ABSTRACT
When hydroquinone-O,O'-diacetic acid is used as a linker arm in solid phase oligonucleotide synthesis, the time for NH4OH cleavage of oligodeoxy- or oligoribonucleotides is reduced to only 2 min. This allows increased productivity on automated DNA synthesizers without requiring any other modifications to existing reagents or synthesis and deprotection methods. The Q-linker may also be rapidly cleaved by milder reagents such as 5% NH4OH, potassium carbonate, anhydrous ammonia, t-butylamine or fluoride ion. However, the Q-linker is sufficiently stable for long-term storage at room temperature without degradation and no loss of material occurs during synthesis. The linker is also reasonably resistant to 20% piperidine/DMF, 0.5 M DBU/pyridine and 1:1 triethylamine/ethanol. The Q-linker can therefore serve as a general replacement for both succinyl and oxalyl linker arms.
Several million oligonucleotides are custom synthesized each year and much effort has been expended to optimize protecting groups, solid phase supports and instrumentation to reduce the time and cost of synthesis. In our studies we have been developing support derivatization methods (1 ,2 ) which will decrease the time required between consecutive rounds of oligonucleotide synthesis. In particular, we have been searching for a better difunctional linker arm to join the terminal nucleoside to the surface of the support. Although a large number of speciality linker arms have been developed (3 ), most oligonucleotide syntheses are still performed using a succinic acid linker arm (4 -10 ). However, hydrolysis of the succinyl linker arm with aqueous NH4OH is slow and most protocols limit this step to only 1 h, which is sufficient for ~80% cleavage. However, the actual assembly time of a typical oligonucleotide is only 1.5-2 h and the 1 h cleavage step accounts for 30-40% of the synthesizer time required. Since most instrumentation with on-line cleavage cannot begin a new synthesis until cleavage is complete, this step represents a significant productivity bottleneck.
Faster cleavage can be obtained using primary amines, such as ethanolamine, with or without hydrazine (11 ), or methylamine (12 ) in either the solution or gas phase (13 ). However, these reagents are not compatible with commonly used N-benzoyl deoxycytidine reagents and there are toxicity and odor problems. Linker arms using silyl (14 -15 ) or disiloxyl groups (16 ) can be rapidly and mildly cleaved using fluoride ion. However, the difficulty of synthesizing these linkages and the need for the extra fluoride reagent remain problems for widespread adoption of these supports. Several other supports have also been prepared which use milder conditions than NH4OH to release the oligonucleotide. A disulfide linkage cleavable with dithiothreitol (17 ), a 4-(2-hydroxyethyl)-3-nitrobenzoic acid linker (1 ) or N-[9-(hydroxymethyl)-2-fluorenyl]-succinamic acid linker (1 ) cleavable with 1,8-diazabicyclo-[5.4.0]undec-7-ene (DBU) and an o-nitrobenzyl linker cleavable by photolysis (20 ) have all been described. However, these linker arms were either difficult to prepare or did not offer any speed advantage.
A satisfactory alternative to the succinic acid linker should not require additional reagents or conditions which cannot be accommodated by present instrumentation. Furthermore, the linker arm should be readily available and compatible with the derivatization methods currently used for attaching nucleosides to solid phase supports. Difunctional carboxylic acids with various [alpha] or [beta] substitutions, to make them more labile than succinic acid, seemed most likely to meet these requirements. In particular, the very labile oxalic acid linker arm 2 seemed quite promising (11 ,19 ,21 -23 ). However, as described in this report, we found the oxalyl linker too unstable to be satisfactory and so a series of five other bifunctional acids were examined. Of these, hydroquinone-O,O'-diacetic acid (HQDA, 4a) was the most satisfactory (1 ). Nucleosides with this linker arm (`Q-linker') can be attached to solid phase supports and used for oligonucleotide synthesis in the same way as succinyl linker arms. However, the time required for NH4OH hydrolysis can be reduced to as little as 2 min. Furthermore, the Q-linker can also be cleaved with milder reagents, such as NH3/MeOH, K2CO3/MeOH, fluoride ion or t-butylamine. Therefore, HQDA can be used as a replacement for both succinyl and oxalyl linker arms since it is sufficiently stable for general use while still labile enough for use with base-sensitive oligonucleotides.
