Maartje M. C.
Bastings
ab,
Tom F. A.
de Greef
ac,
Joost L. J.
van Dongen
c,
Maarten
Merkx
b and
E. W.
Meijer
*abc
aInstitute for Complex Molecular Systems, Eindhoven University of Technology, P.O. Box 513, 5600 MB Eindhoven, The Netherlands. E-mail: e.w.meijer@tue.nl
bLaboratory of Chemical Biology, Eindhoven University of Technology, P.O. Box 513, 5600 MB Eindhoven, The Netherlands
cLaboratory of Macromolecular and Organic Chemistry, Eindhoven University of Technology, P.O. Box 513, 5600 MB Eindhoven, The Netherlands
First published on 18th March 2010
AB type monomers for supramolecular polymers have been developed based on the strong and reversible noncovalent interaction between ribonuclease S-peptide (A) and S-protein (B), resulting in an active enzyme complex as the linking unit. Two AB-type protein constructs are synthesized differing in the length of the flexible oligo(ethylene glycol) spacer separating the two end groups. Using an experimental setup where size exclusion chromatography is directly coupled to Q-TOF mass spectrometry, we have analyzed the self-assembled architectures as a function of concentration. The theory of macrocyclization under thermodynamic control is used to quantitatively analyze the experimental data. Using this theory, we show that AB-type monomers linked by flexible linkers grow reversibly via ring–chain competition. Inherently the formation of linear polymeric assemblies is beyond the capability of these types of building blocks due to concentration limits of proteins. The results therefore contribute to the general understanding of supramolecular polymerization with biological building blocks and demonstrate design requirements for monomers if linear polymerization is desired.
The combination of biological macromolecules with synthetic components to yield semi-synthetic or hybrid molecules, offers the possibility to combine the strengths of biology and chemical synthesis.9 Biological architectures are constructed with high fidelity but the variety of building blocks is limited. Synthetic chemistry on the other hand provides an infinite variation in topology but is less efficient in error-free synthesis. Merging both fields yields a challenging approach to assemble and study semi-synthetic protein architectures. Linking protein and substrate through a flexible linker enables the synthesis of interesting supramolecular building blocks and provides opportunity in the rapidly emerging field of supramolecular protein engineering. Recently, several groups have reported on synthetic supramolecular polymers based on protein–ligand interactions.1j,10,11 For example, Hayashi and co-workers1j,10 reported the formation of linear supramolecular polymers based on heme proteins to which an external cofactor moiety was appended via a flexible spacer. This one-dimensional concept was further expanded to two dimensions by mixing in a heme-triad to enable network formation. Recently Wagner et al.11 presented an in-depth study on the assembly of discrete protein nano-rings by combining a dimeric protein construct with a flexible spacer and a divalent synthetic ligand.
These two examples demonstrate that both linear as well as discrete cyclic assemblies can be obtained using flexible spacers. Indeed, it is well known that the use of flexible linkers inherently brings along the formation of cycles. In the early 1930s Kuhn12 introduced the concept of effective concentration (Ceff) to provide a relation between the mean squared end-to-end length of a linker and the cyclization probability of the end groups when the two ends of the linker are an infinitesimal distance apart. The effective concentration can be thought of as the local concentration of one chain end in the vicinity of the other chain end when these two ends are connected by a linker. As such, the Ceff theoretically quantifies the advantage for an intra- vs. an intermolecular interaction. Because the Ceff depends on the length and conformational flexibility of the linker, changes in these values will have a pronounced effect on the supramolecular polymerization of monomeric building blocks and should therefore be taken into consideration in the design process. Besides the effective concentration, other factors such as the presence of linker strain or specific noncovalent interactions within oligomeric assemblies can also play a crucial role in the supramolecular polymerization process as will be shown in this paper.
