ArticleNo.jmbi.1999.3063availableonlineathttp://www.idealibrary.comon
`
`J. Mol. Biol. (1999) 292, 251–262
`
`Universal DNA Microarray Method for Multiplex
`Detection of Low Abundance Point Mutations
`
`2,JosephDay 1,
`NormanP.Gerry 1,NancyE.Witowski
`RobertP.Hammer 3,GeorgeBarany 2 andFrancisBarany 1*
`
`1Department of Microbiology
`Hearst Microbiology Research
`Center, and Strang Cancer
`Prevention Center, Joan and
`Sanford I. Weill Medical
`College of Cornell University
`1300 York Ave., Box 62, New
`York, NY 10021, USA
`
`2Departments of Chemistry and
`Laboratory Medicine &
`Pathology, University of
`Minnesota, 207 Pleasant Street
`S.E., Minneapolis, MN
`55455, USA
`
`3Department of Chemistry
`Louisiana State University
`232 Choppin Hall, Baton Rouge
`LA 70803, USA
`
`Cancers arise from the accumulation of multiple mutations in genes regu-
`lating cellular growth and differentiation. Identification of such mutations
`in numerous genes represents a significant challenge in genetic analysis,
`particularly when the majority of DNA in a tumor sample is from wild-
`type stroma. To overcome these difficulties, we have developed a new
`type of DNA microchip that combines polymerase chain reaction/ligase
`detection reaction (PCR/LDR) with ‘‘zip-code’’ hybridization. Suitably
`designed allele-specific LDR primers become covalently ligated to adja-
`cent fluorescently labeled primers if and only if a mutation is present.
`The allele-specific LDR primers contain on their 50-ends ‘‘zip-code com-
`plements’’ that are used to direct LDR products to specific zip-code
`addresses attached covalently to a three-dimensional gel-matrix array.
`Since zip-codes have no homology to either the target sequence or to
`other sequences in the genome, false signals due to mismatch hybridiz-
`ations are not detected. The zip-code sequences remain constant and
`their complements can be appended to any set of LDR primers, making
`our zip-code arrays universal. Using the K-ras gene as a model system,
`multiplex PCR/LDR followed by hybridization to prototype 3 (cid:2) 3 zip-
`code arrays correctly identified all mutations in tumor and cell line DNA.
`Mutations present at less than one per cent of the wild-type DNA level
`could be distinguished. Universal arrays may be used to rapidly detect
`low abundance mutations in any gene of interest.
`
`# 1999 Academic Press
`
`*Corresponding author
`
`Keywords: zip-code addressing; DNA hybridization; thermostable DNA
`ligase; ligase detection reaction; single nucleotide polymorphism (SNP)
`
`Introduction
`
`of
`accumulation
`from the
`arise
`Cancers
`mutations in genes controling the cell cycle, apop-
`tosis, and genome integrity. These mutations may
`be inherited or somatic, arising from exposure to
`environmental
`factors or from malfunctions in
`DNAreplicationandrepairmachinery(Fearon,
`1997;Fearon&Vogelstein,1990;Liuetal.,1996;
`Perera,1997).Oncogenesmaybeactivatedby
`point mutations, translocation, or gene amplifica-
`tion, while tumor suppressor genes may be inacti-
`vated by point mutations,
`frameshift mutations
`
`Abbreviations used: LDR, ligase detection reaction;
`FAM, 6-carboxyfluorescein; Mes, 2-(N-morpholino)
`ethanesulfonic acid; SNP, single nucleotide
`polymorphism.
`E-mailaddressofcorrespondingauthor:
`barany@mail.med.cornell.edu
`
`anddeletions(Bishop,1991;DaCostaetal.,1996;
`Venitt,1996).Amajorhurdletodetecting
`in primary
`mutations in these genes is that,
`tumors, normal stromal cell contamination can be
`as high as 70 % of total cells, and thus a mutation
`present in only one of the two chromosomes of a
`tumor cell may represent as little as 15 % of the
`DNA sequence present in a sample for that gene.
`Thus, there is an urgent need to develop technol-
`ogy that can identify accurately one or more low
`abundance mutations, at multiple adjacent, nearby,
`and distal loci in a large number of genes.
`The advent of DNA arrays has resulted in a
`paradigm shift
`in detecting sequence variations
`and monitoring gene expression levels on a geno-
`micscale(Beattieetal.,1995;Brown&Botstein,
`1999;Cheeetal.,1996;Croninetal.,1996;DeRisi
`etal.,1996;Drobyshevetal.,1997;Eggersetal.,
`1994;Gundersonetal.,1998;Guoetal.,1994;
`Hacia,1999;Haciaetal.,1996;Kozaletal.,1996;
`
`0022-2836/99/370251–12 $30.00/0
`
`# 1999 Academic Press
`
`Personalis EX2028
`
`

