A
`
`with
`Sequencing
`DNA
`Rapid
`for
`System
`Fluorescent Chain-Terminating
`Dideoxynuclcotides
`
`JAMES M. PROBER,* GEORGE L. TRAINOR, RUDY J. DAM, FRANK W. HOBBS,
`CHARLES W. ROBERTSON, ROBERT J. ZAGURSIKY ANTHONY J. COCUZZA,
`MARIK A. JENSEN, KIRiK BAUMEISTER
`
`A DNA sequencing system based on the use of a novel set
`of four chain-terminating dideoxynucleotides, each carry-
`ing a different chemically tumed succinyffluorescein dye
`is described.
`distinguished by its fluorescent emission
`Avian myeloblastosis virus reverse transcriptase is used in
`a modified dideoxy DNA sequencing protocol to produce
`a complete set of fluorescence-tagged fragments in one
`reaction mixture. These DNA fragments are resolved by
`polyacrylamide gel electrophoresis in one sequencing lane
`and are identified by a fluorescence detection system
`specifically matched to the emission characteristics of this
`dye set. A scanning system allows multiple samples to be
`run simultaneously and computer-based automatic base
`sequence identifications to be made. The sequence analy-
`sis of M13 phage DNA made with this system is de-
`scribed.
`
`T
`
`FOR DETER-
`METHODS
`OF SEQUENCING
`t HE DEVELOPMENT
`mining the order of nucleotide bases in deoxyribonucleic
`acid (DNA) has led to rapid advances in our understanding
`of the organization and processing of information in biological
`systems. Two methods are available for DNA sequencing: the
`chemical degradation method of Maxam and Gilbert (1), and the
`dideoxy chain termination method of Sanger (2). Traditionally, each
`approach affords four' sets of radioisotopically labeled fragments
`which are resolved according to'their lengths by gel electrophoresis
`and the resulting autoradiographic pattern is used to obtain the
`DNA sequence.
`The Maxam-Gilbert and Sanger techniques, which are conceptu-
`ally elegant and efficacious, are in practice time-consuming and
`labor-intensive, partly because a single radioisotopic reporter is used
`for detection. Using one reporter to analyze each of the four bases
`requires four separate reactions and four gel lanes. The resulting
`autoradiographic patterns, obtained after a delay for exposure and
`development, are complex and require skilled interpretation and
`data transcription.
`These deficiencies can be corrected by switching from a radioiso-
`topic to a fluorescent reporter. We describe here a system for DNA
`sequencing in which four chemically related yet distinguishable,
`fluorescence-tagged dideoxynucleotides are used to label DNA by a
`modified Sanger protocol with' a suitable chain-extending DNA
`
`336
`
`polymerase. The fluorescent sequencing fragments are resolved
`temporally rather than spatially in a single lane by conventional
`polyacrylamide gel electrophoresis. Analysis of the fluorescent emis-
`sion of each fragment permits us to identify the terminating
`nucleotide and assign the sequence directly in real time.
`Fluorescent tags. We have developed a family of fluorescent dyes
`with largely overlapping yet distinct emission bands. These dyes
`are 9-(carbo ethyl)-3-hydroxy-6-oxo-6H-xanthenes or succinyl-
`fluoresceins (SF-xxx, where xxx refers to the emission maximum in
`nanometers) (Fig. 1A). They are readily prepared from succinic
`anhydride and an appropriately substituted resorcinol by a modifica-
`dion of the procedure cited for the parent dye, SF-505 (3). The
`fluorescent forms of these dyes are the dianions, which predominate
`in aqueous solution above pH 7. The dianion of SF-505 absorbs
`maximally at 486 nm, a wavelength that is well suited for excitation
`by an argon ion laser operating at 488 nm. This species shows an
`absorption coefficient of 72,600M' cn-' at the maximum and
`fluoresces with an emission maximum at 505 nm with a quantum
`yield comparable to that offluorescein. The carboxylic acid finction-
`ality is not essential for fluorescence and is used here for covalent
`attachment to the nucleotides by means of standard methodologies.
