Novel Isothermal, Linear Nucleic AcidAmplification Systems for Highly Multiplexed

ApplicationsNurith Kurn,* Pengchin Chen, Joe Don Heath, Anne Kopf-Sill, Kathryn M. Stephens,

and Shenglong Wang

Background: Global analysis of the genome, transcrip-tome, and proteome is facilitated by the recent develop-ment of tools for large-scale, highly parallel analysis.We describe a novel nucleic acid amplification systemthat generates products by several methods. 3�-Ribo-SPIATM primes cDNA synthesis at the 3� polyA tail, andwhole transcript (WT)-Ribo-SPIA primes cDNA synthe-sis across the full length of the transcripts and thusprovides whole-transcriptome amplification, indepen-dent of the 3� polyA tail.Methods: We developed isothermal linear nucleic acidamplification systems, which use a single chimericprimer, for amplification of DNA (SPIA) and RNA(Ribo-SPIA). The latter allows mRNA amplificationfrom as little as 1 ng of total RNA. Amplificationefficiency was calculated based on the delta thresholdcycle between nonamplified cDNA targets and ampli-fied cDNA. The amounts and quality of total RNA andamplification products were determined after purifica-tion of the amplification products. GeneChip® arraygene expression profiling and real-time PCR were usedto test the accuracy and reproducibility of the method.Quantification of cDNA products (before and afteramplification) at the 2 loci along the transcripts wasused to assess product length (for evaluation of the3�-initiated Ribo-SPIA) and equal representationthroughout the length of the transcript (for evaluation ofthe whole transcript amplification system, WT-Ribo-SPIATM).Results: Ribo-SPIA–based global RNA amplificationexhibited linearity over 6 orders of magnitude of tran-

script abundance and generated microgram amounts ofamplified cDNA from as little as 1 ng of total RNA.Conclusions: The described methods enable compre-hensive gene expression profiling and analysis fromlimiting biological samples. The WT-Ribo-SPIA proce-dure, which enables amplification of non–polyA-tailedRNA, is suitable for amplification and gene expressionanalysis of both eukaryotic and prokaryotic biologicalsamples.© 2005 American Association for Clinical Chemistry

High-density microarray platforms enable highly parallelcomprehensive measurements of the components of bio-logical systems, integrating many aspects of a phenotypeas determined by the genome, transcriptome, and pro-teome. The shift in strategy from defined componentanalysis to global and comprehensive approaches re-quires sophisticated analyses and multiplexing of thetargets being queried and is facilitated by the ability toglobally amplify the analytes of choice, and subsequentanalysis at addressable loci.

NuGENTM has developed rapid, isothermal linear nucleicacid amplification systems that use a single chimeric primer,DNA polymerase with strand displacement activity andRNase H: SPIATM (single primer isothermal amplification), aDNA amplification procedure; and Ribo-SPIATM, an RNAamplification procedure. SPIA amplification was imple-mented for global genomic DNA amplification and for theamplification of specific genomic sequences and syntheticoligonucleotide DNA targets. Ribo-SPIA is similarly suit-able for global and target-specific RNA amplification(1–3). Both methods can be used for amplification ofhighly diverse populations of species, such as globaltranscriptome amplification.

Ribo-SPIA enables mRNA amplification from as littleas 1 ng of total RNA. Procedures for global amplificationof all transcripts in both eukaryote and prokaryote totalRNA were developed, enabling gene expression profiling

NuGEN Technologies, Inc., 821 Industrial Rd., Unit A, San Carlos, CA94070.

*Author for correspondence. Fax 650-622-9867; e-mail nkurn@nugeninc.com.

Received April 30, 2005; accepted August 4, 2005.Previously published online at DOI: 10.1373/clinchem.2005.053694

Clinical Chemistry 51:101973–1981 (2005) Oak Ridge Conference

1973

and analysis from minute biological samples. The ampli-fication products are single-stranded cDNA suitable for avariety of detection and quantification platforms, includ-ing high- and low-density arrays, and solution-phasequantification methods for a variety of quantitative real-time PCR chemistries. The linear amplification of largepopulations of nucleic acid species is particularly impor-tant when sample input is limited, as is commonly en-countered in clinical research.

In this study we describe the linearity, sensitivity, andreproducibility of the Ribo-SPIA–based OvationTM sys-tems for gene expression and profiling. The use of theRibo-SPIA technology for global amplification of mRNAin biological samples and subsequent quantification ofdefined sets of transcripts of interest has been validated ina mouse model for study of systemic HIV-1 infection (4 ).Similarly, the Ovation RNA amplification system hasbeen used for mRNA amplification and transcriptionalanalysis of pig embryogenesis (5 ).

