Rachel Quarrell, Timothy D.W. Claridge, George W. Weaver and Gordon Lowe*.
Dyson Perrins Laboratory, Oxford University, South Parks Road, Oxford. OX1 3QY, UK.
SUMMARY
In view of the widespread use of TentaGel resin beads for the synthesis of combinatorial libraries, the properties of TentaGel resin have been examined using a combination of confocal laser microscopy and NMR spectroscopy. Evidence is presented that trypsin, a 23.5 kDa enzyme, can penetrate to the core of 90-µm TentaGel beads, and that the matrix of such beads permits molecular motion at a similar rate to that in solution. The beads act as a separate gel phase rather than as a porous solid. These conclusions have important implications for the bioassay of on-bead combinatorial chemical libraries.
ADDRESSES in full Dr. Rachel Quarrell, Dyson Perrins Laboratory, South Parks Road, Oxford OX1 3QY, United Kingdom. Tel +44-1865-275686, fax +44-1865-275674 and email quarrell@vax.ox.ac.uk Dr. Timothy D.W. Claridge, Dyson Perrins Laboratory, South Parks Road, Oxford OX1 3QY, United Kingdom. Tel +44-1865-275658, fax +44-1865-275674 and email tim.claridge@dpl.ox.ac.uk. Dr. George W. Weaver, Dyson Perrins Laboratory, South Parks Road, Oxford OX1 3QY, United Kingdom. Tel. +44-1865-275686 and fax +44-1865-275674. Professor Gordon Lowe, Dyson Perrins Laboratory, South Parks Road, Oxford OX1 3QY, United Kingdom. Tel +44-1865-275649, fax +44-1865-275674 and email gordon.lowe@dpl.ox.ac.uk.
*To whom correspondence should be addressed.
KEYWORDS
COMBINATORIAL LIBRARIES TENTAGEL RESIN BEADS 31P-NMR SPIN-LATTICE RELAXATION TRYPSIN CONFOCAL SCANNING LASER MICROSCOPE
Abbreviations are noted in brackets after the first use of the phrase in the text.
INTRODUCTION
One of the most extensively-used supports for solid-phase peptide synthesis1 (SPPS) is the beaded resin, comprizing a polystyrene or polyamide microsphere, functionalized with an amine or alcohol anchor for the C-terminus of a polypeptide chain. For standard SPPS, mechanical stability, incompressibility and an efficient loading per gramme are the most important qualities required, leading to rigid beads of average size approx. 50 µm diameter.
In 1991, Furka et. al.,2 Lam et. al.3 and Houghten et. al.4 published the first reports describing a new synthetic approach, producing peptide synthetic combinatorial libraries (SCL's). The strategy for the production of an SCL where each bead carries a single chemical structure has been well documented and reviewed elsewhere.5-8 The microscopic beads of traditional SPPS lend themselves well to the "divide, couple, recombine" approach. However, if advantage is to be taken of this approach in the field of drug discovery the beads must also be compatible with the conditions for bioassay. Excellent results have been claimed for TentaGel (TG) resin,9-11 and we describe here new evidence in support of its unusual properties. A two-part investigation was undertaken. First the penetration of a fluorescein-labelled enzyme into the beads was investigated using a confocal scanning laser microscope (CSLM), and secondly the effects of the physical environment within the beads, by NMR spectroscopy.
TentaGel (Rapp Polymere Ltd) is produced by a process in which polyethylene glycol (PEG) units of approximately 3000 daltons are co-grafted onto a low-cross-linked polystyrene. This results in a relatively hydrophilic support, which swells both in water and in most polar organic solvents. These properties, which are not common to SPPS resins, are ideal for synthesizing combinatorial libraries, where organic solvents are used, followed by bioassay in aqueous media. It is important therefore that the internal sites are accessible for the bioassay as well as for the synthesis. The number of sites in a bead available to a reacting molecule will depend on the loading, the swelling of the matrix, steric hindrance from neighbouring sites and the size of the molecule itself.
