete blocking of RNAito what we observe here with commercially availablephotocleavable and noncleavable 5� phosphate modifica-tion.

It is possible that modifications on the 5� phosphate aretolerated because, although they block one of the chargeson the phosphate, they leave another oxygen unmodified,which may be able to continue to bind to RISC in thesame fashion as unmodified siRNA. In addition, thechain that builds off of the phosphate oxygen may be ableto avoid steric clash with the RISC by adopting an appro-priate conformation. Modifications that more completelyblock this phosphate, through greater bulk, or by modify-ing both charged oxygens, may be more effective atblocking the siRNA’s activity.

Our results both contrast with those of other groups,who have shown that modification of the 5� antisensephosphate group completely eliminates RNAi (Chiu andRana, 2002; Czauderna et al., 2003; Nguyen et al., 2006)and confirm the results of those groups who show thatthis modification incompletely blocks RNAi (Schwarz etal., 2002; Harborth et al., 2003; Jakymiw et al., 2005).Our results are representative and typical of a larger num-ber of experiments that we have performed using our sys-tem and modified oligonucleotides from multiple sources(IDT, Dharmacon) on two different gene targets, GFPand RFP. In these experiments, we consistently show anincomplete blocking of RNAi by modification of the 5�antisense phosphate, even if that modification is non-cleavable by light. It is possible that our observed differ-ences are rooted in subtle differences in the setup of therespective experiments. We are actively exploring thesepossibilities in our program to refine light activated

RNAi into a simple and robust system for biologicalanalysis.

REFERENCES

AMARZGUIOUI, M., HOLEN, T., BABAIE, E., and PRYDZ,H. (2003). Tolerance for mutations and chemical modifica-tions in a siRNA. Nucleic Acids Res. 31, 589–595.

CHIU, Y.L., and RANA, T.M. (2002). RNAi in human cells:Basic structural and functional features of small interferingRNA. Mol. Cell 10, 549–561.

CHOUNG, S., KIM, Y.J., KIM, S., PARK, H.O., and CHOI,Y.C. (2006). Chemical modifications of siRNAs to improveserum stability without loss of efficacy. Biochem. Biophys.Res. Commun. 342, 919–927.

CZAUDERNA, F., FECHTNER, M., DAMES, S., AYGUN,H., KLIPPEL, A., PRONK, G.J., GIESE, K., and KAUF-MANN, J. (2003). Structural variations and stabilising modi-fications of synthetic siRNAs in mammalian cells. NucleicAcids Res. 31, 2705–2716.

DANDE, P., PRAKASH, T.P., SIOUFI, N., GAUS, H., JAR-RES, R., BERDEJA, A., SWAYZE, E.E., GRIFFEY, R.H.,and BHAT, B. (2006). Improving RNA interference in mam-malian cells by 4�-thio-modified small interfering RNA(siRNA): Effect on siRNA activity and nuclease stabilitywhen used in combination with 2�-O-alkyl modifications. J.Medic. Chem. 49, 1624–1634.

DYKXHOORN, D.M., NOVINA, C.D., and SHARP, P.A.(2003). Killing the messenger: short RNAs that silence geneexpression. Nature Rev. Mol. Cell Biol. 4, 457–467.

ELBASHIR, S.M., HARBORTH, J., LENDECKEL, W., YAL-CIN, A., WEBER, K., and TUSCHL, T. (2001a). Duplexesof 21-nucleotide RNAs mediate RNA interference in cul-tured mammalian cells. Nature 411, 494–498.

ELBASHIR, S.M., LENDECKEL, W., and TUSCHL, T.(2001b). RNA interference is mediated by 21- and 22-nucle-otide RNAs. Genes & Dev. 15, 188–200.

ELMEN, J., THORNBERG, H., LJUNBERG, K., FRIEDEN,M., WESTERGAARD, M., XU, Y.H., WAHREN, B.,LIANG, Z.C., URUM, H., KOCH, T., and WAHLESTEDT,C. (2005). Locked nucleic acid (LNA) mediated improve-ments in siRNA stability and functionality. Nucleic AcidsRes. 33, 439–447.

HARBORTH, J., ELBASHIR, S.M., VANDENBURGH, K.,MANNINGA, H., SCARINGE, S.A., WEBER, K., andTUSCHL, T. (2003). Sequence, chemical, and structuralvariation of small interfering RNAs and short hairpin RNAsand the effect on mammalian gene silencing. Antisense &Nucleic Acid Drug Dev. 13, 83–105.

JAKYMIW, A., LIAN, S., EYSTATHIOY, T., LI, S., SATOH,M., HAMEL, J.C., FRITZLER, M.J., and CHAN, E.K.L.(2005). Disruption of GW bodies impairs mammalian RNAinterference. Nature Cell Biol. 7, 1267–1274.

LI, Z.Y., MAO, H.B., KALLICK, D.A., and GORENSTEIN,D.G. (2005). The effects of thiophosphate substitutions onnative siRNA gene silencing. Biochem. Biophys. Res. Com-mun. 329, 1026–1030.

SHAH AND FRIEDMAN42

NGUYEN, Q.N., CHAVLI, R.V., MARQUES, J.T., CONRADII, P.G., WANG, D., HE, W., BELISLE, B.E., ZHANG, A.,PASTOR, L.M., WITNEY, F.R., MORRIS, M., HEITZ, F., DI-VITA, G., WILLIAMS, B.R., and MCMASTER, G.K. (2006).Light controllable siRNAs regulate gene suppression and phe-notypes in cells. Biochim. Biophys. Acta 1794, 394–404.

OLEJNIK, J., KRZYMANSKAOLEJNIK, E., and ROTH-SCHILD, K.J. (1996). Photocleavable biotin phospho-ramidite for 5�-end-labeling, affinity purification and phos-phorylation of synthetic oligonucleotides. Nucleic AcidsRes. 24, 361–366.

OLEJNIK, J., KRZYMANSKA-OLEJNIK, E., and ROTH-SCHILD, K.J. (1998). Photocleavable aminotag phospho-ramidites for 5�-termini DNA/RNA labeling. Nucleic AcidsRes. 26, 3572–3576.

PLASTERK, R.H. (2002). RNA silencing: the genome’s im-mune system. Science 296, 1263–1265.

POTIER, N., VANDORSSELAER, A., CORDIER, Y., ROCH,O., and BISCHOFF, R. (1994). Negative electrospray-ion-ization mass spectrometry of synthetic and chemically modi-fied oligonucleotides. Nucleic Acids Res. 22, 3895–3903.

SCHWARZ, D.S., HUTVAGNER, G., HALEY, B., and

ZAMORE, P.D. (2002). Evidence that siRNAs function asguides, not primers, in the drosophila and human RNAi path-ways. Mol. Cell 10, 537–548.

SHAH, S., RANGARAJAN, S., and FRIEDMAN, S.H. (2005).Light activated RNA interference. Angew Chem. Int. Edit.44, 1328–1332.

TUSCHL, T. (2001). RNA interference and small interferingRNAs. Chembiochem 2, 239–245.

Address reprint requests to:Dr. Simon H. Friedman

Division of Pharmaceutical SciencesSchool of Pharmacy

University of Missouri, Kansas City5005 Rockhill Road

Kansas City, MO 64110

E-mail: friedmans@umkc.edu

Received December 2, 2006; accepted December 14,2006.

MODIFICATIONS OF siRNA 5� ANTISENSE PHOSPHATE 43