Long chain alkylamine controlled pore glass (LCAA-CPG; 120-200 mesh, 500 Å, 90-120 [mu]mol/g NH2 groups) was obtained from CPG Inc. (Lincoln Park, NJ) and underivatized highly cross-linked aminomethyl 1000 Å polystyrene (PS; LU6-009, 36 [mu]mol/g NH2 loading) was donated by Perkin/Elmer Applied Biosystems Division (PE/ABD). A 5'-dimethoxytrityl-2'(3')-O-acetyluridine-derivatized PS support with a succinic acid linker was purchased from PE/ABD. HQDA was obtained from Lancaster Synthesis Ltd (Lancashire, UK). Nucleoside-3'-O-succinates were purchased from Sigma or Chem-Impex (Chicago, IL). Cap A (acetic anhydride/2,6-lutidine/THF 1:1:8), Cap B (N-methylimidazole/THF 16:84) and I2/H2O oxidation (0.02 M I2 in THF/H2O/pyridine 88:2:10) reagents were prepared in house. FAB mass spectrometry was performed using a Kratos MS25RFA mass spectrometer with a nitrobenzyl alcohol matrix. Oligonucleotide synthesis was performed using a PE/ABD 394 DNA synthesizer. Dimethoxytrityl analysis was performed by measuring the absorbance at 505 nm ([epsilon]505 = 76 ml/[mu]mol/cm) in 5% dichloroacetic acid/1,2-dichloroethane (v/v) solution. HPLC analysis (of hemiester 5a and diester) was performed using a Whatman Partisil C-8 cartridge (4.6 * 125 mm) and isocratic elution with either 55 or 60% acetonitrile in water (2 ml/min).
Oxalyl chloride (2.5 mmol, 218 [mu]l) was added to a stirred, room temperature solution of triazole (12.5 mmol, 863 mg) in anhydrous pyridine (25 mmol, 2.0 ml) and anhydrous acetonitrile (17.5 ml) in a septum-sealed 50 ml vial. A solution of protected nucleoside 1 (2.5 mmol) in anhydrous pyridine (2.5 ml) and acetonitrile (17.5 ml) was then added, via syringe. After stirring for 30-45 min the solution was added to LCAA-CPG (10 g) in a 100 ml flask. The slurry was occasionally agitated for 30 min and then anhydrous methanol (50 ml) was added. After 5 min the CPG was filtered off, washed with CHCl3 and dried. The support was capped with 1:1 Cap A/Cap B solutions (30 min), then washed, dried and assayed for DMT content (~40 [mu]mol/g).
A mixture of 3,3'-thiodipropionic acid (28 mmol, 5 g) and trifluoroacetic acid (30 ml) was stirred at room temperature (5 min) and then cooled in ice/water. Aqueous H2O2 (30%, 4.8 ml) was added dropwise. The internal temperature was controlled at ~40oC by repeated immersion in ice. This H2O2 addition was repeated twice or more at 30 min intervals. After the final addition the reaction mixture was stirred (37-40oC, 2 h) and then filtered and washed with water (until the filtrate was neutral). The white solid was dissolved in CH2Cl2, extracted once with 1% aqueous sodium bisulfite and twice with water. The organic phase was dried over magnesium sulfate, filtered and evaporated to dryness to yield 4 g (68%) of crude 4c.