Clearly the penetration of molecules into beads is an important factor controlling the validity and success of bioassays of resin-based combinatorial chemical libraries. Initial experiments had suggested that enzymes as large as porcine pancreatic trypsin (PPT) and human leucocyte elastase (HLE) were capable of penetrating the internal matrix of TG beads.12, 13 We now report experiments using a confocal laser fluorescence microscope to investigate further the penetration of fluorescent molecules into the pores of TG beads.
The confocal microscope, first described in 1961, uses a pinhole aperture to screen reflections of polarized light passing through a translucent specimen, to obtain an image which is limited to a single focal plane.14 The depth of this plane is typically set to around 0.5 µm. The result is that a precise image of the core of a translucent sample may be obtained. This technique was used to study TG resin beads. We had already investigated the binding of fluorescently-labelled PPT by TG beads carrying the sequence of Cucurbita Maxima trypsin inhibitor I (CMTI-I),13, 15 a 29-residue polypeptide with tight-binding trypsin-inhibiting properties.16, 17 This approach has now been extended to study the process of enzyme diffusion into and out of the beads.
We also report on the mobility of a-triphenylphosphonium p-toluic acid bromide salt (Compound P) attached to TG beads. The quaternary phosphonium salt is ideal for 31P NMR spectroscopic studies, giving a single resonance around 23 ppm. Accordingly, the NMR spectrum of Compound P, both in solution and attached to TG beads, has been investigated in different solvents.
MATERIALS AND METHODS
Machine-automated synthesis of the CMTI-I sequence onto TG beads has been described previously.13 High-resolution fluorescence data and images were obtained using a Nikon DIAPHOT-200 confocal scanning laser microscope driven by BioRad COMOS software on a linked PC computer. COMOS images were opened into Adobe Photoshop v. 2.5 for processing, and reproduced as laserprints. Electrospray mass spectra were recorded by Dr. R. Aplin (DP Laboratory, Oxford) on a VG-BIOQ spectrometer. 31P NMR spectra were recorded on a Bruker AM250 multinuclear spectrometer operating at 101.26 MHz and referenced externally to phosphoric acid, or on a Bruker AMX500 at 202.52 MHz, locked to D2O with no internal reference. Reagent-grade chemicals were obtained from Sigma-Aldrich UK Ltd, Lancaster or BDH Ltd, and peptide synthesis reagents from DuPont-NEN UK, Rathburns UK Ltd and Novabiochem Ltd. TentaGel resin (NH2-TG, 90-µm, 0.25-0.29 mmolg-1 NH2) was obtained from Rapp Polymere Ltd, Germany.
Acetyl-TG resin:- TG resin (50 mg, 0.29 mmolg-1 substitution) was placed in a flask and swollen with 2 ml DMF. Acetic anhydride (1.0 ml, 21 mmol) and diisopropylethylamine (DIPEA, 2.0 ml, 5.8 mmol) were added in DMF (20 ml), and the suspension stirred at room temperature for 2 hours. Reagents and solvents were removed by filtration, the Ac-TG resin washed well with MeOH and a ninhydrin test used to determine complete acetylation.
Fluorescein porcine pancreatic trypsin (FITC-PPT):- Porcine pancreatic trypsin (0.20 mg, 7.2 nmoles in 0.2 ml deionised water) was reacted with an excess of fluorescein isothiocyanate isomer 1 (FITC-1) (0.5 mg, 0.64 µmoles in 0.5 ml H2O), by vigorous shaking over 10 mg celite for 10 min. The celite was filtered off and the filtrate applied to a Sephadex G25 gel filtration column (30x1 cm, equilibrated with 20 mM Tris-Cl, pH 7.5). FITC-PPT was eluted with an aqueous NaCl gradient running from 0.05 to 0.2 M, collected and examined by UV at 405 and 492 nm to identify the labelled enzyme. The purity of the labelled enzyme was assessed by analytical HPLC on a reverse-phase 250x4.6 mm C18 Hypersil column, particle size 5 µm, gradient run from 10% to 90% acetonitrile in 0.1% TFA solution over 20 minutes at 0.6 ml/min (RT 18.85 mins), and the labelled enzyme lyophilized and stored at -20šC. ESMS, observed MH+ = 23848.07 daltons (FITC-PPT) from 23458 (pure PPT), calc. 23847.39.
