ABSTRACT
We have synthesized a set of four energy transfer dyes and demonstrated their use in automated DNA sequencing. The donor dyes are the 5- or 6-carboxy isomers of 4'-aminomethylfluorescein and the acceptor dyes are a novel set of four 4,7-dichloro-substituted rhodamine dyes which have narrower emission spectra than the standard, unsubstituted rhodamines. A rigid amino acid linker, 4-aminomethylbenzoic acid, was used to separate the dyes. The brightness of each dye in an automated sequencing instrument equipped with a dual line argon ion laser (488 and 514 nm excitation) was 2-2.5 times greater than the standard dye-primers with a 2 times reduction in multicomponent noise. The overall improvement in signal-to-noise was 4- to 5-fold. The utility of the new dye set was demonstrated by sequencing of a BAC DNA with an 80 kb insert. Measurement of the extinction coefficients and the relative quantum yields of the dichlororhodamine components of the energy transfer dyes showed their values were reduced by 20-25% compared with the dichlororhodamine dyes alone.
Four-color automated DNA sequencing instruments have traditionally utilized single dyes as fluorescent labels (1 -3 ). Recently, several groups have described the use of oligonucleotide backbones as linkers in energy transfer dyes which exhibit brighter fluorescence and longer Stokes' shifts than the individual dye components (4 -8 ). These labeled oligonucleotides can be used in automated sequencing as dye-labeled primers and provide brighter signal than the set of four acceptor dyes alone. However, the 10mer oligonucleotide backbone that separates the donor-acceptor pair prohibits the use of these energy transfer compounds as labels for small molecules such as deoxy- and dideoxynucleotides.
In 1988 we examined a compound named `bifluor-1', a dye dimer consisting of 5-carboxytetramethylrhodamine linked to 4'-aminomethyl-5-carboxyfluorescein (Fig. 1 ). Bifluor-1 was a gift from Richard Haugland of Molecular Probes Inc. (Eugene, OR). The absorbance spectrum of bifluor-1 appeared to be the additive absorbance spectra of both donor and acceptor dyes and its emission spectrum appeared to be only of the longer wavelength acceptor dye, tetramethylrhodamine. The fluorescence of the fluorescein component, the donor dye, was completely quenched. It appeared that bifluor-1 was an energy transfer dye which could prove useful in DNA sequencing. However, bifluor-1 was no brighter than tetramethylrhodamine alone. Our goal was to improve the brightness and spectral properties of the bifluor type of energy transfer dye.
Absorbance measurements were performed on a Perkin-Elmer Lambda-Bio UV-Vis spectrophotomer. Fluorescence measurements were performed on a Perkin-Elmer LS-50B fluorescence spectrophotomer. High pressure liquid chromatography (HPLC) was performed with a Waters 660E pump equipped with a 996 photodiode array detector. Separations utilized an Applied Biosystems Aquapore RP-300 C-8 column.
DNA sequencing reactions were performed using a Perkin-Elmer GeneAmp PCR system 9600 thermal cycler and analyzed using an ABI PRISM 377 DNA sequencer. The standard dye-primer set used 10 nm virtual filters at 536, 559, 586 and 615 nm. The d-rhodamine and fluorescein-d-rhodamine sets used 10 nm filters centered at 540, 570, 595 and 625 nm.
A mixture of 3-dimethylaminophenol (0.25 g, 2.3 mmol), 3,6-dichlorotrimellitic anhydride (0.33 g, 1.3 mmol) and sulfuric acid (1 ml) were combined and heated to 190oC for 12 h. Water (10 ml) was added to the reaction and the black solid separated by filtration. The solid was extracted with acetonitrile (10 ml). The orange solution was concentrated to dryness and the residue dissolved in 250 mM carbonate/bicarbonate buffer, pH 9 (10 ml). The solution was acidified with concentrated HCl and the red precipitate collected by centrifugation and dried in a vacuum centrifuge. A red solid was obtained (32 mg, 5% yield).
The three dyes, dR6G, dTMR and dROX, were each prepared by condensation of substituted 3-aminophenols with 3,6-dichlorotrimellitic anhydride in polyphosphoric acid. A representative preparation of dR6G is described below.
