Real‐Time Monitoring of Enzyme‐Catalysed Reactions using Deep UV Resonance Raman Spectroscopy

Abstract For enzyme‐catalysed biotransformations, continuous in situ detection methods minimise the need for sample manipulation, ultimately leading to more accurate real‐time kinetic determinations of substrate(s) and product(s). We have established for the first time an on‐line, real‐time quantitative approach to monitor simultaneously multiple biotransformations based on UV resonance Raman (UVRR) spectroscopy. To exemplify the generality and versatility of this approach, multiple substrates and enzyme systems were used involving nitrile hydratase (NHase) and xanthine oxidase (XO), both of which are of industrial and biological significance, and incorporate multistep enzymatic conversions. Multivariate data analysis of the UVRR spectra, involving multivariate curve resolution‐alternating least squares (MCR‐ALS), was employed to effect absolute quantification of substrate(s) and product(s); repeated benchmarking of UVRR combined with MCR‐ALS by HPLC confirmed excellent reproducibility.


Reaction sample preparation and monitoring
For all biotransformations, the reaction mixture was focused under the microscope objective in a small, glass Petri dish (see Figure 3.5). At various time points after enzyme addition, UVRR and HPLC spectra were collected throughout the biotransformation. To minimise the risk of reduced focus on the sample through solvent evaporation and removal of volume (for HPLC analysis), the reaction was performed on a 10 mL scale. For biotransformation 1 and 2, at specific time points, 20 µL of sample was removed from the reaction mixture and immediately quenched and diluted with 180 µL with MeOH. The sample was then centrifuged at 21,000 ×g for 6 min. 100 µL of the sample was then transferred to a HPLC vial and subjected to HPLC analysis. For biotransformations 3 and 4, at specific time points, 20 µL of sample was removed from the reaction mixture and immediately diluted to 80 µL with water. The sample was then heated for 5 min at 80°C and centrifuged at 14000 ×g for 4 min. 60 µL of the sample was then transferred to a HPLC vial and subjected to HPLC analysis.

Instrumentation
UV-Vis absorption analysis was carried out using Thermo Biomate 5 (Thermo Fisher Scientific Inc., Massachusetts, USA). 1 mL of the sample was pipetted into a quartz cuvette and inserted into a sample holder. Data were acquired over a wavelength range of 210-350 nm.
HPLC Analysis for biotransformations 1 and 2. HPLC separation was conducted using an Agilent Zorbax Eclipse Plus HPLC system set up for reverse phase separation consisting of a diode array detector. For both biotransformations, the column was a 100 × 4.6 mm, Phenomenex Eclipse Plus® C18 with a 3.5 µm particle size. For each injection, the run time was 12.0 min pumped at a flow rate of 1 mL min -1 and at 30 o C column temperature. The mobile phase consisted of a linear gradient, starting conditions of 5% MeCN/H 2 O (plus 0.05 % TFA) held for 2 min before increasing to 75 % MeCN/H 2 O over 6 min. Prior to washing at 95 % MeCN/H 2 O for 1.5 min and re-equilibration to initial conditions over 2.5 min (total run time 12 min). 5 µL of each sample was introduced using an auto-injector. UV absorbance was detected at 254 nm throughout.
HPLC Analysis for biotransformations 3 and 4. HPLC separation was conducted using an Agilent 1100 series HPLC system set up for reverse phase separation consisting of a diode array detector. For biotransformation 3, the column was 150 × 4.6 mm, Phenomenex Hyperclone C18 with a 5 µm particle size. For each injection, the run time was 10.0 min. The mobile phase was 0.020 M aqueous potassium phosphate buffer at pH 6.5, pumped at a flow rate of 1.0 mL min -1 . For biotransformation 4, the column was a 250 × 4.6 mm, ACE 5 C18-AR (Reading, Berkshire) with a 5 µm particle size. For each injection, the run time was 12.0 min. The mobile phase was 0.047 M aqueous acetic acid buffer at pH 4.65, pumped at a flow rate of 1.0 mL min -1 . For both biotransformations, 50 µL of each sample was introduced using an auto-injector. UV absorbance detection was measured at 254 nm.
UVRR analysis for all biotransformations. UVRR was performed using a Renishaw Raman 1000 system (Renishaw, Wotton-under-edge, Gloucestershire, UK). Approximately ~0.2 mW of power was delivered to the sampling point using a Lexel Model 95 ion laser emitting at 244 nm. The solution was continuously stirred under the laser to avoid photodegradation using a magnetic stirrer plate and magnetic bar. Spectra were collected with an acquisition time of 20 s. Only spectra with no demonstrable photo-degradation and signal from the reaction vessel were used for analysis.