Protected nucleoside 3 (B = N6-benzoyladenine, N4-benzoylcytosine, N2-isobutyrylguanine or thymine, R' = H, 2.0 mmol), HQDA (2.4 mmol), 4-dimethylaminopyridine (DMAP, 0.2 mmol), 1-(3- dimethylaminopropyl)-3-ethylcarbodiimide hydrochloride (DEC, 2 mmol) and triethylamine (0.2 ml) were dissolved in anhydrous pyridine (20 ml) and stirred at room temperature (16 h). The reaction was checked by TLC (10% MeOH/CHCl3) and if any unreacted nucleoside remained, additional DEC was added and the reaction continued. Once nucleoside 3 was consumed the reaction was worked up in one of the following three ways:Simplified work-up. The pyridine solution was evaporated to an oil, redissolved in CHCl3 (50 ml) and washed with H2O (three times). The CHCl3 solution was then evaporated to yield a brown foam containing ~80% hemiester 5a (as the pyridinium salt), ~20% diester and residual HQDA.Work-up with bicarbonate extraction. The pyridine solution was evaporated to an oil, redissolved in CHCl3 (100 ml), washed with H2O (1 * 50 ml) and then either saturated aqueous NaHCO3 or NH4HCO3 (3 * 50 ml). The resulting mixture of liquid phases was separated by centrifugation. The CHCl3 solution was washed again with H2O (1 * 50 ml), dried over anhydrous MgSO4 and evaporated to a foam. The crude mixture of hemiester 5a (as either the Na+ or NH4+ salt) and diester was redissolved in pyridine (10-20 ml) and stirred (5 min) with BioRad AG 50W-X4 H+ form ion exchange resin (~5 ml). The pyridine solution was filtered off and evaporated to a brown foam to yield a mixture of ~80% hemiester 5a (as the more soluble pyridinium salt) and ~20% diester.Work-up with silica gel column purification. The pyridine solution was treated as described in the simplified work-up and the chloroform solution of crude material was applied to a silica gel column (40 g KP-Sil prepacked Flash-40 cartridge). The column was eluted sequentially with CHCl3 (500 ml), 5% EtOH/CHCl3 (500 ml) and 30% EtOH/CHCl3 (500 ml) to yield the pure hemiester 5a (yield ~50%).
When the above reaction was increased in scale (i.e. to 10 mmol) the emulsion formed during the bicarbonate extraction was difficult to separate and was precipitated in hexanes. The sticky white precipitate was filtered off, redissolved in CHCl3, washed again (if necessary), dried over MgSO4 and evaporated to a foam. The Na+ or NH4+ salt was then converted to the more soluble pyridinium salt, as described above.Physical data. For 5a, B = N6-benzoyladenine, R' = H: Rf (20% MeOH/CHCl3) 0.27; UV (EtOH) 229, 282 nm; mass (C48H43 N5O11), calculated 865.90, M+ 865.11. For 5a, B = N4-benzoylcytosine, R' = H: Rf (20% MeOH/CHCl3) 0.31; UV (EtOH) 229, 262 nm; mass (C47H43N3O12), calculated 841.88, MH+ 842.21. For 5a, B = N2-isobutyrylguanine, R' = H: Rf (20% MeOH/CHCl3) 0.19; UV (EtOH) 229, 262, 283 nm; mass (C45H45N5O12), calculated 847.89, M+Na+ 870.32. For 5a, B = thymine, R' = H: Rf (20% MeOH/CHCl3) 0.32; UV (EtOH) 227, 269 nm; mass (C41H40N2O12), calculated 752.79, MH+ 753.24. Satisfactory 1H NMR spectra (200 MHz) were also obtained for the above compounds.
Dicarboxylic acid 4b-d (1 mmol), 5'-dimethoxytritylthymidine (1 mmol, 544 mg), DMAP (0.1 mmol, 12 mg), DEC (1 mmol, 192 mg) and triethylamine (0.08 ml) were dissolved in anhydrous pyridine (25 ml). After 1-2 days stirring the pyridine was removed by evaporation and the residue was redissolved in CHCl3. The solution was washed with water (three times), dried over magnesium sulfate, filtered and evaporated to yield the crude product 5b-d as a light brown solid. The products were coupled to CPG supports without further purification.
Unpurified nucleoside-3'-O-hydroquinone-O,O'-diacetyl hemiesters 5a (B = N6-benzoyladenine, N4-benzoylcytosine, N2-isobutyrylguanine or thymine, R' = H, 2.0 mmol), DMAP (0.5 mmol, 60 mg), DEC (5 mmol, 0.96 g), triethylamine (0.5 ml), LCAA-CPG (5 g) and anhydrous pyridine (~50 ml) were combined in a 100 ml flask. The flask was shaken at room temperature for 4.5 h. The CPG was filtered off, washed with methanol and CH2Cl2 and capped with 1:1 Cap A/Cap B solutions (2 h), washed and dried. Nucleoside loading determined by DMT analysis for dA, dG, dC and T supports respectively were 39, 32, 39, and 34 [mu]mol/g.