`

`252
`
`Universal Array for Multiplex Mutation Detection
`
`Peaseetal.,1994;Schenaetal.,1996;Shalonetal.,
`1996;Southernetal.,1999;Yershovetal.,1996;Zhu
`etal.,1998).DNAchipsdesignedtodistinguish
`single nucleotide differences are generally based
`onhybridizationoflabeledtargets(Beattieetal.,
`1995;Cheeetal.,1996;Croninetal.,1996;
`Drobyshevetal.,1997;Eggersetal.,1994;Guoetal.,
`1994;Haciaetal.,1996;Kozaletal.,1996;Parinov
`etal.,1996;Sapolskyetal.,1999;Wangetal.,1998;
`Yershovetal.,1996)orpolymeraseextensionof
`arrayedprimers(Lockleyetal.,1997;Nikiforov
`etal.,1994;Pastinenetal.,1997;Shumakeretal.,
`1996).WhileDNAchipsbasedonthesetwo
`formats can confirm a known sequence, the simi-
`larities in hybridization profiles create ambiguities
`in distinguishing heterozygous from homozygous
`alleles(Beattieetal.,1995;Cheeetal.,1996;Eggers
`etal.,1994;Kozaletal.,1996;Southern,1996;Wang
`etal.,1998).Toovercomethisproblem,several
`methods have been proposed, including the use of:
`(i)two-colorfluorescenceanalysis(Haciaetal.,
`1996,1998a);(ii)atilingstrategythatuses40over-
`lapping addresses for each known polymorphism
`(Croninetal.,1996);(iii)incorporationofnucleo-
`tideanaloguesinthearraysequence(Guoetal.,
`1997;Haciaetal.,1998b);and(iv)adjacentco-
`hybridizedoligonucleotides(Drobyshevetal.,
`1997;Gentalen&Chee,1999;Yershovetal.,1996).
`A recent side-by-side comparison revealed that the
`use of hybridization chips for nucleotide discrimi-
`nation gave an order of magnitude higher back-
`ground than was observed with the primer
`extension approach, resulting in an increased likeli-
`hoodoffalsepositiveidentifications(Pastinenetal.,
`1997).Nevertheless,solid-phaseprimerextension
`can also generate false positive signals from mono-
`nucleotide repeat sequences,
`template-dependent
`errors,andtemplate-independenterrors(Nikiforov
`etal.,1994;Shumakeretal.,1996).Inaddition,
`neither of these two types of arrays can detect
`cancer mutations when these are present
`in a
`minority of the total target DNA.
`Over the past few years, our laboratories have
`pursued an alternate strategy in DNA array
`design. In concert with polymerase chain reaction/
`ligase detection reaction (PCR/LDR) assays carried
`outinsolution(Barany,1991a,b;Belgraderetal.,
`1996;Dayetal.,1995,1996;Khannaetal.,1999),
`our array concept allows for accurate identification
`of mutations and single nucleotide polymorphisms
`(SNPs). Primary PCR amplification of the gene of
`interest is followed by LDR, which uses a thermo-
`stable Tth DNA ligase that
`links two adjacent
`oligonucleotides annealed to a complementary tar-
`get if and only if the nucleotides are perfectly base-
`pairedatthejunction(Figure1(a)).Sinceasingle-
`base mismatch prevents ligation, it is possible to
`distinguish mutations with exquisite specificity,
`evenatlowabundance(Khannaetal.,1999).Fur-
`thermore, such assays are ideal for multiplexing,
`since several primer sets can ligate along a gene
`without the interference encountered in polymer-
`ase-basedassays(Belgraderetal.,1996;Dayetal.,
`
`1995;Khannaetal.,1999).High-throughputdetec-
`tion of specific multiplexed LDR products is then
`achieved via divergent sequences termed ‘‘zip-
`code’’ complements which guide each LDR pro-
`duct to a designated zip-code address on a DNA
`array(Figure1(b)).Thisconceptisanalogousto
`molecular tags developed for bacterial and yeast
`genetics(Henseletal.,1995;Shoemakeretal.,
`1996).BasedonrecentmultiplexedPCR/LDR
`results from our laboratory,
`the new approach
`should allow detection of: (i) dozens to hundreds
`of polymorphisms in a single-tube multiplex for-
`mat; (ii) small insertions and deletions in repeat
`sequences; and (iii) low abundance mutations in a
`backgroundofnormalDNA(Khannaetal.,1999,
`and unpublished results).
`
`Results and Discussion
`
`Zip-code concept and design
`
`Our approach uses microarrays of unique 24-
`base oligonucleotides that are coupled to a three-
`dimensional polymer at known locations. These
`24-mersorzip-codes(Table1)hybridizespecifi-
`cally to molecules containing sequences that are
`complementary to the zip-codes. By linking the
`zip-code complements to fluorescent primers via a
`tandem PCR/LDR strategy, zip-code microarrays
`can be used to assess the presence and abundance
`of mutations in biological specimens. Importantly,
`because the zip-codes represent unique artificial
`sequences, zip-code microarrays can be used as a
`universal platform for molecular
`recognition
`simply by changing the gene-specific sequences
`linked to the zip-code complements.
`Each zip-code sequence is composed of six tetra-
`mers (designed as described below) such that the
`full-length 24-mers have similar tm values. The 256
`(44) possible combinations in which the four bases
`can be arranged as tetramers were reduced to a set
`of 36; these were chosen such that each tetramer
`differed from all others by at
`least
`two bases
`(Figure2).Tetramercomplements,aswellastetra-
`mers that would result in self-pairing or hairpin
`formation of the zip-codes, were eliminated. Fur-
`thermore,
`tetramers that were palindromic, e.g.
`TCGA, or repetitive, e.g. CACA, were excluded
`(diagonallyhatchedboxesinFigure2).Theindi-
`cated set of 36 tetramers represents just one of the
`possible sets that can be created; alternative sets
`can be developed by starting in any of the unused
`lightgrayboxes(Figure2).
`Six tetramers were chosen from the larger set of
`36 for use in designing the zip-codes for the proto-
`type array. These six tetramers were combined
`such that each zip-code differs from all others by
`atleastthreealternatingtetramerunits(Table1).
`This ensures that each zip-code differs from all
`other zip-codes by at least six bases, thus prevent-
`ing even the closest zip-code sequences from cross-
`hybridizing. The tm values of correct hybridizations
`range from 70 (cid:14)C to 82 (cid:14)C and are at least 24 deg. C
`
`Personalis EX2028
`
`