`The wavelengths of the absorption and emission maxima in the
`succinylfluorescein system are tuned by changing the substituents
`-RI and -R2. The absorption spectra for the four dyes where -R1
`and -R2 are -H or -Me are shown in Fig. IB. Close spacing of the
`absorption maxima results in efficient excitation of all dyes by the
`single emission line of the argon ion laser. (The absorption coeffi-
`cients at 488 nm of the two most disparate dyes, SF-505 -and SF-
`526, differ only by a factor of 2.) These succinylfluoresceins carry
`the same charge and are nearly identical in size, minimizing any
`differential perturbation of the electrophoretic mohilities of the
`DNA fragments to which they are attached. We observe no
`differences in the relative mobilities among identical DNA frag-
`ments tagged with any of the four dyes.
`sequencing involves the template-
`Labeling strategy. Dideo
`directed, enzymatic extension of a short oligonucleotide primer in
`the presence of chain-terminating d'deo ribonucleotide triphos-
`phates (ddNTP's). A nested set of DNA fragments is produced
`
`J. M. Prober, R. J. Da,
`and C. W. Robertson are wit
`the Engineering Physics
`Laboratory 'G. L. Trainor, F. W. Hobbs, R. J. Zagursky, A. J. Cocuzza, and K.
`Baumeister ear with the Central Research and Development Department (contribution
`number 4543); and M. A. Jensen is with the Medical Products Department, all of the E.
`I. du Pont de Nemours & Company (Inc.), Wilmington, DE 19898.
`*To whom correspondence should be addressed.
`
`SCIENCE, VOL. 238
`
`This content downloaded from 128.187.103.98 on Mon, 02 Nov 2015 18:53:03 UTC
`All use subject to JSTOR Terms and Conditions
`
`Page 336
`
`Illumina Ex. 1041
`IPR Petition - USP 10,435,742
`
`

`

`having prImer-defined 5' ends and variable 3' ends determined by
`the positions of incorporation of a given base (as ddNMP).
`Radioisotopic label is generally incorporated by including [(.-
`32P]dNTP or [a-35S]dNTP in the reaction mixture such that
`internal nttcleotides in the newly synthesized portions of the frag-
`ments are labeled.
`The analogous incorporation of fluorescence-tagged nucleotides
`may be difficult to achieve enzymatically, and may adversely affect
`the crucial relation between chain length and electrophoretic mobil-
`ity, which is essential for accurate sequence determination. We have
`chosen instead to incorporate a single fluorescent tag by labeling
`chain-terminating dideoxynucleotide triphosphates. This labeling
`approach offers several advantages. Generation of all four sets of
`DNA sequencing fragments can be carried out simultaneously in a
`single reaction since only the terminating nucleotide carries the tag.
`In addition, many polymerase pausing artifacts are eliminated since
`only those fragments resulting froni bona fide termination events
`carry a fluorescent tag.
`The set of four fluorescence-tagged chain-terminating reagents we
`have designed and synthesized is shown in Fig. 2A. These are
`ddNTP's to which succinylfluoiescein has been attached via a linker
`to the heterocyclic bdse. (These reagents are designated N-xxx where
`N refers to the ddNTP and xxx refers to the SF-xxx portion.) The
`linker is attached to the 5 position in the pyrimidines and to the 7
`position in the 7-deazapurines. The 7-deazapurines were used to
`facilitate stable linker arm attachment at that site. Coupling of the
`dyes to the nucleotides was found to influence both the wavelengths
`of the absorption and emission maxima and their relative fluores-
`cence yields. The dyes and bases were therefore paired for maximum
`distinguishability and for balanced net fluorescence sensitivities.
`The synthetic scheme devised for these reagents is highly conver-
`gent and general. The preparation of T-526 is illustrative. 5-Iodo-
`2',3'-dideoxyuridine was coupled to N-trifluoroacetylpropargyl-
`amine under palladium(O) catalysis in dimethylformamide (4). The
`resulting derivatized nucleoside was converted to its 5'-triphosphate
`(5) and deacylated to afford 5-(3-amino-1-propynyl)-2',3'-dideoxy-
`uridine triphosphate. This amine was coupled with an O-acetyl-
`protected form of the SF-526-sarcosine conjugate and deprotected
`to afford T-526.