Data Processing
All data were exported from the respective instrument operating software and analysed using Matlab R2015a (The Mathworks, Natick, MA, USA).

HPLC data analysis
The peaks of the target analytes were integrated with the results of the HPLC data and served as an external validation data set to independently verify the accuracies of the MCR-ALS models in prediction.

UVRR data analysis
Multivariate curve resolution -alternating least squares (MCR-ALS) was employed due to the fact that the UVRR spectra of all three analytes are highly similar and do not possess characteristic peaks. [1] In MCR-ALS, the UVRR spectra were first baseline corrected, smoothed using wavelet smoothing and then normalised so that the sum of squares of each spectrum equals 1. This is to account for the decreasing signal strength and relatively increasing noise (attributed to the removal of aliquots for HPLC analysis (and specifically for biotransformation 1 and 2 the inherent volatility of the solution).
The spectra of all the monitored time points of a single reaction were then augmented to form a t×n data matrix X where t is the number of time points monitored and n is the number of wavenumbers recorded. Multivariate curve resolution using alternating least squares (MCR-ALS) was performed to deconvolve X into a product of two sub matrices C and S where C contains the "profiles" of the change in the concentrations the reactants during the reaction while S is the matrix storing the resolved spectra the reactants. Non-negativity constraint was applied to both concentration profile C and spectral profile S and each deconvolved pure spectrum had a unit norm (i.e. the sum of squares of each spectrum equals 1). For the time points when HPLC measures were also taken, a linear regression model was built between the concentration profile of each reactant in C and the corresponding concentration measured by HPLC. The regression model was then applied to the whole concentration profile to get UVRR calibrated concentrations of the reactant over the whole monitored period of the reaction.

Figure S1
Annotated instrument set-up to monitor biotransformation using UV resonance Raman spectroscopy. The x40 UVRR objective was carefully focused onto the reaction mixture (10 mL scale), with 100% power on sample (~ 0.2mW at sampling point). Throughout the course of the reaction, samples for HPLC analysis were removed as well as UVRR data collected (20 s spectral acquisitions). The solution was constantly stirred using the magnetic stirrer plate beneath the stage and the magnetic stirrer bar in the reaction vessel.

Photo-degradation of sample:
We initially tried to photodegrade our sample so we were aware of the spectral changes to expect if the sample started to degrade or 'burn'. Using biotransformation 3 as an example, we tried to photo-degrade the starting material, xanthine (in solution). After 30 min of constant interrogation of the laser on the sample, there were no changes in the spectra and thus no photodamage. Our only observation was the evaporation of solvent meaning as time increased, the sample point became out of focus and hence the spectra became nosier. We next looked at photo-degrading the corresponding solid sample, and after 45 min (vastly exceeding the total reaction time) we noticed broadening of peaks around 1550 cm -1 region (from C-C), similar to spectra of graphitic carbon-type species and charcoal, thus indicating 'burning' of the sample. These observations are in agreement with the literature. indicates C-C presence suggesting photo-degradation.