`

`Universal Array for Multiplex Mutation Detection
`
`253
`
`Figure 1. Scheme for PCR/LDR detection of mutations using an addressable array. (a) Schematic representation of
`LDR primers used to distinguish mutations. Each allele-specific primer contains an addressable sequence complement
`(cZ1 or cZ3) on the 50-end and the discriminating base on the 30-end. The common LDR primer is phosphorylated on
`the 50-end and contains a fluorescent label on the 30-end. The primers hybridize adjacent to each other on target
`DNA, and the nick will be sealed by the ligase if and only if there is perfect complementarity at the junction. (b) The
`presence and type of mutation is determined by hybridizing the contents of an LDR to an addressable DNA array.
`The zip-code sequences are designed to be sufficiently different, so that only primers containing the correct
`complement to a given zip-code will remain bound at that address. (c) Schematic representation of chromosomal
`DNA containing the K-ras gene. Exons are shaded and the positions of codons 12 and 13 are shown. Exon-specific
`primers were used to selectively amplify K-ras DNA flanking codons 12 and 13. Primers were designed for LDR
`detection of seven possible mutations in these two codons as described in (a).
`
`higher than that of any incorrect hybridization
`(calculated using Oligo 6.0, Molecular Biology
`Insights, Inc., Cascade, CO). The concept of using
`
`alternating rows and columns of tetramer units
`may be extended to include all 36 tetramers, hence
`creating an array with 1296 divergent addresses.
`
`Personalis EX2028
`
`

