Analytical Biochemistry 390 (2009) 181–188

Contents lists available at ScienceDirect

Analytical Biochemistry

journal homepage: www.elsevier .com/locate /yabio

Reversed-phase ion-pair liquid chromatography analysis and purificationof small interfering RNA

Sean M. McCarthy *, Martin Gilar, John GeblerWaters Corporation, Milford, MA 01757, USA

a r t i c l e i n f o

Article history:Received 5 March 2009Available online 2 April 2009

This work is dedicated to Marianna Kele.

Keywords:OligonucleotideDuplexDouble-strandedRNADNAUltra-performance liquid chromatography(UPLC)HPLCMass spectrometry

0003-2697/$ - see front matter � 2009 Elsevier Inc. Adoi:10.1016/j.ab.2009.03.042

* Corresponding author.E-mail address: sean_m_mccarthy@waters.com (S

1 Abbreviations used: RNAi, ribonucleic acid interferensmall interfering RNA; mRNA, messenger RNA; RISC, RNssRNA, single-stranded RNA; PAGE, polyacrylamide gelgel electrophoresis; AX HPLC, anion-exchange high-peraphy; RP IP LC, reversed-phase ion-pair liquid chrotrometry; HFIP, hexafluoroisopropanol; TEAA, triethexylamine; PDII, phosphodiesterase II; ACN, acetonitliquid chromatography; dsDNA, double-stranded DNAtate; ESI, electrospray ionization; UV, ultraviolet; Tm, m

a b s t r a c t

Small interfering RNA (siRNA)-induced gene silencing shows great promise in genomic research and ther-apeutic applications. siRNA duplexes are typically assembled from complementary synthetic oligonucle-otides. High-purity single-stranded species are required for in vivo applications. Methods for separation,characterization, and purification of short RNA strands have been developed based on reversed-phaseion-pair liquid chromatography. The purification strategies were developed for both single-strandedand duplex RNA species. The method of duplex purification uses on-column annealing of complementaryRNA strands, followed by separation of the target duplex from truncated duplexes and single-strandedRNA forms. The proposed method significantly reduces the purification time of synthetic siRNA.

� 2009 Elsevier Inc. All rights reserved.

Ribonucleic acid interference (RNAi)1 is a recently discovered RNA molecules, including siRNA, for therapeutic use are pro-

mechanism for regulation of gene expression. The RNAi mechanismuses two forms of small RNA molecules: microRNA (miRNA), whichprevents protein synthesis, and small interfering RNA (siRNA), whichdegrades messenger RNA (mRNA). The mechanism was first discov-ered in plants and later discovered in worms and mammals [1–3].The RNAi pathway uses short oligonucleotide sequences (typically21–23 nt) in duplex form referred to as siRNA. The currently ac-cepted mechanism of RNAi describes the incorporation of siRNA intothe RNA-induced silencing complex (RISC). The RISC uses one of thesingle-stranded components of the introduced siRNA to sequestercomplementary mRNA. Following binding, the mRNA is degraded,thereby preventing protein translation [4].

ll rights reserved.

.M. McCarthy).ce; miRNA, microRNA; siRNA,A-induced silencing complex;

electrophoresis; CGE, capillaryrformance liquid chromatog-

matography; MS, mass spec-hylammonium acetate; HA,rile; UPLC, ultra-performance; HAA, hexylammonium ace-elting temperature.

duced synthetically via stepwise synthesis. Although this processis very efficient, with yields as high as 99.5% per coupling step, asimple 21-mer oligo typically used for RNAi therapeutic purposeswould have a maximum purity of 90% without further purification.Most impurities, often referred to as ‘‘failure sequences,” arise dur-ing synthesis [5–7]. The ease of oligonucleotide synthesis, coupledwith its utility in gene silencing, is leading to a dramatic increase ingene silencing experiments in genomic research. siRNA is alsoemerging as a therapeutic strategy for a wide variety of diseases,including cancer, macular degeneration, and viral infection [8–12]. A major challenge in developing siRNA therapeutics remainsassurance of purity to minimize off-target gene silencing.

Gene silencing experiments are often performed with siRNAprepared by hybridization (annealing) of two complementary sin-gle-stranded counterparts. Each strand introduces its own set ofimpurities that can further complicate the siRNA mixture withmismatched sequences and noncomplementary single-strandedsequences. The presence of these impurities in a therapeutic mix-ture may lead to unwanted, and perhaps detrimental, nontargetedgene silencing. The presence of nonhybridized single-stranded RNA(ssRNA) is often associated with a decrease in therapeutic potency,further highlighting the need for efficient purification strategies forboth RNA and siRNA.