The objective of this work is to quantitatively analyze and study the mechanism of the supramolecular polymerization of biological building blocks in which the reversibly associating A and B end groups are separated by a flexible spacer. To this end, an AB monomer was developed based on the strong, noncovalent interaction between the ribonuclease (RNase) S-protein and S-peptide (Scheme 1, Ka ∼ 7 × 106 M−1 in NaOAc at 25 °C),13 thereby constructing supramolecular architectures entirely consisting of enzymatically active proteins as linking units. This RNase S system has previously been successfully used in the noncovalent synthesis of protein dendrimers by our group.14 Both S-protein and S-peptide were connected via a flexible oligo(ethylene glycol) (EG) linker. EG linkers are commonly used to construct multivalent ligands and are known to be resistant to nonspecific protein adsorption.15 This minimizes the influence of the protein fragments on the flexibility and thus the Ceff of the linker. Upon successive addition of monomers to the growing chain, the AB building block can undergo cyclic as well as linear polymerization, the distribution of which is determined by the effective molarity of each oligomer and the overall monomer concentration. The influence of the length of the EG linker on the supramolecular polymerization will be experimentally demonstrated using concentration dependent size exclusion chromatography and the results obtained will be quantitatively analyzed using the theory of macrocyclization under thermodynamic control. Finally, a comparison is made between the experimentally determined product compositions and a reversible isodesmic polymerization model in which no cyclization occurs.
Scheme 1 The formation of ribonuclease S upon cleavage of ribonuclease A by the protease subtilisin. Ribonuclease S can be separated into the S-peptide and S-protein that form a tightly bound supramolecular complex. |
Scheme 2 Synthetic outline of the AB1 (n = 18) and AB2 (n = 5) S-peptide S-protein monomeric building blocks, formed via aniline catalyzed oxime chemistry between the oxidized serine residues on the S-peptide and S-protein and aminooxy end groups on the EG linkers. |
As only a correctly folded and associated RNase S complex is enzymatically active, measurements of the enzymatic activity of the tethered RNase S constructs AB1 and AB2 can reveal whether the protein still contains its native structure after synthesis and that the enzyme can be used to direct self-assembly. The fluorescent 6-FAM-dArUdAdA-6-TAMRA substrate, specifically developed to quantify RNase activity, was used to monitor enzymatic activity.19 At 1–10 nM concentrations, we obtained enzymatic activities of ∼60 and ∼80% of that of commercially available RNase A and S for AB1 and AB2, respectively (see ESI†). Since MS analysis showed no evidence that other parts of the S-protein besides the N-terminus are oxidized during synthesis, this diminished activity could be due to substrate hindrance by the flexible EG linker.15 From the retained enzymatic activity of both AB1 and AB2 we conclude that both ends of the monomer are still able to form a native RNase S complex and can therefore be used as a linking unit for supramolecular assemblies.
AB1 was injected after equilibration in a concentration series of 100 μM, 1 mM and 10 mM. Measurements at higher concentrations could not be included due to practical limitations. At the lowest concentration analyzed (Fig. 1a, dotted line) the TIC trace shows the presence of one main peak and a second smaller peak. Q-TOF MS measurements attributed the first peak to the AB1-monomer (Fig. 1b, c), whereas for the second peak no conclusive MS spectrum was obtained and it was therefore attributed to low molecular weight buffer compounds that are present in relatively large quantities compared to the protein monomer at this concentration. Injection of a 1 mM AB1 solution yielded not only monomeric, but also dimeric architectures (Fig. 1a, dashed line, d + e) and upon further increment of the concentration, the sample of 10 mM showed, besides monomeric and dimeric, also trimeric species (Fig. 1a, solid line, f + g). DLS analysis confirmed that no larger aggregates were present (see ESI†).
Fig. 1 (a) Q-TOF analysis chromatograms of the AB1 SEC runs, with schematic representations of ring sizes at the corresponding peaks; (b)m/z and (c) deconvoluted spectra of the cyclic monomer (MWcalc = 14314 Da, MWcalc - ser = 14227 Da); (d)m/z and (e) deconvoluted spectra of the dimer (MWcalc = 28628 Da, MWcalc - ser = 28541 Da, MWcalc – 2 ser = 28454 Da) and (f)m/z and (g) deconvoluted spectra of the trimer (MWcalc = 42942 Da, MWcalc – ser = 42855 Da, MWcalc – 2 ser = 42768 Da, MWcalc – 3 ser = 42681 Da). |
AB2 was injected in a concentration series of 10 μM, 100 μM, 1 mM and 10 mM after equilibration of the mixtures. For this building block the linker is significantly shorter than for AB1 and therefore the equilibrium between rings and chains is expected to shift. Injection at a concentration of 10 μM yielded only monomeric species (Fig. 2a, dash-dotted line, b + c). Injection at a higher concentration of 100 μM resulted in the appearance of dimeric species (Fig. 2a, dotted line, d + e) while injection of a 1 mM solution resulted in the formation of trimeric species (Fig. 2a, dashed lines, f + g). At a concentration of 1 mM AB2 we observe an increased amount of dimeric species compared to AB1, indicating that the short linker length in AB2 enforces preferential formation of these species when the overall monomer concentration is sufficient. Successive concentration increments up to a final concentration of 10 mM showed, besides monomers, dimers and trimers, also tetrameric species (Fig. 2a, solid line, h + i) and the dimer has become the predominant species. The difference in the distribution of species for the two different AB monomers as a function of concentration clearly demonstrates the influence of linker length on the supramolecular polymerization of the two AB monomers.