`The fluorescence-tagged chain-terminators are accepted as alter-
`native substrates by avian myeloblastosis virus (AMV) reverse
`transcriptase with efficiencies comparable to that of the correspond-
`
`ing unsubstituted ddNTP's. A comparison of the sequencing ladders
`produced using the ddNTP's and their fluorescent counterparts is
`shown in Fig. 3. The fidelity of terminator incorporation is main-
`tained. Comparison of lanes 4 and 5, containing ddCTP and C-5 19,
`respectively, shows that the fluorescence-tagged fragments run
`approximately 2 bases slower than their untagged counterparts. This
`mobility shift is consistent for all four sets of fluorescence-tagged
`fragments, allowing the sequence ladder to be read and qualifying
`this set of reagents for single-lane DNA sequencing. These termina-
`tors are also substrates for modified T7 DNA polymerase (6). They
`are not, however, substrates for the Klenow fragment of DNA
`polymerase I from Escherichia coli.
`Fluorescence detection system. Labeled DNA fragments, pro-
`duced by the enzymatic chain extension reactions, are separated by
`polyacrylamide gel electrophoresis, detected, and identified as they
`migrate past the fluorescence detection system illustrated in Fig. 4.
`High signal-to-noise ratio is achieved in this system through
`efficient excitation, optical filtering, and light collection.
`Strong excitation is obtained with a laser source that provides
`most of its output energy at a single wavelength and by matching
`the dye absorbances to that wavelength. Fluorescence detection is
`enhanced approximately fourfold by depositing a mirror on the
`outside surface of the glass gel support plate furthest away from the
`laser. The excitation beam is thereby returned through the gel, in
`effect doubling the excitation pathlength, and the fluorescence
`headed away from the detectors is also returned, increasing the
`amount of collected light.
`In addition to the desired fluorescence from the dye-labeled
`DNA, light emanating from the excitation region includes scattered
`laser radiation, Raman scattering, and fluorescence from other
`sources. Optical filtering and other means are used to reduce these
`undesired signals. Secondary laser lines are removed with a narrow-
`band interference filter placed at the source. The reflected compo-
`nent of the 488-nm laser light is eliminated by the mirror, which
`provides a return path for the beam out of the detectors' field of
`view. Scattered light is removed by a filter stack consisting of an
`interference filter, a fiber-optic face plate, and a colored glass
`absorbing filter. The interference filter is designed to strongly reject
`the excitation light, pass the dye emission region, and reject Raman
`scattering and fluorescence above 560 nm. At high angles of
`incidence, the filter passband is shifted toward the excitation
`wavelength, thus increasing background levels and associated noise.
`
`A
`
`HO
`
`R2
`
`R
`
`2
`SFR5,9
`
`R,-CH3 R2=H
`
`~~. ~~. ~~~.
`
`COOH
`
`SF-505: R =R =H
`R1=H R =CH3
`SF-512:'
`R2 SF-519: R=CH 3R2 H
`
`SF-526:
`
`R1=R2=CH3
`2
`3
`
`B
`
`?
`
`LU
`
`u
`
`0
`
`4 0 Gil//8/
`
`.6//
`
`4
`
`S44S
`
`100Z
`
`/
`
`\
`'
`
`\
`'
`
`/
`
`~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~DYES
`
`~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~
`/
`.2
`
`4
`
`\
`
`\\
`
`0400
`
`440
`
`480
`
`520
`
`560
`
`WAVELENGTH (nm)
`
`Fig. 1. Succinylfluorescein dyes. (A) Chemical structure of the four dyes
`used to label dideoxynucleotide triphosphates for use as chain-terminators in
`modified dideoxy DNA sequencing protocols. (B) Normalized absorption
`spectra of the dyes shown in (A). Absorption coefficient at the maximum for
`
`SF-505 is 72,600M-1 cm-'. Spectra were measured in pH 8.2, 50 mM
`aqueous tris-HCI buffer. The other dyes have coefficients within 10 percent
`of this value. Vertical bar (laser) indicates the position of the argon ion laser
`line at 488 nm used for fluorescence excitation.
`
`M6 OCTOBER I987
`
`RESEARCH ARTICLES
`
`337
`
`This content downloaded from 128.187.103.98 on Mon, 02 Nov 2015 18:53:03 UTC
`All use subject to JSTOR Terms and Conditions
`
`Page 337
`
`

`

`Thus a fiber-optic face plate with extramural absorber is employed as
`an aperture to restrict entrance angles on the filter. This element
`absorbs incident rays outside the acceptance angle of the fiber, in
`this case about 22 degrees. This aperture appears over any point of
`excitation across the gel.