There have been a number of methods reported for the analysisand purification of oligonucleotides, including polyacrylamide gel

182 RP IP LC analysis and purification of siRNA / S.M. McCarthy et al. / Anal. Biochem. 390 (2009) 181–188

electrophoresis (PAGE), capillary gel electrophoresis (CGE) [13,14],anion-exchange high-performance liquid chromatography (AXHPLC) [15–17], and reversed-phase ion-pair liquid chromatogra-phy (RP IP LC) [5,18,19].

RP IP LC provides good resolution of oligonucleotides and is com-patible with mass spectrometry (MS) detection. Only a few reportshave shown LC MS analysis of siRNA using a triethylamine/hexafluo-roisopropanol (HFIP) ion-pairing system [20,21].

The preparation of siRNA typically requires separate purificationsteps for each complementary RNA strand, followed by annealing ofthe purified products in the final step, yielding the desired siRNA.This article outlines an alternative, less labor-intensive method forpurification of siRNA in duplex form using on-column annealing ofcrude synthetic complementary RNA strands. The efficient methodsfor LC and LC MS analysis of siRNA are also presented.

Materials and methods

Materials

Two molar triethylammonium acetate (TEAA, cat. no. 5204-74-0), hexylamine (HA, cat. no. 203-851-8), acetic acid (cat. no. 200-580-7), and phosphodiesterase II (PDII) were purchased from Sig-ma–Aldrich (St. Louis, MO, USA) and used without further purifica-tion. Acetonitrile (ACN, Optima LC/MS grade, cat. no. 75-05-8) waspurchased from Fisher Scientific (Fair Lawn, NJ, USA). SyntheticRNA oligonucleotide sequences were purchased from IntegratedDNA Technologies (IDT, Coralville, IA, USA). The sequences were asfollows and are labeled arbitrarily for the purpose of clarity as upperand lower strands; however, the reader should understand thatthese distinctions may have important meaning elsewhere. The se-quences were 19-mer RNA with DNA TT overhangs and are listedas complementary sequences: RNAi upper 50-UCG UCA AGC GAUUAC AAG GTT-30, RNAi lower 50-CCU UGU AAU CGC UUG ACG ATT-30, FAS upper 50-CCC UGA GAU CCC AGC GCU GTT-30, and FAS lower50-CAG CGC UGG GAU CUC AGG GTT-30. Sequences used for the du-plex melting study were as follows: low melt duplex (47.5 �C) upper50-TTG TCA CTG TGT TGT AAT C-30 and lower 50-GAT TAC AAC ACAGTG ACA A-30, medium melt duplex (57.0 �C) upper 50-TTG TCCCTG TGC CGT ACT C-30 and lower 50-GAG TAC GGC ACA GGG ACAA-30, and high melt duplex (66.0 �C) upper 50-TTG TCC CGG TGCCGG ACG C-30 and lower 50-GCG TCC GGC ACC GGG ACA A-30. AMilli-Q system (Millipore, Bedford, MA, USA) was used to preparedeionized water (18 MX cm) for HPLC mobile phases.

RNA and siRNA purification with HPLC

Crude synthetic oligonucleotides were reconstituted to yield asolution of approximately 3.5 nmol/ll in 0.1 M TEAA. The resultingmixture was purified using a Waters Alliance Bio 2696 separationmodule with a 996 photodiode array detector (Waters, Milford,MA, USA) scanning from 220 to 350 nm. Oligonucleotides were sep-arated using a 4.6 � 50-mm XBridge C18 OST column packed with2.5-lm particles (average pore diameter 135 Å, Waters). The mobilephases were as follows: 0.1 M TEAA (pH 7.0) (A) and 80/20% (v/v) A/ACN (B). The mobile phases were degassed via sonication for 10 min.The flow rate was 1 ml/min.

The gradient was from 30% to 52.5% B in 30 min (0.15% ACN/min) for ssRNA and from 25% to 75% B in 30 min (0.33% ACN/min) for duplex separations. Single-stranded separations were per-formed at 60 �C, and duplex separations were performed at 20 �C.