A mixture of 3-aminomethyl-p-cresol (100 mg, 0.66 mmol), 3,6-dichlorotrimellitic anhydride (100 mg, 0.38 mmol) and polyphosphoric acid (0.5 g) were combined and heated to 180oC for 12 h. Water (10 ml) was added to the red-brown reaction and the mixture filtered. The solid was taken up in 1 M carbonate/ bicarbonate buffer, pH 9 (25 ml), and filtered. The filtrate was acidified with concentrated HCl and the solid collected by centrifugation and dried. The solid was extracted with acetonitrile (10 ml) and the extracts were concentrated to a sticky solid. The solid was dissolved in dimethylformamide (0.6 ml). The yield was estimated by diluting an aliquot into 40% acetonitrile, 0.1 M triethylammonium acetate (TEAA) buffer and measuring the absorbance in a UV/visible spectrophotometer. Using an extinction coefficient of 50 000/cm/M at the absorbance maximum of 540 nm the yield was estimated to be 9%.
Identities of 5- and 6-d-rhodamine isomers were inferred from HPLC retention times. All 1H NMR-confirmed 6-carboxy isomers of xanthene dyes, e.g. 6-carboxyfluorescein and 6-carboxyrhodamines, elute faster than the 5-carboxy isomers on reverse phase HPLC. The isomers of 5- and 6-carboxy-4,7-dichloroxanthene dyes could not be assigned based on 1H NMR spectral data since there is no splitting of the lone 5- or 6- hydrogen. We assumed the 5- and 6-carboxy-4,7-dichloroxanthene dyes exhibited the same relative HPLC elution order as the 5- and 6-carboxyxanthene dyes.
Activation of 5- and 6-carboxy-d-rhodamines to the succinimidyl esters was performed on the crude mixtures using N-hydroxysuccinimide and 1-(3-dimethylaminopropyl)-3-ethylcarbodiimide hydrochloride. A representative preparation of dR6G-NHS is described below.
A solution of dR6G in dimethylformamide (6 [mu]mol in 0.1 ml) was combined with N-hydroxysuccinimide (22 mg, 0.2 mmol) and 1-(3-dimethylaminopropyl)-3-ethylcarbodiimide hydrochloride (12 mg, 0.06 mmol). Reaction progress was monitored by thin layer chromatography (TLC) on silica gel using a mixture of dichloromethane, methanol and acetic acid (600:60:16) as eluant. After the starting material was consumed, the reaction was diluted with dichloromethane (10 ml) and washed with 5% HCl (10 ml). The material which was insoluble in both phases was discarded. The organic phase was washed successively with 250 mM carbonate/bicarbonate (10 ml) and with 5% HCl (10 ml). The solution was dried (MgSO4) and concentrated to a dark oil. The residue was taken up in dimethylformamide (0.2 ml) and stored at -20oC.
After 3 days at -20oC gold crystals had formed in the dimethylformamide solution. The supernatant was decanted and discarded. The residue was dissolved in methylsulfoxide (0.1 ml). An aliquot was diluted into 40% acetonitrile, 0.1 M TEAA. Using an extinction coefficient of 50 000/cm/M at the absorbance maximum of 548 nm the yield was estimated to be 7%.
Preparation of fluorescein-rhodamine dimers with either the standard rhodamines or the d-rhodamines were performed by reaction of the rhodamine succinimidyl esters (dye-NHS) with 4'-aminomethyl-5-carboxyfluorescein. Preparation of the fluorescein-rhodamine dimers with the linker compound, 4-aminomethylbenzoic acid (B), required reaction of the rhodamine succinimidyl ester with 4-aminomethylbenzoic acid followed by activation of the derivatized rhodamine. Reaction with 4'-aminomethyl-5- carboxyfluorescein followed by another activation reaction provided the succinimidyl ester of the fluorescein-B-rhodamine dimer. A representative synthesis of 5CFB-5TMR and its activation is described below.
A mixture of 4-aminomethylbenzoic acid (3 mg, 19 [mu]mol), 5TMR-NHS (5 mg, 9 [mu]mol) and triethylamine (20 [mu]l) was suspended in dimethylformamide (DMF, 200 [mu]l) in a 1.5 ml Eppendorf tube. The mixture was heated to 60oC for 10 min. Reaction progress was monitored by TLC on silica gel with elution with a 400/30/10 mixture of dichloromethane, methanol and acetic acid. The insoluble 4-aminomethylbenzoic acid was separated by centrifugation and the DMF solution was decanted into 5% HCl (1 ml). The insoluble B-5TMR was separated by centrifugation, washed with 5% HCl (2 * 1 ml) and dried in a vacuum centrifuge. The product was dissolved in DMF (200 [mu]l) and used to prepare B-5TMR-NHS.
