Fast CE for combinatorial catalysis

Two solvent‐modified MEKC methods were developed for the quantitative analysis of heterocyclic amines synthesised using intramolecular ring closure via catalysed hydroamination. The first method was capable of resolving six of the amines (precursors and products) with a sample‐to‐sample injection time of 2 min employing a 20 mM borate buffer, pH 9.2 with 20 mM SDS and 5% v/v n‐butanol (n‐BuOH). A second low‐pH method using 20 mM phosphate buffer, 100 mm SDS, 5% v/v n‐BuOH and 20% v/v iso‐propanol (i‐PrOH) was able to resolve an additional pair of compounds with a sample‐to‐sample time of 3.5 min. Application of the first method to the analysis of a sample containing catalyst as well as amines placed directly in a 96‐well plate showed excellent performance, with migration time and peak height and area reproducibility being less than 0.9 and 9.6%, respectively. The quantity of conversion by catalyst was calculated to be 68 ± 7%, which was in excellent agreement with the 67 ± 5% obtained by more conventional 1H NMR experiments, with the added advantage that this method is also cheaper, quicker and more amendable to high‐throughput screening of combinatorial libraries.


Introduction
Combinatorial approaches to catalyst development are increasingly being employed in the identification of effective homogeneous organometallic catalysts [1]. These approaches rely on accurate and expedient methods of product analysis. Several groups have utilised various high-throughput screening techniques for the identification of active catalysts, unfortunately these techniques are often substrate specific. Wagner and coworkers [2] have utilised immunoassay techniques to screen for enantioselective catalysts. This technique is an elegant approach but is not widely applicable as it is highly selective for the specific substrate. Any change in catalyst precursor/product therefore requires development of a new immunoassay at considerable time and expense. Lavastre and co-workers [3,4] have utilised TLC coupled with image analysis as a quantitative screening methodology. TLC as a screening technique is widely applicable and low cost, however without the quantitative image analysis it is subject to false-positive results due to small quantities of strongly UV active products. Colourimetric approaches have also been utilised as catalyst high-throughput screening methods by several researchers [5,6]. This method frequently relies on the synthesis of dyes incorporating the functional group which is being investigated in the catalysis. Hartwig and coworkers [6] employed this approach for the investigation of the hydrosilylation of alkenes and imines investigating the catalysis of dyes incorporating these moities. Since catalysis is often substrate specific the portions of the substrates imparting the colourimetric change can interfere with the catalysis leading to false positives and negatives. This is of course also limited in that catalysis is not performed on the precursor molecules of interest and if different precursors are of interest, then new coloured chemicals must be synthesised, it can be a lengthy and time-consuming task.
CE is an alternative technique that shows much potential for combinatorial screening. It is rapid and requires only small volumes of sample which can be analysed directly from a 96-well plate making it perfectly compatible with combinatorial screening where only minute quantities of large libraries of samples may be produced. Furthermore, the different separation mechanisms offered by CZE and EKC provide a completely flexible platform that can be adapted for the analysis of almost any set of target analytes. It is worthy to note that other forms of chromatography, namely GC and HPLC, have been restricted in application in this area. This is unsurprising given the long analysis times, expense and difficulty of performing parallel analysis times for HPLC, while for GC faces these same restrictions as well as the need for analytes to be volatile or be made volatile via derivatisation. It is therefore not surprising that CE has been more widely used for screening combinatorial catalysts, using either MEKC or CZE [7][8][9][10][11]. In many of these, throughput was severely restricted due to the long separation times (between 10 and 60 min), although the issue was addressed in the later work by using a 96-capillary array instrument, with a total analysis time of 96 samples within 1 h [8,11]. A cheaper alternative using a single capillary instrument and a rapid 3.5 min injection was developed by Simms et al. [7] making it possible to screen 96 samples in 336 min (5.6 h).
In this paper, we developed a rapid method for application in a single capillary instrument for the quantitative analysis of a range of heterocyclic amines and their precursors synthesised through intermolecular ring formation via catalysed hydroamination. Two rapid methods based on solvent-modified MEKC were developed to allow the rapid separation of the precursor and product pairs and the potential of this approach for quantitative combinatorial screening of transition metal catalysts was demonstrated by application of a sample containing catalyst, precursor and product and comparison with 1 H NMR analysis.