Fractions were collected manually from the peak apex toapproximately 30% of the peak height in all cases. Collected frac-tions were evaporated using a CentriVap (Labconco, Kansas City,MO, USA) and reserved for further analysis and use.

UPLC analysis of single-stranded RNA

The single- and double-stranded nucleic acids were analyzedwith an ACQUITY ultra-performance liquid chromatography(UPLC) system using a Waters ACQUITY OST C18 column packedwith 1.7-lm particles (average pore size 130 Å). The mobile phasewas 100 mM TEAA as mobile phase A and 20% ACN in A with a flowrate of 0.2 ml/min and a gradient of 35–65% B in 10 min. A photo-diode array detector was used for detection at 260 nm (all instru-ments were obtained from Waters). On-column loading of eachsample was 7.5–10 pmol. The temperature was varied between20 and 70 �C in 5� increments (see figure captions).

Following purification, ssRNA, double-stranded DNA (dsDNA),and siRNA were analyzed using a UPLC and single quadrupole massspectrometer. dsDNA was prepared from complementary oligonu-cleotides by heating the sample to 90 �C and cooling slowly toroom temperature over 10 min. The LC MS mobile phases werecomposed of 25 mM hexylammonium acetate (HAA, pH 7.0) asmobile phase A and 100% ACN as mobile phase B. The gradient var-ied depending on the sample and mobile phase used as specified inthe figure legends. The MS instrument was used in electrosprayionization (ESI) negative ion mode, capillary voltage was 3 kV, conevoltage was 28 V, extractor voltage was 3 V, source temperaturewas 150 �C, desolvation temperature was 350 �C, cone gas flowwas 31 L/h, and desolvation gas flow was 700 L/h. MS data weredeconvoluted using the maximum entropy software MaxEnt 1.

Results and discussion

Purification of RNA single strands

The RP IP LC method is well suited for purification of nucleicacids [5]. As shown in Fig. 1, useful amounts of RNA can be purifiedin a single injection using an analytical column. The purification upto 140 nmol of a 21-mer oligonucleotide was performed usingTEAA mobile phases. As the column load increased, typical peakbroadening is observed with a partial retention shift of failed se-quences (selected impurities are labeled with asterisks in Fig. 1).Due to the displacement effect, ‘‘hearth cutting” is a viable strategyto obtain good purity of target oligonucleotides with only a moder-ate sacrifice in product recovery [5].

For all mass loads illustrated in Fig. 1, the main peak was collectedfrom its apex to approximately 30% of the peak height, shown in thefigure for the largest injection by arrows. Typically, 1–2 ml was col-lected as a single fraction depending on the column mass load. Theyield was between 55% and 70% based on the peak area in the col-lected range versus the total peak ultraviolet (UV) area.

UPLC analysis of the collected fractions indicated that the iso-lated oligonucleotides were typically P95% pure (Fig. 1 inset),which is suitable for a variety of molecular biology and therapeuticapplications.

Duplex DNA stability in RP IP LC

RNA can be purified in single-stranded or duplex form. The lat-ter approach is intriguing, potentially shortening the process. Nev-ertheless, several practical problems need to be addressed. Theforemost concern is the stability of siRNA duplexes under RP IPLC conditions. It has been well documented that RP IP LC is suitablefor analysis of 50- to 1000-bp-long dsDNA restriction fragments[22,23]. However, unmodified siRNA species are typically 19 baseslong complementary region, with melting temperatures (Tm) rang-ing between 45 and 65 �C. Some on-column melting of siRNA canpotentially occur during the analysis or purification, especially atelevated temperatures [22].

140 nmol

84 nmol

28 nmol

1.4 nmol

0 18Minutes

RNAImpurities

**

**

2 3.5Minutes

Purified OligonucleotidePurity > 95%

N-1

start of collection

end of collection

Fig. 1. RP IP chromatographic purification of a crude synthetic RNA. The column loading was varied from 1.4 to 140 nmol load. Fractions were collected for each purificationfrom the peak apex to 30% of the peak height. The eluent was monitored at 260 nm. Column: XBridge OST BEH C18, 4.6 � 50 mm, 2.5 lm. Mobile phase A: 100 mM TEAA (pH7.0); mobile phase B: 20% ACN in mobile phase A. Gradient: from 30% B (6% ACN) to 52.5% B (10.5% ACN) in 30 min, 1 ml/min, 60 �C.