Unpurified 5a (B = uracil, R' = O-t-butyldimethylsilyl) was used as above, but on one-fifth scale to yield 6a with a loading of 30 [mu]mol/g, while 5a (B = uracil, R' = OH) yielded a loading of 16 [mu]mol/g.
Good results can also be obtained using only half as much 5a, if the volume of solvent is also reduced by half. For example, 5a (B = thymine, R' = H, 0.2 mmol), DMAP (0.1 mmol), DEC (1.0 mmol), triethylamine (0.1 ml) and LCAA-CPG (1 g) were shaken in pyridine (5 ml) at room temperature (4 h) to yield a nucleoside loading of 46 [mu]mol/g.
5a (B = thymine, R' = H, 0.2 mmol, 0.2 g), DMAP (0.1 mmol, 12 mg), DEC (1 mmol, 0.19 g), triethylamine (0.1 ml), aminomethyl polystyrene (0.5 g) and anhydrous pyridine (7 ml) were shaken at room temperature (16 h). The support was filtered off and washed with methanol and CH2Cl2, washed and dried. Nucleoside loading was determined by DMT analysis to be 31 [mu]mol/g. Ribonucleoside derivatives 5a (B = uracil, R'= O-t-butyldimethylsilyl or OH) were similarly prepared with nucleoside loadings of 18 (R' = OH) and 29 (R' = O-t-butyldimethylsilyl) [mu]mol/g.
A mixture of LCAA-CPG (500 mg), 1-hydroxybenzotriazole (HOBT, 0.015 mmol, 2 mg), diisopropylcarbodiimide (DIC, 0.15 mmol, 24 [mu]l), pyridine (0.1 ml) and anhydrous acetonitrile (2 ml) were shaken in a sealed glass vial at room temperature (20 min). Either 5a or 5e (B = thymine, R' = H, 0.05 mmol) was then added and the mixture shaken overnight. The CPG was filtered off, washed with MeOH and CHCl3, dried and unreacted amino groups were acetylated with Cap A and Cap B solutions (5 ml each, 60 min). DMT analysis indicated nucleoside loadings of 56 and 70 [mu]mol/g for the Q- (6a) and succinyl (6e) linker arms respectively.
Succinic anhydride 7a or diglycolic anhydride 7b (2 mmol), DMAP (0.2 mmol), LCAA-CPG (1 g) and anhydrous pyridine (10 ml) were shaken at room temperature (24 h). The support was filtered off, washed with methanol and CHCl3 and dried. Carboxyl loadings of 90 and 112 [mu]mol/g were obtained for 8a and 8b respectively.
LCAA-CPG (0.5 g), HQDA (2 mmol), DEC (2 mmol), DMAP (0.1 mmol) and triethylamine (0.1 ml) were shaken in anhydrous pyridine (8 ml) at room temperature (5 days). The support was filtered off, washed with MeOH and CHCl3 and dried. A carboxyl loading of 56 [mu]mol/g was obtained.
6-Amino-1-hexanol (20 mmol, 2.42 g) and monomethoxytrityl chloride (20 mmol, 6.37 g) were dissolved in anhydrous pyridine (100 ml) and stirred at room temperature (2 days). The solution was concentrated by evaporation, redissolved in CHCl3 (75 ml), washed with aqueous NaHCO3 (once) and H2O (twice) and dried over MgSO4. The CHCl3 solution was concentrated, applied to a silica gel column and eluted with 5-15% MeOH/CHCl3 to yield the pure product (5.5 g) in 71% yield. Rf (20% MeOH/CHCl3) 0.14; mass (C26H31NO2), calculated 389.54, m/z 389.2; 1H-NMR, [delta]7.50-7.20 (12H, aromatic), [delta]6.90-6.80 (2H, aromatic), [delta]3.80 (3H, -OCH3), [delta]3.07 (2H, -CH2-OH), [delta]2.73 (2H, -CH2-NH), [delta]2.26 (2H, OH/NH), [delta]1.70-1.20 (8H, CH2).