Materials and Methodstotal rna samplesWe used Universal Human Reference (UHR)1 total RNAand HeLa total RNA from Stratagene, human spleen andplacenta total RNA from Ambion, and human skeletalmuscle total RNA from Clontech. RNA quality was de-termined with an Agilent Bioanalyzer. PCR primers werepurchased from Operon Biotechnologies, Inc., or fromIntegrated DNA Technologies.

In vitro–prepared transcripts for supplementation ex-periments were purchased from Affymetrix (GeneChip®

Eukaryotic Poly-A RNA Control Kit).

quantitative pcrQuantification of DNA targets and amplification of prod-ucts, as well as cDNA (before and after amplification),were carried out with an ABI PRISMTM 7700 sequencedetector (Applied Biosystems) or an MJ Opticon (MJResearch). TaqMan® Universal PCR Master Mix, NoAmpErase® UNG (Applied Biosystems), or 2� Thermo-Start Q-PCR Master Mix (ABgene) and QPCR (SYBR®

Green I) reactions were performed with a QiagenQuantiTect SYBR Green PCR Kit. The reactions werecarried out in 96-well optical reaction plates (ABgene),using the recommended thermocycling program (15 minat 95 °C to activate Taq polymerase followed by 40 cyclesof 95 °C for 15 s and 60 °C for 60 s). Primers and probes forquantification of various gene products were designedwith Primer ExpressTM software (Applied Biosystems).Primer and probe sets were selected based on assessmentof their amplification efficiency. Amplification efficiencywas determined from a calibration curve constructed withserially diluted samples.

Human glyceraldehyde-3-phosphate dehydrogenase(GAPDH) gene products were quantified by TaqManquantitative PCR for assessment of amplification effi-ciency, yield of specific amplification products, and repro-ducibility of the new amplification system. Both nonampli-fied and amplified cDNA were interrogated at 2 locationsalong the gene transcript product, the 3� end (at 0.3 kb fromthe polyA tail) and the 5� end (at 1 kb from the polyA tail).The quantification of cDNA products (before and afteramplification) at the 2 loci along the transcripts providesa tool for the assessment of product length (an importantcriterion for evaluation of the 3�-initiated Ribo-SPIA) andequal representation throughout the length of the tran-script (an important criterion for evaluation of the whole-transcript amplification system, WT-Ribo-SPIA).

SPIA and Ribo-SPIAAll amplifications used the Ovation RNA amplificationSystem or Ovation Biotin System (NuGEN Technologies,Inc.), or components thereof, according to manufacturer’sinstructions (http://www.nugeninc.com).

The quantification of target nucleic acids and amplifi-cation, SPIA, and Ribo-SPIA products is expressed as thethreshold cycle (Ct) value as determined by real-time PCR(TaqMan or SYBR Green I). Amplification efficiency wascalculated based on the delta Ct between nonamplifiedcDNA targets and amplified cDNA, with added consid-eration of dilution factors, given the relative differences incDNA concentration. The sizes of amplification productsand of fragmented and biotin-labeled products preparedfor array analysis, after fragmentation and biotin labeling,were determined by Lab-on-a-Chip analysis in a Bioana-lyzer (Agilent).

The amount and quality of total RNA and amplifica-tion products were determined by use of a NanoDrop®

ND-1000 Spectrophotometer (NanoDrop Technologies)after purification of the amplification products. Amountswere calculated with the following formula: 1 A260 unit �33 �g/mL for DNA and 40 �g/mL for RNA.

Nucleic acids were purified with commercial DNApurification reagent sets such as the NucleoSpin® ExtractKit (Clontech), QIAquick® Purification Kit (Qiagen), orDNA Clean & ConcentratorTM (Zymo Research).

gene expression analysis with GeneChip arraysTargets for GeneChip array gene expression analysis wereprepared with the Ovation Biotin reagent set, the WT-Ovation, or the Affymetrix standard protocol. The frag-mented and biotin-labeled cRNA or cDNA targets werehybridized to either HG-U133A or U133A v2 GeneChiparrays, stained with streptavidin–phycoerythrin with an-tibody amplification, and scanned according to the man-ufacturer’s protocols, except that only 2–2.5 �g of Ribo-SPIA cDNA target per array was hybridized (comparedwith 10 �g of cRNA). Denaturation of the cDNA targetsfor 2 min at 99 °C before hybridization was followed by

1 Nonstandard abbreviations: UHR, universal human reference; GAPDH,glyceraldehyde-3-phosphate dehydrogenase; and Ct, threshold cycle.

1974 Kurn et al.: Novel Isothermal Linear Nucleic Acid Amplification

hybridization for 18–20 h. Array data were analyzed byMAS5 software (Affymetrix).