For the hydroamination of EA using the catalyst [Rh(bim)(CO) 2 ]BPh 4 , EA (60.0 mg, 0.512 mmol) and [Rh(bim)(CO) 2 ]BPh 4 (16.1 mg, 0.0246 mmol, 4.8 mol%) were dissolved in 2.1 mL of acetone. The mixture was stirred at reflux (807C) for 7 h. The mixture was cooled and transferred to a 10 mL volumetric flask and diluted to 10 mL with acetone to provide sufficient solution for analysis.
For calculation of the conversion of EA to Indole(I), the sample was dried in vacuo, redissolved in d 6 -acetone and secondary conversion analysis was performed via 1 H NMR. The 1 H NMR spectrum was obtained at 500 MHz, 300 K. Catalytic conversion was calculated by comparison of the 1 H NMR spectra of EA and I with a unique peak selected from each and integrated. The ratio of the two peaks defined the relative amount of EA and I which was then used to calculate a final conversion of 67 6 5%.

Electrophoresis
All method development using purified standards was performed using an Agilent HP 3D CE (Agilent Technologies, Waldbron, Germany) and polyimide-coated fused-silica capillary (Phoenix, AZ, USA) of 50 mm id with a length of 32.5, 26.5 or 8.5 cm to the detector. Detection was performed using the in-built DAD with individual traces monitored at 200, 214, 230 and 254 nm all with a bandwidth of 10 nm.
Application of the developed method to the analysis of real catalyst screening samples was performed using a Beckman P/ACE MDQ (Beckman Coulter, Fullteron, USA) with polyimide coated fused-silica capillary of 50 mm id with a length of 21.5 and 11.5 cm to the detector. Data were collected with both the DAD and individual traces at 214 and 254 nm at a sampling rate of 32 Hz, with the detection trace autozero at 0.25 min. As the sample was in a 96-well plate, injection was performed by voltage, with 11 kV, 10 s. The capillary was conditioned between separations by rinsing with 0.1 M NaOH (20 psi, 1 min) followed by BGE (10 psi, 0.2 min). All separations were performed at 30 kV unless otherwise stated. Identification of the third peak observed during the analysis of real catalyst conversion samples was confirmed by injection of samples containing 0.5 mM NaBPh 4 and 0.5 mM bim, the ligand contained in the catalyst, with the former comigrating with the unknown peak in the catalyst samples. Data in Fig. 8 were plotted using Transform V 3.3 (RMIT University, Melbourne, Australia).
BGEs were prepared by thoroughly mixing appropriate quantities of the aqueous buffer and the SDS stock solution with required amount of water. Once mixed, an appropriate amount of the required organic modifier was added and the solution again mixed thoroughly. For example, for the 5% n-butanol (n-BuOH), 20 mM Borate and SDS BGE: 2 mL of 100 mM sodium tetraborate, 0.4 mL of 500 mM SDS and 7.1 mL of water (Milli-Q) were vigorously mixed. Subsequently, 0.5 mL of n-BuOH was then added and mixed thoroughly to generate the BGE.
Calculation of the catalytic conversion was performed by quantitation based on peak area of EA and I from a five-point calibration curve constructed over the concentration range of EA and I.

Results and discussion
The catalysed reaction that we are interested involves intramolecular ring formation via hydroamination. The heterocyclic molecules obtained from the intramolecular hydro-amination of aminoalkynes often generate the backbone structures of many biologically active molecules. Indole(I) is the core of many therapeutically beneficial compounds including, for example the ergot alkaloid clavicipitic acid [19,20]. As hydroamination can be applied to a wide range of substrates, both aliphatic and aromatic, it was also desirable to have a method that could separate a range of compounds to more thoroughly evaluate catalyst properties. A range of precursor/product pairs was selected as potential targets to evaluate catalyst performance, with names and structures shown in Fig. 1. Estimation of the physical properties of the compounds using the ACD software, indicated that while all of them possessed an amine, it would be almost impossible to protonate them to make them positive, suggesting that they would need to be separated by EKC. The log P values for the compounds ranged predominantly from 3 to 5, suggesting that there may be some difficulty in separating them using MEKC with SDS micelles as they would completely interact with the micelles and thus not separate easily. Experimental data (shown in the bottom trace in Fig. 2) showed that three of the compound pairs could be separated in a relatively simple MEKC electrolyte of 20 mM sodium tetraborate and 20 mM SDS, although PA and PI could not be resolved due to strong interaction with the micelles. While not gauged from the figure, it is important to note that this system was unsuitable for prolonged use as there was a gradual decline in efficiency and reproducibility over several hours. This was later discovered to be due to the insolubility of PA, PI, PBA and PPA in the electrolyte and we speculate that this precipitated onto the capillary walls causing the aforementioned problems. While this could potentially be avoided by the use of more surfactant, this would increase micelle interaction and make separation of PA and PI even more difficult, thus an alternative approach was taken.