For HQDA- or succinyl-derivatized supports carboxyl-CPG (25 mg), N-monomethoxytrityl-6-amino-1-hexanol (10 mg), O-(benzotriazol-1-yl)-1,1,3,3-tetramethyluronium hexafluorophosphate (HBTU, 10 mg), DMAP (3 mg), triethylamine (10 [mu]l) and CH2Cl2 (0.5 ml) were shaken at room temperature (1 h). The CPG was filtered off, washed (MeOH, CH2Cl2) and dried. The absorbance of an accurately weighed sample (4 mg) in 5% dichloroacetic acid/1,2-dichloroethane (v/v, 10 ml) was measured at 480 nm ([epsilon]480 = 56 ml/[mu]mol/cm).
5'-Dimethoxytritylthymidine (0.2 mmol, 0.11 g), DMAP (0.1 mmol, 12 mg), diglycolic acid-CPG 8b (1 g), DEC (1 mmol, 0.19 g), triethylamine (0.1 ml) and anhydrous pyridine (10 ml) were shaken at room temperature (16 h). Pentachlorophenol (0.6 mmol, 160 mg) was added and shaking was continued (24 h). The CPG was filtered off, washed with pyridine, treated with piperidine (5 ml, 5 min), washed and capped with 1:1 Cap A/Cap B (20 ml, 2 h). The fully capped support was washed with CH2Cl2 and dried (loading 56 [mu]mol/g). Similarly, 5'-dimethoxytritylthymidine (0.1 mmol), DMAP (0.1 mmol), HQDA-CPG 8c (0.5 g), DEC (0.5 mmol) and triethylamine (50 [mu]l) in anhydrous pyridine (2.5 ml) were reacted to produce 6g (loading 18 [mu]mol/g).
The stability of linker arms attached to nucleosides was determined by treating portions of support (~10-20 mg) with cleavage reagent for set times. The supports were washed with methanol and CH2Cl2, dried and assayed for DMT content. The percentage cleavage was calculated from the initial nucleoside loading. The stability of linker arms attached to oligonucleotide sequences was obtained by running one of two custom end procedures on the ABI 394 DNA synthesizer. The first end procedure allowed 10 NH4OH or 3:1 NH4OH/ethanol fractions to be consecutively collected at 1 min intervals. The second end procedure produced 12 consecutive fractions at 15 min intervals. The collected fractions were deprotected (16 h, 55oC), evaporated to dryness, redissolved in water and quantitated at 260 nm.
The oxalic acid linker 2 (Scheme 1) was the first investigated. The published synthesis procedure (21 ) was scaled up to a 10 g scale and supports with all four of the common deoxynucleosides were prepared. Cleavage of the oxalyl linker by NH4OH at room temperature was very fast. More than 90% cleavage occurred within the first 20 s and hydrolysis was complete within 60 s or less. The oxalyl-CPG was used to satisfactorily prepare >3000 oligonucleotides. However, there were two major problems with the oxalyl linker.
The preliminary tests clearly indicated that the Q-linker was the most satisfactory and so the properties of this linker arm were investigated in more detail. The first step in synthesis was formation of a nucleoside-3'-O-hemiester (5a). Unlike succinic acid, HQDA was not available as an anhydride. However, attachment of the commercially available diacid to a nucleoside was easily performed using DEC and DMAP to yield a mixture of the desired hemiester and a lesser amount of inert diester. The diester, which was readily detected as a faster moving product on TLC, was a consequence of the bifunctional nature of HQDA. However, use of only 1 equiv. DEC and a slight excess of HQDA (1.2 equiv.) limited the amount of nucleoside converted to diester to ~20% or less (as determined by HPLC).
The diester byproduct was unreactive in subsequent couplings and we found that chromatographic separation of the hemiester from the diester was unnecessary. This was convenient because nucleosides with the Q-linker (5a) were more polar than the corresponding nucleoside-3'-O-succinates (5e). For example, on silica gel TLC plates in 20% MeOH/CHCl3 the relative mobilities (Rf) were ~0.2-0.3 and 0.6-0.8 for 5a and 5e respectively. Instead, work-up with an aqueous wash to remove the urea byproduct from the DEC coupling reagent and a bicarbonate wash to remove residual HQDA was sufficient. However, the extraction with bicarbonate was troublesome because a difficult to break emulsion formed and because the resulting Na+ or NH4+ salts were less soluble. Therefore, preliminary experiments omitted the bicarbonate extraction and subsequent preparations used an ion exchange resin to convert the Na+ or NH4+ salts into the more soluble pyridinium form.