Ribo-SPIA linearityThe linearity of Ribo-SPIA RNA amplification was dem-onstrated by quantification of transcripts added to acomplex background of total RNA. The in vitro–preparedmixture of 4 transcripts used for assessment of RNAamplification and subsequent analysis by either quantita-tive PCR or array-based gene expression analysis wasfrom Affymetrix (GeneChip Eukaryotic Poly-A RNA Con-trol Kit). Although not intended for this use, dilution ofthe transcript mixture into a known amount of total RNA(5 ng of HeLa total RNA in our study) provides a systemfor the assessment of the linearity and sensitivity of theamplification process. The experimental design was asfollows: The transcript mixture was added to a HeLa totalRNA sample (5 ng) at various dilutions of the initialmixture, and the number of molecules of each of the 4added transcripts, as well as its relationship to the totalmRNA background, was calculated. The number of eachof the added transcripts was calculated from the givenconcentration. The initial dilution of the mixture was1:2000 into the RNA background sample and was furtherdiluted to cover an input range of 106 (see Figs. 3 and 4 inthe Data Supplement that accompanies the online versionof this article at http://www.clinchem.org/content/vol51/issue10/). The lowest number of input transcriptmolecules that could be well detected and quantified afteramplification determined the detection limit of the ampli-fication system. As few as 10 transcript molecules in abackground of 108 sample mRNA molecules are shown tobe amplified linearly by the Ribo-SPIA system. PCRprimers were designed for quantification of each of theadded transcripts in the various HeLa RNA samples. Theamplified cDNA of each of the transcripts in the totalamplification products was quantified by quantitativePCR (Ct; SYBR Green I). Triplicate independent reactionswere carried out for each of the samples on the OvationBiotin System.

Results and DiscussionA schematic presentation of the Ribo-SPIA RNA amplifi-cation method is shown in Fig. 1. The novel linear andisothermal amplification method comprises steps forcDNA synthesis followed by SPIA-based DNA amplifica-tion to generate single-stranded cDNA amplificationproduct.

The initial steps of Ribo-SPIA RNA amplification aredirected toward the formation of a unique double-stranded cDNA that is a substrate for subsequent SPIAamplification, which generates multiple copies of single-stranded DNA products complementary to the samplemRNA. First-strand cDNA synthesis is carried out with achimeric primer comprising a 3� DNA portion that canhybridize to the RNA to be amplified and a 5� RNAportion. The 5� RNA portion is a unique sequence that is

not complementary to the sample RNA and does nothybridize to it. The 3� DNA portion of the first-strandchimeric primer can be designed to vary in length andcomposition to accommodate the desired priming speci-ficity. Thus, this portion may be longer when primingfrom the polyA tail or shorter when designed to primerandomly along the full length of the transcripts (forexample, random hexamer). The 5�-RNA portion se-quence is designed to accommodate the full length of thechimeric amplification primer. Reverse transcriptase isused to generate first-strand cDNA by extension of thepartially hybridized chimeric primer along the RNA tem-plate. The RNA template is then partially degraded in aheating step that also serves to denature the reversetranscriptase. DNA polymerase is added to the reactionmixture to carry out second-strand cDNA synthesis alongthe first-strand cDNA product of the first step. RNA-dependent DNA polymerase activity elongates the prod-uct along the RNA portion of the chimeric primer, form-ing a double-stranded cDNA with a unique RNA/DNAheteroduplex at one end. This unique product serves as asubstrate for the subsequent SPIA DNA amplificationstep.

The amplification step is initiated by the addition of areaction mixture containing a chimeric primer, a DNApolymerase with strong strand-displacement activity andRNase H. The RNase H cleaves the RNA portion of theheteroduplex at one end of the double-stranded cDNA,thus generating a unique partial duplex cDNA with a

Fig. 1. Schematic representation of the 3�-initiated Ribo-SPIA process.The WT-Ribo-SPIA is similarly carried out except for the first strand-synthesisstep, which is carried out with chimeric primer with a randomized 3� DNA portionto effect random priming across the full length of the transcript. RT, reversetranscriptase.