Fig. 2 (a) Q-TOF analysis chromatograms of the 10 mM, 1 mM, 100 μM and 10μM AB2 SEC runs, with schematic representations of ring sizes at different peaks; (b)m/z and (c) deconvoluted spectra of the monomer (MWcalc = 13740 Da, MWcalc - ser = 13653 Da); (d)m/z and (e) deconvoluted spectra of the dimer (MWcalc = 27480 Da, MWcalc - ser = 27393 Da, MWcalc – 2 ser = 27306 Da); (f)m/z and (g) deconvoluted spectra of the trimer (MWcalc = 41220 Da, MWcalc – ser = 41133 Da, MWcalc – 2 ser = 41046 Da); (h)m/z and (i) deconvoluted spectra of the tetramer (MWcalc = 54960 Da, MWcalc – ser = 54873 Da, MWcalc – 2 ser = 54786 Da). |
The kinetic stability of the aggregates during the course of the 15 min time interval (flow rate = 0.1 ml min−1) required for SEC-MS was probed by performing SEC at different flow rates. We slowed down the flow rate to 0.075 ml min−1, resulting in an increase in the length of time the complex spent on the column by a factor of 1.33. Integration of the peak areas of the UV chromatogram at this lower flow rate resulted in the same distribution of species observed for the higher flow rate (see ESI†), indicating the kinetic stability of the aggregates is high enough not to be affected by any dilution effects that occur upon injection of the self-assembled architectures to the SEC.
We have calculated the effective concentration of the two different AB monomers by assuming that the end-to-end displacement vector for the ethylene glycol linker separating the two end groups has a Gaussian probability density:26
(1) |
Ceff(d) = p(d)/NAV | (2) |
Assuming that the buffers used conform to θ conditions, the root-mean-square distance <r2> can be estimated assuming a three dimensional random flight model:28
<r2> = Cnnl2 | (3) |
Combining eqn (2) with eqn (3), relates the effective concentration to the amount of PEG repeats in a linker, for different values of the distances, d. The distance d between the end of the S-peptide and S-protein is approximately 25 Å31 and the number of PEG units in the linker is 19 and 6 for AB1 and AB2, respectively. Using these values and a characteristic ratio of 4.1, the calculated effective concentrations then become 8 mM for AB1 and 0.7 mM for AB2.32
As most supramolecular polymerizations occur in relatively dilute solutions, the model proposed by Ercolani et al.,33 is eminently suited to describe the equilibrium between cyclic and linear species in these equilibrium polymerizations. The ring–chain model developed by Ercolani et al. is characterized by two distinct thermodynamic constants (Fig. 3) i.e. an intermolecular binding constant (Kinter) and the intramolecular binding constant for i-th ring closure (Kintra(i)).
Fig. 3 Schematic representation of the ring–chain supramolecular polymerization of RNase S building blocks in which Kinter (M−1) represents the intermolecular binding constant for bimolecular association and Kintra(i) represents the dimensionless intramolecular equilibrium constant for i-th ring closure. |
Under the fulfilment of conditions (i)–(v), the EMi values for i > 1 can be conveniently written as a function of EM1 (the effective molarity of the bifunctional AB monomer):
(4) |
In such a case, the mass-balance equation takes the following form:33
(5) |
[Ci] = EM1i−5/2xi | (6) |
The percentage of cyclic oligomers is then calculated as:
(7) |
Under the conditions that eqn (4) applies (vide supra), the yield of cyclic monomer is always higher than any other cyclic oligomer (Fig. S9†).