LCAA-CPG supports with the Q-linker and all four common deoxynucleosides and two ribonucleosides (6a, B = uracil, R' = O-t-butyldimethylsilyl or O-acetyl) were prepared as shown in Scheme 2. In addition, highly cross-linked aminomethyl 1000 Å polystyrene (PS), was also derivatized with the Q-linker and 5'-dimethoxytritylthymidine and the above uridine ribonucleosides. Attachment of the nucleoside-3'-O-hemiester (5a) to the amino-functionalized surface was readily accomplished using the same base-catalyzed carbodiimide coupling reaction (30 ) as previously employed for nucleoside-3'-O-succinyl hemiesters (5e). 2'-Deoxynucleoside loadings of ~30-40 [mu]mol/g and 2'-O-t-butyldimethylsilylated ribonucleoside loadings of 30 [mu]mol/g were obtained on LCAA-CPG (6a)when 0.4 mmol nucleoside/g CPG was employed in early experiments. Similar results were also obtained when the PS support was used, even though the PS surface amino functionalization was much less (36 [mu]mol/g).
A number of experiments were conducted to optimize the support derivatization with respect to speed and nucleoside consumption. The most important parameter, aside from coupling reagent, was nucleoside concentration. This should be maintained as high as possible but is limited by the volume required to suspend the insoluble support (~4-5 ml/g CPG). For example, an overnight reaction using only 0.1 mmol nucleoside 5a/g CPG and the maximum nucleoside concentration (0.025 M) produced a loading of 35 [mu]mol/g. However, use of either 0.2 or 0.4 mmol 5a/g CPG at 0.04 and 0.08 M respectively produced faster results (36-40 [mu]mol/g within 1-2 h) and higher nucleoside levels after overnight coupling (46-48 [mu]mol/g). Use of a large excess of nucleoside but at a more dilute concentration (0.013 M) was less efficient than a smaller excess of nucleoside at a higher concentration. The DEC/DMAP coupling reaction could also be performed in dichloromethane or anhydrous acetonitrile instead of pyridine, but supports derivatized in pyridine gave the best results.
Recently a very efficient derivatization procedure using DIC and an acid catalyst, HOBT, has been described (31 ) for producing high loaded succinylated CPG supports. When this method was evaluated using both 6a and 6e (B = thymine, R' = H) respective loadings of 56 and 70 [mu]mol/g were obtained. These high nucleoside loadings confirmed the effectiveness of the DIC and HOBT reagent combination.
An alternative route to immobilized nucleosides involves the formation of a carboxyl functionalized support (Scheme 3, 8a-c). The reaction of LCAA-CPG with succinic anhydride is easy (24 ) and high carboxyl functionalization (~100 [mu]mol/g) is readily obtained. However, because HQDA was not available as an anhydride, a carbodiimide coupling reagent was required. Attempts to determine the COOH functionalization, using derivatization with p-nitrophenol (8 ,24 ) were not successful and we had to develop a better derivatization assay for carboxyl groups. This involved the synthesis of N-monomethoxytrityl-6-amino-1-hexanol from monomethoxytrityl chloride and 6-amino-1-hexanol. Esterification of this MMT-labeled alchohol to an immobilized carboxyl group was performed using the uronium coupling reagent HBTU and DMAP. This indicated an HQDA loading of 56 [mu]mol/g. However, coupling of 5'-dimethoxytritylthymidine to support 6g using DEC/DMAP only produced a relatively low nucleoside loading (18 [mu]mol/g). Therefore, this approach using the Q-linker was not as satisfactory as with the succinyl linker.
Samples of 2 and 6a were treated with either capping or oxidation reagents to confirm that premature cleavage would not occur during synthesis. The oxalyl linker arm showed losses of 10-20 and 20-30% upon overnight treatment with the capping and oxidation reagents respectively. The Q-linker was not affected at all by similar exposure to the capping reagent and only a small (8%) loss occurred with the oxidation reagent. Since exposure of the linker arm to these reagents in a typical oligonucleotide synthesis is ~100-200 times less than in the above test, the amount of premature cleavage will be negligible, especially when the Q-linker is used. More importantly, CPG samples of 6a derivatized with either dA, dC or T showed no detectable decrease in loading after 2 years storage at room temperature.