Clinical Chemistry 51, No. 10, 2005 1975

single-stranded DNA tail at the 3� end of the second-strand cDNA. This tail is the priming site for the SPIAamplification step. The sequence of the SPIA amplifica-tion primer, a chimeric DNA/RNA primer, is comple-mentary to the sequence of the single-stranded 3� end ofthe second-strand cDNA in the partial duplex. ThisDNA/RNA chimeric primer is composed of a DNAsequence at the 3� end and an RNA sequence at the 5� end.DNA amplification is carried out by extension of thisprimer, when hybridized to target DNA, by a DNApolymerase with strand-displacement activity and bycleavage of the RNA portion of the primer in the RNA/DNA heteroduplex created by primer hybridization to thetarget by RNase H. Cleavage of the 5� RNA portion of theprimer annealed at the priming site clears this site forhybridization of a new primer molecule, which is ex-tended along the template DNA by DNA polymerase.Strand-displacement DNA synthesis leads to displace-ment of the previous primer extension product away fromthe template DNA. This cycle of primer binding, exten-sion, displacement, and cleavage causes efficient genera-tion of multiple copies of the amplification product. Theamplification step is carried out at a constant temperature(between 47 and 50 °C). The peak of size distribution(Bioanalyzer) of the single-stranded cDNA amplificationproducts, generated under the reaction conditions usedfor the 3�-initiated Ribo-SPIA procedure, was between 500and 1000 nucleotides (as shown in Fig. 4). Of note, theinitiation of primer extension to generate single-strandedDNA amplification product is not dependent on thecompletion of a previous primer extension step. Thus,multiple primer extension products are generated bymultiple DNA polymerase molecules moving along thesame template DNA. This process is rapid, efficient,isothermal, and linear. In addition, cleavage of the 5�portion of the primer extension product (the amplificationproduct) by RNase H renders it nonamplifiable in thegiven amplification system, thus eliminating containmentrequirements, and reduces potential deleterious effects ofamplification product contaminating naı̈ve samples, acommon cause of false-positive results in tests conductedwith PCR amplification systems.

Global RNA amplification using the Ribo-SPIA methodcan be initiated either from consensus sequences or ran-domly across all transcripts. The 3� polyadenylated tail iscommonly used for priming cDNA synthesis by reversetranscription as well as for linear amplification of mRNAby various established procedures. The most commonlyused method for RNA amplification and gene expressionanalysis is that first described by Van Gelder et al. (6 ) andEberwine et al. (7 ) and further optimized by others (8 ). AT7-based method is also used by Affymetrix for samplepreparation for GeneChip arrays (http://www.affymetrix.com/support/technical/technotes/smallv2_technote.pdf).The use of this method for amplification of RNA fromvery small samples requires one or more additionalrounds of T7-RNA polymerase transcription to generate

sufficient amounts of cRNA required for microarray-based analysis. WT-Ribo-SPIA allows comprehensive, lin-ear, global amplification of the total transcriptome in botheukaryotes and prokaryotes and is independent of thepresence or absence of the 3� polyA tail. To ensurewhole-transcript amplification without loss of the 3� end,which is likely to occur when using random primers forinitiation of first-strand cDNA synthesis, we combinedchimeric primers comprising polyT with those compris-ing random hexamer DNA portions for priming of first-strand cDNA synthesis. This is particularly important forthe successful eukaryotic whole-transcript amplification.

characterization of Ribo-SPIA productsThe continuous nature of the Ribo-SPIA method, both the3�- and randomly initiated RNA amplification methods,makes it possible to generate microgram amounts of DNAfrom �5 ng of total RNA input in �4 h. The generation oftargets for gene expression analysis by high-densityGeneChip arrays requires further fragmentation and la-beling of the amplification products. The Ovation BiotinSystem provides reagents and protocols for fragmentationand biotin labeling of the amplification products. ThecDNA fragmentation and biotin-labeling process requirestwo 30-min incubation steps, and the entire protocol fromtotal RNA to fragmented, biotin-labeled cDNA targetsready for hybridization to GeneChip arrays can be com-pleted in a single day. Alternatively, direct incorporationof Aminoallyl-dUTP to amplified cDNA enables prepara-tion of Cy3/Cy5-labeled cDNA product ready for hybrid-ization to any of a variety of spotted arrays (not de-scribed).

Ribo-SPIA reproducibility, linearity, andreliabilityThe Ribo-SPIA RNA amplification process for bothpolyA- and WT-Ribo-SPIA was assessed by quantificationof nonamplified (products generated after second-strandcDNA synthesis) and amplified cDNA. Quantificationwas by quantitative real-time PCR. TaqMan PCR wasused for quantification of cDNA products generated bythe 3�-initiated amplification, and SYBR Green I real-timePCR was used for quantification of cDNA generated byWT-Ribo-SPIA. The absolute yield of amplification prod-ucts was assessed by spectroscopic measurement of puri-fied products, as described in the Materials and Methods.

The reproducibility of RNA amplification was mea-sured by quantitative PCR and the yield of amplificationproducts by spectrophotometry (see Figs. 1 and 2 in theonline Data Supplement). Quantification of the GAPDHgene product was adopted as a routine measure ofamplification efficiency, to assess successful amplificationacross the length of the transcript, and to determinereproducibility of the amplification system across time,reagent lots, and by various users. Primers and probes forquantitative PCR were designed at 2 locations along the

1976 Kurn et al.: Novel Isothermal Linear Nucleic Acid Amplification

transcript: the 3� end portion (330 bp from the polyA tail)and the 5� portion (1 kb from the polyA tail). Quantifica-tion of amplification products obtained from 88 indepen-dent 3�-Ribo-SPIA amplifications of 20 ng of UniversalHuman RNA (Stratagene) yielded mean (SD) Ct values of15.1 (0.5) for the 3� location of the human GAPDHtranscript and 16.1 (0.5) for the 5� location of the tran-script. Similar reproducibility was achieved with WT-Ribo-SPIA (see Fig. 2 in the online Data Supplement). TheOvation Biotin System (3�-initiated amplification) rou-tinely generates �7 �g of purified amplification productsfrom 5 to 100 ng of total RNA samples (88 independentamplification reactions). Similarly reproducible yield ofamplification products is achieved with the WT-Ribo-SPIA (data not shown).