Microemulsion EKC emerged in the late 1990s as a popular alternative for the separation of highly hydrophobic compounds, such as those encountered in this work. Subsequent work by Honoré Hansen et al. [21] suggest that the immiscible liquid used to make the emulsion only plays a minor role in solubility and selectivity of the system and it is actually the addition of high concentrations of organic solvents and cosurfactants that provide the improved performance. Consideration of a number of these solvents revealed that n-BuOH provides excellent separations of basic hydrophobic drugs [22] and was therefore selected for use in this work. The influence of n-BuOH on the separation of the four precursor/product pairs is shown in Fig. 2. There are a number of points to note about the separations in Fig. 2. First, that with the addition of n-BuOH it was possible to separate all the eight compounds in a single separation, with the optimal separation being near 4% v/v n-BuOH. Second, the selectivity was different with the addition of n-BuOH particularly with regard to PBA and PPA. Third, the migration times of the early migrating analytes (EA and I) increased over the entire n-BuOH range examined, while PA and PI increased up to 4% v/v n-BuOH, after which point they decreased. Consideration of the nature of the chemicals and the effect of n-BuOH can provide some explanation for these results. The elongated migration times for EA and I are indicative of a reduction in the EOF as a result of changes in viscosity and dielectric constant. The reduction in migration times for PA and PI can be attributed to a reduction of the interaction with the SDS micelles at high concentrations of n-BuOH, most likely due to variation of the phase ratio due to an increase in the CMC of SDS under these conditions. This is also supported by consideration of PBA and PPA. pK a estimated using the ACD software indicated that PBA and PPA have values of 9.6 and 8.6, respectively, and will be positively charged in the electrolyte used and there will therefore be a significant electrostatic interaction with the negatively charged SDS micelles. This is supported by the migration order, with PPA having a much longer migration time than PBA. Because of this strong electrostatic interaction, any variation in micelle concentration and structure is likely to have a highly significant influence on the migration of these two analytes. While these results are interesting and provide potential insight into the behaviour of the SDS system in these conditions, no further attempt was made to understand this result as the emphasis of this work was to develop a system to allow rapid screening in combinatorial catalysis.
Having developed appropriate conditions to separate the four combinatorial precursor/product pairs, it was desirable to make the separation occur as quickly as possible to allow rapid screening of a combinatorial plate without using a 96capillary array instrument. To reduce the analysis time, injections were made from the short end of the capillary, giving an effective length of 8.5 cm to the detector. Separation of the four precursor/product pairs are given in Fig. 3. It can be seen that three of the pairs are separated within 1.0 min, with PA and PI difficult to resolve under these conditions due to the strong interaction with the micelles. Similar short-end separation of PA and PI were made using 6 and 8% n-BuOH but these failed to provide acceptable resolution of these two analytes. The sample-to-sample time was 2.0 min, mainly due to the time required for the instrument to move the vials for preconditioning and injection. A full evaluation of the method was performed by analysis of 96 consecutive samples in 192 min (3.2 h) with an overlay of every 12th separation shown in Fig. 4. Excellent migration time reproducibility was observed, with the migration time variation less than 0.2% RSD over the entire 96 separations. Peak area and peak height reproducibility was slightly worse being less than 4.5% RSD. These results indicate the method is stable and suitable for three of the combinatorial catalyst precursor/product pairs, however the method was unsuitable for the rapid separation of PA and PI with the shortest separation time obtained using these conditions being 5 min with a 6 min sample-to-sample time. Given the large number of samples to be analysed, a quicker method was desired to allow more rapid screening.