RP IP LC analysis and purification of siRNA / S.M. McCarthy et al. / Anal. Biochem. 390 (2009) 181–188 183

To investigate the stability of short duplexes, we used a set ofDNA duplexes with various melting points. Three duplexes wereprepared from purified single-stranded counterparts with meltingpoints of 47.5 �C (low Tm), 57.0 �C (medium Tm), and 66.0 �C (highTm). The variation in melting point was accomplished by strategicreplacement of T and/or A bases with G and/or C bases that offergreater thermal stability due to increased H bonding. The duplexeswere prepared with a 50% excess of upper strand so as to monitorRP IP LC retention of both single-stranded oligonucleotide anddsDNA species.

Figs. 2A–C illustrate the analysis of duplexes at various temper-atures. Approaching the melting temperature, dramatic peakbroadening occurs, indicating the on-column duplex melting. ThedsDNA melting is accompanied by an appearance of complemen-tary oligonucleotides. The Tm values observed in Fig. 2 roughly cor-respond to the calculated dsDNA values. When the columntemperature reached and exceeded the melting point, the duplexesrapidly melted on injection on column and eluted from column as apair of sharp and symmetrical single-stranded peaks. Retentiontimes and peak shape of the single-stranded oligonucleotides in-jected separately are provided for comparison as supplementalmaterials (see Supplemental Figs. 1–3 in supplementary material).

It is worth pointing out that retention of oligonucleotides isnoticeably affected by their sequence (nucleobase composition)[5,24], whereas all dsDNA species elute in nearly identical reten-tion time. The RP IP LC retention of duplexes is driven less by se-quence and more by charge-to-charge interaction (dsDNA ismore retained than its corresponding oligonucleotides).

UPLC analysis of siRNA

Based on the experiment shown in Fig. 2, we selected 20 �C as ageneric separation temperature for further siRNA experiments. Theduplex siRNA samples were prepared with an excess of eitherupper or lower strand to demonstrate the resolution of duplex siR-NA from its single-stranded counterparts.

As shown in Figs. 3A and B, the RP IP LC clearly separates full-length duplex from single-stranded oligonucleotide ‘‘impurities.”No apparent deterioration of duplex peak shape is observed, indi-cating little or no on-column melting of siRNA duplex.

Fig. 3C shows the analysis of siRNA duplex prepared by annealingnearly equimolar amounts of complementary RNA oligonucleotides.A small excess of the lower RNA strand was present, whereas thepeak corresponding to the upper strand was not detected. This sup-ports the conclusion that siRNA duplex is stable at given chromato-graphic conditions. Minor peaks eluting prior to the target siRNApeak are presumably partially truncated duplex species. The abilityof the RP IP LC method to resolve siRNA from its truncated dupleximpurities is discussed in more detail in the following section.

Resolution of partially truncated siRNA species

‘‘Mismatched” truncated duplexes arise when annealing oligo-nucleotides are contaminated with shorter failed sequences. Littleis known about the performance of RP IP LC for the separation oftruncated duplex species. Naturally, the purification of siRNA induplex form is a viable strategy only if an efficient separation oftarget duplex from truncated species is achieved.

To investigate the ability of our method to afford these separa-tions, we prepared a 50 truncated ladder of one complementarystrand by its partial digestion with the exonuclease PDII. We thenannealed this mixture of truncated sequences with a full-lengthcomplementary strand. The resulting mixture mimics a crude syn-thetic mixture that is often contaminated with failure sequences(Scheme 1).

Full-length truncated duplexes and remaining single-strandedspecies were separated using UPLC with 25 mM HAA mobile phase,which provides suitable chromatographic resolution and adequateMS sensitivity. Data shown in Fig. 4 suggest that the RP IP LC meth-od employed resolved the truncated duplexes.

The retention order of peaks in Fig. 5 was confirmed by MS. Thetraces of selected ion chromatograms were generated by extractingthe –3 or –4 charge state mass for the appropriate oligonucleo-tides, and the results are shown in Fig. 5A. Interestingly, wheninspecting the MS data under detected peaks, we do not observethe expected masses of duplexes. Instead, the MS data reveal thepresence of two complementary oligonucleotides correspondingto the original duplexes. Apparently, the siRNA species survivethe chromatographic separation, but siRNA duplexes are meltedin the ESI source upon ionization.