Equilibration of FITC-PPT with beads (general method):- To a solution of the fluoresceinated enzyme (1.12 x 10-8 moles) in 500 µL buffer (50 mM Tris-Cl pH 8.0 with 20 mM CaCl2) was added a sample of approximately 750 beads (Ac-TG, blank TG or CMTI-TG), and the suspension allowed to equilibrate at 20šC for several hours. To quench and wash the beads, excess 40% aq. methanol was added to the reaction vessel, it was rapidly vortexed, the beads allowed to settle and the supernatant removed by pipette. This was repeated twice and the beads transferred by microcapillary into the wells of a Terasaki plate for examination. Alternative washes used included 1.0 M brine, acidified brine of pH 2.0, and the serine proteinase inhibitor phenylmethanesulfonyl fluoride (PMSF) in methanol (5 mg in 5 ml).
CSLM examination of bead samples (general):- Beads were examined in Terasaki microplate dishes (well capacity 10 µL), routinely suspended in 40% aq. methanol to aid matrix swelling and chemical solvation. Laser power was set to high, iris 2.0, zoom = 1.0, lexcitation = 488 nm (blue), lemission = 525 nm, and photomultiplier gain 1200. No other variable parameters affect fluorescence levels, with the exception of laser level, which is indicated. Fluorescence values were obtained using the COMOS software of the microscope, and were reproducible to + 5 fluorescence units.
a-Triphenylphosphonium-p-toluic acid bromide (Compound P):- Triphenyl phosphine (0.6295 g, 2.4 mmole) and a-bromo-p-toluic acid (0.43 g, 2.0 mmole) were dissolved together in solvent grade toluene (50 ml) and refluxed for 18 hr at 110šC. Cooling to room temperature yielded a white solid which was isolated by filtration, washed with toluene (5 ml) and dried under vacuum. The crude product (yield 96%, 1.028 g) was recrystallized from boiling water to give fine needles in a yield of 76%, m.p. 315šC. d1H(DMSO-d6) = 7.63-7.93 (m, 19H Ar), 5.30 (2H, CH2, JH,P=16.4), acidic proton not observed. d13C {1H}(DMSO-d6) = 167.5 (CO2H), 130-135 (22C, m, Ar groups), 21.8 (CH2). d31P (D2O) = 22.62 (s). ESMS, M+=397 (lacking Br -), calc. 397.1 daltons.
Symmetrical anhydride of Compound P:- a-Triphenylphosphonium-p-toluic acid bromide (0.191 g, 0.4 mmole) was dissolved in DCM:methanol (100:1, 25 ml) and dicyclohexylcarbodiimide (0.0454 g, 0.22 mmole) was added. Precipitation began at once and the reaction mixture was stirred at room temperature for a further 15 min, then the solvent removed by evaporation and the residue triturated twice with DCM. The solid was resuspended in DCM (10 ml), filtered through a sinter, and washed (3 x 5 ml DCM) to remove the urea by-product. The single white product released CO2 upon addition of aqueous NaHCO3, but did not react with thionyl chloride, indicating the anhydride functionality. This was confirmed by IR peaks at 1786 and 1713 cm-1 (C=O stretches). Yield 0.17 g, 88%, mp 343-344šC. The anhydride was used without further purification.
a-Triphenylphosphonium-p-tolyl-TG-resin:- The anhydride of Compound P (11 mg, 12 µmole) was dissolved in DMF (2.5 ml) and the solution circulated through TG resin (20 mg of 0.29 mmoleg-1 amine, 5.8 µmoles total) for 24 h until a Kaiser test showed complete coupling, followed by DCM/DMF washing. 31P NMR of the resin suspended in D2O (external reference against H3PO4) gave a signal at 22.6 ppm corresponding to +P-CAr.