`

`254
`
`Universal Array for Multiplex Mutation Detection
`
`Table 1. Zip-code sequences used in prototype array
`
`Zip#Tetramerorder
`
`a
`
`Zip-codesequence(5 0 !3 0)b
`
`Zip11-6-3-2-6-3TGCG-ACCT-CAGC-ATCG-ACCT-CAGC-spacer-NH
`2
`Zip3
`3-6-5-2-2-3
`CAGC-ACCT-GACC-ATCG-ATCG-CAGC-spacer-NH2
`Zip5
`5-6-1-2-4-3
`GACC-ACCT-TGCG-ATCG-GGTA-CAGC-spacer-NH2
`Zip11
`1-4-3-6-6-1
`TGCG-GGTA-CAGC-ACCT-ACCT-TGCG-spacer-NH2
`Zip13
`3-4-5-6-2-1
`CAGC-GGTA-GACC-ACCT-ATCG-TGCG-spacer-NH2
`Zip15
`5-4-1-6-4-1
`GACC-GGTA-TGCG-ACCT-GGTA-TGCG-spacer-NH2
`Zip21
`1-2-3-4-6-5
`TGCG-ATCG-CAGC-GGTA-ACCT-GACC-spacer-NH2
`Zip23
`3-2-5-4-2-5
`CAGC-ATCG-GACC-GGTA-ATCG-GACC-spacer-NH2
`Zip25
`5-2-1-4-4-5
`GACC-ATCG-TGCG-GGTA-GGTA-GACC-spacer-NH2
`a Order of tetramer oligonucleotide segments in the corresponding zip-code sequence. Six tetramers were
`chosen from the full set of 36 to prepare the zip-codes for the prototype array. The six tetramers which were
`renumbered for ease of use are: 1, TGCG; 2, ATCG; 3, CAGC; 4, GGTA; 5, GACC; and 6, ACCT. Closely related
`sequences, (Zip1, 3, 5), (Zip11, 13, 15) and (Zip21, 23, 25) differ at the first, third, and fifth tetramer positions,
`but are identical at the second, fourth, and sixth tetramer positions.
`b spacer-NH2 (cid:136) -O(PO2)O-(CH2CH2O)6-PO2-O(CH2)3NH2.
`
`Array preparation
`
`Optimization of hybridization conditions
`
`Numerous types of two and three-dimensional
`matrices were examined with respect to: (i) ease of
`preparation of the surface; (ii) oligonucleotide load-
`ing capacity; (iii) stability to conditions required
`for coupling of oligonucleotides, as well as for
`hybridization and washing; and (iv) compatibility
`with fluorescence detection. Our currently favored
`methodology to construct zip-code arrays involves
`initial creation of a lightly crosslinked film of acryl-
`amide/acrylic acid copolymer on a glass solid
`support; subsequently, the free carboxyl groups
`dispersed randomly throughout the polymeric sur-
`face are activated with N-hydroxysuccinimide, and
`amine terminated zip-code oligonucleotide probes
`are added to form covalent amide
`linkages
`(Figure3(a)).Thedescribedcouplingchemistryis
`rapid, straightforward, efficient, and amenable to
`both manual and robotic spotting. Both the acti-
`vated surfaces and the surfaces with attached
`oligonucleotides are stable to long-term storage.
`
`Hybridizations of a fluorescently labeled 70-mer
`probe onto model zip-code arrays were measured
`as a function of buffer, metal cofactors, volume,
`pH,time,andthemechanicsofmixing(Table2).
`Even with closely related zip-codes, cross-hybridiz-
`ation was negligible or non-existent, with a signal-
`to-noise ratio of at least 50:1. Our experiments
`suggest
`that different zip-codes hybridize at
`approximately the same rate, i.e. the level of fluor-
`escent signal is relatively uniform when normal-
`ized for the amount of oligonucleotide coupled per
`address (data not shown). Magnesium ion was
`obligatory to achieve hybridization, and less than
`1 fmol of probe could be detected in the presence
`ofthisdivalentcation(Table2andFigure4).The
`hybridization signal was doubled upon lowering
`the pH from 8.0 to 6.0, most likely due to masking
`of negative charges (hence reducing repulsive
`interactions with oligonucleotides) arising from
`uncoupled acrylic acid groups in the bulk polymer
`
`Table 2. Effect of hybridization conditions on hybridization signal
`
`HybridizationbufferVol.(ml)Mixing
`
`a
`
`Time(minutes)Relativesignal
`
`Buffer A
`Buffer A minus MgCl2
`Buffer A
`Buffer B
`Buffer B
`Buffer B
`Buffer B
`Buffer A (cid:135) Capped Surface
`Buffer B minus MgCl2
`Buffer B
`
`55
`55
`20
`55
`20
`55
`55
`55
`55
`55
`
`Inter.
`Inter.
`Inter.
`Inter.
`Inter.
`Cont.
`Cont.
`Cont.
`Cont.
`Cont.
`
`30
`30
`30
`30
`30
`30
`60
`60
`60
`180
`
`1
`<0.01
`2.5
`2
`3
`4
`8
`8
`<0.01
`10
`
`Following general procedures described in Materials and Methods, hybridizations were carried out with 1 pmol of FAMcZip13-Prd
`and 3 (cid:2) 3 manually spotted arrays. Buffers were: buffer A, 300 mM bicine (pH 8.0), 10 mM MgCl2, 0.1 % SDS; buffer B, 300 mM
`Mes (pH 6.0), 10 mM MgCl2, 0.1 % SDS.
`a Mixing was as follows: intermittent (Inter.), manual mixing of the sample once every ten minutes; continuous (Cont.), mixing of
`sample at 20 rpm in a hybridization oven.
`
`Personalis EX2028
`
`