Fig. 4 (a) Gaussian peak deconvolution of the Q-TOF TIC trace of AB1 at 10 mM. Trimeric rings are represented by the dark grey peak (diagonal filling), dimeric rings by the medium grey peak (horizontal filling) and monomeric rings by the light grey peak (checker filling). With these four distinct graphs, the original curve is accurately reproduced (red line). (b) Gaussian peak fits to the Q-TOF TIC trace of AB2 at 10 mM. Tetrameric rings are represented by the black peak (diagonal filling). Trimeric rings are represented by the dark grey peak (diagonal filling), dimeric rings by the medium grey peak (horizontal filling) and monomeric rings by the light grey peak (checker filling). With these distinct graphs, the original curve is accurately reproduced (red line). |
We have compared the experimentally determined product distribution with the calculated product distribution obtained using the previously discussed ring–chain competition model and a standard isodesmic polymerization model in which no cycle formation occurs.35 For both AB monomers an intermolecular equilibrium constant of 7 × 106 M−1 was used and EM1 values were based on the calculated Ceff using the Gaussian chain model, thus for AB1EM1 = 8 mM and for AB2EM1 = 0.7 mM. Comparison of the experimentally determined product distribution with the calculated product distribution determined using the ring–chain model (eqn (5)–(7)) at various total concentrations of AB1 (10, 1 and 0.1 mM) shows good correspondence (Table 1). The small deviation between theory and experiment for the AB1 monomer is most likely due to the excluded volume effects between the protein and ligand which have been neglected in the calculation of the Ceff but have been shown to be important in other studies on reversible cyclizations.24,26 Comparison of the experimentally determined mole fractions to the calculated mole fractions obtained using an isodesmic polymerization model, in which only linear association with equilibrium constant 7 × 106 M−1 occurs, clearly shows that this model is unable to describe the supramolecular polymerization of monomer AB1. Hence, for this monomer, the experimental data closely obey the Jacobson–Stockmayer theory. As a consequence, the mole fraction of cyclic monomer is always higher than the mole fraction of any other cyclic oligomer below the critical concentration while for concentrations higher than the critical concentration, which are beyond experimental boundaries here, rapid polymerization into linear chains occurs.
10 mM | AB1 | Ring–chain | Isodesmic | AB2 | Ring–chain | Isodesmic |
---|---|---|---|---|---|---|
71 ± 2.5 | 64 | ≪1 | 39 ± 1 | 7 | ≪1 | |
28 ± 1 | 18 | ≪1 | 43 ± 0.5 | 2.5 | ≪1 | |
1 ± 0.2 | 8 | ≪1 | 9 ± 1 | 1.5 | ≪1 | |
0 | 4 | ≪1 | 4 ± 1 | <1 | ≪1 | |
Linear | 0 | 0 | 100 | 0 | 84 | 100 |
DPn | — | — | 265 | — | 10 | 265 |
1 mM | AB1 | Ring–chain | Isodesmic | AB2 | Ring–chain | Isodesmic |
---|---|---|---|---|---|---|
80 ± 4 | 95 | <1 | 80 ± 2 | 60 | <1 | |
20 ± 0.6 | 4 | <1 | 18 ± 5 | 18 | <1 | |
0 | <1 | <1 | 1.5 ± 6 | 8 | <1 | |
0 | ≪1 | <1 | 0 | 4.5 | <1 | |
Linear | 0 | 0 | 100 | 0 | 0.5 | 100 |
DPn | — | — | 84 | — | 1.4 | 84 |
0.1 mM | AB1 | Ring–chain | Isodesmic | AB2 | Ring–chain | Isodesmic |
---|---|---|---|---|---|---|
100 | >99 | 1 | 89 ± 1 | 95 | 1 | |
0 | ≪1 | 1 | 11 ± 1 | 5 | 1 | |
0 | ≪1 | 1 | 0 | 0.5 | 1 | |
0 | ≪1 | 1 | 0 | <1 | 1 | |
Linear | 0 | 0 | 96 | 0 | 0 | 96 |
DPn | — | — | 27 | — | — | 27 |
Comparison of the experimentally determined product distribution of the AB2 monomer at various monomer concentrations with the Jacobson–Stockmayer theory, shows a good correspondence for the lower concentrations (0.1 mM and 1 mM). However, large deviations between the experimentally determined product composition and calculated values are observed at a concentration of 10 mM. Where the experimental data indicate the formation of oligomeric assemblies with the dimer as most abundant species, the ring–chain competition model, in which all cycles are assumed to be strainless, suggests the initiation of linear polymers with a DPn of 10. Comparison between the experimental data and values obtained using an isodesmic polymerization model show that this model cannot describe the experimental data as it predicts a number averaged degree of polymerization of 265 at this concentration. A possible explanation for the failure of the Jacobson–Stockmayer theory to describe the experimental data at higher concentrations can be found in the shorter EG linker of AB2. Discrepancy between the Jacobson–Stockmayer theory and experimental data has been observed in other studies in which short linkers have been used, for example during ring formation in covalent polymers,36–38 cyclization of short DNA fragments39 and cyclization of synthetic supramolecular polymers.40 It has been suggested that the origin of this discrepancy is due to the fact that short chains are inherently strained in their cyclic conformation and therefore do not obey Gaussian statistics, an important assumption in the derivation of eqn (4). As a result, the effective molarity of the cyclic AB2 monomer is close to, or lower than the effective molarity of the cyclic dimer and the mole fraction of cyclic dimer can surpass the mole fraction of cyclic monomer33 as is also observed experimentally. This effect is most notable at concentrations slightly above EM1, in our case at concentrations around 10 mM.