`The dyes that we used are distinguished by small differences in
`their absorption and emission spectra. Close spacing of the spectra
`facilitates efficient excitation and simultaneous detection. This con-
`trasts with the use of large differences in emission spectra that one
`would intuitively select to facilitate discrimination. Figure 2B shows
`the emission spectra of the fluorescence-tagged chain-terminators
`that occur in a polyacrylamide gel. In our system, two photomulti-
`plier tube detectors, each with a different filter stack, view the
`fluorescence simultaneously; one collects light at the low-wave-
`length side of the emission bands and the other collects light at the
`high wavelength side. When, for example, G-505-terminated frag-
`ments pass the detectors, one detector registers a large signal relative
`to the other; when T-526-terminated fragments pass, the reverse
`occurs. The A-512- and C-519-terminated fragments give specific
`detector signal ratios lying between these two extremes. The ratio of
`baseline-corrected peak intensities is used to determine the base
`assignment. This method provides very efficient light collection and
`obviates the need for more refined spectral analysis.
`The scanning mechanism shown in Fig. 4 directs the laser beam to
`the gel through a periscope mounted on the shaft of a stepper
`motor. By stepping the periscope, any point in the detection region
`can be accessed. The detectors span the full width of a gel so that
`fluorescence is collected and analyzed from wherever the laser is
`pointed. Directing only the laser beam, rather than any of the major
`optical elements, results in a low-noise scanning system that is
`mechanically simple and stable.
`Data processing and analysis. Processing of the raw detector
`data and subsequent nucleotide base assignment is accomplished in
`several steps. Digitized signals from both detectors are subjected to a
`peak-finding algorithm to determine the time positions of the DNA
`bands. Once the DNA peaks are located, the ratios of baseline-
`corrected peak intensities are calculated along with a statistical
`calculation of the certainty of those ratios. The 15 percent most
`certain ratios are displayed as a histogram in Fig. 5. These fall
`naturally into clusters centered about four values, each defining one
`of the bases. These reference values can differ slightly from lane to
`
`lane due to geometrical effects of the wavelength selection filters. To
`allow for these variations, the reference values are determined
`independently for each lane.
`The four reference values correspond to peaks in the histogram.
`Measurement of scatter about the reference values allows the
`determination of boundaries which are used for base assignment.
`When DNA peaks are no longer resolvable or when their amplitude
`falls below the inherent noise level of the detection system, no
`further bases are assigned. In the current system, this usually occurs
`between 300 to 400 bases from the 3' terminus of the primer.
`Deconvolution, predictive modeling, and digital signal processing
`are being investigated to extend the readable sequence range.
`sequence determination. Raw detector data ob-
`Nudeotide
`tained from the nucleotide sequence of the phage M13mpl8 during
`a 6-hour run are shown in Fig. 6. A modified, two-stage dideoxy
`chain termination protocol (6) was used to generate the DNA
`fragments for fluorescence-based sequence analysis. The data shown
`correspond to bases 3 to 234 from the 3' terminus of the primer.
`Reference values for this sequencing run are established in Fig. 5,
`which illustrates the principles discussed above for base assignment.
`Regions of the sequence are shown in greater detail in Fig. 7.
`in sequencing. As illustrated by recent discus-
`Improvements
`sions of human genome sequencing, large-scale sequencing projects
`would require massive resources with current technology (7).
`Programs of intermediate scope, crucial in fundamental research as
`well as in the development of DNA probes, clinical reagents, and
`useful genetically engineered organisms, are likewise limited by the
`speed and ease of DNA sequencing.
`Efforts to improve the technology of radioisotopic sequencing
`have been reviewed (8). In one example, a technique of continuous
`DNA blotting during electrophoresis is described in which the
`DNA fragments resulting from conventional dideoxy sequencing
`are separated by electrophoresis and allowed to elute from the
`bottom of the gel onto a moving membrane (9). The membrane is
`subsequently exposed to film and the resulting autoradiographic
`band patterns are interpreted in the normal fashion. This technique
`offers some advantages in readability of the band patterns and in the
`potential for a degree of automation in sequence analysis. However,
`the inherent disadvantages of using isotopically labeled reporters in
`the traditional manner remain and will limit the ultimate utility of
`this approach. A newer method with potential for genomic sequenc-
`
`AB
`
`0
`
`-o
`OH3 ~ ~ ~ OH3 ~ T-526
`
`0
`
`-o
`
`01'\
`
`C-519
`
`H
`
`N~~~~~~~~~~'
`
`~~~~~~~~~~~~~~~~~~~~~.8/
`
`C.