`

`Universal Array for Multiplex Mutation Detection
`
`255
`
`Figure 2. Design of tetramers for use in zip-code arrays. The checkerboard pattern shows all 256 possible tetramers.
`A given square represents the two bases on the left followed by the two bases on the top of the checkerboard. To be
`included, each tetramer must differ from all others by at least two bases, and be non-complementary. The chosen
`tetramers are shown in the white boxes, while their complements are listed as (number)0. Thus, as an example, the
`complementary sequences GACC (20) and GGTC (200) are mutually exclusive in this scheme. In addition, tetramers
`that are palindromic, e.g. TCGA (off-diagonal hatched boxes) or repetitive, e.g. CACA (hatched boxes on diagonal
`from upper left to lower right) have been eliminated. All other sequences which differ from the 36 tetramers by only
`one base are shaded in light gray. Four potential tetramers were not chosen as they are either all A(cid:1) T or G(cid:1) C bases
`(open boxes).
`
`matrix. To confirm this hypothesis, the free car-
`boxyl groups on arrays to which zip-codes had
`already been attached were capped with ethanol-
`amine
`under
`standard
`coupling
`conditions.
`Hybridizations of the capped arrays at pH 8.0 gave
`results comparable to hybridizations at pH 6.0 of
`the same arrays without capping. Continuous mix-
`ing proved to be crucial for obtaining good hybrid-
`ization, and studies of the time-course led us to
`choose one hour at 65 (cid:14)C as standard. Reducing
`the hybridization volume improved the hybridiz-
`ation signal due to the relative increase in probe
`concentration. Further
`improvements may be
`achieved using specialized small volume hybridiz-
`ation chambers that allow for continuous mixing.
`
`Array hybridization of K-ras LDR products
`
`PCR/LDR amplification coupled with zip-code
`detection on an addressable array was tested with
`the K-ras gene as a model system. Exon-specific
`PCR primers were used to selectively amplify
`
`K-ras DNA flanking codons 12 and 13. LDR
`primers were designed to detect the seven most
`common mutations found in the K-ras gene in
`colorectalcancer(Figure1(c)andTable3).For
`example, the second position in codon 12, GGT,
`coding for glycine, may mutate to GAT, coding for
`aspartate, which is detected by ligation of
`the
`allele-specific primer (containing a zip-code comp-
`lement, cZip3, on its 50-end, and a discriminating
`base, A, on its 30-end) to a fluorescently labeled
`commonprimer(Figure1(c)).
`PCR/LDR was carried out on nine individual
`DNA samples derived from cell lines or paraffin-
`embedded
`tumors
`containing
`known K-ras
`mutations (as described in Materials and Methods).
`An aliquot (2 ml) was taken from each reaction and
`electrophoresed on a sequencing apparatus to con-
`firm that LDR was successful (data not shown).
`Next, the different mutations were distinguished
`by hybridizing the LDR product mixtures on 3 (cid:2) 3
`addressable DNA arrays (each zip-code address
`was spotted in quadruplicate), and detecting the
`
`Personalis EX2028
`
`

`

`256
`
`Universal Array for Multiplex Mutation Detection
`
`Figure 3. Detection of K-ras mutations on a DNA array. (a) Schematic representation of gel-based zip-code array.
`Glass microscope slides treated with g-methacryloyloxypropyltrimethoxysilane are used as the substrate for the
`covalent attachment of an acrylamide/acrylic acid copolymer matrix. Amine-modified zip-code oligonucleotides are
`coupled to N-hydroxysuccinimide-activated surfaces at discrete locations (see Materials and Methods). Each position
`in the 3 (cid:2) 3 grid identifies an individual zip-code address (and corresponding K-ras mutation or wild-type sequence).
`(b) Each robotically spotted array was hybridized with an individual LDR and fluorescent signal detected as
`described in Materials and Methods using a two second exposure time. All nine arrays identified the correct mutant
`and/or wild-type for each tumor (G12S, G12R, and G12C) or cell line sample (Wt, G12D, G12A, G12V, and G13D).
`The small spots seen in some of the panels, e.g. near the center of the panel containing the G13D mutant, are not
`incorrect hybridizations, but noise due to imperfections in the polymer.
`
`positionsoffluorescentspots(Figure3(b)).The
`wild-type samples, Wt(G12) and Wt(G13), each
`displayed four equal hybridization signals at Zip1
`and Zip25, respectively, as expected. The mutant
`samples each displayed hybridization signals cor-
`responding to the mutant, as well as for the wild-
`type DNA present in the cell line or tumor. The
`sole exception to this was the G12V sample, which
`
`was derived from a cell line (SW620) homozygous
`for the G12V K-ras allele. The experiment was
`repeated several times, using both manually and
`robotically spotted arrays, and LDR primers
`labeled with either fluorescein or Texas Red. False-
`positive or false-negative signals were not encoun-
`tered in any of
`these experiments. A minimal
`amount of noise seen on the arrays can be attribu-
`
`Figure 4. Determination of zip-code array capture sensitivity using two different detection instruments. Quadrupli-
`cate hybridizations were carried out on manually spotted arrays as described in Materials and Methods. The graphs
`depict quantification of the amount of captured 70-mer complement using either a fluorimager (left) or an epifluores-
`cence microscope/CCD (right). Each symbol represents hybridizations to an individual array. The filled square on
`each graph is the average of the backgrounds from all four arrays.
`
`Personalis EX2028
`
`