The experimental data on the AB2 monomer elegantly show that linker composition and concentration can be used to tune the yield of a specific cyclic supramolecular biological assembly. Whereas in the AB1 monomer the linker connecting the protein and ligand is large enough for the formation of a cyclic monomer, in the AB2 monomer the formation of the cyclic monomer is hindered, resulting in high yields of the cyclic dimer at concentrations close to the critical concentration. Alternatively, the higher effective molarity of cyclic AB2 dimer formation compared to AB2 cyclic monomer formation can also be caused by additional noncovalent interactions between the two protein–ligand complexes in the cyclic dimer of AB2.
By combining experimental data with theoretical modeling, valuable insights were obtained into the supramolecular polymerization mechanisms and design criteria for protein based supramolecular polymeric architectures. The experimentally obtained product distribution could be quantitatively described using the theory of reversible macrocyclization and shows the relation between linker length, the effective molarity of the monomer and the concentration of higher molecular weight cyclic oligomers and as such confirms that the supramolecular polymerization occurs via ring–chain competition. Consequently, the mole fraction of cyclic monomer is always higher than other cyclic oligomers below and close to the critical concentration.
The shorter linker in AB2 results in a decreased effective concentration for this monomer. With this decreased Ceff the presence of various sized cyclic species as well as linearly polymerized architectures was predicted by the ring–chain competition model at the highest concentration. In sharp contrast, the experimental data clearly show the presence of oligomeric species instead of larger linear supramolecular polymers. The higher yield of the dimeric cycle compared to the monomeric cycle at concentrations close to the critical concentration indicates that the linker in the monomeric cycle of AB2 is either strained, resulting in an enthalpic contribution to the cyclization constant, or cannot be described as a random Gaussian coil. Hence, the effective molarity of cyclic AB2 dimer formation must be close to or even higher than the effective molarity of cyclic monomer formation, resulting in the preferred formation of the cyclic dimer at concentrations close to the critical concentration.
The described system is characterized by synthetic ease and can be analyzed in great detail. It therefore is promising in the further study of self-assembly directed by protein and peptide interactions to form active biological objects. Besides the here described AB homo-polymerization, the system is very suitable to study AA–BB hetero-polymerization in a comparable fashion as well as to analyze multivalency when used in combination with more branched linker structures.
When dealing with AB type protein monomers in which the A and B end groups are separated by a flexible tether, formation of linear polymers is excluded by solubility limitations and inherently only cyclic assemblies can be obtained. As implied by the ring–chain mechanism, only with an effective concentration close to 0, where cyclization is excluded, combined with a high association constant between the A and B units that is in the order of 1011 M−1, is linear polymerization reachable in biologically workable concentrations (<10−3 M). This is the case in several examples of supramolecular polymers found in Nature like actin filaments and microtubuli, but is hard to design for synthetic protein based AB building blocks. Finally, longer linear polymers can be obtained when the growth of the linear chains is coupled to growth in the lateral direction resulting in fiber like aggregates as in such a case a first order nucleated transition can occur.41 Lateral association due to the presence of a hydrophobic linker has been suggested to occur in the growth of serpin polymers in which competition between rings and chains also takes place.8
When aiming for linear polymers, cycle formation needs to be considered for all systems with flexible linkers. The ring–chain competition mechanism therefore sets out clear guidelines for the design criteria of monomeric building blocks to be used in supramolecular polymerization, either cyclic or linear.