`
`"
`
`TERMINATORS
`
`.~~~
`~~~~~~~~~~~4 TRANSMISSION
`
`FILTER
`
`BFILTER
`TRANSMISSION
`
`H
`
`N
`
`0NH/
`
`0
`
`N
`
`0
`
`CH3
`
`4E3N
`
`-3H09P30
`
`0
`0
`-0
`OH3 ~~~~~OH3
`
`A-512
`
`NH
`CH3O
`06
`4Et3NH- 3O~3
`N'-~O
`O~~~~~~~~~~~~~~~~~~~~~~
`I,
`-O ~~~~~~~~~~~~~~0A
`0
`0
`.0
`G-505/
`
`0
`
`N
`
`H3 0
`
`4Et3NH''N4Et3NH-
`4Et3NH~3H0 p
`
`0
`~~~NH,2
`
`N
`
`N~'
`
`HoN
`3H09P3
`
`/
`
`...,-,'NH
`
`AH
`
`.2/
`
`40
`
`520
`
`560
`
`600
`
`WAVELENGTH (nm)
`
`Fig. 2. Fluorescence-tagged chain-terminating reagents. (A) Chemical struc-
`tures of the reagents used in modified dideoxy reactions for DNA sequenc-
`ing. (B) Normalized fluorescence emission spectra of the reagents, measured
`in an 8 percent w/v polyacrylamide gel (19:1 w/w acrylamide:bis) under
`electrophoresis conditions similar to those of Fig. 6. The excitation wave-
`length was 488 nm. The absolute emission intensity values of the four
`
`compounds varied relative to each other by less than a factor of 2.
`Superimposed on the emission spectra are transmission functions of the
`interference filters used in the fluorescence detection system (Fig. 4). The
`nucleotide base assignment for each band is achieved by measuring the
`relative fluorescence signal in two detectors with spectral responses defined
`primarily by these filter functions.
`
`338
`
`SCIENCE, VOL. 238
`
`This content downloaded from 128.187.103.98 on Mon, 02 Nov 2015 18:53:03 UTC
`All use subject to JSTOR Terms and Conditions
`
`Page 338
`
`

`

`ing employing radioactively labeled probes has also been described
`(10). This technique may address the problem of low throughput by
`enabling multiple sequences to be read from a single gel. The
`method does reduce the amount of gel preparation and electropho-
`resis required for a given sequencing task; however, sequence data
`are still obtained from autoradiographs which must be analyzed and
`transcribed in separate steps.
`Advances in fluorescent sequencing have also been reported. The
`synthesis of oligonucleotides labeled at 5' ends with fluorescent
`labels and useful for dideoxy sequencing (11), and several systems
`based on fluorescent primer technology (12-14), have been de-
`scribed. These systems successfully reduce or eliminate a number of
`the deficiencies of the current radioisotope technology. However,
`
`the disadvantages of primer labeling remain. In addition to the
`polymerase pausing problem noted previously, the use of labeled
`primers requires that oligonucleotides be custom synthesized and
`purified for each set of sequencing reactions. This restricts the use of
`sequencing strategies which employ multiple priming sites or differ-
`ent doning vectors.
`
`MIRRORED SCAN GEL LABELED
`DNA BANDS
`SURFACE
`LINE
`
`Terminator
`
`O
`
`v
`
`<
`
`Lane
`
`1
`
`2
`
`3
`
`0n
`
`4
`
`0H
`
`5
`
`<6
`
`7
`
`8
`
`6
`
`TIME
`
`
`
`O CTIME
`
`I~~~~~~
`
`,I
`
`/
`
`LINE
`~~~FILTER
`I
`\
`
`LASER
`
`BEAM EXPANDER
`
`STAC
`
`z
`
`G
`G
`45~~~~~~~
`C
`
`55CC
`
`5 0T
`
`A
`
`T
`
`T
`
`A
`G A
`
`C
`
`4 Gc
`
`40 A C
`
`CSG
`
`a
`
`C
`T
`
`35 C
`5?TC Ga
`
`_
`
`.