Ribo-SPIA amplification across 6 orders of magnitudeprovided strong evidence of the linearity and sensitivityof Ribo-SPIA (Fig. 2). The linear correlation factors of thecalculated log10 of the input number of molecules of thespecific transcripts added relative to the obtained Ct valuefor the 4 added transcripts in the various samples (R2)were 0.995, 0.980, 0.975, and 0.998 for Lys, Phe, Thr, andDap, respectively (also see Fig. 3 in the online DataSupplement). An example of the linear correlation plot ofCt values and input number of molecules for Lys, whichrepresents the lowest to highest addition, is shown in Fig.2A (each point represents the mean of the 3 independentamplification reactions).

The accuracy and linearity of the WT-Ribo-SPIA werealso assessed. As shown in Fig. 3, good linear correlation(R2 � 0.85) was obtained for the quantification (Ct values)of amplified and nonamplified cDNA for HeLa total RNAsamples (20 ng input) subjected to WT-Ribo-SPIA ampli-fication. Each of the points in Fig. 3 represents the mean of4 independent reactions. A total of 27 gene products werequantified in each of the samples (before and after theamplification step) by use of PCR primers designed forquantification at different distances from the 3� end of thespecific transcripts. It is important to note here that thisgroup of genes consisted of several housekeeping genesand other transcripts, most of which are in the very lowtranscript abundance range. This group of genes (seeTable 1 in the online Data Supplement) was selected forevaluation of the system in the low expression range oftranscripts, where gene expression analysis is of particu-lar interest. These results demonstrate the linearity ofamplification across the transcript length and a widerange of abundance (�1000-fold of low and mediumtranscript abundance, as determined by quantification ofnonamplified cDNA). Similar results (data not shown)were obtained for the Ovation RNA amplification system,with the exception that the amplification products gener-ated by the 3�-initiated Ribo-SPIA method represent the 3�portion of the transcriptome (limited to �1.2 kb from the3� end).

Ribo-SPIA for gene expression analysis onAffymetrix high-density GeneChip arraysUnlike the T7-based RNA amplification systems (IVT)that generate cRNA products, Ribo-SPIA generates single-stranded DNA products. The methods commonly usedfor the fragmentation of cRNA, as required for efficienthybridization and gene expression analysis by use ofoligonucleotide microarrays, are not suitable for cDNAproducts. In the Ribo-SPIA, fragmentation and end label-ing of cDNA amplification products were accomplishedby two 30-min, simple reagent addition and incubationsteps, which involve enzymatic and chemical reactions.The electrophoretic mobility profile (or size distribution)

Fig. 2. Linearity and sensitivity of the Ribo-SPIA process.(A), quantification of cDNA products from amplification of total RNA samples (5ng of HeLa total RNA) to which mixtures of 4 in vitro–prepared RNA transcriptswere added to yield samples with transcript input copy numbers covering 10orders of magnitude. Amplified cDNA of Lys in vitro–prepared RNA transcript,which represents the lowest of the 4 transcripts tested, was quantified byreal-time PCR (SYBR Green I). (B), targets prepared by the Ovation Biotin Systemfrom duplicate independent reactions were hybridized to U133A v2 GeneChiparrays, and array signals were calculated by the MAS5 software package.

Clinical Chemistry 51, No. 10, 2005 1977

of amplified products before and after fragmentation andbiotin labeling, analyzed on a Bioanalyzer (Agilent), areshown in Fig. 4. Targets prepared by use of the OvationBiotin System are suitable for use with high-densityGeneChip arrays under the manufacturer-recommendedconditions for hybridization; however, this approach uses2–2.5 �g of cDNA target per array compared with 15 �gof cRNA target (Affymetrix protocol).