A second approach was undertaken to develop a rapid method for the separation of PA and PI based on using a low pH electrolyte and reversed polarity with the analytes migrating past the detector by association with the micelles. This type of system has been shown to be exceptionally powerful for analytes that interact strongly with the micelles as relatively minor difference in micelle interaction result in significant differences in migration due to the absence of EOF [23]. Experiments using exactly the same conditions developed above, namely 20 mM SDS and 5% v/v n-BuOH but in 20 mM phosphate buffer (pH 2.2) failed to provide any resolution of PA and PI. Furthermore, baseline instabilities and a loss in efficiency were again observed suggesting that under these conditions that PA and PI were also slightly insoluble. To counteract this problem, the SDS concentration was increased to 100 mM and a second solvent, iso-propanol (i-PrOH), was added to improve the stability of the system. Separation of PA and PI in differing amounts of BGE with 0-20% v/v i-PrOH are shown in Fig. 5. As anticipated, a higher concentration of i-PrOH improved the separation of PA and PI, with baseline separation obtained with 10% v/v i-PrOH within 4 min. While this system provided excellent resolution, it was undesirable for two reasons. First, the total analysis time was 5 min per sample, which was longer than desired, and second, the high concentration of SDS and moderate concentration of organic solvent resulted in high currents of 2140 mA which caused several problems with obtaining a stable current and a stable system, with regular current breakdowns during a sequence. Higher concentrations of i-PrOH were found to provide much lower currents, with a current of280 mA at 20% v/v i-PrOH, as well as providing an increase in resolution between PA and PI. However, this was achieved at the cost of analysis time, with the separation under these conditions taking 10 min. Again, using injections from the short-end of the capillary with the 20% v/v i-PrOH BGE, it was possible to significantly de-  crease the analysis time, with separations under the same electrolyte conditions, but with an effective length of 8.5 cm, shown in Fig. 6. While the resolution is inferior to that from the long end, the separation time is slightly over 2.5 min, making a sample-to-sample time of 3.5 min requiring total analysis time of 336 min or 5.6 h for the analysis of a 96-well plate. The main advantage that this system has over the others providing a similar analysis time was the stability. Again, performing a sequence of 96 replicates the migration time variation was less than 2.5% RSD and peak area and peak height reproducibility less than 10% RSD.  To demonstrate the potential of the developed methods to high-throughput combinatorial catalyst screening, the catalytic conversion of EA to I was performed in bulk and aliquots transferred into 80 out of the 96 wells in a 96-well plate. Catalytic conversion was performed off-plate to provide sufficient sample to allow comparison and secondary determination of the amount of conversion by the more commonly accepted 1 H NMR experiments. Eight of the remaining wells were filled with a control sample with the other eight used for calibration to calculate the concentrations of EA and I. This was analysed using a Beckman P/ACE MDQ CE which has the capability to sample directly from the 96-well plate. A representative separation from one of the 80 wells is shown in Fig. 7, where it can be seen that good resolution between the EA and I was obtained and that there were no interfering peaks. A large peak was observed at approximately 0.7 min, which was later identified to be the catalyst counter ion, BPh 4 2 , while several smaller as yet unidentified peaks were observed preceeding EA. The separation times were slightly longer than those obtained previously due to the slightly longer capillary length to the detector in the Beckman instrument in comparison to the Agilent (11.5 cm compared to 8.5 cm, respectively) this gave a total separation time of 1 min and a sample-to-sample time of 2 min, meaning that a complete 96-well plate can be analysed in 288 min. Visualisation of the entire sample set analysed on the 96-well plate is shown in a 2-D format in Fig. 8. One standard and a control was analysed every 11th and 12th run, respectively, which Figure 7. Rapid separation of a combinatorial catalyst sample using the Beckman P/ACE MDQ instrument. All conditions as in Fig. 3 with the exception of the capillary length (21.5 cm total, 11.5 cm to detector) and injection directly from a 96-well plate. can be seen most easily by the absence of the BPh 4 2 peak at 0.7 min. From the figure, it can be seen that migration time reproducibility was excellent, with the times of EA and I within 0.9% RSD. Peak area reproducibility for the catalyst sample over 80 injections was 9.6% RSD for EA and 9.3% RSD for I, and the amount conversion of EA to I was calculated to be 68 6 7%. This calculated conversion is in excellent agreement with the 67 6 5% obtained by 1 H NMR spectroscopy. CE at approximately 3 min per sample is much faster than performing the analysis via 1 H NMR spectroscopy; it also is much cheaper as expensive deuterated solvents are not required and requires minimal sample preparation as it is possible to directly analyse the samples from a 96-well plate. Work is currently underway to validate these methods for the combinatorial screening of catalyst.