`
`low
`
`C
`
`-
`
`_
`
`o
`
`30CG
`45G
`
`C
`T T=
`C
`
`25 A
`
`A
`
`20 T
`
`C
`C
`
`A
`
`C
`
`C
`
`Fig. 3. Comparison of electrophoretic mobilities of DNA sequencing
`fragensterminated with ddNTP's and with fluorescence-tagged chain-
`terminators. The dideoxy method of Sanger (2) was used to sequence a
`region of M13mp18. A 17-bp oligonucleotide corresponding to the (-40)
`region of the template was end-labeled with [-y-32P]ATP and polynucleotide
`kinase. Each reaction mixture contained template (0.1 Rtg), end-labeled
`primer (2.5 ng, annealed by heating to 95C for 2 minutes and slowly cooled
`to room temperature), 60 mM tris-HCI at pH 8.5, 7.5 mM MgCI2, 75 mM
`NaCl, 0.5 mM dithiothreitol, AMV reverse transcriptase (NEN, 20 units),
`and nudleotides at the following concentrations: lane 1, 2.5 pM ddGTP, 12
`p.M dGTP, 2.5 pM dATP, 50 p.M dCIP and dfl'P; lane 2, 0.25 pM
`ddATP, 2.5 pM dATP, 50 pM dGTP, dC1TP, and dfTP; lane 3, 3.0 pM
`dd"IT, 12 pM dUTP, 2.5 pM dATP, 50 pM dCIP and dGTP; lane 4, 2.5
`pM ddC`FP, 12 pM dXrP, 2.5 pM dATP, 50 pM dGTP and dTTP; lane 5,
`as in lane 4 except ddGTP has been replaced with 5.0 pM C-519; lane 6, as
`in lane 3 except ddTITP has been replaced with 6.0 pM T-526; lane 7, as in
`lane 2 except ddATP has been replaced with 2.0 pM A-512; lane 8, as in lane
`1 except ddGTP has been replaced with 8.0 pM G-505. The primer
`extensions were carried out at 42C for 10 minutes, and then unlabeled
`dNTP's (100 pM) were added and the reaction proceeded for 10 minutes at
`420C. The reactions were stopped by diluting to 60 percent (v/v) fonnamide
`and denatured at 65C for 10 minutes. A portion of each sample (1/8) was
`loaded onto an 8 percent (w/v) polyacrylamide gel containing urea (8M) and
`TBE buffer (100 mM tris-HCI, 83 mM boric acid, 1 mM Na2EDTA, pH
`8.1). The electrophoresis was carried out at 1600 volts.
`
`16 O~rOBER 1987
`
`PMTB
`
`IL
`
`U.
`
`IM
`TIME
`
`SCANNING
`~~~~~~OPTICS
`Fig. 4. Fluorescence detection system. Schematic drawing of the optical
`system used for scanning excitation and for measuring fluorescence from
`multiple sequencing lanes in an electrophoresis gel. Light from the argon ion
`laser is filtered to isolate the 488-nm emission line. The beam is deflected by
`a mirror into the scanning optics which arc mounted on the shaft of a
`digitally controlled stepper motor. A lens focuses the beam into a spot in the
`plane of the gel. A second mirror directs the beam to a position on the scan
`ine defined by the rotational position of the motor shaft. Sequencing of
`multiple samples is achieved by directing the beam sequentially to each of the
`sequencing lanes on the gel. Upon entering the gel, the beam excites
`fluorescence in the terminator-labeled DNA. Fluorescence is detected by two
`elongated, stationary photomultiplier tubes (PMI A and PMT B) which
`span the width of the gel. In front of each PMT, a filter stack is placed with
`one of the complementary transmission functions (Fig. 2B). Baseline-
`corrected ratios of signals in the PMTs are used to identify the labeled DNA
`fragments currently in the excitation region. Excitation efficiency and
`fluorescence collection arc increased by the mirrored outer surface of the
`glass plate in the clectrophorcsis gel assembly.
`
`- T -
`
`4
`
`C -
`
`4
`
`A -4
`
`-
`
`G -
`
`LEi
`
`z
`
`-2.30
`
`-1.56
`
`-0.87
`
`-0.26
`
`Ln (R)
`Fig. 5. Histogram of the 15 percent most certain ratios of baselinc-corrected
`peak intensities of detector A to detector B for the run shown in Fig. 6. The
`natural logarithm of the ratio is used to enhance visual presentation. The
`ratios duster about four reference values corresponding to the four bases as
`shown. The intervals used to assign the base identity of each peak are shown
`above the histogram. Data acquisition and analysis were performed by a
`Hewlett-Packard 9000 Model 500 computer.