Targets prepared by Ribo-SPIA amplification of small

samples of total RNA and the NuGEN fragmentation andbiotin-labeling system provide highly reproducible geneexpression analysis with the Affymetrix GeneChip array.Reproducible performance of the Ovation-generated tar-gets on human U133A GeneChip arrays, across a range ofinput total RNA (1–100 ng of total UHR RNA), wasdemonstrated. The correlations of signals generated onreplicate arrays hybridized with independently preparedtargets from samples containing 1–100 ng of total RNA(UHR), as well as call concordance (generated by theAffymetrix software package), are shown in Table 1.Signal correlation (R2) and call concordance were wellmaintained across the 5- to 100-ng total RNA input range.Targets generated from samples containing 1 ng of inputtotal RNA yielded somewhat lower detection of presentcalls (expressed as the percentage of genes detected aspresent) and lower signal correlation coefficients. Thesignal correlation coefficient for arrays hybridized withtargets generated from 1-ng samples compared withtarget prepared from 100-ng samples was 0.93 comparedwith 0.97 for arrays hybridized with targets generatedfrom 1 ng of input each. Similar results were observed forcall concordance. These results indicate the high repro-ducibility and array performance of the Ovation BiotinSystem, which allows single-round target preparationfrom very small input RNA samples. The simple andrapid target preparation system is easily automated andsuitable for high-throughput applications such as in clin-ical research.

The linearity and accuracy of gene expression profilingby the Ribo-SPIA method for amplification of small totalRNA samples and analysis on GeneChip arrays wasfurther demonstrated with samples to which in vitro–prepared transcripts were added, as described above. TheGeneChip arrays contained probe sets for specific hybrid-ization of targets generated with each of the in vitro–prepared transcripts and allowed interrogation at 3 loca-tions along the in vitro–prepared transcripts (3�, middle,and 5�). This array design further enabled the assessmentof amplification linearity and amplification productlength (�1.2 kb). The correlation of array signals andtranscript input into the sample (log10 input copy num-

Fig. 3. Reliability of WT-Ribo-SPIA as determined by quantification ofcDNA products (20 ng of HeLa total RNA).The quantification of a set of amplified and nonamplified cDNA productsrepresenting 27 gene products quantified at different locations along thetranscript length was carried out by real-time PCR (SYBR Green I; Ct values). Eachdata point is the mean of results from duplicate Ribo-SPIA reactions.

Fig. 4. Bioanalyzer traces for Ribo-SPIA cDNA prod-ucts before and after fragmentation and biotin la-beling.The internal standard peaks are at 0.2, 0.5, 1, 2, 4, and 6kb (plus an added oligonucleotide at 19 s).

1978 Kurn et al.: Novel Isothermal Linear Nucleic Acid Amplification

ber) for the Lys transcript, which includes the lowestinput number of added molecules, is shown in Fig. 2B.Each point represents the mean signals from two U133Av2 GeneChip arrays. Signal linearity of the amplifiedtranscripts over a �1000-fold range of transcript copynumber input was demonstrated for all 4 transcripts (seeFig. 4 in the online Data Supplement) and reflects theexpected lower detection limit of the platform comparedwith quantitative PCR (shown above). The signals gener-ated across the length of the individual transcript prod-ucts, as indicated by the 3�-, middle-, and 5�-located probesets, were almost identical for the 4 transcripts, indicatingthe high accuracy of amplification across the length of thetranscripts and the compatibility of the Ovation-gener-ated targets from low total RNA input samples with geneexpression profiling on GeneChip arrays.

Gene expression profiles obtained with targets pre-pared with the Ovation Biotin System and HG-U133AGeneChip arrays were comparable to those obtained fromthe same RNA samples prepared by the Affymetrixstandard target protocol. Three independent IVT targetpreparation reactions of UHR total RNA (Stratagene) runby an independent laboratory starting with 10 �g of totalRNA per amplification gave a present call of 51.5% �1.7%. A total of 13 Ribo-SPIA amplifications with 20 ng oftotal RNA input per amplification gave a present call of53.6% � 2.7%. Call concordance between the 2 methodswas 86%. Thus, the 2 methods detect comparable num-bers of transcripts.

An important advantage offered by the Ovation RNAamplification and target preparation systems is the gen-eration of single-stranded cDNA products that provide ahighly specific hybridization target for oligonucleotidearrays in general and GeneChip arrays in particular. Thehigher specificity afforded by DNA/DNA hybridizationcompared with cRNA/DNA heteroduplex formation hasbeen suggested previously (9–11). This was further veri-

fied by analysis of GeneChip array data for the OvationBiotin System–generated targets (3 ). The high sensitivityand specificity afforded by this single-round amplifica-tion method, compared with the standard protocol(higher input of total RNA) and other commerciallyavailable 2-round small-sample T7-based protocols (20 ngtotal RNA input), were reported recently (12 ). It should benoted that highly accurate gene expression profiling fromnanogram amounts of input total RNA by optimizedprotocols for 2 rounds of T7-based amplification wasachieved (13 ).