Experimental evaluation of T1:- T1 relaxation times were obtained by the inversion-recovery sequence (180x-z-90x-acquire) performed on a Bruker spectrometer at 202.52 MHz, field strength 11.7 Tesla. Free compound P was dissolved in 90% isopropyl alcohol at 10.1 mg/ml unless otherwise stated, with 10% D2O as an internal lock, and the same solvent system was used as a suspension medium for on-bead acquisitions. All solutions were degassed by 3 cycles of freeze-pump-thaw. Time delays t were set as detailed in Figures 7 and 8, and the spin-lattice relaxation time T1 determined from fitting the resonance intensities to Equation 1 (Table 2), with standard Bruker T1 software routines (DISNMR or WINNMR).
I = I0.[1-2.Ae-t/T1] Equation 1 I = intensity of observed peak with delay t I0 = intensity of observed peak with t >> 5T1 (equilibrium intensity) A = constant to compensate for imperfect initial inversion of magnetization.
EXPERIMENTAL RESULTS
Confocal microscopy
Untreated TentaGel 90-µm beads (blank TG), together with 90-µm beads which had been acetylated (Ac-TG) were used as controls to identify non-specific binding to the bead matrix. TentaGel 90-µm beads onto which the 29-residue sequence of a natural proteinaceous proteinase inhibitor, Cucurbita Maxima trypsin inhibitor I (CMTI-1),16, 17 had been synthesized (CMTI-TG) were also investigated. This tightly-coiled molecule is a slow, tight-binding serine proteinase inhibitor, with specificity for pancreatic trypsin, against which it has a Ka of 3.2 x 1011 M-1.18 The binding enzyme was porcine pancreatic trypsin, labelled with one molecule of FITC-1 per molecule of enzyme, which did not interfere with its enzymic activity.
Background fluorescence levels of each sample of beads were first determined and the fluorescence levels across each set of beads found to be reasonably consistent, confirming that the bead matrix and loading was homogeneous (Figure 1). It has been noted before that treatment with chemical reagents can alter the background fluorescence properties of the resin, but they were all considerably lower than the fluorescence levels observed after treatment with FITC-reagents. Background levels did not vary more than + 2 fluorescence units in different solvents.
Next each set of beads was incubated separately with FITC-1. This reacts rapidly with free amino functions at room temperature, especially on a solid surface.19 A sample of 20% FITC-1 in methanol was employed. The reaction of FITC-1 with blank TG beads, which have free amino groups, gave a very high fluorescence level immediately, as did the reaction with CMTI-TG beads. Reaction of FITC-1 with Ac-TG gave rise to a lower level of labelling due to penetration of the fluorescent solution inside the bead matrix.
Next, samples of each set of beads were equilibrated for up to 72 hours with a large excess of FITC-PPT in 50 mM Tris-Cl, pH 8.0 + 20 mM CaCl2. Several beads were extracted from each set, washed with neutral brine and examined at various time intervals. Figure 2 shows the beads after washing had no further effect on the fluorescence intensity. While no binding was observed to the Ac-TG beads (not shown), the blank TG beads became modestly fluorescent at 10% laser intensity whereas the CMTI-TG beads became intensely fluorescent even at 3% laser intensity. These images clearly show the full penetration of FITC-PPT, an enzyme with molecular weight in excess of 23.5 kDa, to the centre of 90-µm beads. The fluorescence was removed by rinsing with acidified brine, which denatures the enzyme.20
A series of images was obtained of FITC-PPT penetrating into the beads (Figure 3), showing that the spread of the fluorescent enzyme between beads occurs very rapidly by gel-phase diffusion when beads are in contact, and by diffusion through the solution.