`
`RESEARCH ARTICLES
`
`339
`
`This content downloaded from 128.187.103.98 on Mon, 02 Nov 2015 18:53:03 UTC
`All use subject to JSTOR Terms and Conditions
`
`Page 339
`
`

`

`The use of fluorescence-labeled chain-terminators for DNA se-
`quencing, on the other hand, offers several distinct advantages.
`Labeled chain terminators afford complete flexibility of sequencing
`strategy and choice of vectors and allow the sequencing reactions to
`take place in one vessel. In this method, only true dideoxy termina-
`tions result in detectable fluorescent bands. Chain synthesis artifacts
`may still occur, but they will not be detected since these products are
`not fluorescent. This results in simplified elution patterns and
`facilitates data analysis.
`We have succeeded in using fluorescence-tagged chain-terminat-
`ing reagents for DNA sequencing. Structurally, each of the four
`reagents described here (Fig. 2A) consists of a succinytfluorescein
`attached to a dideoxynucleoside triphosphate through a novel
`acetylenic linker. The linker is both sterically compact and syntheti-
`cally accessible for all four bases. The linker is attached to the 5
`position of the pyrimidines and the 7 position of the 7-deazapurines.
`These positions have been reported in some cases to be acceptable
`for attaching substituents to chain-propagating substrates (that is,
`deoxynucleotide triphosphates) for DNA polymerases (15, 16). The
`chain-terminators reported here are incorporated by both AMV
`reverse transcriptase and modified T7 DNA polymerase (6), but
`they are not accepted by the Klenow fragment of DNA polymerase I
`from Escherichia coli. The reasons for the inability of these termina-
`tors to serve as substrates for the Klenow fragment are not
`understood.
`Fluorescent reporters have the potential for imparting differential
`mobility shifts to the DNA fragments, thus complicating their
`electrophoretic elution pattern and subsequent base assignments. In
`the systems previously described employing more than one reporter,
`this effect is aggravated by the use of structurally dissimilar dyes
`
`(12). Perturbations to the linker arm structure and software algo-
`rithms were used to reconstruct the correct elution order of the
`fragments (12). Since the dyes reported here belong to a single
`family with only minor substituent differences on the succinyl-
`fluorescein moieties, they impart no significant differential mobility
`shifts to the DNA fragments and the correct elution order is
`observed in the raw data.
`A fundamental criterion for an effective sequencing system is
`accuracy. Accuracy in detection and data analysis derives primarily
`from high sensitivity to the reporter, and in systems using multiple
`reporters, the ability to discriminate one label from another. A
`dideoxy sequencing system must be capable of detecting 10-15 to
`10-16 mol of DNA per band (12). For a given number of reporter
`molecules, the measured fluorescence intensity is determined by the
`incident laser power, the optical properties of the reporter, and the
`efficiency of the detection system. This signal is superimposed on a
`background of Raman and elastic scattering arising from the
`interaction of the laser beam with the gel matrix and the glass plates
`of the electrophoresis assembly. Variation in this background
`determines the inherent noise in the total signal observed. Depend-
`ing on lane geometry, gel thickness, and other factors, our system is
`capable of detecting, in real time, between 10 18 and 10- 17 mol of
`succinyifluorescein reporter under sequencing conditions. Previous-
`ly reported sensitivities for fluorescence-based sequencing systems
`are 3 x 10-18 mol (14) and from 10-17
`to 5 x 10-17 mol (12).
`Possible errors in sequencing include misidentifying a detected
`base, inserting a base, or failing to detect a base at all. To eliminate
`insertions and deletions, advantage can be taken of the uniform local
`temporal spacing of the DNA bands as they pass the detectors (9).
`The prediction of peak positions results in simplified detection and
`
`6.0
`
`-
`
`1
`1
`
`l
`1
`~~~~~50
`
`l
`
`100
`
`150
`
`200
`
`0
`
`1.5
`
`0
`
`TIME, MIN.