The performance of targets generated by WT-Ribo-SPIA and the NuGEN fragmentation and biotin-labelingreagents and protocols, on U133A v2 GeneChip arrays, issummarized in Table 2. Targets were prepared from HeLatotal RNA samples (20 ng) and applied to the GeneChiparrays by the same procedure as target prepared with theOvation Biotin System. As shown in Table 2, array per-formance and signal reproducibility (R2) were very simi-lar for the 2 Ribo-SPIA RNA amplification and targetpreparation methods, as would be expected given thedesign of the array (biased toward the 3� portion of thetranscripts). However, we expect that the WT-Ribo-SPIARNA amplification system will be more suitable for geneexpression analysis of the full length of the transcripts(whole-transcript amplification) and provide a new toolfor transcriptome analysis from small samples, thus en-abling analysis of the full transcriptome repertoire, en-compassing all splice variants.

differential gene expression analysis using theRibo-SPIA rna amplification methodDifferential gene expression analysis is a criterion mostrelevant for most applications in which global changes intranscriptome composition are analyzed. Accurate deter-mination of differential gene expression in small samplesrequires highly reproducible linear RNA amplification orsample preparation. These requirements are particularly

Table 1. Signal correlation coefficients (R2) and callconcordance (%) between independent

Ribo-SPIA–generated cDNA products hybridized onHG–U133A GeneChip arrays, as a function of total RNA

input (UHR).Signal correlation, R2

1 ng 5 ng 20 ng 100 ng

1 ng 0.985 ng 0.98 0.99

20 ng 0.96 0.98 0.99100 ng 0.91 0.94 0.96 0.98

Call concordance, %

1 ng 5 ng 20 ng 100 ng

1 ng 87.35 ng 87.4 89.8

20 ng 85.9 88.7 90.9100 ng 85.8 88.6 89.5 89.3

Table 2. Array performance for duplicate independentlyprepared targets hybridized to U133A v2 GeneChip Arrays.a

Array performance

SFb Bkgd %P3�/5�

GAPDH3�/5�Actin

R2 forsignal

Ovation biotinsystem

1 44 58.2 1.39 9.4 92.5

0.87 50 59.3 1.64 10.5

WT-Ovation biotinsystem

1.33 45 61.4 0.96 6.7 91.5

1.48 82 57.3 0.89 7.58a Targets were prepared from HeLa total RNA samples (20 ng) with the 3� or

WT-Ribo-SPIA.b SF, scale factor; Bkgd, background signal; %P, percentage of genes detected

as present; 3�/5�, ratio of signal obtained for probe sets complementary to the3� region of the transcript vs signals obtained for probe sets complementary tothe 5� region of the transcript.

Clinical Chemistry 51, No. 10, 2005 1979

important because studies of transcription regulation areincreasingly conducted in more homogeneous cell popu-lations [such as laser capture microdissection (14 ) orsorted cell samples]. The accuracy and reproducibility ofdifferential gene expression analysis with samples ampli-fied with Ribo-SPIA was assessed by either quantificationof products by real-time PCR or high-density GeneChiparrays.

The linearity of the Ovation RNA amplification systemwas evaluated by determination of differential gene ex-pression between total RNA samples, UHR (referencehuman RNA), and human skeletal muscle (each at 20 ngof total RNA). The amounts of 40 gene products (see Table2 in the online Data Supplement) were determined byTaqMan PCR in the nonamplified and amplified cDNA.Differential expression between the 2 RNA samples wasdetermined (expressed as the difference in Ct). The linearrelationship for log2 of relative expression before or afteramplification (expressed as difference in Ct) over a rangeof nearly 20 Ct, or 6 orders of magnitude, was excellentwith a correlation coefficient (R2) of 0.97 (Fig. 5). Thus,Ribo-SPIA provides a reliable representation of changesin transcript abundance in nonamplified mRNA.

We also evaluated differential gene expression deter-mined with the Ovation Biotin system and GeneChiparrays. The results obtained from replicate independentsamples (human liver and UHR at 20 ng of total RNAinput each) prepared with the Ovation Biotin Systemamplification and target systems were compared withthose obtained by the standard T7-based protocol usinghigher input of the same total RNA samples. The corre-lation coefficients (R2) obtained were 0.94 for the repro-ducibility of the Ovation Biotin System and the standard

T7 protocol and 0.83 for the comparison between the 2methods. These results demonstrate the accuracy of theOvation RNA amplification system relative to the com-monly used target preparation method, which is carriedout with �100-fold more total RNA per sample (standardprotocol).

ConclusionsThe Ribo-SPIA isothermal linear RNA amplification sys-tems are highly reproducible and sensitive (5–100 ng oftotal RNA input), fast, and simple to operate. Both whole-transcript amplification and 3�-initiated RNA amplifica-tion reproducibly generate high-quality cDNA amplifica-tion products suitable for transcriptome analysis on high-density oligonucleotide arrays (GeneChip arrays) or forreal-time quantitative PCR (TaqMan or SYBR Green I).These systems provide a means for global gene expressionanalysis from very small samples and should be particu-larly useful for clinical research in which supplies ofbiological samples are limited. In addition, this protocolmakes these methods suitable for RNA amplification ofall samples irrespective of RNA amount.