Fig. 2. Stability of DNA duplexes in RP IP LC. Duplex melting is induced at highercolumn temperature. Duplexes have calculated melting points of 37.5 �C (A), 45.7 �C(B), and 53.9 �C (C). Duplex melting point was increased by increasing relative G/Ccontent in DNA single strands, yielding greater stability due to increased H bonding.Mobile phase A: 100 mM TEAA (pH 7.0); mobile phase B: 20% ACN in mobile phaseA. Gradient: from 35% B (7% ACN) to 65% B (13% ACN) in 10 min, 0.2 ml/min. Thecolumn temperature was varied from 20 to 70 �C in 5 �C increments.

184 RP IP LC analysis and purification of siRNA / S.M. McCarthy et al. / Anal. Biochem. 390 (2009) 181–188

The MS data for truncated U21/L20 duplex peak is shown inFig. 5B. The MS spectra arising from multiply charged species weredeconvoluted, yielding the expected masses of 21 nt upper strandand 20 nt lower strand oligonucleotide (Fig. 5C). This confirms thatthe RP IP LC method is capable of resolving closely related trun-cated duplex species.

We investigated the prospects for MS detection of the intact du-plex mass by decreasing both the desolvation and source temper-atures. Although we were able to produce spectra that containedthe duplex along with some single-stranded components, these

species were also present with a large percentage of hexylamineadducts (data not shown). We investigated the use of ammoniumacetate for intact duplex MS analysis with greater success. Wefound that the use of ammonium acetate yielded ionization ofthe intact duplex using normal MS operating conditions, but thechromatographic elution was reversed in this non-ion-pairing buf-fer (duplex species eluting prior to the single-stranded species) andthe separation selectivity suffered.

MS signal of oligonucleotides typically contains alkali ionadductation. Fig. 5C shows a significant iron adductation in thepresented case. The source of iron is unclear; we believe that tracesof iron are present in the mobile phase.

Closer inspection of Fig. 4 indicates that as the lower standlength decreases from 21- to 16-mer and lower, incomplete duplexformation and/or on-column melting become apparent. We calcu-lated the theoretical melting points for truncated duplexes [25].The Tm values decreased from 49 �C for the full-length duplex to34 �C for the U21/L16 mismatched duplex (only 14 base pairs con-tribute to annealing). This dramatic decrease in melting point isthe likely source of incomplete duplex formation and melting forthe shortest duplexes even when using a 20 �C separationtemperature.

Purification of siRNA from crude oligo mixtures

After demonstrating the stability of siRNA duplexes and theirresolution from truncated failure sequence products, we investi-gated the RP IP LC method utility for duplex purification. Similarto our purification strategy with single-stranded oligonucleotides,we evaluated the ability of our method to yield purified siRNA ina limited number of steps.

First, we annealed crude complementary oligonucleotide mix-tures off-line to prepare the duplex siRNA. The expected productimpurity profile is more complex than the one obtained by anneal-ing a partially digested single strand with a full-length comple-mentary strand shown in Scheme 1. Nevertheless, the HPLCpurification of the resulting mixture was successful, and the meth-od yielded good separation of single-stranded and mismatched du-plexes from the desired full-length duplex (data not shown). Thefull-length duplex was collected by appropriate hearth cutting ofthe peak, and subsequent analysis via UPLC revealed that thefull-length duplex purity was P98%.

Following our success in duplex purification using a separateannealing step, we investigated whether the method can be furtherimproved and shortened. This was prompted by the observationthat when injecting two complementary RNA strands on chro-matographic column at conditions below melting temperature,they spontaneously anneal and elute as duplex. This is the caseeven when including the 8 M urea in the sample as denaturant. Itappears that as the urea is washed away by the mobile phaseand the complementary strands are focused on the head of column,they spontaneously and effectively anneal to the correspondingduplex.