Finally, a sample of CMTI-TG beads which had been fluorescently labelled with FITC-PPT were treated with the irreversible serine proteinase inhibitor PMSF. This binds covalently to the labelled trypsin and prevents it from binding to the beads. Figure 4 shows the departure of FITC-PPT from the CMTI-TG beads into the solution.
NMR studies
Coupling of a a-triphenylphosphonium-p-toluic acid bromide (Compound P) (as its anhydride), formed from the condensation of triphenylphosphine with a-bromo-p-toluic acid,21 (Figure 5), allowed the environment within the TentaGel beads to be investigated by NMR spectroscopy.
31P NMR spectra of a-triphenylphosphonium-p-toluic acid bromide in various solvents and coupled directly to the amino function of TG resin in the same solvents were obtained on a Bruker AM250 spectrometer operating at 101.26 MHz. Chemical shifts (d) and the linewidth at half height (Dn1/2) are shown in Table 1. The values of d and Dn1/2 do not appear to be affected greatly by the choice of solvent. This allowed an optimum solvent mixture of 90% isopropanol with 10% D2O to be used for the investigation of the spin-lattice relaxation using a Bruker AMX500 spectrometer.
Linewidths
The phosphonium salt in free solution clearly has a narrower linewidth than when resin bound (Figure 6); the solvent effects are small. However, the identical line width of the 31P resonance of free compound P in the presence of beads compared with when it is covalently tethered to beads suggests that the line broadening is predominantly a function of the magnetic field inhomogeneity caused by the beads.22 Nevertheless, line-widths have been used to investigate the mobility of compounds attached to beads, using 13C NMR spectroscopy.23 It is clear from Figure 6 that the mobility of a tethered compound cannot be determined in this way. A much better guide to the mobility of molecules tethered to beads is the spin-lattice (longitudinal) relaxation time, T1.24
Spin-lattice relaxation times
The relaxation of spin systems when perturbed from equilibrium is dependent on the availability of local fluctuating magnetic fields, which in turn are sensitive to the motion of molecules. A comparison of relaxation rates (or times) for the free and resin-bound compound P, determined under identical conditions of field, temperature and solvent, can provide an insight into the relative motional freedom it possesses in the two states. The values presented in Table 2 suggest that no significant differences between the spin-lattice relaxation times of the free and bound molecules are observed. Figures 7 and 8 show typical Mz against t plots for these experiments, and Figure 9 shows the fitting of data to Equation 1, from which T1 values are derived.
Deuterium exchange
Spin-lattice relaxation experiments were carried out on one solution of compound P several times over the period of a month. In the later spectra two "decomposition" peaks were observed very close to the original signal (Figure 10). The chemical shifts and relaxation times of these two species are also worthy of note (Table 2, impurities 1 and 2). The shift differences from compound P are small (0.06 and 0.12 ppm respectively) suggesting that there are no gross differences in chemical environment for the phosphorus nucleus. Such small changes are consistent with H/D isotope shifts,25 which suggested the slow replacement of protons on the neighbouring CH2 with deuterons from the 10% D2O in solution, probably via the ylid. Mass spectral analysis of the sample (Figure 11) confirmed this hypothesis, showing a ratio of the CH2:CHD:CD2 isotopomers to be 30:20:50 at that point, which agrees with the composition estimated from the NMR spectrum. This suggests, as expected, that the exchange is subject to a positive kinetic isotope effect. There is also a steady increase in T1 values for the three species, suggesting the reduction of a relaxation pathway for the phosphorus nucleus. The dominant relaxation mechanisms for 31P nuclei are dipole-dipole interactions, chemical shift anisotropy and spin rotation.25 The increase in T1 is, therefore, consistent with the replacement of a proton with a deuterium at the adjacent CH2, with an associated reduction in dipole-dipole relaxation. The increase in T1 is even more marked in the CD2 species. This suggests that the dipole interaction with the adjacent CH2 group provides a substantial relaxation pathway for phosphorus in this molecule.