The ease of hydrolysis of the Q-linker also allows it to be used in applications where rapid removal of N-protecting groups is required or where NH4OH is too harsh. Examples of this latter category include synthesis of oligonucleotides with methylphosphonate (22 ,23 ,33 ) and sulfonate (34 ) linkages, fluorescent rhodamine dyes (35 ) and 2-pyrimidinone bases (23 ), partially N- or P-protected oligonucleotide sequences (21 ,22 , 36 ,37 ) and strategies which use fluoride ion to cleave deprotected (14 ,15 ) or partially purified oligonucleotides (16 ) from the support. Therefore, the rate of nucleoside cleavage from a Q-linker was determined using some of the conditions employed in the above studies (Table 2 ).
Not surprisingly, the rate of cleavage using the potent AMA mixture (methylamine/NH4OH) was very fast and rapid base deprotection as well as cleavage from the support was possible with this reagent (12 ). However, of greater interest was the suitability of the Q-linker in synthesis of more base-sensitive oligonucleotides. For example, methylphosphonate or methylphosphotriester backbone-modified oligonucleotides are rapidly degraded upon treatment with NH4OH and methanolic potassium carbonate has been used instead (36 ,37 ). When this reagent was used with the Q-linker hydrolysis was surprisingly fast (15 s). For comparison, cleavage of a levulinyl ester, similar to a succinyl linker, with 0.05 M K2CO3 required 4 h (36 ). Cleavage of the Q-linker with other mild bases was also easily accomplished. Room temperature cleavage of the support was possible using t-butylamine in only 5 min. Similar treatment of a succinylated linker for 60 min only resulted in 71% cleavage (35 ). Almost as fast cleavage (15 min) was also possible using anhydrous ammonia in methanol, a reagent preferred for oligoribonucleotide deprotection (38 ). Finally, the 98% cleavage of the Q-linker obtained using 5% NH4OH/MeOH for 60 min was comparable to the 70 min treatment used to cleave and deprotect base-sensitive fluorescent 2-pyrimidinone bases from oxalyl-linked supports (23 ). Therefore, the Q-linker should be a satisfactory replacement in any of the above applications.
In the past cleavage of a linker arm with fluoride ion has been limited to oxalyl- or silyl-linked supports because the succinyl linker arm is cleaved very slowly (~30% cleavage after 1 h in 1 M TBAF/THF). Therefore, it was very interesting to find that a 15 min treatment with 1 M TBAF completely cleaved the Q-linker. Faster cleavage was possible using neat triethylamine trihydrofluoride, but this reagent also dissolved the CPG particles upon prolonged (~1 h) exposure. Since supports with silyl linker arms are generally difficult to synthesize, the more easily prepared Q-linker arm is a better choice for applications requiring fluoride ion release of base-sensitive oligonucleotides.
Finally, it was interesting to determine the stability of the Q-linker to certain mild reagents (Table 2 , last three entries). In the first case resistance to cleavage by either piperidine or DBU was examined to determine if the Q-linker could be used in combination with the base-labile Fmoc protecting group. We found that the Q-linker was only slowly hydrolyzed by 20% piperidine (t1/2 ~ 3 h) or 0.5 M DBU (t1/2 ~ 16 h). Since Fmoc cleavage occurs much faster (39 ), selective Fmoc removal should be possible in the presence of a Q-linker. This would be useful in the synthesis of oligonucleotide-peptide chimeras (40 ,41 ).
Table 2
The Q-linker was more resistant to DBU cleavage than the succinyl linker, because the larger size and decreased flexibility of the Q-linker makes intramolecular nucleophilic attack of the deprotonated amide onto the ester more difficult. For example, on LCAA-CPG a 6 h treatment with 0.5 M DBU/pyridine caused 21 and 66% cleavage (19 ) respectively of the Q- and succinyl linkers. Use of the Q-linker instead of a succinyl linker should therefore improve the NPE/NPEOC strategy (42 ), which uses DBU to remove p-nitrophenylethyl (NPE) and p-nitrophenylethoxycarbonyl (NPEOC) base protecting groups from oligonucleotides prior to a 2 h NH4OH cleavage from the support.