We scaled up the siRNA purification using the on-columnannealing with the following strategy. The first strand was injectedat initial gradient conditions, followed by injection of the comple-mentary strand. The gradient elution was started immediatelyafterward. We observed quantitative formation of the duplex siR-NA and its elution at expected elution time. The purification strat-egy was scalable (Fig. 6), yielding results similar to those found forssRNA. Expected single-stranded and double-stranded impuritieswere eluting mostly prior to the target peak. The collection wasperformed by hearth cutting the main peak from the apex to 30%of the peak height, as indicated in Fig. 6 by the arrows (for the85-nmol purification scale). Collected fractions were analyzed viaUPLC. Fig. 7A shows a typical UPLC chromatogram of a purified du-

0 8Minutes

A

B

C

single strandedupper

single strandedlower

single strandedlower

mismatchduplexes

mismatchduplexes

mismatchduplexes

Duplex

Duplex

Duplex

Fig. 3. UPLC separation of siRNA duplex from excess of upper chain of ssRNA (A) and excess of single-stranded lower chain (B). (C) Separation of siRNA duplex prepared byannealing nearly stoichiometric ratio of complementary RNA oligonucleotides. Column: Waters OST BEH C18, 2.1 � 50 mm, 1.7 lm. Mobile phase A: 100 mM TEAA (pH 7.0);mobile phase B: 20% ACN in mobile phase A. Gradient: from 35% B (7% ACN) to 85% B (13% ACN) in 10 min, 0.2 ml/min, 20 �C. The eluent was monitored at 260 nm.

Scheme 1. Potential mismatched duplexes arising from annealing of a mixture offull-length upper strand (U21) and complementary truncated lower strand (L21, L20,L19, . . .) RNA sequences.

RP IP LC analysis and purification of siRNA / S.M. McCarthy et al. / Anal. Biochem. 390 (2009) 181–188 185

plex at 20 �C overlaid with crude upper and lower strands analyzedat the same conditions. These data confirm the purity of the col-lected duplex. Two small peaks of single-stranded oligonucleotideswere detected at 0.6–0.8% levels, suggesting that minor on-columnduplex melting occurs on injection at a given low-ionic-strength LCMS mobile phase system (25 mM HAA). MS detection confirmedthe identity of the main duplex peak. We also analyzed the purifiedduplex at 60 �C, conditions where the duplex is completely dena-tured. Fig. 7B shows that the duplex is composed of an equimolarratio of the complementary RNA strands. Little or no single-stranded failed product impurities were detected in thisexperiment.

Limitations of the method

We investigated the generality of the RNA purification methodsby purifying multiple sets of RNA oligonucleotides. Interestingly,we found that oligonucleotides with strong secondary structureyield broad fronting and tailing peaks and could be difficult to pur-ify in single-stranded form even at 60 �C. This may present a prob-lem for scaling up the method to a preparative scale, where thecolumn is typically kept at ambient temperature. Occasionally,the on-column formation of intramolecular structures may inter-fere with purification even for oligonucleotides with relativelyweak secondary structure.

To overcome this issue, we annealed a complementary pair ofoligonucleotides with significant secondary structure off-columnand performed purification in duplex form (see SupplementalFig. 4 in supplementary material). The purification was successful;the secondary structure impact apparently was eliminated by theformation of duplex. On the other hand, the on-column annealingmethod was less efficient; besides the duplex peak, we detected in-creased amounts of non-annealed single-stranded oligonucleo-tides. It appears that secondary structure can, to a certain extent,prevent the on-column formation of duplex. The duplex purifica-tion was still possible, but the yield was reduced to approximatelyhalf for the given set of oligonucleotides. For oligonucleotides withstrong secondary structure, it is advantageous to use a separate off-column annealing step if high purification yields are desired.

Conclusions

The methods presented in this article allow semipreparativescale purification of RNA in both single-stranded and duplex forms.The proposed methods are rapid and gentle while offering impres-sive resolution of both single-stranded and duplex RNA impurities.

ssRNAi siRNA Duplex

Minutes0 8

lower strand (PDII digest)upper strand

Fig. 4. UPLC separation of single-stranded oligonucleotides and truncated siRNA duplexes from full-length siRNA duplex. Lower traces show analysis of partially fragmentedlower strand and upper strand prior to annealing. Upper trace shows separation of the mixture after annealing (upper strand was provided in excess). Column: Waters OSTBEH C18, 2.1 � 50 mm, 1.7 lm. Mobile phase A: 25 mM HAA (pH 7.0); mobile phase B: 100%. Gradient: from 30% B to 40% B in 10 min, 0.2 ml/min, 10 �C. The eluent wasmonitored at 260 nm.