A solution of B-5TMR in DMF (125 [mu]l), diisopropylethylamine (10 [mu]l) and disuccinimidylcarbonate (10 mg) was combined in a 1.5 ml Eppendorf tube and heated to 60oC. Reaction progress was monitored by TLC on silica gel with elution with a 600/60/16 mixture of dichloromethane, methanol and acetic acid. After 5 min the reaction appeared to be complete. The solution was diluted into methylene chloride (3 ml) and washed with 250 mM carbonate/bicarbonate buffer, pH 9 (4 * 1 ml), dried (Na2SO4) and concentrated to dryness in a vacuum centrifuge. The solid was dissolved in DMF (100 [mu]l). The yield was determined by diluting an aliquot into pH 9 buffer and measuring the absorbance at 552 nm. Using an extinction coefficient of 50 000/cm/M, the concentration of 5TMR-B-NHS was 4.8 mM. Yield from 5TMR-NHS was 8%.
A solution of B-5TMR-NHS (1 [mu]mol in 250 [mu]l DMF) was combined with a solution of 4'-aminomethyl-5-carboxyfluorescein (5CF, 2.2 [mu]mol in 100 [mu]l DMSO) and triethylamine (20 [mu]l) in a 1.5 ml Eppendorf tube. The reaction was monitored by HPLC using a C8 reverse phase column with an elution gradient of 15-35% acetonitrile versus 0.1 M TEAA. HPLC analysis indicated that 5TMR-B-NHS was consumed, leaving the excess, unreacted 5CF. The reaction was diluted with 5% HCl (1 ml) and the product separated by centrifugation, leaving the unreacted 5CF in the aqueous phase. The solid was washed with 5% HCl (4 * 1 ml), dried in a vacuum centrifuge and taken up in DMF (300 [mu]l). The yield was quantitative.
A solution of 5CFB-5TMR (0.6 [mu]mol in 100 [mu]l DMF), 1-(3-dimethylaminopropyl)-3-ethylcarbodiimide hydrochloride (DEC, 2 mg) and N-hydroxysuccinimide (4 mg) were combined in a 1.5 ml Eppendorf tube. The mixture was sonicated briefly and heated to 60oC. The reaction was monitored by TLC on silica gel with elution with a 600/60/16 mixture of dichloromethane, methanol and acetic acid. The reaction was complete in 30 min and diluted with 5% HCl. The product was separated by centrifugation and dried in a vacuum centrifuge. The activated dye was dissolved in DMF (20 [mu]l).
Both d-rhodamine labeled primers and fluorescein-d-rhodamine labeled primers were synthesized by reaction of succinimidyl esters of the dyes with 5'-aminohexyl-derivatized universal primer (-21 M13). A representative preparation of dR6G-labeled primer is described below.
A solution of 5'-aminohexyl-functionalized oligonucleotide (10 [mu]l, 1 mM) and dR6G-NHS (10 [mu]l, 12 mM in methylsulfoxide) and carbonate/bicarbonate buffer (2 [mu]l, 1 M) were combined. After 10 min at room temperature the solution was subjected to gel filtration on Sephadex G-25 to separate free dye. The fraction containing dye-labeled oligonucleotide and unlabeled oligonucleotide was collected and subjected to HPLC purification on a reverse phase column. The unlabeled oligonucleotide and each dye isomer of dye-labeled oligonucleotide were separated using an elution gradient of 10-30% acetonitrile versus 0.1 M TEAA. The solutions containing dye-labeled oligonucleotide were concentrated in a vacuum centrifuge and redissolved in TE buffer.
Dye-primer cycle sequencing was performed using AmpliTaq DNA polymerase, FS, according to the ABI PRISM 377 DNA Sequencer User's Manual (revision B, January 1995, Chapter 2; Perkin-Elmer Corp., Foster City, CA).
A necessary feature of the energy transfer fluorescein-rhodamine dye is the bifunctionality of the donor dye. One functional group is used to attach the acceptor and the other functional group is used to attach the oligonucleotide or nucleotide. Fluorescein and its derivatives are ideal as bifunctional donor dyes because their structure is readily modified both before and after construction of the chromophore. The 4'-aminomethyl group of bifluor-1 is added to the fluorescein dye component by a Mannich reaction of hydroxymethylacetamide and 5'-carboxyfluorescein (9 ,10 ). The resulting acetamidide is hydrolyzed by HCl to provide the amine.