Comprehensive amplification of sample transcriptomefor the generation of targets suitable for large-scale geneexpression analysis has been achieved. The Ribo-SPIAsystem is highly sensitive while maintaining reliableamplification and linearity, and it enables gene expressionprofiling of minute amounts (5–100 ng) of total RNA. Thesimple “add and incubate” protocol makes the methodsuitable for automation, which is of great advantage forhigh-throughput applications. Most importantly, the useof a single method for all samples streamlines acquisitionof gene expression data across all samples.

We acknowledge the contributions of the following scien-tists to the development of the new amplification systems:Dr. Alan Dafforn, Dr. Glenn Deng, Dawn Iglehart, Dr.Susan Lato, Susheela Pillarisetty, Reshma Purohit, Kori-tala Sriveda, and Dr. Martin Wang. We thank Dr. AndrewBrooks for helpful discussions.

References1. Dafforn A, Chen P, Deng G, Herrler M, Iglehart D, Koritala S, et al.

Linear mRNA amplification from as little as 5 ng total RNA forglobal gene expression analysis. Biotechniques 2004;37:854–7.

2. Singh R, Maganti R, Jabba SV, Wang M, Deng G, Heath JD, et al.Microarray-based comparison of three amplification methods fornanogram amounts of total RNA. Am J Physiol Cell Physiol2005;288:1179–89.

3. Barker CS, Griffin C, Dolganov GM, Hanspers K, Yang JYH, et al.Increased DNA microarray hybridization specificity using sscDNAtargets BMC Genomics 2005, 6:57 (accessed April 22, 2005).

4. Potash MJ, Chao W, Bentsman G, Paris N, Saini M, Nitkiewicz J, etal. A mouse model for study of systemic HIV-1 infection, antiviralimmune responses, and neuroinvasiveness. Proc Natl Acad SciU S A 2005;102:3760–5.

5. Whitworth KM, Agcar C, Kim J-G, Paterl RV, Springer GK, Bivens N,et al. Transcriptional profiling of pig embryogenesis by using a 15K

Fig. 5. Correlation between differential expression in skeletal musclevs UHR total RNA measured by quantification of nonamplified andamplified cDNA products of the 3�-initiated Ribo-SPIA amplification.Log2 differential expression is represented as the difference between Ct valuesmeasured by TaqMan QPCR for each gene. Each data point is the mean of resultsfrom duplicate Ribo-SPIA reactions.

1980 Kurn et al.: Novel Isothermal Linear Nucleic Acid Amplification

member unigene set specific for pig reproductive tissues andembryos. Biol Reprod 2005;72:1437–51.

6. Van Gelder RN, von Zastrow ME, Yool A, Dement WC, Barchas JD,Eberwine JH. Amplified RNA synthesized from limited quantities ofheterogeneous cDNA. Proc Natl Acad Sci U S A 1990;87:1663–7.

7. Eberwine J, Yeh H, Miyashiro K, Cao Y, Nair S, Finnell R, et al.Analysis of gene expression in single live neurons. Proc Natl AcadSci U S A 1992;89:3010–4.

8. Baugh LR, Hill AA, Brown EL, Hunter CP. Quantitative analysis ofmRNA amplification by in vitro transcription. Nucleic Acids Res2001;29:e29.

9. Casey J, Davidson N. Rates of formation and thermal stabilities ofRNA:DNA and DNA:DNA duplexes at high concentrations of form-amide. Nucleic Acids Res 1977;4:1539–52.

10. Sugimoto N, Nakano M, Nakano S-I. Thermodynamics—structure

relationship of single mismatches in RNA/DNA duplexes. Bio-chemistry 2000;39:11270–81.

11. Rosenow C, Saxena RM, Durst M, Gingeras TR. Prokaryotic RNApreparation methods useful for high density array analysis: com-parison of two approaches. Nucleic Acids Res 2001;29:e112.

12. Viale A, Tiesman J, Brooks A, Hester S, Massimi A, Griffin C, et al.;MARG Research Study Group. Evaluation of small sample nucleicacid amplification technologies for gene expression profiling.http://www.abrf.org/ResearchGroups/Microarray/EPosters/MARG_Research_Study_2005.pdf (accessed August 9, 2005).

13. Kenzlmann M, Klaren R, Hergenhahn M, Bonrouhi M, Grone HJ,Schmid W, et al. High-accuracy amplification of nanogram totalRNA amounts for gene profiling. Genomics 2004;83:550–8.

14. Emmert-Buck MR, Bonner RF, Smith PD, Chuaqui RF, Zhuang Z,Goldstein SR, et al. Laser capture microdissection. Science1996;27:998–1001.

Clinical Chemistry 51, No. 10, 2005 1981