M/Z600 2000 6000 7000

TIC5 8Minutes

Deconvolute-5

-4 Lower6301.04

Upper6692.63

}

}

+ Fe + Fe

U21/L20U21/L19U21/L18U21/L17U21/L16U21/L15U21/L14U21/L13

Mass

A

B C

Fig. 5. Identification of siRNA impurities by LC MS. (A) Total ion chromatogram and extracted ion chromatograms for selected impurities. Selected ion species typicallycorresponded to [M – 3H]–3 charge state for truncated RNA species. (B) MS spectrum for U21/L20 truncated duplex. (C) Deconvoluted MS data. Two dominant mass signalscorrespond to the expected mass of 21 nt upper strand and 20 nt lower strand. UPLC conditions are the same as those listed in the Fig. 3 caption. The MS was operated at ESInegative mode, capillary voltage was 3 kV, cone voltage was 28 V, extractor voltage was 3 V, source temperature was 150 �C, desolvation temperature was 350 �C, cone gasflow was 31 L/h, and desolvation gas flow was 700 L/h.

186 RP IP LC analysis and purification of siRNA / S.M. McCarthy et al. / Anal. Biochem. 390 (2009) 181–188

The scalability of the LC purification method allows scaling up to apreparative scale. The proposed methods are fast, efficient, andcost-effective, and they represent a significant advance in siRNApurification strategies. In addition, we developed the method forLC MS analysis of ssRNA and dsRNA using the nondenaturing mo-bile phase consisting of HAA. This concurrent analysis via UV and

MS reduces analysis times considerably compared with othermethods. The use of HAA is also cost-effective compared withHFIP-based mobile phases, making it an attractive alternative forroutine duplex analysis. The mobile phase system was shown tobe nondenaturing for duplex analysis, as evidenced by intact du-plex elution.

0 Minutes 25

ssRNAiimpurities

RNAi dupleximpurities

85 nmol

15 nmol

1.5 nmol

end of collection

start of collection

Fig. 6. RP IP HPLC purification of siRNA duplex formed on column after injecting two complementary synthetic RNA oligonucleotides. Column: XBridge OST BEH C18,4.6 � 50 mm, 2.5 lm. Mobile phase A: 100 mM TEAA (pH 7.0); mobile phase B: 20% ACN in mobile phase A. Gradient: from 25% B (5% ACN) to 75% B (12% ACN) in 30 min,1 ml/min, 20 �C. The eluent was monitored at 260 nm.

A

B

0 10Minutes

0 10Minutes

Fig. 7. UPLC verification of purified siRNA quality. (A) UPLC analysis of purified siRNA duplex and each crude complementary strand at 20 �C. (B) UPLC analysis of purifiedsiRNA and each crude complementary strand at 60 �C. Column: Waters OST BEH C18, 2.1 � 50 mm, 1.7 lm. Mobile phase A: 25 mM HAA (pH 7.0); mobile phase B: 100%.Gradient: from 30% B to 40% B in 10 min, 0.2 ml/min.

RP IP LC analysis and purification of siRNA / S.M. McCarthy et al. / Anal. Biochem. 390 (2009) 181–188 187

188 RP IP LC analysis and purification of siRNA / S.M. McCarthy et al. / Anal. Biochem. 390 (2009) 181–188

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.ab.2009.03.042.

References

[1] A. Fire, S. Xu, M.K. Montgomery, S.A. Kostas, S.E. Driver, C.C. Mello, Potent andspecific genetic interference by double-stranded RNA in Caenorhabditis elegans,Nature 391 (1998) 806–811.

[2] S.M. Elbashir, J. Harborth, W. Lendeckel, A. Yalcin, K. Weber, T. Tuschl,Duplexes of 21-nucleotide RNAs mediate RNA interference in culturedmammalian cells, Nature 411 (2001) 494–498.

[3] R. Jorgensen, Altered gene expression in plants due to transinteractions between homologous genes, Trends Biotechnol. 8 (1990)340–344.

[4] D. Grimm, M.A. Kay, Therapeutic application of RNAi: is mRNA targeting finallyready for prime time?, J Clin. Invest. 117 (2007) 3633–3641.

[5] M. Gilar, Analysis and purification of synthetic oligonucleotides by reversed-phase high-performance liquid chromatography with photodiode array andmass spectrometry detection, Anal. Biochem. 298 (2001) 196–206.

[6] M. Gilar, E.S.P. Bouvier, Purification of crude DNA oligonucleotides by solid-phase extraction and reversed-phase high-performance liquidchromatography, J. Chromatogr. A 890 (2000) 167–177.