DISCUSSION
The experiments described here reveal physical properties of TentaGel beads which have not been previously studied. It is plain from the confocal microscope study with fluorescent probes of various sizes that not only small molecules but also biological molecules of considerable size (23,800 daltons) are capable of penetrating to the core of the bead. Moreover, they remain active while within the bead. Indeed to release the fluorescein-labelled trypsin from the CMTI-TG beads required strongly denaturing conditions. This confirms the potential of TG resin for on-bead biological assays, although care should be taken to eliminate background effects.
The fact that trypsin is able to penetrate TG beads should not be a surprise, but neither should it be taken as an indication that all enzymes or larger biomolecules will penetrate TentaGel. In their current form, the TG beads have an average loading of around 0.25 mmoles per g. On the 90-µm beads used for this work, this gives a total of 6.02 x 1011 sites within 2.7 ¼ x 10-14 m3, i.e. an approximate distance between two adjacent sites of 5 nm when the beads are unswollen and approx. 7.5 nm when swollen. Since each site is at the end of a long PEG spacer, this is a crude estimate of the maximum space between chains and therefore of the pore size. However, these calculations and our observations with FITC-PPT appear to be at variance with the conclusions drawn by Barany and co-workers,26 who claimed that a-chymotrypsin hydrolyses a TG-bound substrate solely at the external sites. a-Chymotrypsin is ellipsoidal with dimensions 51Åx40Åx40Å27, but ß-trypsin has similar molecular dimensions.28, 29 a-chymotrypsin, however, is known to dimerize at high enzyme concentration and ionic strength.30, 31 Most of the associated species are dimers, but higher polymeric species also exist.32 Since Barany et al.26 used a 0.1 M ammonium carbonate buffer it seems highly probable that their "shaving" experiments succeeded due to dimerization of the a-chymotrypsin; these species, which have unimpaired activity, being unable to penetrate the bead pores in accord with the above approximate calculation of pore size.
A comparison of the 31P NMR spectra of compound P, resin-bound compound P and free compound P in the presence of blank beads (Figure 6) demonstrates that magnetic susceptibility differences at the bead:solvent interface degrade the magnetic field homogeneity and give rise to line broadening. Linewidths are not, therefore, considered to be a realistic guide to the molecular mobilities within a bead. The similar spin-lattice relaxation times for the free and bound phosphonium salts suggests that the free and bound ligands have similar mobility, presumably due to essentially unrestricted segmental motion at the end of the PEG spacer.
These experiments give us further evidence that TentaGel resin provides an environment which does not interfere with the normal behaviour of biomolecules in aqueous solution,33 Indeed the environment within the bead is clearly dominated by the solvent used and not by the cross-linked polymer matrix. However, the diffusion of the fluorescent label between beads (Figure 3) shows that the solvated bead acts as if it were a second liquid phase, allowing enhanced rates of diffusion when beads are in contact.
CONCLUSIONS
The experiments presented here contain direct evidence for the penetration of large biomolecules into TentaGel beads, and that the environment within the beads is similar to that in solution. These conclusions have important implications for the burgeoning field of combinatorial chemistry, since there is widespread concern that the solid support may interfere with a bioassay if the SCL is assayed while attached to beads. These concerns may be justified if the bioassay involves the use of very large biomolecules, but for bioassays involving macromolecules of less than about 50 Å diameter the advantages of on-bead SCL screening should not be abandoned lightly. The unusual PEG-PS structure of TG beads appears to be well adapted to such procedures.
ACKNOWLEDGEMENTS
The authors would like to thank the EPSRC and BBSRC for support of this work.