`

`Universal Array for Multiplex Mutation Detection
`
`257
`
`Table 3. Primers designed for K-ras mutation detection by PCR/LDR/array hybridization
`Sequence (50 ! 30)
`
`Primer
`
`K-ras exon 1 forward
`K-ras exon 1 reverse
`
`ATAAGGCCTGCTGAAAATGACTGAA
`CTGCACCAGTAATATGCATATTAAAACAAG
`
`cZip1-K-ras c12.2WtG
`cZip3-K-ras c12.2D
`cZip5-K-ras c12.2A
`cZip11-K-ras c12.2V
`K-ras c12 Com-2
`
`cZip13-K-ras c12.1S
`cZip15-K-ras c12.1R
`cZip21-K-ras c12.1C
`K-ras c12 Com-1
`
`cZip23-K-ras c13.4D
`cZip25-K-ras c13.4WtG
`K-ras c13 Com-4
`
`GCTGAGGTCGATGCTGAGGTCGCAAAACTTGTGGTAGTTGGAGCTGG
`GCTGCGATCGATGGTCAGGTGCTGAAACTTGTGGTAGTTGGAGCTGA
`GCTGTACCCGATCGCAAGGTGGTCAAACTTGTGGTAGTTGGAGCTGC
`CGCAAGGTAGGTGCTGTACCCGCAAAACTTGTGGTAGTTGGAGCTGT
`pTGGCGTAGGCAAGAGTGCCT-fluorescein
`pTGGCGTAGGCAAGAGTGCCT-Texas Red
`CGCACGATAGGTGGTCTACCGCTGATATAAACTTGTGGTAGTTGGAGCTA
`CGCATACCAGGTCGCATACCGGTCATATAAACTTGTGGTAGTTGGAGCTC
`GGTCAGGTTACCGCTGCGATCGCAATATAAACTTGTGGTAGTTGGAGCTT
`pGTGGCGTAGGCAAGAGTGCC-fluorescein
`pGTGGCGTAGGCAAGAGTGCC-Texas Red
`GGTCCGATTACCGGTCCGATGCTGTGTGGTAGTTGGAGCTGGTGA
`GGTCTACCTACCCGCACGATGGTCTGTGGTAGTTGGAGCTGGTGG
`pCGTAGGCAAGAGTGCCTTGAC-fluorescein
`pCGTAGGCAAGAGTGCCTTGAC-Texas Red
`
`The PCR primers were specifically designed to amplify exon 1 of K-ras without co-amplifying N and H-ras.
`The allele-specific LDR primers contained 24-mer zip-code complement sequences on their 50-ends (bold) and
`the discriminating bases on their 30-ends (underlined). The common LDR primers contained 50-phosphate groups
`and either a fluorescein or a Texas Red label on their 30-ends.
`
`ted to dust, scratches, and/or small bubbles in the
`polymer. These flaws are
`readily recognized
`because they are weak and sporadic, rather than
`reproducing the quadruplicate spotting pattern; we
`expect such noise will be minimized with more
`stringent manufacturing conditions. Ultimately,
`these protocols are amenable to quantifying the
`relative amounts of each allele, and work is
`currently in progress to convert our quantitative
`PCR/LDR protocols for K-ras mutations from
`gel-based detection to
`array-based detection
`(unpublished results).
`
`side) on three out of the four arrays; the signal to
`noise was 2:1 on the fourth array. For a given array,
`with fluorescence quantified by either instrument,
`the captured counts varied linearly with the amount
`of labeled FAMcZip13-Prd added. Rehybridization
`of the same probe, at the same concentration, to the
`same array, was reproducible within (cid:6)5 % (data not
`shown). Variations in fluorescent signal between
`arrays may reflect variations in the amount of zip-
`code oligonucleotide coupled, due to the inherent
`inaccuracies of manual spotting and/or variations
`in polymer uniformity.
`
`Array capture sensitivity
`
`After an LDR, the successfully ligated and fluor-
`escently labeled LDR product competes with an
`excess of unligated discriminating primer
`for
`hybridization to the correct zip-code address
`on the array. To determine capture sensitivity,
`DNA arrays were hybridized in quadruplicate,
`under standard conditions, with from 100 amol
`((cid:136) 1/90,000) to 30 ((cid:136) 1/300) fmol of a labeled syn-
`thetic 70-mer, FAMcZip13-Prd (this simulates a full-
`length LDR product; see Materials and Methods for
`the sequence), in the presence of a full set of K-ras
`LDR primers (combined total of 9000 fmol of discri-
`minating and common primers). Array analyses
`withaFluorImager(Figure4,left-side)indicatethat
`a signal-to-noise ratio of greater than 3:1 can be
`achieved when starting with a minimum of 3 fmol
`((cid:136) 1/3,000) of FAMcZip13-Prd-labeled probe in the
`presence of 4500 fmol of FAM-labeled LDR primers
`and 4500 fmol of zip-code complement primers in
`the hybridization solution. Results using micro-
`scope/CCD instrumentation to quantify fluor-
`escence were even more striking: a 3:1 signal-to-
`noise ratio was maintained starting with 1 fmol
`((cid:136)1/9,000)oflabeledproduct(Figure4,right-hand
`
`Detection of low abundance mutations by
`PCR/LDR/array hybridization
`
`To determine the limit of detection of low-level
`mutations in wild-type DNA using PCR/LDR/
`array hybridization, a dilution series was set up
`and analyzed. PCR-amplified pure G12V DNA
`was diluted into wild-type K-ras DNA in ratios
`ranging from 1:20 to 1:500. Duplicate LDRs were
`carried out on 2000 fmol of total DNA, using a
`two-primer set consisting of 2000 fmol each of the
`discriminating and common primers for the G12V
`mutation. It proved possible to quantify a positive
`hybridization signal at a dilution of 1:200 with a
`signal-to-noiseratioof2:1(Figure5).Asignalwas
`distinguishable
`even at a dilution of 1:500,
`although noise levels due to dust or bubbles in the
`polymer prevented us from accurately quantifying
`the results. A control of pure wild-type DNA
`showed no hybridization signal. These results indi-
`cate clearly that zip-code array hybridization,
`when coupled with PCR/LDR, may detect poly-
`morphisms present at less than 1 % of the total
`DNA. These results are consistent with our earlier
`work showing that PCR/LDR, using a 26-primer
`set and analyses based on gel electrophoreses of
`
`Personalis EX2028
`
`