[7] J. Temsamani, M. Kubert, S. Agrawal, Sequence identity of the n – 1 product of asynthetic oligonucleotide, Nucleic Acids Res. 23 (1995) 1841–1844.

[8] R. Agami, RNAi and related mechanisms and their potential use for therapy,Curr. Opin. Chem. Biol. 6 (2002) 829–834.

[9] J.M. Alisky, B.L. Davidson, Towards therapy using RNA interference, Am. J.Phamacogenomics 4 (2004) 45–51.

[10] B.L. Bass, Double-stranded RNA as a template for gene silencing, Cell 101(2000) 235–238.

[11] G. Randall, A. Grakoui, C.M. Rice, Clearance of replicating hepatitis C virusreplicon RNAs in cell culture by small interfering RNAs, Proc. Natl. Acad. Sci.USA 100 (2003) 235–240.

[12] J.M. Silva, S.M. Hammond, G.J. Hannon, RNA interference: a promisingapproach to antiviral therapy, Trends Mol. Med. 8 (2002) 505–508.

[13] A.S. Cohen, A.J. Bourque, B.H. Wang, D.L. Smisek, A. Belenky, A nonradioisotopeapproach to study the in vivo metabolism of phosphorothioateoligonucleotides, Antisense Nucleic Acid Drug Dev. 7 (1997) 13–22.

[14] M. Gilar, A. Belenky, D.L. Smisek, A. Bourque, A.S. Cohen, Kinetics ofphosphorothioate oligonucleotide metabolism in biological fluids, NucleicAcids Res. 25 (1997) 3615–3620.

[15] A.J. Bourque, A.S. Cohen, Quantitative analysis of phosphorothioateoligonucleotides in biological fluids using fast anion-exchange chromatographyand capillary gel electrophoresis, J. Chromatogr. B 662 (1994) 343–349.

[16] G.S. Srivatsa, P. Klopchin, M. Batt, M. Feldman, R.H. Carlson, D.L. Cole,Selectivity of anion exchange chromatography and capillary gelelectrophoresis for the analysis of phosphorothioate oligonucleotides, J.Pharm. Biomed. Anal. 16 (1997) 619–630.

[17] W.J. Warren, G. Vella, Principles and methods for the analysis and purification ofsynthetic deoxyribonucleotides by high-performance liquid chromatography,Mol. Biotechnol. 4 (1995) 179–199.

[18] K.J. Fountain, M. Gilar, J.C. Gebler, Analysis of native and chemically modifiedoligonucleotides by tandem ion-pair reversed-phase high-performance liquidchromatography/electrospray ionization mass spectrometry, Rapid Commun.Mass Spectrom. 17 (2003) 646–653.

[19] M. Gilar, K.J. Fountain, Y. Budman, J.L. Holyoke, H. Davoudi, J.C. Gebler,Characterization of therapeutic oligonucleotides using liquid chromatographywith on-line mass spectrometry detection, Oligonucleotides 13 (2003) 229–243.

[20] Y. Zou, P. Tiller, I.W. Chen, M. Beverly, J. Hochman, Metabolite identification ofsmall interfering RNA duplex by high-resolution accurate mass spectrometry,Rapid Commun. Mass Spectrom. 22 (2008) 1871–1881.

[21] M. Beverly, K. Hartsough, L. Machemer, Liquid chromatography/electrospraymass spectrometric analysis of metabolites from an inhibitory RNA duplex,Rapid Commun. Mass Spectrom. 19 (2005) 1675–1682.

[22] W. Xiao, P.J. Oefner, Denaturing high-performance liquid chromatography: areview, Hum. Mutat. 17 (2001) 439–474.

[23] C.G. Huber, P.J. Oefner, E. Preuss, G.K. Bonn, High-resolution liquidchromatography of DNA fragments on non-porous poly(styrene-divinylbenzene) particles, Nucleic Acids Res. 21 (1993) 1061–1066.

[24] M. Gilar, K.J. Fountain, Y. Budman, U.D. Neue, K.R. Yardley, P.D. Rainville, R.J.Russell, J.C. Gebler Jr., Ion-pair reversed-phase high-performance liquidchromatography analysis of oligonucleotides: retention prediction, J.Chromatogr. A 958 (2002) 167–182.

[25] W.A. Kibbe, OligoCalc: an online oligonucleotide properties calculator, NucleicAcids Res. 35 (2007) W43–W46.