Finally, an improved procedure for the synthesis of long oligonucleotides has recently been reported (16 ). In this procedure 1:1 triethylamine/ethanol was used (3 h, room temperature) to cleave apurinic sites in oligonucleotides immobilized on a CPG support via a disiloxyl linker arm. Less than 10% loss of intact oligonucleotides from the support was reported and selective hydrolysis greatly simplified the subsequent isolation of full-length oligonucleotide. Since this selective cleavage of apurinic sites has general utility, we also tested the Q-linker under these conditions. As shown in Table 2 , hydrolysis with 1:1 triethylamine/ethanol was very slow and we found that even after 3 h treatment only 6% cleavage had occurred. Therefore, the Q-linker was also sufficiently stable to permit on-column cleavage of apurinic sites and the lengthy procedure required to synthesize the disiloxyl linker arm can be avoided.
A major advantage of the hydroquinone-O,O'-diacetic acid linker was the fact that no other changes in the synthesis and deprotection procedures, other than a decreased cleavage time, were required. Unlike changes to the deprotection reagent, which may cause base modifications (11 ,12 ,18 ,19 ), substitution of the linker cannot have any effect on the oligonucleotide structures assembled beyond the first base. Thus coupling yields for phosphoramidite reactions were unaffected and we have prepared >8000 different oligonucleotides satisfactorily. Furthermore, to confirm that no modification or blocking of the 3'-terminus occurred a M13 universal sequencing primer (dCGCCAGGGTTTTCCCAGTCACGAC) was prepared using 6a. After synthesis, deprotection and purification the sequencing primer was tested in an automated DNA sequencing protocol. The results (not shown) indicated sequencing efficiency equivalent to primers made on 6e (99% accuracy).
Hydroquinone-O,O'-diacetic acid is a suitable replacement for both succinyl and oxalyl linker arms. This linker arm is commercially available, inexpensive and easily coupled to insoluble supports using the same coupling methods currently used for succinyl linker arms. The Q-linker is readily cleaved by a variety of mild bases but is sufficiently stable that no accidental cleavage, during either oligonucleotide synthesis or storage, occurs. The short (2 min) cleavage time and the fact that no other changes, to either reagents, synthesis or deprotection protocols, are required will allow a significant increase in synthesis productivity.
We would like to thank Alex Andrus (Perkin-Elmer Applied Biosystems Division) for donation of aminomethyl polystyrene and Maria Loskot, Michael Nodwell and Wade Stout for technical assistance.
*To whom correspondence should be addressed. Tel: +1 403 220 4277; Fax: +1 403 283 4907: Email: rtpon@acs.ucalgary.ca
The four dicarboxylic acids 4a-d in Scheme 2 were coupled to DEC/DMAP to yield the 3'-O-hemiesters 5a-d, which were then attached to LCAA-CPG. This coupling reaction produced nucleoside loadings of between 24 and 35 [mu]mol/g for compounds 5a-d. In another approach (Scheme 3) succinic (7a) and diglycolic anhydride (7b) were reacted first with LCAA-CPG (24 ) and then DEC/DMAP was used to attach 5'-dimethoxytritylthymidine to the carboxyl-CPGs (8a-b).
Reagent (room temperature)
Amount of cleavage (%)
t1/2
1 min
5 min
15 min
60 min
40% Aqueous MeNH2/NH4OH (1:1, AMA)
100
<10 s
0.05 M K2CO3/MeOH
100
<10 s
Triethylamine trihydrofluoride (neat)
46
100
~1 min
1 M Tetrabutylammonium fluoride/THF
36
83
99
~2 min
t-Butylamine/MeOH/H2O (1:2:1)
66
97
~1 min
NH3/MeOH (saturated)
27
75
98
~3 min
5% NH4OH/MeOH
9
27
63
98
~11 min
20% Piperidine/DMF
15
~3 h
0.5M (7.5%) DBU/pyridine
7
~16 h
Triethylamine/EtOH (1:1)
2
4
5
REFERENCES