`

`258
`
`Universal Array for Multiplex Mutation Detection
`
`hybridizations at temperatures from 0 (cid:14)C to 44 (cid:14)C.
`The result is increased background noise and false
`signals due to mismatch hybridization and non-
`specific binding, for example, on small insertions
`anddeletionsinrepeatsequences(Croninetal.,
`1996;Haciaetal.,1996;Southern,1996;Wangetal.,
`1998).Incontrast,ourapproachallowsmultiplexed
`PCRinasinglereaction(Belgraderetal.,1996),
`does not require an additional step to convert pro-
`duct into single-stranded form, and can readily dis-
`tinguish all point mutations including slippage in
`repeatsequences(Dayetal.,1995;Khannaetal.,
`1999).AlternativeDNAarrayssufferfromdifferen-
`tial hybridization efficiencies due
`to
`either
`sequence variation or to the amount of target pre-
`sent
`in the sample. By using our approach of
`designing divergent zip-code sequences with simi-
`lar thermodynamic properties, hybridizations can
`be carried out at 65 (cid:14)C, resulting in a more strin-
`gent and rapid hybridization. The decoupling of
`the hybridization step from the mutation detection
`stage offers the prospect of quantification of LDR
`products, as we have already achieved using gel-
`basedLDRdetection(Khannaetal.,1999).
`Arrays spotted on polymer surfaces provide sub-
`stantial improvements in signal capture, as com-
`pared with arrays spotted or synthesized in situ
`directlyonglasssurfaces(Drobyshevetal.,1997;
`Parinovetal.,1996;Yershovetal.,1996).However,
`the polymers described by others are limited to
`using 8 to 10-mer addresses, while our polymeric
`surface readily allows 24-mer zip-codes to pene-
`trate and couple covalently. Moreover, LDR pro-
`ducts of length 60 to 75 nucleotide bases are also
`found to penetrate and subsequently hybridize to
`the correct address. As additional advantages, our
`polymer gives little or no background fluorescence
`and does not exhibit non-specific binding of fluor-
`escently labeled oligonucleotides. Finally, zip-codes
`spotted and coupled covalently at a discrete
`address do not ‘‘bleed over’’ to neighboring spots,
`hence obviating the need to physically segregate
`sites, e.g. by cutting gel pads.
`
`Summary and Conclusions
`
`Here, we describe a strategy for high-throughput
`mutation detection which differs substantially from
`other array-based detection systems presented pre-
`viously in the literature. In concert with a polymer-
`ase chain reaction/ligase detection reaction (PCR/
`LDR) assay carried out
`in solution, our array
`allows
`for accurate detection of
`single base
`mutations, whether inherited and present as 50 %
`of the sequence for that gene, or sporadic and pre-
`sent at 1 % or less of the wild-type sequence. We
`achieve this sensitivity because thermostable DNA
`ligase provides the specificity of mutation discrimi-
`nation, while the divergent addressable portions
`(zip-codes) of our LDR primers guide each LDR
`product
`to a designated address on the DNA
`array. Since the zip-code sequences remain con-
`
`Figure 5. Detection of minority K-ras mutant DNA in
`a majority of wild-type DNA using PCR/LDR with zip-
`code array capture. DNA from cell
`line SW620,
`containing the G12V mutation, and DNA from normal
`lymphocytes were PCR amplified in exon 1 of the K-ras