The original bifluor-1 dye used tetramethylrhodamine (TMR) as the acceptor dye in the energy transfer system. TMR is a member of a set of four spectrally resolved rhodamine dyes which are useful in automated DNA sequencing, particularly as part of a set of dye-labeled dideoxynucleotides. We will refer to this set as the `standard' set of rhodamine dyes, consisting of R110, R6G, TMR and ROX (11 ). The dyes are attached to the dideoxynucleotide or oligonucleotide by either the 5- or 6-carboxy isomer of each dye. Both isomers provide the same fluorescent emission spectra, but a pure isomer must be used because of differences in electrophoretic mobility between the two isomers. The structures and normalized fluorescence emission spectra of the set of rhodamine dyes attached to oligonucleotides are shown in Figures 2 and 3 .
Four-color sequencing relies on the spectral resolution of the set of four dyes. The degree of resolution of the standard rhodamine dyes can be seen in the points of intersection of their normalized emission spectra. A perfectly resolved set of dyes would intersect at an ordinate value of 0 and have zero overlap of emission spectra.
The standard rhodamine dyes have emission spectra that intersect at values of 0.75-0.85. The original dye set for dye-primer sequencing, FAM, JOE, TMR and ROX, includes two of the standard rhodamine dyes, TMR and ROX, and has a similar set of fluorescence emission spectra. Bodipy dyes have intersection values of 0.35-0.6 (12 ). Most of the useful fluorescent dyes (fluoresceins, rhodamines, bodipies) have Stokes' shifts of 15-25 nm and absorbance spectra which are mirror images of the emission spectra. Therefore, dye sets which are better resolved and have narrow emission spectra also have lower extinction coefficients at excitation wavelengths useful for exciting the shortest wavelength dye of the set. However, in energy transfer systems the absorbance at shorter wavelengths is provided by the donor dye. Energy transfer systems which utilize well-resolved acceptor dyes should provide both bright and well-resolved signals.
Xanthene dyes with narrow fluorescence emisson spectra can be obtained by the addition of chlorides to the phenyl group. For example, the emission spectra of 4,7-dichlorofluorescein dyes are 10-15% narrower and have emisson maxima at 10-15 nm longer wavelengths than the unsubstituted fluoresceins (13 ). By adding the 4,7-dichloro- substitution to the rhodamine dyes, we synthesized a set of four dichlororhodamine dyes, which we refer to as d-rhodamines, which have narrower emissions and emission maxima at longer wavelengths than the standard rhodamine set. The structures and normalized fluorescence emission spectra of the d-rhodamine dyes are shown in Figures 2 and 3 . The intersection values of the d-rhodamines are 0.65-0.68, an improvement over the standard rhodamine set.
In an initial attempt to improve the brightness of bifluor-1 we synthesized several fluorescein-tetramethylrhodamine adducts with different linking groups between the fluorescein and rhodamine components. Two non-rigid amino acid linkers, glycine and lysine, produced fluorescein-tetramethylrhodamine dimers with decreased brightness compared with bifluor-1 (data not shown). A rigid linker, 4-aminomethylbenzoic acid, produced energy transfer compounds with increased brightness.
A comparison of the brightness of fluorescein-rhodamine dimers with variable acceptors, acceptor isomers and linkers is shown in Figure 4 . Energy transfer fluorescein-rhodamine dyes were constructed using both isomers of 4'-aminomethyl-5 (or 6)- carboxyfluorescein (5CF or 6CF) as the donor dye and 5-carboxy- R6G (5R6G) or 5- or 6-carboxy-dR6G (5dR6G or 6dR6G) as the acceptor dye. The linker, if used, was 4-aminomethylbenzoic acid (B). The fluorescence emission spectra were measured using dye solutions with equivalent absorbances at the acceptor dye absorbance maxima. Assuming equivalent losses in extinction coefficients, the relative peak heights can be regarded as the relative molar brightness of the various energy transfer compounds.