`Fig. 6. Detection of fluorescent terminator-labeled DNA fragments from a
`region of M13mp18. Sum of the detector outputs for PMT A and PMT B (in
`arbitrary units of fluorescence intensity) versus time are shown for a 6-hour
`sequencing run. The fragments were generated in a two-stage reaction. In
`one reaction tube were added 3 fig of M13mp18 single-stranded DNA and
`60 ng of primer (17 bp). This mixture was heated at 950C for 2 minutes and
`then placed on ice for 5 minutes to anneal. In the primer extension reaction,
`250 pmol each of dATP, dCTP, dTTP, and c7dGTP were added and the
`reaction incubated at 420C for 12 minutes in the presence of 60 mM tris-
`HCI, pH 8.3, 7.5 mM MgCI2, 75 mM NaCl, 0.5 mM dithiothreitol, and 17
`units of AMV reverse transcriptase. The extension reaction was stopped by
`the addition of a mixture containing 100 pmol of G-505, 800 pmol of A-
`512, 200 pmol of C-519, and 800 pmol of T-526 and then incubated for an
`additional 30 minutes at 420C. Unincorporated fluorescent terminators were
`removed by gel filtration using a Sephadex G-25 spin column (5' to 3' Inc.).
`The effluent was dried under vacuum, washed with 70 percent ethanol, dried
`again, and resuspended in 5 [I of 90 percent formamide containing 11 mM
`Na2EDTA. The sample was then heated to 650C for 7 minutes and loaded
`onto an 8 percent w/v polyacrylamide gel (19: 1 acrylamide: bis, 20 cm by 40
`cm by 0.3 mm) containing 7M urea, 100 mM tris-HCI, pH 8.3, 83 mM
`boric acid, and 11 mM Na2EDTA. The gel was electrophoresed for about 6
`hours at 27 watts of constant power. Markers at top show the locations of
`bands corresponding to the bases.
`
`387
`
`Al
`1.30-
`
`o
`
`if
`
`24
`A
`
`23
`
`2127
`
`T
`
`IT
`
`B
`-4.80
`
`C
`
`A
`
`G~~~22
`
`A
`
`A I
`0.4
`
`T
`219 22
`A
`
`21
`25 T
`
`I
`
`B1
`
`-
`-1.62
`
`the
`
`0
`-,
`base
`
`363.7 ~ ~ 2
`
`f
`
`27
`
`m
`
`(m
`
`In)
`
`4353
`c
`
`3B
`
`19220
`A
`
`I
`
`I
`1226221T6238
`
`\
`
`G
`\I\3351225G
`
`227
`
`A~~~~~~~
`
`12
`
`39A.
`
`~3
`2
`r j~~~~~~~~~~~~~l1
`~~~33
`4
`231
`
`232
`
`S
`
`V
`
`363.7
`
`Tim
`(mnI8.
`
`fuorseneitest
`
`units is plteda31fntino
`
`tie1cl
`
`atr
`
`o
`
`base 3 to base27234.
`
`LL
`
`C
`
`230 k G~~~SIECE VL.23
`
`340
`
`This content downloaded from 128.187.103.98 on Mon, 02 Nov 2015 18:53:03 UTC
`All use subject to JSTOR Terms and Conditions
`
`Page 340
`
`

`

`identification of low-intensity peaks. The variation of detector
`signals across a band helps to identify peaks which are not fully
`resolved.
`The nucleotide sequence of the MI3mpl8 region shown in Fig. 6
`was determined by computer algorithms from base 3 to base 234
`with no wrong base assignments and no missing or extra bases. Peak
`intensity variations result from sequence-dependent differences in
`rates of base incorporation and termination. This distribution is also
`affected by relative concentrations of dNTP's in the extension
`reaction and of the terminators in the final chain termination step, as
`well as by reaction times and conditions. More sequencing experi-
`ence is required to determine the inherent accuracy in base assign-
`ment, but we are encouraged by the absence of errors for peaks
`above a minimum intensity threshold for discrimination.
`In practice there are regions of DNA which are difficult to
`sequence due to aberrations in electrophoretic mobility caused by
`secondary structure (17). The data analysis system allows the
`location and extent of such regions to be identified so that flanking
`sequences remain in frame. 2'-Deoxy-7-deazaguanosine triphos-
`phate has been used (c7dGTP) in place of dGTP to minimize these
`effects. A similar technique has been used in fluorescent primer-
`based sequencing (18). For the run of Fig. 6, two regions of GC
`overlap were assigned by inspection.
`A fundamental measure of the utility of an automated sequencer is
`raw throughput, defined as the sequencing rate per lane times the
`number of lanes per instrument. In our system 12 lanes are practical.
`After an initial period of electrophoresis during which the first DNA
`bands reach the d