`gene. Mixtures containing 10, 20, 40, or 100 fmol of
`G12V-amplified fragment plus 2000 fmol of PCR-ampli-
`fied wild-type fragment were prepared, and the pre-
`sence of mutant DNA determined by LDR using
`primers specific for the G12V mutation (2000 fmol each
`of discriminating and common primer). Images were
`collected by CCD using exposure times from five to
`25 seconds. Data were normalized by dividing fluor-
`escent signal
`intensity by acquisition time. Each data
`point represents the average hybridization signal from
`four independent robotically spotted arrays. The average
`background signal from all four spots at each address
`following hybridization of pure wild-type control (880
`average fluorescent counts) was subtracted from the
`mutant signal.
`
`products, can detect any K-ras mutation in the pre-
`senceofuptoa500-foldexcessofwild-type,with
`asignal-to-noiseratioofatleast3:1(Khannaetal.,
`1999).
`
`Comparison of universal array to
`gene-specific arrays
`
`Our approach to mutation detection has three
`orthogonal components: (i) primary PCR amplifica-
`tion; (ii) solution-phase LDR detection; and (iii)
`solid-phase hybridization capture. Therefore, back-
`ground signal from each step can be minimized
`and, consequently, the overall sensitivity and accu-
`racy of our method are significantly enhanced over
`those provided by other strategies. For example,
`hybridization of labeled target methods require: (i)
`multiple rounds of PCR or PCR/T7 transcription;
`(ii) processing of PCR amplified products to frag-
`ment them or render them single-stranded; and
`(iii) lengthy hybridization periods (ten hours or
`more)whichlimitsthroughput(Cheeetal.,1996;
`Croninetal.,1996;Guoetal.,1994;Haciaetal.,
`1996;Schenaetal.,1996;Shalonetal.,1996;Wang
`etal.,1998).Additionally,sincetheimmobilized
`probes on the aforementioned arrays have a wide
`range of tm values, it is necessary to perform the
`
`Personalis EX2028
`
`

`

`Universal Array for Multiplex Mutation Detection
`
`259
`
`stant and their complements can be appended to
`any set of LDR primers, our zip-code arrays are
`universal. Thus, a single array design can be pro-
`grammed to detect a wide range of genetic
`mutations.
`the rapid detection of
`for
`Robust methods
`mutations at numerous potential sites in multiple
`genes hold great promise to improve the diagnosis
`and treatment of cancer patients. Non-invasive
`tests for mutational analysis of shed cells in saliva,
`sputum, urine, and stool could significantly sim-
`plify and improve the surveillance of high risk
`populations, reduce the cost and discomfort of
`endoscopic testing, thus leading to more effective
`diagnosis of cancer in its early, curable stage.
`Although
`the
`feasibility
`of detecting
`shed
`mutations has been demonstrated clearly in
`patients with known and genetically characterized
`tumors(Caldasetal.,1994;Hasegawaetal.,1995;
`Nollauetal.,1996;Sidranskyetal.,1992;Wuetal.,
`1994),effectivepresymptomaticscreeningwill
`require that a myriad of potential low frequency
`mutations be identified with minimal false-positive
`and false-negative signals. Furthermore, the inte-
`gration of
`technologies for determining genetic
`changes within a tumor with clinical information
`about the likelihood of response to therapy could
`radica