Bathochromic shifts of XO analytes
Photo-degrading the sample led to an interesting observation regarding the UVRR spectra of xanthine. Notably, there were significant changes in spectral band positions and intensity between the solid spectra from xanthine and that in solution. Changes in environmental conditions (e.g. temperature, solvent, pH etc.) can lead to changes in some vibrational frequencies especially as for some molecules, these functional groups will interact differently with the solvent (through H-bonding, as well as acidic/basic ions). This led to a pH investigation looking at the associated UV-Vis absorption of xanthine and UVRR spectra, with the results indicating the observation of a bathochromic shift on increasing pH. In acidic medium (i.e. low pH), the nitrogen atom is free to lose its lone pair of electrons thus decreasing the delocalisation in the ring, leading to a decrease in conjugation, as observed in Figure 3.7, consequently, the molecule becomes less energetic. As the energy needed for excitation is higher, there is a shift to absorbing at shorter wavelengths. Conversely, in alkaline medium (i.e. high pH), the opposite phenomenon occurs, and the oxygens lone pairs increase delocalisation and conjugation to the ring, (as observed in Figure 3.7) meaning the molecule becomes more energetic, so less energy is required for excitation thus shifting to absorbing at longer wavelengths. This is supported by the increase in wavelength absorbance from λ max 267 to 272 nm when going from pH 3.6 to 9.2. Figure 3.7 c and d) highlight key peaks that are affected by changing the pH of the solution. A similar, less pronounced effect was observed for hypoxanthine and uric acid (see Figure 3.8 and 3.9). [5] Raman Shift ( Figure S3. a) Average UVRR spectra (n=5) and b) UV-Vis absorbance spectra of xanthine at various pH: 3.6, 5.0, 7.0, 7.6, 8.0, and 9.2. A bathochromic shift in the UV-Vis absorption spectra was observed on increasing the pH from pH 3.6 to 9.2, consequently, the UVRR spectra of xanthine changed due to it being in different ionisation states. c) The intensity difference of the key peaks that change and d) a plot of the centre of the peaks that shift on increasing the pH of the solution from pH 3.6 to 9.2.  Figure S4. a) Average UVRR spectra (n=5) and b) UV-Vis absorbance spectra of hypoxanthine at various pH: 3.6, 5.0, 7.0, 7.6, 8.0, and 9.2. A bathochromic shift in the UV-Vis absorption spectra was observed on increasing the pH from pH 3.6 to 9.2, consequently, the UVRR spectra of hypoxanthine changed due to being in different ionisation states. c) The intensity difference of the key peak that changes and d) a plot of the centre of peak that shifts on increasing the pH of the solution from pH 3.6 to 9.2.  Figure S5. a) Average UVRR spectra (n=5) and b) UV-Vis absorbance spectra of uric acid at various pH: 3.6, 5.0, 7.0, 7.6, 8.0, and 9.2. A bathochromic shift in the UV-Vis absorption spectra was observed on increasing the pH from pH 3.6 to 9.2, consequently, the UVRR spectra of uric acid changed due to being in different ionisation states. c) The intensity difference of the key peaks that change and d) a plot of the centre of peak that shifts on increasing the pH of the solution from pH 3.6 to 9.2.

Raw UVRR spectra
Baseline correction of UVRR data

Normalisation of UVRR data
Smoothing of UVRR data using wavelet smoothing

MCR-ALS
Deconvolved spectra of individual components i.e. analytes Figure S7. An MCR-ALS model was applied to the UVRR data where it successfully deconvolved spectra into its pure components a) benzonitrile (substrate) and b) benzamide (product) as shown for biotransformation 1.  Figure S9. A MCR-ALS model was applied to the UVRR data where it successfully deconvolved spectra into its pure components a) p-tolunitrile (green) and b) p-toluamide (orange) as shown for biotransformation 2 c) Shows the reaction dynamics from real-time UVRR measurements (denoted by outlined symbols) and off-line HPLC data (denoted by solid symbols) as a function of time for the conversion of p-tolunitrile to p-toluamide. UVRR spectra were obtained for 20 s with baseline correction, normalisation and smoothing applied (as detailed in 'Materials and methods: data processing').

Figure S10
An MCR-ALS model was applied to the UVRR data where it successfully deconvolved spectra into its pure components a) xanthine (substrate) and b) uric acid (product) as shown for biotransformation 3.  Table S4. A summary of the R 2 co-efficients across all three replicates for the conversion of hypoxanthine to xanthine to uric acid (biotransformation 4).

Replicate
Hypoxanthine R  Figure S12. a) Overall schematic illustrating the two known pathways to catalyse the conversion of nitrile containing compounds into their corresponding carboxylic acid, either in a single step (nitrilase) or a multicomponent process (nitrile hydratase and amidase). Plots bd) show average UVRR spectra (n=5) of each analyte: b) pure spectra of benzonitrile, benzamide and benzoic acid (12.5 mM, pH 7.2) c) pure spectra of 3-pyridinecarbonitrile, nicotinamide (vitamin B 3 ) and nicotinic acid (25 mM, pH 7.2) and d) pure spectra of pyrazinecarbonitrile and pyrazinamide (anti-tuberculosis drug) (25 mM, pH 7.2). Characteristic peaks are annotated. UVRR spectra were obtained for 20 s with baseline correction, normalisation and smoothing applied (as detailed in 'Materials and methods: data processing').