(8) |
Azido-PEG(5)-amine was reduced with Pd/C and H2 gas to obtain the diamine prior to use. PEG(5)-diamine (195 mg) was reacted with NHS-activated tBoc-protected aminooxy (381 mg) and DIPEA (500 μL) in dry CH2Cl2 (4 mL) overnight. After solvent evaporation, the product was dissolved in 40 mL CHCl3 and washed 2× with 40 mL NaHCO3 and once with 40 mL brine. After drying with MgSO4 and filtration, the product was purified on a silica column with DCM/MeOH (20/3 mL). Deprotection of the tBoc groups was achieved upon dilution in TFA/H2O (5/0.25 mL) at 0 °C for 1 hour. After cold solvent evaporation the product was dissolved in 20 mL water and washed with 2 × 10 mL diethyl ether. After removal of the diethyl ether the product was lyophilized.
1H NMR (CDCl3-d1, 400 MHz): tBoc-protected PEG18 linker δ = 8.2 (s, 2H, CH2–NH–CO), δ = 7.8 (bs, 2H, O–NH–CO), δ = 4.6 (s, 4H, O–CH2–CO), δ = 3.6–3.7 (80H, O–CH2–CH2–O), δ =1.5 (18H, C–(CH3)3); 1H NMR (CDCl3-d1, 400 MHz): unprotected PEG18 linker δ = 8.1 (bs, 2H, CH2–NH–CO), δ = 4.7 (bs, 4H, O–CH2–CO), δ = 3.6–3.8 (80H, O–CH2–CH2–O); 13C NMR (CDCl3-d1, 400 MHz): tBoc-protected PEG18 linker δ = 169.01, 157.39, 82.42, 75.85, 70.57, 69.60, 38.92, 28.20; 13C NMR (CDCl3-d1, 400 MHz): unprotected PEG18 linker δ = 168.70, 71.64, 69.49, 69.29, 68.52, 38.62; MALDI-TOF MS protected PEG18 linker MWcalc = 1243.43 Da, MWobs = 1265.70 Da (Na+); MALDI-TOF MS unprotected PEG18 linker MWcalc = 1043.20 Da, MWobs = 1043.62 Da.
1H NMR (CDCl3-d1, 400 MHz): tBoc-protected PEG5 linker δ = 8.2 (s, 2H, CH2–NH–CO), δ = 7.8 (bs, 2H, O–NH–CO), δ = 4.4 (s, 4H, O–CH2–CO), δ = 3.6–3.7 (m, 24H, O–CH2–CH2–O), δ = 3.5 (q, 4H, NH–CH2–CH2–O), δ = 1.5 (18H, C–(CH3)3); 1H NMR (CDCl3-d1, 400 MHz): unprotected PEG5 linker δ = 8.0 (bs, 2H, CH2–NH–CO), δ = 4.5 (4H, O–CH2–CO), δ = 3.6–3.7 (24H, O–CH2–CH2–O), δ = 3.5 (m, 4H, NH–CH2–CH2–O), δ = 1.9 (4H, O–NH2); 13C NMR (CDCl3-d1, 400 MHz): tBoc-protected PEG5 linker δ = 169.06, 157.37, 82.42, 75.79, 70.53, 69.56, 38.93, 28.20.
Flow speed analysis was performed on a Superdex 75 column coupled to UV-vis (Shimadzu LC-20AD liquid chromatography, Shimadzu SPD-M20A prominence diode detector, 230 nm). Flow speed was varied from 0.1 ml min−1 to 0.075 ml min−1. Manual injections were performed with 4 μL samples in 0.1 M ammonium acetate, pH 4.5.
Footnote |
† Electronic supplementary information (ESI) available: Detailed experimental procedures. Analytical characterization data of all compounds. DLS. Extended simulations on ring–chain equilibria. Detailed SEC flow-speed analysis. See DOI: 10.1039/c0sc00108b |
This journal is © The Royal Society of Chemistry 2010 |