The choice of dye sets affects three parameters in automated dye-primer sequencing: the mobility shift, the multicomponent matrix and the overall signal. The mobility shift refers to the effect of the dye on the relative electrophoretic mobility of oligonucleotide fragments. Attachment of a dye causes a change in the mobility of the fragment compared with unlabeled DNA and the relative change in mobility among the set of four dyes is the basis of the mobility shift correction. The correction can be a constant, or `parallel', shift or it can be a variable, or `curved', shift. A parallel mobility shift implies a constant displacement for every fragment length labeled with one dye, relative to the same fragment labeled with another dye. A curved mobility shift applies a greater correction to the shorter fragments than the longer ones. Dye sets composed of dyes of a single chemical class are usually corrected with parallel shifts. A collective mobility shift of less than half a base pair for all four dyes will not require new mobility corrections for small changes in gel composition, temperature or read length.
The multicomponent matrix refers to the 4 * 4 matrix which is applied to the fluorescence signal to quantitate the contribution of each dye. The matrix contains the relative contributions of each of the four dyes to each of the four `bins', or virtual filters, of collected light. Each `bin' is a 10 nm wide window of light collection by the CCD detector. Since the `bins' can vary among instruments, the matrix is both machine and dye set dependent. A useful metric to compare matrices is the condition number, which is the ratio of the largest and the smallest eigenvalues of the inverse matrix (14 ). The condition number can vary between 1 and [infinity], where a value of 1 indicates a perfect matrix, which could be obtained with a dye set with ordinate intersection values of 0. The multicomponent noise, or the artifacts which arise after application of the 4 * 4 matrix to the raw data, is proportional to the condition number (15 ).
The d-rhodamine and fluorescein-d-rhodamine primers were used to compare the extinction coefficients and quantum yields of the dye components (Table 1 ). Extinction coefficients of the fluorescein and d-rhodamine components of the energy transfer dyes could be calculated using the nucleic acid absorbance at 260 nm as a measure of the concentration of the primers. The extinction coefficient of the dR110 component of the fluorescein-dR110 energy transfer dye could not be determined because the energy transfer dye gave only one broad absorbance. Ratios of extinction coefficients of the d-rhodamine component of the energy transfer dye to d-rhodamine alone were calculated [[epsilon](ET)/[epsilon](d-Rhod)]. Solutions of d-rhodamine and fluorescein-d-rhodamine dye-primers which gave equal absorbances at the d-rhodamine absorbance maxima were excited at the d-rhodamine absorbance maxima and the fluorescence emission measured to determine their relative quantum yields [[Phi](ET)/[Phi](d-Rhod)].
Each energy transfer dye had both lower extinction coefficients and lower relative quantum yields than the corresponding d-rhodamine dye alone. These results show that although the energy transfer dyes are effectively brighter than the d-rhodamine dyes with 488/514 nm excitation and although the fluorescein component appears to be completely quenched, the energy transfer system is only ~60% of maximal brightness. Additional combinations of orientation and linker composition could produce energy transfer systems with increased brightness.
Our energy transfer system provides enhancements in sequencing data that compares favorably with the brightness, spectral resolution and mobility corrections of energy transfer systems based on oligonucleotide backbones. The use of a compact dye allows labeling of any primer sequence, regardless of the availability of appropriately spaced linker arm thymidines. Perhaps our biggest advantage, however, is the utility of these energy transfer dyes in labeling small molecules. Future papers will describe the use of the d-rhodamine dyes and the fluorescein-d-rhodamine energy transfer dyes on dye-labeled terminators.
Table 1
We thank Kip Connell and John Shigeura for help in defining and discussing the use of condition numbers and Muhammad Sharaf and Dan Allison for help in matrix and mobility file corrections. We are grateful to the following scientists for their help in optimizing the syntheses and use of the fluorescein-d-rhodamine dyes: Paul Kenney, Brian Evans, Mary Fong, Pete Theisen, Sandy Koepf, Iqbal Zaidi and David Yu. Finally, we would like to thank Richard Henfrey for his enthusiasm in the early stages of the project.
*To whom correspondence should be addressed. Tel: +1 415 638 5990; Fax: +1 415 572 2743; Email: leelg@perkin-elmer.com
Dye
[lambda] (nm)
[epsilon] (per cm/M)
[epsilon](ET)/[epsilon](d-Rhod)
[Phi](ET)/[Phi](d-Rhod)
dR110
522
92 000
5CFB-dR110
504
115 000
-
0.78
dR6G
548
120 000
6CFB-dR6G
502
80 000
548
89 000
0.74
0.82
dTMR
577
120 000
5CFB-dTMR
499
80 000
577
90 000
0.75
0.76
dROX
602
144 000
5CFB-dROX
501
87 000
605
115 000
0.80
0.79
REFERENCES